Financial integration of the European transition

ASERS

Journal of Advanced Studies in Finance

J

ournal of Advanced Studies in Finance

Biannually Volume V Issue 1(9) Summer 2014 ISSN 2068 – 8393 Journal DOI http://dx.doi.org/10.14505/jasf 2

Summer Journal2014 of Advanced Studies in Finance Volume V Issue 1(9) Editor in Chief Laura GAVRILĂ (ŞTEFĂNESCU) Spiru Haret University and Association for Sustainable Education Research and Science, Romania

Contents: A tale of two Euro-zones. Banks’ funding, sovereign risk

1 & unconventional monetary policies Nicolas FULLI-LEMAIRE

Co-Editor Rajmund MIRDALA Technical University of Kosice, Slovak Republic

Editorial Advisory Board Mădălina Constantinescu Spiru Haret University and Association for Sustainable Education Research and Science, Romania

Microeconomic determinants of losses in financial

2 institutions during the crisis Alessandra CEPPARULO Antonio FORTE

Does payment method of mergers and acquisitions Chengkui YE Hao YUAN

4

Lean Hooi Hooi Universiti Sains Malaysia, Malaysia Terence Hung United International College, Hong Kong

Ivan Kitov Russian Academy of Sciences, Russia Piotr Misztal The Jan Kochanowski University in Kielce, Faculty of Management and Administration, Poland

…37

The GINI coefficient. Decomposition and overlapping Augustine C. ARIZE, Paraskevas BAKAREZOS, Krishna M. KASIBHATLA, John MALINDRETOS, Alex PANAYIDES

Effect of inlation on nominal and real stock returns. A Andrey KUDRYAVTSEV, Eyal LEVAV Shosh SHAHRABANI

6

Cross-border listings and price discovery: evidence from Chinese companies triple-listed in Shanghai, Hong Kong and New York Xiaoou YANG Peng KUN

Daniel Stavarek Silesian University, Czech Republic Laura Ungureanu Spiru Haret University and Association for Sustainable Education Research and Science, Romania

…56

…66

The carry trade on the Euro and the European stock

7 market

Fabio PARLAPIANO

…103

Andreea Pascucci University of Bologna, Italy Rachel Price-Kreitz Ecole de Management de Strasbourg, France

…47

5 behavioral view

Renata Karkowska Faculty of Management, University of Warsaw, Poland Kosta Josifidis University of Novi Sad, Serbia

…27

3 matter? An examination of the medical industry in China

Rosaria Rita Canale University of Naples Parthenope, Italy Francesco P. Esposito Allied Irish Bank, Group Market Risk Management

…5

ASERS Publishing http://www.asers.eu/asers-publishing ISSN 2068-8393 Journal's Issue DOI: http://dx.doi.org/10.14505/jasf.v5.1(9).0

2

Journal of Advanced Studies in Finance Call

for Papers

Volume V, Issue 2(10), Winter 2014


The Journal aims to publish empirical or theoretical articles which make significant contributions in all areas of finance, such as: asset pricing, corporate finance, banking and market microstructure, but also newly developing fields such as law and finance, behavioral finance and experimental finance. The Journal will serve as a focal point of communication and debates for its contributors for the better dissemination of information and knowledge on a global scale. The primary aim of the Journal has been and remains the provision of a forum for the dissemination of a variety of international issues, empirical research and other matters of interest to researchers and practitioners in a diversity of subjects linked to the broad theme of finance. The Editor in Chief would like to invite submissions for the 5th Volume, Issue 2(10), Winter 2014 of the Journal of Advanced Studies in Finance (JASF). Journal of Advanced Studies in Finance is indexed in EconLit, RePEC, CABELL's Directories, EBSCO, ProQuest, CEEOL databases. All papers will first be considered by the Editors for general relevance, originality and significance. If accepted for review, papers will then be subject to double blind peer review. Deadline for Submission:

20th November, 2014

Expected Publication Date:

December, 2014

Web:

www.asers.eu/journals/jasf/

E-mails:

[email protected]

4

[email protected]

Volume V Issue 1(9) DOI: http://dx.doi.org/10.14505/jasf.v5.1(9).01

A TALE OF TWO EURO-ZONES. BANKS’ FUNDING, SOVEREIGN RISK AND UNCONVENTIONAL MONETARY POLICIES Nicolas FULLI-LEMAIRE1 Amundi Asset Management, Paris II University, France [email protected] Suggested Citation: Fulli-Lemaire, N. (2014). A tale of two Euro-zones. Banks’ funding, sovereign, risk and unconventional monetary policies, Journal of Advanced Studies in Finance, (Volume V, Summer), 1(9):5-27. Doi:10.14505/jasf.v5.1(9).01. Available from: http://www.asers.eu/journals/jasf/curent-issue. Article’s History: Received November, 2013; Revised January, 2014; Accepted July, 2014. 2014. ASERS Publishing. All rights reserved.

Abstract: The admission by the Greek government on October 18, 2009, of large-scale accounting fraud in its national accounts sparked an unprecedented sovereign debt crisis that rapidly spread to the Eurozone’s weakest member states. As the crisis increasingly drove a wedge between a seemingly resilient Eurozone core and its faltering periphery, its first collateral victims were the private banks of the hardest-hit sovereigns. They were rapidly followed by the rest of the Eurozone’s banks as a result of their large exposure to not only their home country’s sovereign debt, but also to the debt securities of other member states. Measuring each bank’s precise exposure to every sovereign issuer became a key issue for credit analysis in the attempt to assess the potential impact of a selective sovereign default if worse came to worst. Yet, finding that information in a timely manner is hardly an easy task, as banks are not required to disclose it. Building on the efficient market hypothesis in the presence of informed traders, we tested the sensitivity of each of the largest Eurozone private banks’ CDSs to sovereign CDSs using a simple autoregressive model estimated by time-series regressions and panel regressions, comparing the results to news releases to assess its reliability. Eventually, we used the Oaxaca Blinder decomposition to measure whether the unconventional monetary policies, namely the LTRO and the OMT, that the ECB has implemented to stem the crisis have helped banks directly or whether banks were actually helped by the reduction in sovereign CDS spreads. Keywords: private banks, central banks, sovereign debt risk, OMT, LTRO, non-conventional monetary policies, Eurozone’s sovereign debt crisis, Oaxaca-blinder decomposition. JEL classification: C58, D82, E52, G01, G14, G15, G21, G24, N14. 1. Introduction Had the European Central Bank (ECB) not staged a massive intervention, the tenth anniversary of the Eurozone could easily have coincided with the implosion of the single currency. Few economists still doubt the actions of the ECB were both appropriate and sufficient. In fact, the ECB’s announcement of massive long term refinancing operations in December 2011 is the most likely cause of the ensuing significant decrease of sovereign credit default swap (CDS) spreads, which had spiked to unprecedented levels. This signaled, if not the end of the crisis, then at least the end of its most acute phase. Over the course of the following months, further easing of collateral requirements and the announcement of a whole new set of unconventional measures on secondary sovereign debt markets firmly reinstated the belief that the euro had truly been made “irreversible”, as stated by ECB President Mario Draghi when providing details on the Outright Monetary Transaction (OMT) mechanism later that year (Draghi, 2012). Yet the apparent resolution of the sovereign debt scare propelled another interesting debate into the public sphere: had the ECB’s interventions been more helpful to the Treasuries of peripheral Eurozone member states or to private banks holding vast amounts of government bonds. In other words, were taxpayers again at This document presents the ideas and the views of the author only and does not reflect Amundi’s opinion in any way. It does not constitute investment advice and is for information purpose only. 1

5


risk of bailing out large financial institutions less than two years after the whole sector had been either recapitalized or fully nationalized by states and therefore by taxpayers’ money? As austerity bites down on some of the hardest-hit countries in the Eurozone’s periphery, this interesting economic discussion gained a whole new dimension in the public sphere. This paper investigates this issue. Banks are affected by a deterioration in the creditworthiness of their home countries in more than one way, as evidenced by (Panetta, et al., 2011): the first one is that corporate CDS are mostly traded as spreads on their home country sovereign debt, thus upward movement on the base CDS generally affects the spreads of all the securities based on them. Secondly, states offer implicit guaranties for the bank creditors (“too big to fail” or “too interconnected too fail”) as the sequence of crises in the last decade starkly reminded us. Thus, an apparent decrease in the creditworthiness of the state is interpreted as a reduction in the value of the insurance, thus also a decrease in the creditworthiness of the bank incorporated in that country. And it is possible that, as we recently witnessed in Cyprus with the failure of Laiki Bank and the bail-in/bail-out of Bank of Cyprus, some banks hoard vast quantities of government bonds from their home state (Zoli, 2013) and (Bofondi, Carpinelli, & Sette, 2012), or from abroad (Greece in the case of Cypriot banks) for which the value decreases as their mark-to-market values adjust to movements in the discount rates or in the expected recovery rates. In the most extreme cases, debt renegotiations such as those for Greece (Petrakis & Christie, 2012) can directly imperil the principal of those securities (even if it usually implies a narrowing of the CDS spread and thus a reduction of the associated discount factor). To this day, there is no requirement for private banks to disclose their exposure to foreign sovereign bonds. Information about bank holdings of various government bonds is thus the result of either voluntarily disclosure or exceptional disclosures, as occurred during the EU-wide bank stress tests sponsored by the European Banking Authority (EBA), European Central Bank (ECB) and national supervisory bodies in July 2010 and July 2011, or during the capital exercises in December 2011 (Bischof & Daske, 2012). Assessing the sovereign risk embedded in individual banks is thus a difficult exercise. Yet there is probably a large pool of informed traders dealing in CDS markets, which should thus signal to it the relative sensitivity of individual banks to a given sovereign credit risk. Following the work of (Maloney & Mulherin, 2003) on price formation in the presence of informed traders, we investigate this intuition using a very simple autoregressive (AR) model to test the market-implied sensitivities of banks’ CDSs to sovereign CDSs. Out of simplicity considerations, in this study we chose to measure bank risk and sovereign risk using CDS spreads as in (Chiaramontea & Casu, 2012), even if there is an active academic debate on whether bond and CDS markets share the same information as in (Arce, Mayordomo, & Peña, 2012), (O’Kane, 2012) or (Palladini & Portes, 2011). However, in this paper we chose to focus on the largest European banks, which are precisely those that have the most liquid CDSs according to (Markit). We should thus expect liquidity problems to have a lesser distortionary impact on our sensitivity measurements. This paper addresses two main research questions: firstly, do informed traders enable us to extract sovereign risk sensitivities for individual banks from market quotations, thus giving us hints regarding their real exposure to individual sovereign risks? Secondly, building upon the sensitivity analysis conducted in the first step, can we break down the impact of unconventional monetary policies implemented by the ECB into, on the one hand, the “pure impact” of the ECB’s measures (i.e. independent from sovereign risk considerations) and, on the other hand, the relaxation of their funding stress attributable to relaxation in sovereign funding? In other words, did the ECB’s action help banks directly or indirectly by relaxing sovereign funding conditions? To address the first research question, we calibrated a conventional autoregressive model of order one AR(1) in the manner of that is usually used for sovereign CDS spreads, like (Sgherri & Zoli, 2009) or (Schuknecht, Hagen, & Wolswijk, 2011), assuming a relatively strong stationary hypothesis unlike the nonstationary co-integrated panel model used by (Giordano, Pericoli, & Tommasino, 2012). The calibration of the autoregressive parameter upholds this hypothesis, as we can consistently reject the integration hypothesis – except for the Greek banks without the inclusion of the Greek sovereign CDS – thus leading us to believe that we can assume the data to be sufficiently stationary for the purpose of this paper, consistently with prior literature on the subject of determinants of CDS spreads. We then compare the results of the calibration, namely the parameters that are statistically significant with both information that was public at the time regarding sovereign asset holdings and a map of their known wholly owned foreign retail banking subsidiaries inside the Eurozone. Overall, the results we get seem consistent with both controls. Using the model calibrated previously, we then proceed in the manner of (Giordano, Pericoli, & Tommasino, 2012) following the model established by (Eichengreen & Mody, 2000) to disaggregate the role of “Pure”, “Shift” and “Wake-up-Call” contagions in emerging economies. The “Shift” contagions arise from changes

6

Volume V Issue 1(9) in the level of fundamental explanatory variables assuming constant sensitivities. The “Wake-up-Call” contagion is due to changes in the sensitivity towards explanatory exogenous variables of the model. Lastly, the residual “Pure” contagion cannot be attributed in any way to changes in the level of or sensitivities to the exogenous variables in the model. In our case, we know that the ECB’s actions acted as a “reverse contagion”, but we partially ignore the channel through which it operated, which brings us back to our research question regarding whether the observed reduction in bank funding costs immediately after the announcement of the LTRO and OMT operations is attributable to either “Pure” or “Wake-up-Call’ reverse contagion, or whether it should be attributed to “Shift” reverse contagion. The first two types of effects would uphold the belief that unconventional monetary policies had a direct impact on bank funding stress, while the last one would tend to uphold the indirect channel hypothesis. The calibration of the modified model to perform the Oaxaca-Blinder decomposition as in (Giordano, Pericoli, & Tommasino, 2012) requires a very large number of parameters to be estimated. Achieving statistical significance with the Matlab routines used, we had to include a minimum of 300 trading days (roughly a year) before and after the LTRO event. This had two major unintended consequences: firstly, the 600 trading-day window was too large to establish the effect of the LTRO, which seemed to have been much shorted according to a basic analysis of CDS data. Secondly, our dataset does not extend a year after the announcement of the OMT, thus making it difficult for us to perform the computation on this event. To circumvent partially this time-window and data-availability problem, in a second step we proceed to a pooled OLS estimation of our model, which allowed us to focus on a one- to three-month data sample window. This approach has its own caveats: firstly, for short horizons, disentangling the noise of daily data from fundamental CDS movements seems more arduous, thus leading to weak parameter estimates except the firmspecific effects. Secondly, by using a homogeneous assumption regarding the value of the sensitivities of bank CDSs to sovereign CDSs and financial market proxies, we lose greatly in terms of model precision, thus also in terms of parameter determination, which could be detrimental to the strength of our findings. The paper is constructed as following: the first part presents the dataset used to perform the analysis, the second one presents the model and estimates it in Time Series format. Lastly, a third part addresses the paneldata approach to compute Oaxaca-Blinder decomposition over a shorter horizon. A discussion of the main results concludes. 2. Data 2.1. Bank credit default Swap We tried to achieve the most comprehensive Eurozone bank dataset possible. Since we are investigating the role of international linkages, in particular the reverse contagion from peripheral Eurozone sovereigns to core private banks, we aimed to include the most systematically important financial institutions (SIFIs): those institutions are particularly active on the global financial markets and have liquid CDSs and equity stock prices. To avoid selection bias, we referred to the EBA’s assessment of the Eurozone SIFIs (EBA, 2011). The downsizing of investment banks throughout Europe has lead a significant proportion of those banks to drop off the list, but since the current sovereign crisis and resultant banking crisis was caused by investment decisions taken prior to those restructuring events, we chose to take the list of banks that were significant at that time. Out of the EBA’s list of 30 financial institutions, we immediately excluded the seven non-Eurozone incorporated banks. Of the remaining 23 banks, we further excluded the three public or non-listed banks present in the sample: the Bayerishe Landesbank (Germany), the Caixa General de Depositos (Portugal) and Rabobank (Netherlands). Major debt restructuring, government guarantees or outright default lead to a lack of data available and forced us to exclude four other banks: Dexia (France), Anglo Irish Bank (Ireland), Caixa de Barcelona (Spain) and Eurobank EFG (Greece) had to be removed. Reflecting its national SIFIs status, the Banco Espirito Santo (Portugal) was added to the sample. We therefore have a sample of 16 large, private, Eurozone incorporated banks for which daily data is available on our entire test period ranging from January 2008 to April 2013. Since Greece technically defaulted in February 2012, we were not able to include the Greek SIFIs in our sample. We nonetheless created a supplementary restricted sample from November 2008 to February 2012, which includes both Alpha Bank (in place of Eurobank EFG) and National Bank. The five-year Senior CDS premia for all banks in the sample were retrieved in daily close format from Thomson Reuters DataStream. Most of these CDS were for the “Modified Restructuring” (MM) type of credit events except for Intesa Sanpaolo and ING Bank which were for “Full Restructure” (CR) and Banco Santander which was for “No Restructure” (XR).

7


Table 1. Bank CDS's characteristics Shorthand ERS RAI

Full Bank Name ERSTE GROUP BANK AG RAIF ZNTRLBK OSTER AG

Home Country #Obs. CDS Type Austria (AT) 1369 SNR MM 5Y E - CDS PREM. MID (AT) 1369 SNR MM 5Y E - CDS PREM. MID

KBC

KBC BANK

Belgium

(BE)

1369 SNR MM 5Y E - CDS PREM. MID

DEU COM

DEUTSCHE BANK AG COMMERZBANK AG

Germany

(DE) (DE)

1369 SNR MM 5Y E - CDS PREM. MID 1369 SNR MM 5Y E - CDS PREM. MID

BSN BBV

BANCO SANTANDER, SA BBV ARGENTARIA SA

Spain

(SP) (SP)

1369 SNR XR 5Y E - CDS PREM. MID 1369 SNR MM 5Y E - CDS PREM. MID

SOC BNP CRE

SOCIETE GENERALE BNP PARIBAS CREDIT AGRICOLE SA

France

(FR) (FR) (FR)

1369 SNR MM 5Y E - CDS PREM. MID 1369 SNR MM 5Y E - CDS PREM. MID 1369 SNR MM 5Y E - CDS PREM. MID

NAT ALP

NAT BK OF GREECE SA ALPHA BANK A.E.

Greece

(GR) (GR)

852 852

BOI

THE GOVERNOR AND CO BOI

Ireland

(IE)

1369 SNR MM 5Y E - CDS PREM. MID

INT UNI

INTESA SANPAOLO SPA UNICREDITO ITALIANO SPA

Italy

(IT) (IT)

1369 SNR CR 5Y E - CDS PREM. MID 1369 SNR MM 5Y E - CDS PREM. MID

ING

ING BANK N.V.

Netherlands (NL)

1369 SNR CR 5Y E - CDS PREM. MID

BCP BES

BANCO COMR PORTUGUES SA Portugal BANCO ESPIRITO SANTO SA

(PT) (PT)

SNR MM 5Y E - CDS PREM. MID SNR MM 5Y E - CDS PREM. MID

1369 SNR MM 5Y E - CDS PREM. MID 1369 SNR MM 5Y E - CDS PREM. MID

3000 ERS RAI 2500

KBC DEU

COM BSN

2000

BBV SOC BNP

1500

CRE

NAT ALP BOI

1000

INT UNI ING BCP

500

BES

0 2008

2009

2010

2011

2012

2013

Source: Datastream.

Figure 1. Five-Year credit default swap for the Eurozone’s large private banks 2.2. Explanatory variables: sovereign credit default swap Since our model tries to estimate the impact of sovereign credit deteriorations throughout the Eurozone on bank funding conditions, and in particular to disentangle the LTRO’s impact on their refinancing, the main explanatory variable of our model consists in the sovereign CDS spreads of the main Eurozone economies. We again chose the five-year sovereign CDS spreads of a selected group of sovereigns. Out of the current Eurozone 17member club, we selected 11 by excluding Finland, Luxembourg, Cyprus, Slovenia and Malta because of their very small sovereign debt and Slovakia and Estonia because they joined the club only in the first half of the sample period (2011). It thus includes the five “core” economies (Austria, Belgium, Germany, France and the Netherlands) and the five “peripheral” countries (Portugal, Ireland, Italy, Greece and Spain) that have come to be known pejoratively as Europe’s “PIIGS”. We thus collected again from Thomson Reuters DataStream the daily closes of the five-year CDS for all countries except Greece from January 2008 to April 2013. CDSs for Greece stopped trading on February 22, 2012, when it restructured its sovereign debt.

8

Volume V Issue 1(9) 1600

16 000

1400

14 000

1200

12 000

1000

10 000

800

8 000

600

6 000

400

4 000

200

2 000

AT

BE DE SP FR IE IT NL PT

0 2008

GR (RHS)

0 2009

2010

2011

2012

2013

Source: Datastream.

Figure 2. Eurozone's sovereign credit default swap 2.3. Explanatory and control variables To control for firm-specific characteristics, we included in our analysis two control variables: its Share Price and its Senior Debt Long Term Rating. Daily closes of the share prices were downloaded from Thomson Reuters DataStream and so were Standard & Poor’s long-term rating. To include this variable in the regression, we transformed the letter indicators into numeric input by linearly linking from D to AAA numbers from 1 to 22. To account for the rating outlook, we enhanced the rating by ¼ point for a “Positive Outlook” and by a ½ point for a “Positive Watch”. Symmetrically, ratings were reduced by a ¼ point for “Negative Outlook” and by a ½ point for “Negative Watch”. 120

O:ERS

100

O:RAI B:KB D:DBK

D:CBK

80

E:SCH E:BBVA F:BNP 60

F:SGE F:CRDA

G:ETE G:PIST

40

BKIR I:UCG I:ISP H:ING

20

P:BCP P:BES 0 2008

2009

2010

2011

2012

2013

22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 2008

AAA AA+ AA AAA+ A ABBB+ BBB BBBBB+ BB BBB+ B BCCC+ CCC CCCCC C SD D 2009

2010

2011

2012

ERS RAI KBC DEU COM

BSN BBV SOC BNP CRE NAT ALP BOI

INT UNI ING BCP BES

2013

Source: Bloomberg and Standard & Poor's .

Source: Datastream.

Figure 3. Bank’s share prices and S&P's long term ratings As (Moody, 2009) showed, the CDS spreads of both corporates and sovereigns in the pre-crisis era were determined mostly by global risk-aversion factors such as the VIX index (Chicago CBOE index on implied volatility on S&P500 Options), itself closely influencing credit markets. To focus explicitly on the Eurozone, we chose to control for Global Risk Aversion using the VSTOXX index (STOXX index of implied volatility on EUROSTOXX 50 options). To account for the global credit risk-aversion factor, we included the Meryl Lynch Euro BBB Corporate Government Option Adjusted Spread (Er40_GOAS) in our analysis. To account for the state of the Eurozone’s interbank market, we included the Euro’s BOR-OIS Spread (Hull White), which is considered a good proxy. We obtained it by subtracting the three-month Euro-OIS from the three-month EURIBOR rate. All of these control variables were retrieved in daily close format from Thomson Reuters DataStream.

9


9%

90%

8%

80%

7%

70%

6%

60%

5%

50%

4%

40%

3%

30%

2%

20%

1%

10%

0% 2008

0% 2009 2010 ML_BBB Corp GOAS

2011 Eur_BOR-OIS

2012 2013 VSTOXXI (RHS)

Source: Datastream.

Figure 4. Global risk aversion, credit risk and Interbank Market confidence proxies Eventually, we introduced our three specific-event dummy variables: the first one being the currently commonly accepted beginning of the Eurozone’s sovereign crisis, that is, the official reckoning by the Greek government of massive accounting frauds on October 18, 2009. The second event-specific dummy we introduced was the announcement of the launch of LTROs by the ECB on December 8, 2011. Thirdly, we added a dummy to account for the ECB’s change of policy announced on August 2, 2012, regarding OMT transactions. Unreported tests on different event-specific dummies proved inconclusive: the effect of LTROs might have been enhanced by including the two effective operation dates (December 20, 2011, and February 28, 2012) or the Greek technical sovereign default (or debt renegotiation) on March 9, 2012. The closeness of all those events leads to a restricted choice of only three: the G (Greek dummy), L (LTRO dummy) and the O (OMT dummy). All of the dummy variables are worth zero before the triggering event and 1 after it. 3. Time series regressions 3.1. Time series model In accordance with the current financial literature, we wish to fit on our data the following AR(1) model: ( ⏟

) ( )

( ⏟

)

( ⏟

)

( ⏟

)

(

)

⏟

We ran two different regressions according to our sample period: in the first one we excluded Greek banks and the Greek sovereign from our dataset and we then ran a specific regression for the Greek banks on the restricted sample previously exposed, trivially omitting the OMT’s dummy variable. In order for our model to be stable (stationary), we need the estimated coefficients to respect the boundary condition: Since Greek banks’ CDSs were available in a much larger sample than the Greek sovereign’s, we could have included Greek banks but excluded the Greek sovereign on a much larger dataset. Yet, unsurprisingly, the specificity of the Greek sovereign CDS is so large that if we omitted it, the autoregressive parameter would reach a value significantly higher than 1 (reassuringly for our model). We thus excluded both from our main study. We thus fitted on the Greek banks’ CDS data the following reduced model without the dummy variable since our sample does not extend up to the triggering event.

10

Volume V Issue 1(9)

3.2. Time series results First of all, in order to be able to use our model, we must conduct a test on our boundary condition for our autoregressive parameters: since our estimated coefficients are close to one, we clearly have a near-integrated process. We ran a Fisher test on whether the estimated coefficients are significantly different from one. The results for both of our datasets are presented in Tables 6, 7 and 8. Reassuringly, the estimated coefficients are statistically always significantly different from 1 for all banks tested in both of our datasets. We can thus apply our model to study the impact of the LTRO on the refinancing of private banks. The first obvious comment that arises from the analysis of the longer dataset excluding Greek banks is that the dummy coefficient is statistically significant for all the banks present in the sample at the 99% level and negative. This first result is in itself not surprising considering the importance of the impact of this single triggering event. The second conclusion we can draw from this regression is that for all the banks incorporated in one of the four-week sovereigns included in the sample, the home-sovereign-CDS parameter was always significant at the 99% level and positive. For SOC (French), the Belgian and both of the Austrian banks, it was also positive and highly significant. For the Dutch, German and two of the French banks (BNP & CRE) it was not significantly different from zero. The most probable explanation for these phenomena is that, on the one hand, the deterioration of the perception of the creditworthiness of the peripheral sovereigns led to a contagion to the banks incorporated on their territories, thus linking the CDSs of sovereigns and private banks. On the other hand, banks incorporated in countries where creditworthiness was not significantly altered during the “sovereign crisis”, such as France or Germany, but where private banks’ financial health were severely threatened by the rapid deterioration of the Eurozone’s peripheral countries, gained little from the stability of their home sovereign but were negatively impacted by the deterioration of the peripheral sovereigns. For example, French, German and Dutch banks all have a significant and positive coefficient on at least Italy (some also have significantly positive coefficients on Spain, Portugal or Ireland). This is clearly not the case for the Belgian or Austrian banks. The answer to this question must therefore lie in the composition of their books and in their respective footprints in core versus peripheral countries. While the Greek crisis dummy had an unquestionable effect on all of the Eurozone banks in the sample, thus highlighting its systemic nature, this is clearly not the case for the dummy: on the one hand, the LTRO’s dummy coefficient for the Austrian, Belgian and Portuguese banks was not significant and it was barely so (90% significance only) for the Irish banks. On the other hand, the LTRO’s dummy coefficient for Spanish, Italian, Dutch, and French banks (albeit with a lower 95% significance for BNP) was highly significant. As for German banks, interestingly so, the coefficient is not significant for DEU, while it is highly significant for COM. These results should really come as no surprise: the banks whose head offices were incorporated in countries where sovereigns were the worst affected by the crisis, like Ireland or Portugal (and Greece), and where the banking sector was already either receiving massive international aid, like the Greek bank’s Hellenic Financial Stability Fund, or had already been recapitalized by their home-country states therefore had little to gain from these refinancing operations. Meanwhile, banks incorporated in countries where sovereigns were slightly less affected by the crisis and which were still able to issue without guarantees on the market, albeit with some restrictions, like Spain or Italy, were the ideal candidates to tap the ECB’s new facility and thus enhance their creditworthiness as a result. Importantly, the ECB’s facility was aimed at bailing out banks not only in the periphery of the Eurozone, but also in its core: French, Dutch and German banks all possibly had good reasons to use the ECB’s facility. Indeed, the motives for using the ECB’s facility could be twofold: firstly, by using the ECB’s facility as a credit line to invest in government bonds yielding a positive carry (typically Italian BTPs) and secondly to help refinance banks that were put in a precarious situation as a result of the sovereign crisis due to their exposure to either (or both) their investment portfolio earning a negative carry as a result of surging funding costs or because of difficulties refinancing their branches operating in the periphery. Spanish and Italian banks have benefited massively from the LTROs (Reuters, 2012), and it is interesting to note that, for those countries in particular, it is known that a significant fraction of those funds went directly into government securities carrying a significant positive carry (Reuters, 2012). INT’s Chairman Andrea Beltratti was explicit: “The new funds, which come with a 1% interest rate, will be used in part “for a profitable trading strategy regarding Italian government bonds,” (Enrich, 2012). While COM seems to have clearly benefited from the LTRO, its national peer DEU did not. The most probable answer is that COM and DEU were in very different financial positions at the time of the LTRO announcement: while the latter seemed to be in a strong financial position, the former seemed closer to distress,

11


as attested by the fact that barely a few days before the first LTRO round, COM was facing outright nationalization (Wiesmann & Wilson, 2011) as a result of a €5.3bn capital gap identified a few days before by regulators during the EBA’s 2011 stress tests (Jenkins & Atkins, European banks have €115bn shortfall, 2012). Furthermore, as DEU’s then-Chairman Ackermann put it: “The fact that we have never taken any money from the government has made us, from a reputational point of view, so attractive to so many clients in the world that we would be very reluctant to give that up”. He also rejected the ECB-sponsored sovereign carry trade “I’m normally not a friend of carry trades, and I don’t think that we would borrow money to buy sovereign risks even if there is an attractive spread.” (Comfort & Kirchfeld, 2012). Yet, those arguments failed to prevent DEU from participating in the second auction (Jenkins, 2013). Why has the LTRO had no impact on the Austrian and Belgian banks? The basic explanation should lie in the fact that both of the banks in the sample made very limited use of the new facility: RAI didn’t participate in the first round and tapped the facility for a very limited amount in the second round (Global Banking News, 2012). ERS seemed to have participated in both LTRO rounds but also took only a limited amount from it (Dow Jones Newswires Reporters, 2012). The most probable explanation lies in the fact that, though highly internationally diversified, Austrian banks have limited exposure to peripheral Eurozone countries (most of their foreign exposure is in the CEE/CIS region) as is attested by (Moody's, 2013). KBC did participate in both of the LTRO rounds, yet indicated that most of it had been to refinance its Irish subsidiary. Moreover, it used Irish collateral for the operation (KBC, 2013). That might explain why its global creditworthiness at the group level remained pretty much unchanged, thus yielding an insignificant LTRO dummy coefficient. It is particularly interesting to note that the significance of sensitivity to sovereign CDSs of individual banks in Tables 6 and 7 closely reflects the mapping of their foreign subsidiaries. Table 2 was constructed using the latest annual report of each bank to pinpoint the localization of their wholly-owned retail-banking subsidiaries. As a caveat, we must mention that the definition might not be entirely consistent EU-wise and some significant exposure to a given country might have gone unnoticed: for example, some banks seem to have large wholesale corporate lending activities in specific countries where they do not necessarily have a large retail footprint, which are thus harder to track compared to retail activities. Table 2. Matrix of foreign wholly-owned retail subsidiaries AT BE DE ES FR GR IE IT NL PT

ERS RAI KBC

DEU COM BSN

BBV

SOC

BNP

Yes Yes Yes Yes -

Yes -

Yes -

Yes Yes * -

Yes Yes Yes Yes Yes Yes ** Yes Yes Yes -

Yes -

*: Untill 14-Dec-2012.

Yes Yes

CRE

NAT

ALP

BOI

INT

UNI

ING

BCP

BES

Yes -

Yes -

Yes -

Yes Yes Yes -

Yes Yes Yes -

Yes Yes

Yes Yes Yes

**: Untill 1-Feb-2013.

Importantly, those findings are consistent with the BIS report on the “Euro area bank use of ECB facilities” (BIS, 2012). It appeared that the largest users of the two LTRO facilities provided by the ECB were Italian and Spanish banks, followed by Belgian and French banks. Inversely, Finnish, German and Luxembourg banks continued to stay away from those facilities while Greek, Irish and Portuguese banks, which had largely benefited from the previous refinancing operations, did not appear to have engaged any further in the two exceptional ones. The lack of additional available assets to pledge at the ECB is but one hypothesis. It also appeared that after the two rounds of refinancing, the Finnish, German and Luxembourg banks greatly increased their deposits at the ECB, while Belgian, French, Italian and Spanish banks did so, on a much more modest scale. Greek, Irish and Portuguese bank deposits at the ECB remained close to nil. Furthermore, there is also evidence in the report that, between the two LTRO rounds, the banking sector shorted their German and French government bond positions and greatly increased their Spanish and Italian government debt holdings. A small increase was also measured for holdings of Irish government debt securities.

12

Volume V Issue 1(9) 3.3. Oaxaca-Blinder decomposition for time series Going back to our AR(1) model:

( ⏟

)

(

)

Which we will modify in the following way: Let be the LTRO dummy and remove the (

)

and

(

)

(

)

⏟

⏟

dummies, which are constant throughout our period:

)(

(

)(

(

)

⏟

⏟

( )

yields:

(

)

)(

(

)

)

Taking the conditional expectation of the CDS innovations according to our LTRO dummy variable, it Let

(

represents the innovation at date t of the bank I CDS’s:

|

(

)

(

|

)

(

(

)

)

(

|

(

)

(

|

(

(

)

|

)}

)

)

(

|

)}

{

(

13

(

) (

)

|

)

|

|

By independence of the errors, ( | rewrite our problem in the following way: ( with:

{

(

) and:

|

(

)

|

{

|

(

)

)

|

. Combining both equations, we can | )

(

|

)

(

)

|

)}


(

|

)

(

|

)

(

|

)

In the following specification, focusing only on the exogenous innovations part, we can explain the three factors in the following way:  represents the “Pure” LTRO effect that cannot be attributed to any exogenous explanatory variable or coefficient changes.  represents the “Shift” effect of the LTRO on the exogenous variables.  represents the “Wake-up-Call” effect of the LTRO on sensitivities to our exogenous variables and controls. Under this specification, should represent the way bank funding were enhanced by the LTRO through a relaxation of the sovereign scare, while and should represent the direct impact of the LTRO on bank funding. 2.4. Time series decomposition results The TS approach we tried as a first step falls short of a severe caveat, namely the dimension problem: the numerous dummy variables included to measure the LTRO’s effect require a high number of observations to achieve a reasonable statistical significance. In practice, considering our 30 explanatory regressors, we measured the minimal sample of 600 points. Such a lengthy horizon would more than span the LTRO’s maximum effectiveness period: considering that Eurozone bank CDSs spiked in December 2011, at the time of our sample period, going 300 points (roughly a year) before and after that date should give us a measurement of the LTRO’s effectiveness. Yet, barely a quarter after the LTRO became effective, bank CDSs shot back upward as a result of resurging sovereign default fears triggered by Spain’s predicted budgetary deficit slippage and increasing contestation of austerity measures around Europe, thus complicating measurement of the LTRO’s long-term effect, which by the look of the TS results appears short-lived. Table 3. Oaxaca-Blinder time series decomposition factor estimation for the LTRO Effect

Pure A

Changes with Cst. Coefficients B

BS

BP

C

CS

CC

CP

Total A+B+C

LTRO +/- 300 Trading Days (≈ 12M)

Bank ERS 492.23 RAI 121.60 KBC 202.09 DEU -274.88 COM -309.25 BSN -172.62 BBV -22.25 SOC -97.67 BNP 73.95 CRE -85.66 BOI -3 200.66 INT 651.82 UNI 139.32 ING -488.23 BCP 109.21 BES 59.53

BC

New Coefficients post LTRO

12.19 1.78 -3.01 11.78 1.84 37.67 37.77 0.82 8.44 14.25 58.77 -61.37 7.14 -1.39 -9.92 35.38

3.82

-2.48

10.84

-0.11

-1.03

2.91

-5.59

-2.99

5.58

14.71

-6.83

3.91

-4.51

-6.06

12.40

5.24

-2.40

34.83

17.03

-3.31

24.05

3.48

-8.78

6.12

15.96

-7.06

-0.46

14.96

-9.58

8.86

3.32

-15.84

71.29

20.52

-7.52

-74.38

26.79

-8.24

-11.41

1.11

-3.02

0.52

-1.17

-2.86

-5.89

16.68

-2.98

21.67

-496.96 -120.84 -191.18 267.16 317.49 168.61 14.24 121.98 -64.10 94.56 3 095.27 -550.12 -118.89 499.01 -92.01 -93.11

-6.94

15.10

-505.13

-8.23

27.94

-140.56

-7.07

12.42

-196.54

-31.72

46.03

252.85

-15.36

49.54

283.31

-37.04

53.44

152.21

-43.57

17.17

40.64

-16.31

29.45

108.84

-26.69

30.28

-67.69

-16.68

32.99

78.25

-201.94

194.42

3 102.79

-30.79

52.47

-571.80

-44.93

44.88

-118.85

-20.80

-3.56

523.37

-77.53

58.64

-73.12

-85.50

48.73

-56.35

7.46 2.54 7.90 4.07 10.08 33.66 29.76 25.13 18.28 23.15 -46.63 40.33 27.56 9.39 7.28 1.79

Overall, the Oaxaca-Blinder decomposition of the time series data in Table 3 yields a large “Pure” effect and an equivalently large but contradictory “Wake-up-Call” effect. The “Shift” effect appears negligible. In terms of aggregate effects, only the BOI’s average CDS level diminished after the LTRO. The Sovereign “Wake-upCall” effect (CS) is interestingly always negative and large for the banks, which in the prior regression analysis had proved to have a significant LTRO or OMT dummy. Considering the above-mentioned caveats, any further interpretation of the results in terms of the ways of action of the LTRO would seem farfetched. In order to at least partially overcome those caveats, we proceed to an analysis in panel in the next section.

14

Volume V Issue 1(9) 3. Panel data analysis 3.1. Pooled OLS regression Since the estimation of the Least-Square Dummy Variable model (LSDV) previously proposed to perform the estimation of the Oaxaca-Blinder decomposition requires a sample period for which the length greatly exceeds our target range (300 trading days versus 25 to 150), we proceed to an analysis in panel. Going back to the AR(1) model previously used for our time series regressions: (

)

(

)(

(

)(

)

)(

(

)

)

We modify it by adding a firm-specific fixed effect ( ) through the use of a set of dummy variables ( (

)

(

)(

(

)(

)

(

)

)(

(

):

)

)

Since our estimation windows can be narrowed to 25-150 trading days, in this panel framework we can also estimate the Oaxaca-Blinder decomposition around the OMT event, which was previously impossible: (

)

(

(

)(

)(

)

(

)

)(

(

)(

)

)

3.2. Fixed effects pooled OLS results The results from the pooled-OLS regressions for the LTRO and the OMT are presented respectively in Table 11 and Table 12. Consistently with the prior time-series regression we performed as a first step, we presented the results of the calibration of our model centered on the LTRO dummy trigger date, with a sample size of +/- 25, 75 and 150 trading days, which roughly represent one, three and six months before and after the event. The one-month-sampling calibration exercise aimed to establish the most immediate short-term effect of the LTRO on an aggregated sample of 816 points. Yet, even if we have a relatively low autoregressive parameter (0.62) and a relatively good R² (99.66%), the only significant parameters at this stage are the firmspecific control variables and only those without the LTRO dummy multiplier). It is relatively unsurprising that with such a short calibration horizon there would be a high level of heterogeneity in the sensitivities to sovereign credit risk, thus a weak significance parameter. The three-month and six-month sampling calibration exercises yield a more balanced picture as some sovereign risks like AT, SP, IE and PT, respectively DE, SP, IE and IT, achieve statistical significance with various levels of confidence, although the firm-specific dummy variables still dominates. Those results would tend to confirm that the LTRO had a systematic effect on at least some of the sovereign risk sensitivities over a slightly longer period. The results of the calibration exercises centered on the OMT announcement, with the same sample sizes as before, are much clearer regardless of the horizon: even though the control variables, both the firm-specific ones and the general risk-factor ones, have a consistently high statistical significance, the sovereign risk dummies, both with and without the OMT dummy multiplier, achieve a high level of statistical significance. We should thus be able to have a reliable horizon-dependent analysis of the OMT’s effect with the Oaxaca-Blinder Decomposition as the OMT’s effect on sovereign-risk sensitivities appears systematic compared to the LTRO’s

15


effect which appeared, at least in the very short term, to be highly firm-specific and thus more difficult to measure reliably with a pooled regression approach. 3.3. Oaxaca-Blinder decomposition on panel data To account for the fixed effect, we modify the A factor by including the firm fixed-effect coefficient. Thus, the B and C factors previously introduced in the Oaxaca-Blinder decomposition should be affected only by sensitivities towards systematic risk components. Firm-specific sensitivities should only be reflected in the A factor. Nonetheless, the B and C factors are not purely systematic as the value of the firm-specific control variables (share price and long-term rating) are by definition idiosyncratic: the BP and CP sub-factors are thus partly firm-specific because of the common sensitivities used in their computation. The results for the OaxacaBinder decomposition around the LTRO and OMT events for our three sample sizes previously introduced are presented respectively in Table 9 and Table 10. For the LTRO, and because of the previously exposed caveats regarding the significance of the sensitivities to sovereign risks, the results of the one-month exercise should be discarded. At the three-month level, the “Pure” effect is still strongly negative while the “Wake-up-Call” and “Shift” effects are positive. At the six-month level, the “Pure” effect diminished greatly in absolute value and becomes positive in sign. The “Wakeup-Call” effect diminished and remains positive. Interestingly, the “Shift” effect becomes dominant and negative. Moreover, the sub-components leading to the negative values are firstly those of the sovereign shift (CS) and secondly of the control variables (CP). We should thus conclude that, for the LTRO, we have weak evidence that initially the channel through which it reduced bank funding stress was not intrinsically linked to a reduction in perceived sovereign risk (on the contrary) but rather a reduction in the banks’ perceived idiosyncratic risk. Later, the dominating channel becomes the reduction in the sensitivities to sovereign risk, consistently with the findings of (Acharya, Drechsler, & Schnabl, 2011). As for the OMT, regardless of the horizon considered, the “Pure” effect is consistently large and negative. Inversely, the “Wake-up-Call” effect is also consistently positive and large. The “Shift” effect is also significant and negative, and its size shrinks as the maturity lengthens. The “Sovereign-Shift” (BS) sub-effect is consistently negative, but its size reduces as the maturity increases. Interestingly, the “Sovereign Wake-up-Call” sub-effect (CS) is initially positive and then becomes negative. Overall, the OMT’s effect is strongest in the short term and then diminishes with maturity. At the six–month horizon, the only large effects left are those for the Irish and Portuguese banks. It is somewhat unsurprising since the OMT action plan requires that the country undergoes an EU-commission plan before the ECB is allowed to intervene in the secondary sovereign bond markets of that country. Considering that both Portugal and Ireland are the only countries under a joint EU-Commission/ ECB/ IMF (“Troika”) assistance plan, they are the two countries most susceptible to benefit from the OMT in the short term. As for the channel of action, in the case of the OMT, it seems that both the idiosyncratic reduction in perceived bank risk and the reduction in sovereign risk (also in terms of both sensitivities and levels) were effective, with the latter kicking in after the former. Conclusion The dramatic events of the last five years have led to a complete redesign of the rulebook of central banking all over the world as the macro-prudential stability objective seemed to have become the driving factor for essentially every central bank’s policy decisions of late. As central banks took the front stage to spearhead the initial fight against the banking crisis then the sovereign debt scare contagion, the traditional objectives of price stability – and employment for some – seemed to have moved backstage. This new course in policy has been probably driven by a mix of imperatives and pragmatism as central banks rediscovered the full extent of the notion of “lender of last resort” when interbank markets froze in 2008, or when the sovereign funding channels dried up in 2010 and the specter of a cascade of outright sovereign defaults in advanced economies surfaced in the Eurozone, thereby jeopardizing the very existence of the single-currency monetary union. Of all the unconventional monetary policies implemented during those difficult years, none was more controversial than the long-term refinancing operations of private banks: at a time when Europe’s hardest-hit countries were feeling the full extent of the austerity measures imposed either by international creditors trying to shore up the public finances of several Eurozone members or by the governments of other member states desperately trying to avoid suffering the same fate as Greece, public resentment against the banks accused of having wrecked the economy was rife. These unconventional monetary policies have led to a string of existential controversies both inside and outside of central banks, but it was nowhere as acute as within the European Central Bank: the relatively young institution had to navigate treacherous waters, balancing the heritage of

16

Volume V Issue 1(9) northern hawks with the pressing needs of its peripheral members, eventually edging toward a major board reshuffling to accompany the change in course set by its new pragmatic president Mario Draghi. As the debate moved from policy-makers to commentators and politicians, many researchers both in academia and in the industry have started to work on those complex issues and this work follows their steps. The main objective of the paper was to establish whether the unconventional monetary policies implemented by the ECB in the wake of the Eurozone’s sovereign crisis, namely the LTRO and the OMT, had benefited banks directly or indirectly through a relaxation of the sovereign credit-risk scare. As a minor objective, this paper looked at whether the market perception of bank credit risk accurately reflected public information on their exposures to sovereign risk, thereby potentially providing both a market-based information set on banks’ exposure to sovereign risks. Most importantly it provides an operational framework in which we could run the procedure in order to disentangle the various contributions of the relaxation in bank funding conditions and thus provide an answer to our main research question. By running a dummy variable least-square regression on each of the time series of the CDSs of the most important private banks of the Eurozone on which public data was available, we measured the explanatory power and significance of the estimated parameters of the sensitivities to individual sovereign risk measured also by their respective CDSs. Included in the analysis were conventional control variables on both firm-specific characteristics and more general market-level factors. Dummy variables were included to sort out the direct impact of three specific events: the onset of the sovereign scare and the announcement of both the LTROs and the OMT. The analysis of the results seems strongly consistent with both news releases and the map of banks’ retail operations in the countries present in our sample, our control for direct operational exposure to given sovereign risks. This is different from the financial asset holding exposure, on which much less information is available even if there is potentially a strong link between both. Overall, those results support the use of the model to proceed with the Oaxaca-Blinder decomposition. Following the paper of (Giordano, Pericoli, & Tommasino, 2012) using the decomposition technique of (Eichengreen & Mody, 2000) to study the channels of “contagion” of the Eurozone sovereign debt crisis sparked by the surprise Greek admission of window dressing of their national accounts, we endeavor to establish the “reverse-contagion” channels stemming from the unconventional monetary actions undertaken by the ECB. The model allows us to filter the “reverse contagion” between the “Pure” effect, the “Shift” in the levels of the explanatory variables (i.e. the sovereign CDSs and the control variables) and the “Wake-up-Call” effect of changes in the sensitivities toward the explanatory variables. Out of the three principal factors, the “Shift” effect directly exposes the indirect channel of relaxation of bank funding conditions, while the other two mostly reflect the direct effect of the operations. The results of the Oaxaca-Blinder decomposition performed on individual banks’ time series of CDSs could not reveal much, as the length of the sample required to calibrate the model exceeds one year before and after the critical date, which is much longer than the effective impact of the LTRO. The OMT’s impact could not even be assessed because of data availability issues, even though it probably lasted sufficiently long for the time-series methodology to be effective. To overcome that issue, we proceeded in a panel data approach using a pooled-OLS regression methodology and obtained the following results for both operations: At an aggregate level, the LTRO’s impact on bank funding conditions appears to have followed a two-step dynamic, consistently with (Acharya, Drechsler, & Schnabl, 2011). Initially, the “Pure” factor leads the fall in banks’ CDS levels while the other two factors slow the reduction. As we increase the horizon at which the effect is measured, we observe a reversal of the factors: the “Shift” effect dominates while the other two are smaller and positive. Moreover, the leading negative sub-factor is the “Sovereign Shift”. We can thus conclude that the effect of the LTRO on easing bank funding stress had initially been direct, and thus independent of any sovereign risk consideration, and then became indirect as the transmission channel became the source of the reduction of sovereign risks. For the OMT’s impact, the effect seems relatively horizon-independent: the “Wake-up-Call” is consistently large and positive while the other two effects are negative. The “Shift” effect does decrease over time. It should be noted that overall the aggregate effect also decreases with time and only remains outsized for the Irish and the Portuguese banks. In terms of transmission channels, we can thereby conclude that the indirect effect of the OMT decreases over time while the direct effect dominates. In a nutshell, we cannot definitively affirm that the dominant channel of action for the ECB’s unconventional monetary policies has been either the direct or the indirect channel. Both have clearly been active as banks benefited from both an exogenous enhancement of their credit perception by the market and from the relaxation of the sovereign funding stress. The most surprising result has been that even though the

17


LTRO was enacted primarily to stem the bank funding stress while the OMT was primarily directed at reducing sovereign funding stress, the effect has been relatively equivalent in terms of sovereign funding impact on banks. Even more surprising, the OMT seems to have had a longer lasting “pure” impact on bank funding conditions, independently of any sovereign funding considerations. In terms of potential improvements, we can already identify the following, clearly non-exhaustive, list: to enhance the significance of the sensitivities and thus the power of the decomposition, a more firm-specific parsimonious model could have been implemented by, for example, omitting all the variables that appear not to be insignificant at the individual bank level in a second-stage regression. It would also have been interesting to test our model on public banks, for which state support is even more immediate, even if it would require adjustments in the control variables as share prices, for example, are rarely available for such banks. Also, it could have been interesting to identify firm-specific breakpoints in the sensitivities arising from mergers, acquisitions or divestment by running the time-series regression with a reduced timespan to account for changes in the sensitivities, which are here averaged because of our time-constant restrictive assumption. Those issues are left for future research. The logical conclusion that should be drawn from the results of this paper is that from the central bank’s point of view, it appears virtually impossible to act on bank funding or on sovereign funding conditions alone without simultaneously affecting the other. The unconventional monetary policies undertaken by the ECB, regardless of their intended targets, have been mutually beneficial over the medium to long term. It would be wise to remember this when future interventions are debated, regardless of the politically damaging but rapidly subsiding short-term differentiating effect, which seemed to have benefited banks more than sovereign issuers. The symbiotic responses of states and banks to central bank interventions identified in this paper strongly uphold the EU’s proposed Banking Union, which would have banks supervised by the ECB (Asmussen, 2013). References [1] Acharya, V. V., Drechsler, I., and Schnabl, P. (2011). A Pyrrhic Victory? Bank Bailouts and Sovereign Credit Risk. NBER Working Paper, June 2011. [2] Arce, O., Mayordomo, S., and Peña, J. I. (2012). Credit-Risk Valuation in the Sovereign CDS and Bonds Markets: Evidence from the Euro Area Crisis. February 7th, SSRN WP: 1-37. [3] Asmussen, J. 2013. Building Banking Union. Speech by Jörg Asmussen at the Atlantic Council. July 9th (ECB, Intervieweur). [4] Bischof, J., and Daske, H. 2012. Mandatory supervisory disclosure, voluntary disclosure, and risk-taking of financial institutions: Evidence from the EU-wide stress-testing exercises. August 15th WP: 1-63. [5] Bofondi, M., Carpinelli, L., and Sette, E. 2012. Credit Supply During a Sovereign Crisis. Bank of Italy Working Paper: 1-47. [6] Chiaramontea, L., and Casu, B. 2012. The determinants of bank CDS spreads: evidence from the financial crisis. The European Journal of Finance, January 3rd. [7] Comfort, N., and Kirchfeld, A. 2012. Ackermann Shuns ECB Loans Because of ‘Reputational’ Risk for Deutsche Bank. February 3rd Bloomberg News. [8] Draghi, M. 2012. Verbatim of the remarks made by Mario Draghi. July 26th (ECB, Intervieweur). [9] Eichengreen, B., and Mody, A. 2000. What Explains Changing Spreads on Emerging debt. Dans Capital flows and the emergin economies: theory, evidence and controversies. Chicago: University of Chicago Press. [10] Enrich, D. 2012. Intesa Sanpaolo Taps Up LTRO for €24 Billion. Wall Street Journal February 29th. [11] Giordano, R., Pericoli, M., and Tommasino, P. 2012). 'Pure' or 'Wake-Up-Call' Contagion? Another Look at the EMU Sovereign Debt Crisis. Bank of Italy Working Paper, August 2012. [12] Jenkins, P. 2013. Deutsche Bank tapped ECB for up to €10bn. Financial Times, March 9th. [13] Jenkins, P., and Atkins, R. 2012. European banks have €115bn shortfall. Financial Times, February 8th. [14] Maloney, M. T., and Mulherin, J. 2003. The complexity of price discovery in an efficient market: the stock market reaction to the Challenger crash. Journal of Corporate Finance 9(4): 453–479. [15] O’Kane, D. 2012. The Link between Eurozone Sovereign Debt and CDS Prices. EDHEC-Risk Institute Working Paper, January 2012.

18

Volume V Issue 1(9) [16] Palladini, G., and Portes, R. 2011. Sovereign CDS and Bond Pricing Dynamics in the Euro-area. NBER Working Paper, November 2011. [17] Panetta, F. et al. 2011. The impact of sovereign credit risk on bank funding conditions. BIS - CGFS Papers, July 2011. [18] Petrakis, M., and Christie, R. 2012. Greece Pushes Bondholders Into Record Debt Swap. Bloomberg, March 9th. [19] Schuknecht, L., Hagen, J. V., and Wolswijk, G. 2011, March. Government Bond Risk Premiums in the EU Revisited: The Impact of the Financial Crisis. European Journal of Political Economy 27(1): 36-43. [20] Sgherri, S., and Zoli, E. 2009. Euro Area Sovereign Risk During the Crisis. IMF Working Paper: 1-25. [21] Wiesmann, G., and Wilson, J. 2011. Commerzbank faces bail-out by state. Financial Times. [22] Zoli, E. (2013, April). Italian Sovereign Spreads: Their Determinants and Pass-through to Bank Funding Costs and Lending Conditions. IMF Working Paper December 14th. *** BIS. 2012. International banking and financial market developments. March 2012 BIS Quarterly Review: 1-63. *** Dow Jones Newswires Reporters. 2012. Which Banks Took Up Second Round of LTRO. Récupéré sur Wall Street Jounal: http://blogs.wsj.com/eurocrisis/2012/02/29/which-banks-took-up-second-round-of-ltro/ February 29th. *** EBA. 2011. European Banking Authority 2011 EU-Wide Stress Test Aggregate Report. EBA. *** Global Banking News. 2012. Raiffeisen Bank International participates in ECB's LTRO. Global Banking News, February 29th. *** KBC. 2013. KBC Group repays 2011/2012 LTRO to ECB. Récupéré sur http://hugin.info/133947/ R/1673126/544602.pdf, January 25th. *** Markit. (s.d.). Free CDS Pricing Report. Récupéré sur Free CDS Pricing Report. *** Moody's. 2013. Moody's downgrades Austrian banks; ratings carry stable or negative outlooks. Global Credit Research, June 6th. *** Reuters. 2012. Italian, Spanish banks stock up on government debt - ECB data. Reuteurs, February 27th. *** Reuters. 2012. Italian, Spanish banks stock up on government debt in Jan -ECB data.

19


APPENDIX Table 4. Testing the unit boundary condition for the AR(1) parameter for all private banks except Greeks AR(1) Unit Root test for near-integreted series ERS 0.945 *** -7.772 7.59458E-15 RAI 0.933 *** -8.497 2.54378E-17 KBC 0.942 *** -7.159 6.63567E-13 DEU 0.876 *** -12.652 4.65847E-35 COM 0.908 *** -10.519 3.16313E-25 BSN 0.820 *** -17.244 1.11114E-60 BBV 0.828 *** -18.001 2.08836E-65 SOC 0.804 *** -18.638 1.76279E-69 BNP 0.820 *** -16.849 2.89835E-58 CRE 0.821 *** -17.602 6.70642E-63 BOI 0.903 *** -8.415 4.95107E-17 INT 0.849 *** -16.043 1.87855E-53 UNI 0.878 *** -14.400 3.8081E-44 ING 0.924 *** -9.437 8.10489E-21 BCP 0.918 *** -9.929 8.95957E-23 BES 0.909 *** -13.515 1.97009E-39 */**/***: Denote the significance at the 90%/95%/99% level.

Table 5. Testing the unit boundary condition for the AR(1) parameter for the Greek banks AR(1) Unit Root test for near-integreted series Bk. AR(1) T-stat p-value NAT 0.934 *** -6.248 7.30845E-12 ALP 0.827 *** -10.270 7.24337E-16 */**/***: Denote the significance at the 90%/95%/99% level.

20

Volume V Issue 1(9) Table 6. All banks except Greeks and without Greek sovereign Bank #Obs. AdjR² Fisher ERS 1369 98.88% 6 682.15 +++

Itr. -49.88 -0.80 42.108%

RAI

KBC

1369 98.92% 6 955.39 +++

1369 99.21% 9 523.89 +++

COM

1369 98.04% 3 805.70 +++

1369 99.02% 7 715.91 +++

0.93 *** 118.39

2.380%

0.000%

73.41 *** 2.66

26.29 ** 2.14

1369 99.29% 10 657.94 +++

1369 99.51% 15 569.64 +++

1369 99.34% 11 426.91 +++

0.91 *** 103.37

32.42 *** 3.31

33.03 *** 3.33

62.97 *** 3.97 0.007%

BNP

1369 99.28% 10 542.19 +++

-1.69 -0.14 88.856%

CRE

1369 99.14% 8 780.14 +++

-56.01 ** -2.13 3.346%

BOI

1369 98.24% 4 248.67 +++

132.89 *** 2.88 0.403%

INT

1369 99.57% 17 712.58 +++

27.17 1.48 13.798%

UNI

1369 99.58% 18 203.94 +++

13.15 0.72 47.348%

ING

BCP

BES

1369 99.03% 7 765.17 +++

1369 99.77% 33 317.78 +++

1369 99.82% 41 389.06 +++

0.88 *** 89.76

-64.79 -1.21

0.089%

SOC

0.000%

0.000%

0.095%

BBV

0.94 *** 116.74

3.260%

22.472%

BSN

0.000%

-66.02 ** -2.26

0.790%

DEU

Lg1 0.94 *** 133.22

0.000%

0.82 *** 78.55 0.000%

0.83 *** 86.77 0.000%

0.80 *** 76.62 0.000%

0.82 *** 76.81 0.000%

0.82 *** 81.01 0.000%

0.90 *** 78.72 0.000%

0.85 *** 90.46 0.000%

0.88 *** 103.76 0.000%

-60.36 *** -4.98

0.92 *** 113.95

0.000%

0.000%

-36.59 * -1.67

0.92 *** 110.64

9.501%

0.000%

-36.76 ** -2.57

0.91 *** 135.20

1.015%

0.000%

AT 0.08 *** 3.79 0.016%

0.12 *** 6.29 0.000%

0.01 0.58 56.481%

0.03 ** 2.22 2.657%

-0.01 -0.47 64.118%

-0.02 -0.80 42.482%

0.02 0.87 38.656%

-0.01 -0.56 57.604%

0.01 0.95 34.304%

0.03 1.44 15.008%

BE 0.03 1.52

DE 0.02 0.32

SP -0.01 -1.16

FR 0.04 1.07

IE 0.00 0.77

12.947%

75.116%

24.692%

28.588%

43.968%

6.984%

74.592%

0.02 1.34

0.05 0.98

-0.01 -1.02

0.00 -0.06

0.00 1.12

0.01 1.24

-0.02 -0.67

18.004%

32.779%

30.780%

94.890%

26.442%

21.663%

50.497%

0.09 *** 4.07 0.005%

0.03 0.54

0.00 -0.11

-0.07 -1.60

59.161%

91.294%

10.886%

0.01 0.75

-0.06 -1.59

45.479%

11.207%

0.06 *** 3.51 0.046%

0.08 *** 4.29 0.002%

0.11 *** 5.33 0.000%

0.12 *** 6.69 0.000%

0.07 *** 4.75 0.000%

0.07 *** 3.90 0.010%

0.04 0.92 36.019%

-0.05 -1.03 30.391%

-0.14 *** -2.69 0.725%

-0.12 *** -2.59 0.973%

-0.09 ** -2.30 2.167%

-0.07 -1.49 13.566%

0.01 ** 2.03 4.289%

0.02 ** 2.04 4.184%

0.10 *** 8.04 0.000%

0.12 *** 9.14 0.000%

-0.03 *** -2.73 0.644%

0.06 ** 2.09 3.685%

-0.01 -0.33 73.849%

-0.11 *** -2.61 0.926%

-0.11 ** -2.57 1.038%

0.10 *** 2.83 0.474%

-0.01 -0.65

0.04 1.27

51.409%

20.596%

0.06 *** 5.92 0.000%

0.00 -0.07 94.816%

0.14 0.96

0.15 1.05

0.40 1.05

0.09 1.20

0.26 0.88

33.526%

29.362%

29.478%

23.119%

38.026%

0.00 -0.07 94.378%

0.03 1.21 22.634%

0.01 0.82 41.195%

-0.16 *** -2.73 0.642%

-0.14 *** -3.61 0.032%

0.10 *** 4.35 0.001%

0.08 *** 3.49 0.049%

0.03 ** 2.09 3.684%

0.22 *** 3.49 0.050%

0.22 *** 5.65 0.000%

-0.04 -0.57

0.01 1.01

56.816%

31.092%

-0.01 -0.11

0.02 1.32

91.579%

18.614%

-0.09 *** -2.71 0.675%

0.00 0.41 68.052%

-0.12 *** -2.60 0.940%

-0.18 *** -3.79 0.016%

0.06 ** 2.47 1.362%

0.04 0.29

0.01 0.33

0.11 0.97

76.854%

73.994%

33.329%

0.14 1.54 12.311%

0.17 *** 8.56 0.000%

-0.20 *** -2.81 0.500%

21

0.01 * 1.86 6.273%

-0.01 ** -2.42 1.557%

0.01 * 1.86 6.330%

0.00 -0.82 41.320%

0.00 -0.70 48.672%

0.00 -1.37 17.155%

-0.01 ** -2.12 3.388%

-0.01 *** -3.08 0.210%

0.12 *** 3.68 0.024%

0.01 ** 2.00 4.552%

0.01 ** 2.45 1.434%

0.00 1.15 25.048%

0.03 *** 2.94 0.331%

0.02 ** 2.35 1.890%

IT 0.02 * 1.81

NL -0.01 -0.32

0.02 1.49

0.00 -0.01

13.630%

98.914%

0.03 *** 4.08 0.005%

0.03 *** 2.94 0.332%

0.03 *** 3.11 0.190%

0.03 *** 2.75 0.611%

0.09 *** 9.14 0.000%

0.08 *** 9.34 0.000%

0.05 *** 5.76 0.000%

-0.05 -0.68 49.544%

0.15 *** 10.75 0.000%

0.15 *** 9.75 0.000%

0.02 *** 2.67 0.763%

0.07 ** 2.45 1.450%

-0.04 ** -2.45 1.442%

PT 0.00 * -1.77 7.736%

-0.01 ** -2.01 4.456%

-0.01 *** -2.60 0.931%

SPi 4.39 1.17 24.243%

4.66 ** 2.57 1.015%

-3.66 ** -2.24 2.537%

RTi -0.33 *** -3.20

VSX 0.18 *** 3.35

0.143%

0.083%

-0.10 ** -2.55 1.102%

-0.12 *** -3.77 0.017%

-0.02 -0.65

0.00 0.49

-1.05 -1.50

0.02 0.41

51.848%

62.133%

13.495%

67.847%

-0.06 * -1.80 7.255%

0.00 -1.47

4.31 1.32

-0.07 -0.43

14.118%

18.562%

66.673%

0.04 1.20

0.00 -1.19

22.884%

23.573%

1.21 ** 2.44 1.485%

0.02 0.55

0.00 -1.60

0.56 1.22

58.563%

11.053%

22.361%

-0.11 *** -3.18 0.153%

-0.05 * -1.72 8.478%

-0.07 ** -2.23 2.594%

-0.72 *** -2.64 0.839%

0.00 1.62 10.491%

0.00 ** 2.12 3.386%

0.01 *** 2.77 0.562%

-0.03 -1.23 22.029%

-1.65 ** -2.04 4.125%

0.85 1.38 16.705%

4.06 *** 2.85 0.443%

-8.00 *** -2.92 0.354%

-3.91 *** -7.36 0.000%

-2.33 *** -6.40 0.000%

-0.29 *** -4.06 0.005%

-0.09 ** -2.13 3.357%

-0.33 ** -2.03 4.209%

0.00 -1.60 11.018%

-0.07 -0.65 51.756%

0.07 1.03 30.200%

0.05 *** 5.67 0.000%

0.04 *** 7.71 0.000%

3.84 *** 5.27 0.000%

2.61 * 1.92 5.482%

3.59 *** 4.33 0.002%

0.32 *** 8.54 0.000%

0.42 *** 9.46 0.000%

-7.79 *** -4.75 0.000%

-9.64 *** -5.94 0.000%

-10.23 *** -7.52 0.000%

-13.68 *** -9.19 0.000%

-0.02 *** -2.75 0.611%

-0.07 *** -8.48 0.000%

-0.06 *** -7.47 0.000%

-0.04 *** -6.21 0.000%

-0.03 *** -4.59 0.000%

-0.03 *** -5.34 0.000%

-0.01 -0.26

0.07 0.40

0.00 0.16

0.49 *** 10.40 0.000%

-4.53 *** -2.81 0.508%

-0.02 *** -3.61 0.031%

79.361%

68.570%

87.379%

0.31 *** 6.05 0.000%

-6.27 *** -3.88 0.011%

-0.03 *** -2.98 0.296%

8.97 0.78

-1.24 -1.09

0.25 0.24

0.22 *** 4.17 0.003%

-4.68 *** -3.60 0.033%

0.916%

43.476%

27.396%

80.673%

0.24 *** 5.35 0.000%

0.69 0.34 73.752%

0.000%

-0.02 *** -2.61

0.02 0.07

0.00 -1.45

-0.01 * -1.92

0.32 *** 8.65 0.000%

0.829%

0.000%

94.762%

14.614%

5.533%

0.25 *** 4.54 0.001%

-4.68 *** -2.64

GD -7.27 *** -5.62

2.77 1.01

0.06 1.40

0.06 1.43

0.739%

10.580%

BBB -0.05 *** -5.56

31.333%

16.147%

15.269%

0.14 *** 2.68

BOI -2.87 -1.62

-2.01 * -1.95 5.152%

-0.42 *** -2.67 0.759%

0.36 *** 6.26 0.000%

0.40 *** 6.76 0.000%

0.21 *** 5.67 0.000%

-5.07 ** -2.49 1.286%

-5.61 *** -2.89 0.387%

-6.33 *** -5.11 0.000%

-4.09 -0.63

0.23 1.64

-4.83 -1.12

52.815%

10.034%

26.105%

-3.68 *** -2.72 0.660%

0.20 ** 2.18 2.921%

-2.67 -0.94 34.662%

-0.05 *** -5.97 0.000%

-0.07 *** -7.91 0.000%

-0.02 *** -4.34 0.002%

-0.04 * -1.89 5.926%

-0.05 *** -4.00 0.007%

-3.96 *** -3.03 0.252%

-6.79 *** -4.96 0.000%

-5.81 *** -6.12 0.000%

-7.76 *** -6.21 0.000%

-6.08 *** -5.49 0.000%

-6.27 *** -5.80 0.000%

-11.13 *** -9.92 0.000%

-7.28 *** -8.05 0.000%

-10.65 *** -10.50

LD 0.77 0.43

OD 1.71 1.15

66.530%

24.838%

1.40 0.90 36.649%

3.61 1.56

0.66 0.44

11.999%

66.119%

-1.78 -1.46

0.63 0.59

14.513%

55.765%

5.25 *** 3.14 0.174%

7.22 *** 3.79 0.015%

7.53 *** 3.61 0.032%

6.05 *** 3.77 0.017%

3.39 ** 2.48 1.321%

8.96 *** 4.76

0.000%

0.000%

-24.20 ** -2.25

-21.65 * -1.65

2.471%

9.863%

-9.55 *** -6.24 0.000%

-10.85 *** -7.66 0.000%

-2.17 ** -2.31 2.113%

-9.02 *** -2.61 0.917%

-9.37 *** -4.12 0.004%

3.35 ** 2.57 1.022%

8.81 *** 3.81 0.015%

7.21 *** 3.04 0.244%

3.10 *** 2.82 0.483%

-1.68 -0.27 78.743%

-1.21 -0.33 74.102%

0.53 0.40 68.624%

3.04 * 1.65 9.906%

3.79 ** 2.04 4.189%

-0.42 -0.33 74.503%

0.36 0.31 75.704%

-2.30 -1.52 12.835%

20.25 * 1.78 7.536%

7.37 *** 3.90 0.010%

4.05 ** 1.98 4.798%

-0.01 -0.02 98.743%

17.51 *** 3.84 0.013%

13.47 *** 4.23 0.002%


Table 7. All banks except Greeks and without Greek sovereign (Only significant parameters left) Bank #Obs. AdjR² Fisher p-Value ERS 1369 98.88% 6 682.15 +++ 0.00

Itr.

Lg1 0.94 *** 133.22

AT 0.08 *** 3.79 0.02%

0.93 *** 118.39

0.12 *** 6.29 0.00%

RAI

1369 98.92% 6 955.39 +++

0.00 -66.02 ** -2.26 2.38%

KBC

1369 99.21% 9 523.89 +++

0.00 73.41 *** 0.94 *** 2.66 116.74 0.79%

DEU

1369 98.04% 3 805.70 +++

0.00 26.29 ** 2.14 3.26%

COM

1369 99.02% 7 715.91 +++

0.00

BSN

1369 99.29% 10 657.94 +++

BBV

0.88 *** 89.76

BE

DE

SP

FR

IE

IT 0.02 * 1.81 6.98%

NL

PT 0.00 * -1.77 7.74% -0.01 ** -2.01 4.46%

0.09 *** 4.07 0.01%

0.01 * 1.86 6.27%

0.03 ** 2.22 2.66%

0.01 ** 2.03 4.29%

0.06 ** 2.09 3.68%

-0.10 ** -2.55 1.10%

LD

OD

0.14 *** -4.68 *** -0.02 *** -3.96 *** 2.68 -2.64 -2.61 -3.03 0.74% 0.83% 0.92% 0.25%

-0.12 *** 0.25 *** -3.77 4.54 0.02% 0.00%

3.35 ** 2.57 1.02%

-0.03 *** -6.79 *** -2.98 -4.96 0.30% 0.00%

0.03 *** 4.08 0.00%

0.32 *** -4.68 *** -0.02 *** -5.81 *** 8.65 -3.60 -3.61 -6.12 0.00% 0.03% 0.03% 0.00%

0.01 * 1.86 6.33%

0.03 *** -0.06 * 2.94 -1.80 0.33% 7.25%

0.24 *** -6.27 *** -0.02 *** -7.76 *** 5.25 *** 5.35 -3.88 -2.75 -6.21 3.14 0.00% 0.01% 0.61% 0.00% 0.17%

0.02 ** 2.04 4.18%

0.00 32.42 *** 3.31 0.09%

0.82 *** 78.55

0.08 *** 4.29 0.00%

0.10 *** -0.11 *** 8.04 -2.61 0.00% 0.93%

1369 99.51% 15 569.64 +++

0.00 33.03 *** 3.33 0.09%

0.83 *** 86.77

0.11 *** -0.14 *** 0.12 *** -0.11 ** 5.33 -2.69 9.14 -2.57 0.00% 0.72% 0.00% 1.04%

0.03 *** 2.75 0.61%

SOC

1369 99.34% 11 426.91 +++

0.00 62.97 *** 3.97 0.01%

0.80 *** 76.62

0.12 *** -0.12 *** -0.03 *** 0.10 *** 6.69 -2.59 -2.73 2.83 0.00% 0.97% 0.64% 0.47%

0.09 *** -0.11 *** 9.14 -3.18 0.00% 0.15%

BNP

1369 99.28% 10 542.19 +++

0.00

0.82 *** 76.81

0.07 *** -0.09 ** 4.75 -2.30 0.00% 2.17%

CRE

1369 99.14% 8 780.14 +++

0.00 -56.01 ** -2.13 3.35%

0.82 *** 81.01

0.07 *** 3.90 0.01%

BOI

1369 98.24% 4 248.67 +++

0.00 132.89 *** 2.88 0.40%

0.90 *** 78.72

INT

1369 99.57% 17 712.58 +++

0.00

0.85 *** 90.46

0.10 *** 4.35 0.00%

-0.12 *** 0.01 ** -2.60 2.00 0.94% 4.55%

0.15 *** 10.75 0.00%

UNI

1369 99.58% 18 203.94 +++

0.00

0.88 *** 103.76

0.08 *** 3.49 0.05%

-0.18 *** 0.01 ** -3.79 2.45 0.02% 1.43%

0.15 *** 9.75 0.00%

ING

1369 99.03% 7 765.17 +++

0.00 -60.36 *** 0.92 *** -4.98 113.95 0.00%

0.06 ** 2.47 1.36%

0.02 *** 2.67 0.76%

BCP

1369 99.77% 33 317.78 +++

0.00 -36.59 * -1.67 9.50%

0.92 *** 110.64

-0.16 *** 0.22 *** -2.73 3.49 0.64% 0.05%

BES

1369 99.82% 41 389.06 +++

0.00 -36.76 ** -2.57 1.01%

0.91 *** 135.20

-0.14 *** 0.22 *** -3.61 5.65 0.03% 0.00%

0.03 *** 3.11 0.19%

-0.01 ** -2.12 3.39%

1.21 ** 2.44 1.49%

0.08 *** -0.05 * 9.34 -1.72 0.00% 8.48%

-0.04 ** -2.45 1.44%

-3.91 *** 0.22 *** -4.53 *** -0.07 *** -6.08 *** 7.22 *** -7.36 4.17 -2.81 -8.48 -5.49 3.79 0.00% 0.00% 0.51% 0.00% 0.00% 0.02%

3.04 * 1.65 9.91%

-2.33 *** 0.31 *** -7.79 *** -0.06 *** -6.27 *** 7.53 *** -6.40 6.05 -4.75 -7.47 -5.80 3.61 0.00% 0.00% 0.00% 0.00% 0.00% 0.03%

3.79 ** 2.04 4.19%

-0.29 *** 0.49 *** -9.64 *** -0.04 *** -11.13 *** 6.05 *** -4.06 10.40 -5.94 -6.21 -9.92 3.77 0.01% 0.00% 0.00% 0.00% 0.00% 0.02% -0.09 ** -2.13 3.36%

0.01 *** 4.06 *** -0.33 ** 2.77 2.85 -2.03 0.56% 0.44% 4.21%

-0.72 *** -2.64 0.84%

0.03 *** 0.07 ** 2.94 2.45 0.33% 1.45% 0.17 *** -0.20 *** 0.02 ** 8.56 -2.81 2.35 0.00% 0.50% 1.89%

-1.65 ** -2.04 4.13% 0.00 ** 2.12 3.39%

-0.01 *** 0.05 *** -0.07 ** -3.08 5.76 -2.23 0.21% 0.00% 2.59% 0.12 *** 3.68 0.02%

-0.09 *** -2.71 0.68%

4.66 ** 2.57 1.01%

BBB GD -0.05 *** -7.27 *** -5.56 -5.62 0.00% 0.00%

-0.01 ** -2.42 1.56%

0.06 *** 3.51 0.05%

0.03 ** 2.09 3.68%

RTi VSX BOI -0.33 *** 0.18 *** -3.20 3.35 0.14% 0.08%

-0.01 *** -3.66 ** -2.60 -2.24 0.93% 2.54%

0.91 *** 103.37

0.06 *** 5.92 0.00%

SPi

0.32 *** -10.23 *** -0.03 *** -7.28 *** 3.39 ** 8.54 -7.52 -4.59 -8.05 2.48 0.00% 0.00% 0.00% 0.00% 1.32% 0.42 *** -13.68 *** -0.03 *** -10.65 *** 8.96 *** 9.46 -9.19 -5.34 -10.50 4.76 0.00% 0.00% 0.00% 0.00% 0.00%

-8.00 *** -2.92 0.35%

-24.20 ** -2.25 2.47% 0.36 *** -5.07 ** 6.26 -2.49 0.00% 1.29%

-0.01 * -1.92 5.53%

-2.01 * -1.95 5.15%

-21.65 * -1.65 9.86%

20.25 * 1.78 7.54%

-0.05 *** -9.55 *** 8.81 *** -5.97 -6.24 3.81 0.00% 0.00% 0.01%

7.37 ** 3.90 0.01%

0.40 *** -5.61 *** -0.07 *** -10.85 *** 7.21 *** 6.76 -2.89 -7.91 -7.66 3.04 0.00% 0.39% 0.00% 0.00% 0.24%

4.05 ** 1.98 4.80%

3.84 *** -0.42 *** 0.21 *** -6.33 *** -0.02 *** -2.17 ** 5.27 -2.67 5.67 -5.11 -4.34 -2.31 0.00% 0.76% 0.00% 0.00% 0.00% 2.11%

3.10 *** 2.82 0.48%

0.05 *** 2.61 * 5.67 1.92 0.00% 5.48%

-0.04 * -1.89 5.93%

-9.02 *** -2.61 0.92%

17.51 ** 3.84 0.01%

0.04 *** 3.59 *** -3.68 *** 0.20 ** 7.71 4.33 -2.72 2.18 0.00% 0.00% 0.66% 2.92%

-0.05 *** -9.37 *** -4.00 -4.12 0.01% 0.00%

13.47 ** 4.23 0.00%

Table 8. Greek bank restricted sample regression Bank #Obs. AdjR²

Fisher

Itr.

NAT 852 99.9% 41 341.06 +++

Lg1

68.59 *** 2.47 1.381%

ALP

852 99.8% 18 975.78 +++ 158.55 *** 4.30

FR

GR

IE

0.08 0.44

0.00 * -1.77

-0.03 * -1.95

76%

0.40 1.60

44%

7.757%

0.01 *** 5.28 0.000%

IT 0.16 *** 3.03

NL -0.23 * -1.80

5.131%

0.256%

7.181%

-0.02 -0.94

0.10 1.47

0.01 0.08

34.501%

14.228%

93.499%

AT

0.93 *** 89.12 0.000%

-0.05 -0.67

BE 0.17 *** 1.98

-0.08 -0.44 65.976%

-0.13 -1.33

0.18 1.47

0.11 0.39

-0.07 -1.01

0.40 1.60

69.502%

31.377%

11.044%

BBB

GD

0.83 *** 48.96

PT

SPi

RTi

VSX

-3.42 ** -2.40 1.660%

-8.64 *** -3.94 0.009%

-0.03 -0.55 58.285%

14.103%

0.482%

0.31 1.52 12.931%

18.236%

0.07 *** 2.83

FR

4.810%

0.000%

0.074%

SP

50.098%

0.002%

0.06 *** 3.39

DE

BOI

-0.71 -1.45

-0.21 -0.85

4.24 0.53

0.00 0.08

0.14 0.03

14.875%

39.822%

59.440%

93.631%

97.829%

0.09 0.26

2.83 0.26

79.119%

79.375%

-4.72 *** -3.50 0.049%

22

0.07 * 1.82 6.957%

15.89 ** 2.28 2.264%

GR

IE

0.00 * -1.77

-0.03 * -1.95

7.757%

0.01 *** 5.28 0.000%

LD 13.96 * 1.67 9.499%

10.34 0.89 37.227%

5.131%

-0.02 -0.94 34.501%

Volume V Issue 1(9) Table 9. Oaxaca-Blinder Decomposition Factor Estimation for the LTRO Oaxaca-Blinder Decomposition Factor Estimation for the LTRO Effect

Pure A


BS

BP

C

CS

CC

CP

Total A+B+C

LTRO +/- 25 Trading Days (≈ 1M)

Bank ERS RAI KBC DEU COM BSN BBV SOC BNP CRE BOI INT UNI ING BCP BES

BC

New Coefficients post LTRO

-141.67 -138.56 -99.15 -140.05 -135.68 -130.18 -123.31 -123.94 -126.57 -133.82 -224.17 -143.51 -152.13 -134.55 -145.25 -161.69

128.10 134.28 111.20 122.94 120.71 55.25 38.32 107.61 115.52 99.15 121.34 80.23 82.35 121.13 37.65 58.23

-9.89

131.71

6.28

-9.89

131.71

12.46

-9.89

131.71

-10.62

-9.89

131.71

1.12

-9.89

131.71

-1.11

-9.89

131.71

-66.57

-9.89

131.71

-83.50

-9.89

131.71

-14.21

-9.89

131.71

-6.29

-9.89

131.71

-22.67

-9.89

131.71

-0.48

-9.89

131.71

-41.59

-9.89

131.71

-39.47

-9.89

131.71

-0.69

-9.89

131.71

-84.17

-9.89

131.71

-63.59

6.35 4.57 9.96 -2.02 9.91 14.42 19.83 1.36 -3.99 7.68 38.06 19.07 19.53 2.04 51.92 48.06

66.37

37.47

-97.49

66.37

37.47

-99.27

66.37

37.47

-93.88

66.37

37.47

-105.86

66.37

37.47

-93.93

66.37

37.47

-89.42

66.37

37.47

-84.01

66.37

37.47

-102.48

66.37

37.47

-107.83

66.37

37.47

-96.16

66.37

37.47

-65.78

66.37

37.47

-84.77

66.37

37.47

-84.31

66.37

37.47

-101.80

66.37

37.47

-51.92

66.37

37.47

-55.79

25.03

75.72

-3.61

25.03

75.72

-2.11

25.03

75.72

-3.23

25.03

75.72

-1.66

25.03

75.72

-6.47

25.03

75.72

-5.47

25.03

75.72

-4.91

25.03

75.72

-0.64

25.03

75.72

-3.49

25.03

75.72

-5.97

25.03

75.72

-4.71

25.03

75.72

-5.44

25.03

75.72

-5.83

25.03

75.72

-6.17

25.03

75.72

-3.72

25.03

75.72

-3.87

-64.40

22.87

-29.22

-64.40

22.87

-28.51

-64.40

22.87

-27.94

-64.40

22.87

-29.94

-64.40

22.87

-30.53

-64.40

22.87

-28.52

-64.40

22.87

-26.61

-64.40

22.87

-28.22

-64.40

22.87

-31.92

-64.40

22.87

-30.74

-64.40

22.87

-21.52

-64.40

22.87

-27.23

-64.40

22.87

-27.41

-64.40

22.87

-32.41

-64.40

22.87

-16.98

-64.40

22.87

-18.15

-7.23 0.29 22.01 -19.13 -5.06 -60.52 -65.16 -14.97 -15.04 -26.99 -64.76 -44.20 -50.25 -11.38 -55.68 -55.41

LTRO +/- 75 Trading Days (≈ 3M) ERS RAI KBC DEU COM BSN BBV SOC BNP CRE BOI INT UNI ING BCP BES

-161.49 -158.45 -150.37 -159.81 -157.30 -148.74 -144.60 -158.97 -156.81 -151.18 -216.92 -159.32 -160.09 -157.15 -185.97 -161.17

62.63 61.50 56.15 60.21 63.59 41.39 35.36 54.88 56.09 57.15 63.70 52.67 50.69 63.17 39.47 45.53

53.11

10.75

-1.23

53.11

10.75

-2.36

53.11

10.75

-7.71

53.11

10.75

-3.64

53.11

10.75

-0.27

53.11

10.75

-22.47

53.11

10.75

-28.50

53.11

10.75

-8.97

53.11

10.75

-7.77

53.11

10.75

-6.71

53.11

10.75

-0.16

53.11

10.75

-11.19

53.11

10.75

-13.17

53.11

10.75

-0.69

53.11

10.75

-24.39

53.11

10.75

-18.33

97.14 98.64 97.52 99.08 94.27 95.27 95.84 100.10 97.26 94.78 96.04 95.31 94.92 94.58 97.03 96.88

-1.72 1.69 3.30 -0.51 0.57 -12.08 -13.40 -3.98 -3.46 0.74 -57.17 -11.35 -14.48 0.60 -49.47 -18.76

LTRO +/- 150 Trading Days (≈ 6M) ERS RAI KBC DEU COM BSN BBV SOC BNP CRE BOI INT UNI ING BCP BES

22.21 21.49 26.20 22.18 23.49 35.11 37.37 25.03 28.35 31.13 -60.17 30.30 30.29 26.31 -5.19 9.35

37.79 37.80 35.09 37.06 37.74 26.69 23.79 34.97 34.75 34.82 37.98 32.38 30.61 37.66 26.81 29.37

31.62

6.21

-0.04

31.62

6.21

-0.04

31.62

6.21

-2.74

31.62

6.21

-0.78

31.62

6.21

-0.09

31.62

6.21

-11.14

31.62

6.21

-14.04

31.62

6.21

-2.87

31.62

6.21

-3.09

31.62

6.21

-3.02

31.62

6.21

0.14

31.62

6.21

-5.46

31.62

6.21

-7.22

31.62

6.21

-0.17

31.62

6.21

-11.03

31.62

6.21

-8.47

-70.75 -70.04 -69.47 -71.48 -72.06 -70.05 -68.14 -69.75 -73.45 -72.27 -63.05 -68.76 -68.95 -73.94 -58.51 -59.68

-10.75 -10.75 -8.18 -12.24 -10.83 -8.25 -6.98 -9.75 -10.36 -6.31 -85.24 -6.09 -8.04 -9.97 -36.90 -20.96

Table 10. Oaxaca-Blinder Decomposition Factor Estimation for the OMT 23


Effect

Pure A


BS

BP

C

CS

CC

Total

CP

A+B+C

OMT +/- 25 Trading Days (≈ 1M)

Bank ERS RAI KBC DEU COM BSN BBV SOC BNP CRE BOI INT UNI ING BCP BES

BC

New Coefficients post OMT

-75.37 -86.23 -89.04 -78.58 -71.39 -87.56 -93.98 -106.89 -91.48 -76.53 -101.00 -97.16 -108.77 -63.39 -98.76 -103.15

-58.68 -63.10 -53.79 -56.77 -75.41 -120.65 -119.69 -51.56 -74.14 -73.22 -73.91 -71.54 -73.29 -75.33 -73.92 -73.25

-66.22

-7.74

15.28

-66.22

-7.74

10.86

-66.22

-7.74

20.17

-66.22

-7.74

17.19

-66.22

-7.74

-1.45

-66.22

-7.74

-46.69

-66.22

-7.74

-45.73

-66.22

-7.74

22.40

-66.22

-7.74

-0.18

-66.22

-7.74

0.74

-66.22

-7.74

0.05

-66.22

-7.74

2.43

-66.22

-7.74

0.67

-66.22

-7.74

-1.37

-66.22

-7.74

0.04

-66.22

-7.74

0.71

61.53 68.75 65.27 67.86 46.36 56.88 60.78 76.93 61.88 49.56 60.72 54.18 52.30 46.83 69.24 67.48

25.89

71.29

-35.65

25.89

71.29

-28.42

25.89

71.29

-31.91

25.89

71.29

-29.31

25.89

71.29

-50.82

25.89

71.29

-40.30

25.89

71.29

-36.39

25.89

71.29

-20.25

25.89

71.29

-35.30

25.89

71.29

-47.62

25.89

71.29

-36.46

25.89

71.29

-43.00

25.89

71.29

-44.87

25.89

71.29

-50.34

25.89

71.29

-27.93

25.89

71.29

-29.69

-64.45

121.92

23.75

-64.45

121.92

28.37

-64.45

121.92

24.36

-64.45

121.92

30.57

-64.45

121.92

13.89

-64.45

121.92

14.12

-64.45

121.92

13.84

-64.45

121.92

33.85

-64.45

121.92

27.73

-64.45

121.92

16.33

-64.45

121.92

9.33

-64.45

121.92

13.49

-64.45

121.92

12.29

-64.45

121.92

17.43

-64.45

121.92

7.14

-64.45

121.92

8.10

-40.55

81.26

25.99

-40.55

81.26

30.96

-40.55

81.26

26.63

-40.55

81.26

33.36

-40.55

81.26

15.37

-40.55

81.26

15.56

-40.55

81.26

15.23

-40.55

81.26

36.87

-40.55

81.26

30.30

-40.55

81.26

18.00

-40.55

81.26

10.33

-40.55

81.26

14.88

-40.55

81.26

13.59

-40.55

81.26

19.21

-40.55

81.26

7.91

-40.55

81.26

8.96

-72.52 -80.57 -77.57 -67.49 -100.44 -151.34 -152.89 -81.53 -103.74 -100.19 -114.19 -114.52 -129.76 -91.89 -103.43 -108.92

OMT +/- 75 Trading Days (≈ 3M) ERS RAI KBC DEU COM BSN BBV SOC BNP CRE BOI INT UNI ING BCP BES

-50.56 -54.82 -52.43 -56.44 -42.96 -46.98 -48.61 -63.47 -56.93 -48.69 -52.95 -49.08 -51.07 -44.70 -64.96 -58.54

-26.18 -26.20 -26.20 -26.32 -26.66 -22.86 -22.83 -25.13 -24.22 -26.11 -26.79 -26.69 -26.76 -26.29 -26.79 -26.76

-67.21

40.42

0.61

-67.21

40.42

0.59

-67.21

40.42

0.59

-67.21

40.42

0.47

-67.21

40.42

0.13

-67.21

40.42

3.93

-67.21

40.42

3.96

-67.21

40.42

1.66

-67.21

40.42

2.57

-67.21

40.42

0.68

-67.21

40.42

0.00

-67.21

40.42

0.10

-67.21

40.42

0.03

-67.21

40.42

0.50

-67.21

40.42

0.00

-67.21

40.42

0.03

81.22 85.84 81.83 88.04 71.36 71.59 71.31 91.32 85.20 73.80 66.80 70.95 69.76 74.90 64.61 65.57

4.48 4.82 3.21 5.28 1.75 1.75 -0.12 2.72 4.04 -1.00 -12.94 -4.81 -8.07 3.92 -27.13 -19.73

OMT +/- 150 Trading Days (≈ 6M) ERS RAI KBC DEU COM BSN BBV SOC BNP CRE BOI INT UNI ING BCP BES

-56.13 -59.49 -58.40 -62.02 -46.04 -53.14 -53.87 -68.02 -61.49 -51.57 -58.80 -48.43 -48.19 -49.18 -69.88 -63.69

-16.42 -16.57 -17.34 -15.59 -15.03 -8.53 -8.58 -16.90 -15.30 -15.13 -15.19 -14.50 -14.40 -15.00 -14.44 -14.75

-9.12

-6.03

-1.27

-9.12

-6.03

-1.42

-9.12

-6.03

-2.19

-9.12

-6.03

-0.44

-9.12

-6.03

0.12

-9.12

-6.03

6.62

-9.12

-6.03

6.57

-9.12

-6.03

-1.75

-9.12

-6.03

-0.15

-9.12

-6.03

0.02

-9.12

-6.03

-0.05

-9.12

-6.03

0.65

-9.12

-6.03

0.75

-9.12

-6.03

0.15

-9.12

-6.03

0.71

-9.12

-6.03

0.40

24

66.69 71.67 67.34 74.06 56.07 56.27 55.94 77.57 71.01 58.70 51.04 55.59 54.30 59.91 48.62 49.67

-5.85 -4.39 -8.40 -3.55 -4.99 -5.41 -6.51 -7.34 -5.79 -8.00 -22.95 -7.34 -8.29 -4.26 -35.70 -28.77

Volume V Issue 1(9)

Table 11. Pooled OLS results for the LTRO’s effect LTRO +/- 25 Trading Days (≈ 1M) #Obs.

Itr.

AT

BE

DE

SP

816

418.98 1.61

0.64 1.58

0.39 1.11

-0.99 -0.65

-0.26 -0.80

10.706%

11.346%

26.707%

51.859%

42.620%

lAT

lBE

lDE

lSP

-0.69 -0.87

-0.15 -0.13

-1.15 -0.49

0.63 1.20

38.724%

89.373%

62.751%

23.149%

F1

F2

F3

F4

AdjR² 99.66%

Lg1 0.62 *** 20.87 0.000%

Fisher

LD

2 163.90 -161.69 +++ -0.43 66.963%

p-Value 0

FR 1.07 ** 2.22 2.684%

lFR -1.55 ** -2.24 2.565%

F5

IE

IT

-0.24 -0.59

0.11 0.29

55.217%

77.329%

lIE

lIT

NL -0.39 * -1.78 7.551%

lNL

PT -0.10 -1.23

SPi 22.48 *** 3.39

RTi

VSX

1.34 0.82

-2.48 -1.29

22.059%

0.075%

41.009%

19.598%

lRTi

lVSX

lPT

lSPi

0.50 0.90

-0.12 -0.26

0.34 1.13

-0.09 -0.31

-5.61 -1.52

-0.20 -0.22

4.06 1.63

37.007%

79.141%

25.787%

75.930%

13.013%

82.706%

10.312%

F6

F7

F8

F9

F10

F11

F12

-405.96 *** -437.30 *** -391.48 *** -502.74 *** -420.97 *** -453.28 *** -428.99 *** -458.43 *** -492.16 *** -449.33 *** -9.93 -9.65 -9.77 -8.27 -11.09 -9.51 -9.40 -7.41 -8.88 -10.20

BOI -210.24 ** -2.40 1.686%

lBOI 387.03 * 1.83 6.821%

F13

BBB 0.22 0.76 44.848%

lBBB -0.62 -1.37 17.026%

F14

F15

97.72 *** -340.08 *** -310.88 *** -476.39 *** 181.58 *** 7.97 -9.86 -9.38 -10.47 10.47

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

F1-l

F2-l

F3-l

F4-l

F5-l

F6-l

F7-l

F8-l

F9-l

F10-l

F11-l

F12-l

F13-l

F14-l

F15-l

20.02 1.18

23.13 1.25

62.54 *** 3.74

21.63 0.92

26.01 1.43

31.51 1.43

38.38 * 1.88

37.75 1.55

35.11 * 1.65

23.762%

21.019%

0.020%

35.639%

15.186%

15.424%

AT

BE

DE

SP

FR

6.100%

12.133%

9.929%

27.87 1.40 16.236%

-62.48 *** -4.21 0.003%

18.18 1.05

9.56 0.53

27.14 1.34

16.43 1.28

29.407%

59.890%

18.092%

20.217%

VSX

BOI

BBB

LTRO +/- 75 Trading Days (≈ 3M) #Obs. 2416

Itr. 149.32 *** 3.12 0.185%

AdjR² 99.28%

Lg1 0.87 *** 100.15 0.000%

Fisher

LD

5 531.67 -161.17 ** +++ -2.52 1.173%

p-Value 0

0.18 ** 1.78 7.596%

lAT -0.67 ** -2.06 3.992%

F1

0.03 0.29

-0.15 -0.53

77.382%

59.781%

lBE

lDE

0.36 *** 3.35 0.082%

lSP

-0.06 -0.25 80.048%

lFR

IE -0.12 *** -3.42 0.063%

lIE

0.27 1.46

-0.32 -0.56

0.28 1.57

-0.42 -1.41

0.07 0.89

14.468%

57.400%

11.750%

15.900%

37.431%

F2

F3

F4

F5

F6

IT

NL

0.00 -0.06

-0.09 -0.56

95.056%

57.537%

lIT

lNL

-0.20 * -1.76 7.873%

F7

PT -0.05 * -1.68 9.266%

lPT

SPi 6.52 ** 2.26

RTi -0.38 -0.72

2.394%

47.397%

lSPi

lRTi

1.86 *** 3.90 0.010%

lVSX

0.23 1.18

0.04 1.20

-0.40 -0.17

0.17 0.26

0.05 0.06

23.695%

23.156%

86.516%

79.301%

95.427%

F8

F9

F10

F11

F12

-114.77 *** -119.82 *** -112.22 *** -134.69 *** -128.73 *** -139.82 *** -133.24 *** -116.07 *** -139.19 *** -135.02 *** -7.29 -7.18 -7.30 -6.57 -8.93 -6.95 -6.76 -5.49 -6.48 -7.85

-23.54 -0.94 34.477%

lBOI 163.64 *** 3.48 0.051%

F13

-0.31 *** -5.13 0.000%

lBBB 0.05 0.34 73.364%

F14

44.24 *** -104.20 *** -101.73 *** -141.64 *** 5.94 -7.56 -7.14 -8.13

F15 64.65 *** 9.29

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

F1-l

F2-l

F3-l

F4-l

F5-l

F8-l

F9-l

F10-l

F11-l

F12-l

F13-l

F14-l

F15-l

F6-l

F7-l

-0.31 -0.03

2.72 0.21

10.80 1.00

1.36 0.08

3.87 0.32

12.43 0.83

16.57 1.18

2.21 0.13

4.36 0.30

9.99 0.78

97.794%

83.420%

31.575%

93.433%

74.698%

40.489%

24.001%

90.051%

76.545%

43.480%

AT

BE

DE

SP

FR

0.07 0.69

-0.04 -0.39

48.787%

69.920%

lAT

lBE

-55.75 *** -5.71 0.000%

1.85 0.17

1.08 0.09

4.02 0.31

86.255%

92.571%

75.959%

BOI

BBB

-24.80 *** -3.17 0.154%

LTRO +/- 150 Trading Days (≈ 6M) #Obs.

Itr.

4816

69.14 * 1.91 5.645%

AdjR² 98.68%

Lg1 0.90 *** 146.62 0.000%

Fisher

LD

6 010.81 +++

9.35 0.21 83.430%

p-Value 0

-0.13 -0.66 50.787%

F1

0.33 ** 2.35 1.877%

F2

0.46 * 1.80 7.146%

lDE

0.30 *** 4.26 0.002%

lSP

-0.61 -1.38

0.08 0.71

16.905%

47.781%

F3

F4

-0.18 -1.13 25.829%

lFR -0.42 * -1.80 7.230%

F5

IE -0.07 *** -3.86 0.012%

lIE 0.04 0.85 39.615%

F6

IT 0.13 * 1.85 6.428%

lIT -0.24 ** -2.22 2.679%

F7

NL

PT

-0.09 -0.54

0.02 0.82

SPi 3.01 1.14

RTi -0.09 -0.39

59.220%

41.191%

25.320%

69.994%

lNL

lRTi

VSX 0.50 * 1.70 8.836%

lVSX

-15.15 -0.71 47.849%

lBOI

-0.28 *** -4.72 0.000%

lPT

lSPi

0.29 1.48

-0.03 -1.07

-1.84 -0.78

0.08 0.15

13.881%

28.612%

43.686%

87.944%

0.064%

0.000%

F8

F9

F10

F11

F12

F13

F14

F15

-92.60 *** -6.12

58.42 *** 8.66

-76.48 *** -6.20

-73.76 *** -5.52

-97.12 *** -6.41

37.32 *** 7.16

1.74 *** 164.04 *** 3.42 5.23

lBBB -0.24 ** -2.10 3.576%

-85.26 *** -6.31

-87.61 *** -6.52

-80.50 *** -6.11

-96.37 *** -5.84

-88.32 *** -7.26

-92.85 *** -5.02

-88.18 *** -4.81

-84.48 *** -4.98

-96.79 *** -4.96

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

F1-l

F2-l

F3-l

F4-l

F5-l

F6-l

F7-l

F8-l

F9-l

F10-l

F11-l

F12-l

F13-l

F14-l

F15-l

12.86 1.09

12.14 0.91

16.85 1.49

12.82 0.76

14.13 1.22

25.75 * 1.66

28.01 * 1.85

15.68 0.92

18.99 1.19

27.708%

36.524%

13.542%

44.468%

22.114%

35.854%

23.403%

9.664%

6.469%

25

21.78 * 1.67 9.443%

-69.52 *** -7.63 0.000%

20.94 * 1.95 5.082%

20.94 * 1.77 7.760%

16.96 1.28 20.079%

-14.54 ** -2.10 3.609%


Table 12. Pooled OLS results for the OMT’s effect OMT +/- 25 Trading Days (≈ 1M) #Obs. 816

Itr. 0.00 *** 0.00 0.000%

AdjR² 99.66%

Lg1 0.45 *** 14.01 0.000%

Fisher

OD

16 561.99 -103.15 +++ -0.82 41.299%

p-Value 0

AT 0.69 ** 1.70 8.884%

BE -0.11 -0.33 73.936%

lAT

lBE

-3.18 *** -2.72

-0.36 -0.68

0.672%

F1

49.952%

F2

DE -2.34 *** -2.65 0.818%

lDE 3.76 *** 2.63 0.868%

F3

SP 0.67 *** 2.74 0.636%

lSP -0.84 ** -2.37 1.819%

F4

FR -1.07 *** -3.95 0.009%

lFR 1.03 ** 2.09 3.651%

F5

IE

IT

0.13 1.00

-0.29 -0.96

31.546%

33.833%

lIE

lIT

0.39 1.31

0.99 ** 2.29

19.078%

2.202%

F6

F7

NL 0.46 * 1.87 6.172%

lNL -3.15 *** -3.71 0.022%

F8

PT -0.02 -0.16

SPi 35.76 *** 3.96

RTi 2.48 *** 2.91

VSX 1.27 ** 2.07

87.559%

0.008%

0.374%

3.923%

lPT

lSPi

lRTi

lVSX

0.01 0.06 95.421%

F9

-3.11 ** -2.33

0.78 * 1.91

2.030%

5.607%

F10

F11

-609.56 *** -649.29 *** -542.09 *** -700.84 *** -552.02 *** -473.28 *** -401.34 *** -630.50 *** -686.69 *** -561.67 *** -8.71 -8.98 -8.64 -8.65 -8.04 -8.06 -8.18 -8.49 -7.85 -7.95

1.42 1.33 18.252%

F12

BOI 109.30 ** 2.26 2.420%

lBOI 284.29 * 1.77 7.705%

F13

BBB -0.20 -0.74 46.221%

lBBB 0.01 0.01 98.858%

F14

-80.82 *** -376.31 *** -343.24 *** -644.05 *** -4.33 -7.75 -7.16 -8.00

F15 75.24 *** 8.51

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.000%

0.002%

0.000%

0.000%

0.000%

0.000%

F1-l

F2-l

F3-l

F4-l

F5-l

F6-l

F7-l

F8-l

F9-l

F10-l

F11-l

F12-l

F13-l

F14-l

F15-l

27.78 *** 3.28

16.92 * 1.70

14.11 * 1.74

24.57 ** 2.42

31.76 *** 3.25

0.109%

8.915%

8.154%

1.581%

0.121%

AT

BE

DE

SP

FR

15.59 * 1.92 5.559%

9.17 1.28

-3.74 -0.31

11.67 1.11

19.921%

75.600%

26.782%

26.62 *** 2.78 0.549%

2.15 0.38

5.99 0.81

-5.62 -0.72

70.069%

42.007%

46.960%

RTi

VSX

BOI

39.76 *** 3.82 0.014%

4.39 0.82 41.202%

OMT +/- 75 Trading Days (≈ 3M) #Obs. 2416

AdjR² 99.58%

Itr. 157.04 *** 5.01 0.00% Lg1

0.58 *** 4.08 0.00%

-0.44 *** -3.61 0.03%

-1.30 *** -4.87 0.00%

0.15 ** 2.40 1.64%

-0.33 *** -2.75 0.60%

lAT

lBE

lDE

lSP

lFR

0.92 *** 0.08 114.33 0.23 0.000% 81.58%

Fisher

OD

9 477.56 +++

-58.54 * -1.67 9.59%

F1

IE

IT

0.15 *** 0.12 3.32 1.60 0.09% 10.93% lIE

lIT

NL

PT

SPi

0.35 *** 0.01 3.15 0.80 0.16% 42.46%

-2.44 * -1.82 6.82%

lNL

lSPi

lPT

0.82 *** 1.54 *** -0.46 *** 0.68 *** -0.35 *** 0.46 *** -1.82 *** -0.11 *** 0.79 4.07 3.48 -3.72 3.53 -5.57 3.62 -6.96 -5.78 0.81 0.005% 0.050% 0.021% 0.042% 0.000% 0.030% 0.000% 0.000% 41.75% F2

F3

F4

F5

F6

F7

F8

F9

F10

0.10 *** 0.28 77.73%

BBB

0.73 *** 192.69 *** 2.61 6.43 0.91% 0.00%

-0.78 *** -8.52 0.00%

lRTi

lVSX

lBOI

lBBB

0.50 1.39 16.46%

-0.62 -1.26 20.90%

68.05 1.16 24.77%

F11

F12

F13

F14

F15

-16.38 ** -2.28 2.29%

-12.09 * -1.70 8.95%

-29.92 *** -2.59 0.97%

10.75 *** 3.62 0.03%

F12-l

F13-l

F14-l

F15-l

13.84 * 1.94 5.25%

-6.42 * -1.79 7.34%

-31.69 *** -2.91 0.36%

-35.01 *** -2.76 0.58%

-27.69 *** -2.79 0.54%

-35.69 ** -2.47 1.35%

-27.56 *** -2.75 0.60%

-22.87 *** -2.60 0.93%

-19.52 *** -2.60 0.93%

-28.41 ** -2.10 3.59%

-27.02 ** -2.06 3.96%

-24.82 ** -2.45 1.44%

0.89 0.26 79.64%

p-Value

F1-l

F2-l

F3-l

F4-l

F5-l

F6-l

F7-l

F8-l

F9-l

F10-l

F11-l

0

7.98 1.42 15.64%

3.71 0.50 61.52%

6.11 1.13 25.79%

2.09 0.26 79.67%

15.58 ** 2.24 2.52%

11.55 ** 2.12 3.38%

9.93 ** 2.14 3.27%

-4.93 -0.53 59.30%

1.60 0.23 81.66%

9.85 1.53 12.69%

5.59 1.50 13.31%

AT

BE

DE

SP

SPi

RTi

-2.62 *** -4.12

-0.24 -1.30

9.46 * 1.91 5.65%

7.47 1.41 15.75%

0.49 *** 3.57 0.036%

OMT +/- 150 Trading Days (≈ 6M) #Obs.

Itr.

4816

93.72 *** 6.44 0.000%

AdjR² 99.56%

Lg1 0.93 *** 208.18 0.000%

Fisher

OD

18 009.58 -63.69 *** +++ -3.87 0.011%

p-Value 0

0.02 0.33 74.147%

lAT 0.94 *** 4.19 0.003%

F1

0.19 *** 4.45 0.001%

-0.11 -0.80 42.616%

0.37 *** 9.86 0.000%

lBE

lDE

lSP

-0.02 -0.13

-0.23 -0.70

-0.39 *** -5.38

89.836%

48.531%

F2

F3

0.000%

F4

FR -0.52 *** -8.67 0.000%

lFR 0.48 *** 3.54 0.041%

F5

IE

IT

-0.08 *** -4.08 0.005%

-0.09 ** -2.00 4.555%

lIE

lIT

-0.03 -0.92

0.30 *** 4.33

35.553%

0.002%

F6

F7

NL 0.20 *** 4.56 0.001%

PT 0.01 1.24 21.669%

0.004%

19.339%

lNL

lPT

lSPi

lRTi

-0.94 *** -4.67

-0.07 *** -4.63

0.000%

F8

0.000%

F9

0.87 1.18 23.656%

-0.74 -0.36

-17.34 *** -4.53

-15.69 *** -4.15

-23.78 *** -4.25

0.087%

0.019%

0.635%

0.000%

0.045%

0.046%

6.108%

1.175%

0.004%

F1-l

F2-l

F3-l

F4-l

F5-l

F6-l

F7-l

F8-l

F9-l

F10-l

17.65 *** 3.26

10.55 ** 2.44

80.665%

0.113%

1.472%

0.950%

26

55.616%

71.252%

0.50 *** 7.83

-20.62 *** -4.14

0.008%

26.529%

lBBB

F14

-16.72 ** -2.52

47.748%

lBOI

-1.28 *** -141.03 *** -4.59 -4.50

0.000%

-13.72 * -1.87

12.602%

lVSX

F13

-15.21 *** -3.50

2.20 0.37

0.000%

0.001%

-17.19 *** -3.51

-4.33 -0.59

0.000%

F12

-25.08 *** -5.13

9.82 *** 2.59

0.000%

0.000%

-20.87 *** -2.73

1.67 0.24

-0.47 *** -8.93

F11

-18.73 *** -3.74

5.29 1.11

BBB

2.103%

-21.31 *** -3.33

4.20 0.71

BOI

1.51 *** 141.18 *** 8.75 14.71

F10

-21.61 *** -3.96

7.57 1.53

0.54 ** 2.31

VSX

12.12 ** 2.36 1.844%

72.133%

F11-l 4.89 * 1.78 7.532%

F15 8.52 *** 4.12

0.001%

0.003%

0.002%

0.004%

F12-l

F13-l

F14-l

F15-l

15.26 *** 3.76 0.017%

15.50 *** 3.64 0.027%

14.51 ** 2.54 1.126%

-6.19 ** -2.28 2.278%

Volume V Issue 1(9) Summer 2014 DOI: http://dx.doi.org/10.14505/jasf.v5.1(9).02

MICROECONOMIC DETERMINANTS OF LOSSES IN FINANCIAL INSTITUTIONS DURING THE CRISIS Alessandra CEPPARULO University of Urbino, Faculty of Economics, Italy [email protected] Antonio FORTE University of Bari, Italy [email protected] Suggested Citation: Cepparulo, A. (2014). Microeconomic determinants of losses in financial institutions during the crisis, Journal of Advanced Studies in Finance, (Volume V, Summer), 1(9):27-37. Doi:10.14505/jasf.v5.1(9).02. Available from: http://www.asers.eu/journals/jasf/curent-issue. Article’s History: Received March, 2014; Revised April, 2014; Accepted July, 2014. 2014. ASERS Publishing. All rights reserved.

Abstract: This paper tries to find out the determinants of bank losses and their extent during the sub-prime crisis, analysing banks structure and performance. The results of the proposed probit models confirm the new tendencies of the international regulations inspired both by the Basel III Accord and the Financial Stability Board: sound patrimonial ratios together with safer assets structures are prerequisites for the “too big to fail” financial institutions and they favor a decrease in the probability of suffering losses. Keywords: bank, tier1, financial crisis, losses JEL Classification: G21, G01 1. Introduction The sub-prime crisis caused financial turmoil all over the World. Its impact, as known, not only affected financial markets but had deep consequences on real economy too. This double impact produced severe aftermath on banking sectors, especially in the US and European markets. Financial institutions recorded unprecedented losses and the banking sector was particularly menaced, due to large write-downs on risky mortgage related positions. Indeed, from the beginning of the financial crisis (second quarter of 2007) to the spring of 2011, worldwide banking credit losses and write downs related to sub-prime crisis were $1.5 trillion as reported in the Write-downs and Credit Losses (WDCI)2 list by Bloomberg. Given this scenario, this paper investigates the possible determinants of bank performance, measured by write downs and credit losses during the crisis, using banks’ balance sheets and profitability characteristics for a sample of large banks. The rest of the paper is organized as follows: section 2 shows data on write downs and credit losses. Section 3 presents the literary background while section 4 and 5 describe the banking institution sample and the determinants selected variables. Section 6 and 7 show the empirical method and the results respectively. Section 8 draws conclusions and proposes policy recommendations. 2. Write downs and credit losses A large heterogeneity in losses distribution can be observed (Table 1 and Table 2). Banks operating in North America reported the highest level of write-downs among the institutions included in the WDCI list. In order to compare data among continents, the ratio between write-downs and total assets is also taken into account.

2

WDCI list by Bloomberg collects sub-prime related write-downs by banks worldwide from the second quarter 2007 ahead. 27

Journal of Advanced Studies in Finance This analysis confirms the fact that the sub-prime crisis had its deepest effect on the banks located in North America. Indeed, as shown in Table 1, this ratio is equal to 5.2% in North America, to 1.4% in Europe and to 0.5% in Asia. This primacy is valid not only in terms of total losses but also in terms of unit value of write-downs, as Europe gathers the highest number of banks (59 banks). Table 1. Write downs and Credit losses by Continents (Q2 2007-May 2011) Losses ($ billions)* Losses to total assets** Number of banks North America 801,1 5,2% 34 Europe° 654,7 1,4% 59 Asia 34,6 0,5% 10 Worldwide°° 1492,4 2,2% 105 Source: Elaboration on Bloomberg WDCI and banks web sites data. Notes: * Data on losses are summed up and then harmonized using the exchange rates of national currencies against the US dollar in May 2011, as reported by Bloomberg; ** Data on total assets are referred to the end of 2007. We transformed European and Asian data on total assets in US dollars in order to use the same currency on numerator and denominator of the ratios in column two; Europe includes the following countries banks: UK, Germany, France, Italy, Spain, Austria, the Netherlands, Denmark, Ireland, Greece, Belgium, Sweden, Switzerland, and Norway. We disregarded Landesbank Sachsen Bank, because of the impossibility to find its total assets at the end of 2007 and we also disregarded the losses of the “Other Asian banks” because of the vagueness of this notation; Worldwide also includes Island and Israel.

By focusing on European countries, see Table 2, Ireland, Switzerland and Greece show the highest losses to total assets ratio. Moreover, among the biggest economies, the lowest values have been recorded in France (0.8%) and Italy (1.2%), while the highest in UK (2.0%) and Spain (1.8%). Table 2. Credit losses and Write downs inside Europe (Q2 2007-May 2011) Losses Losses to total Number of ($ billions )* assets** banks UK 211,2 2,0% 9 Germany° 112,5 1,3% 15 Switzerland 82 2,6% 2 France 62,4 0,8% 5 Belgium-the Netherlands 57,2 0,9% 7 Spain 39,4 1,8% 3 Italy°° 37,5 1,2% 4 Ireland 24,2 3,5% 3 Denmark 10,6 1,6% 3 Greece 9,3 2,5% 4 Austria 5,1 1,5% 2 Norway 2,8 1,0% 1 Sweden 0,5 0,1% 1 Europe 654,7 1.4% 59 Source: Elaboration on Bloomberg WDCI and banks web sites data. Notes: * Data on losses are summed up and then harmonized using the exchange rates of national currencies against the US dollar in May 2011, as reported by Bloomberg; ** Data on total assets are referred to the end of 2007. We transformed European data on total assets in US dollars in order to use the same currency on numerator and denominator in column two; We disregarded Landesbank Sachsen bank because of the impossibility to find its total assets at the end of 2007; Italy includes HVB.

3. Literature review The bursting of the housing bubble forced banks to write down billion dollars due to subprime crisis. The economic literature not only tries to understand if write-downs were timely and value-relevant (Beltratti, Spear and Szabo 2010; ABA, 2009; Hellwig, 2009) but it also tries to draw an identikit of a poor performance bank during the

28

Volume V Issue 1(9) Summer 2014 crisis, as if certain banks can be more predisposed to experience large losses. This strand advocates the role for several determinants: regulation (Dooley, Folkerts-Landau, Garber, 2009; Stiglitz, 2010), bank characteristics (Acharya and Schnabl, 2009; Adrian and Shin, 2008; Brunnermeier, 2009; Gorton, 2010), bank governance (Diamond and Rajan, 2009; Bebchuk and Spamann, 2010; Kirkpatrick, 2008), country characteristics (Beck and Levine, 2004; Levine, 2004; Stulz and Williamson,2003) and macroeconomic conditions (Arpa et al., 2001; Bikker and Hu, 2002). Empirical results show how the role of banks structure before the crisis, more than governance and/or regulation, is the key player in explaining the performance of banks during the crisis, measured mainly by buy-andhold dollar returns and write-downs (Beratti and Stulz, 2009, 2011). By focusing on bank characteristics, although the selection of variables is always conditioned by data availability, the following six are the main observed (Beltratti and Stulz, 2009, 2011): the bank capital consistence, the reliance on short term finance, the liabilities composition, the bank’s risk, the dimension and the bank ownership. As expected, empirical results confirm the positive role of banks’ capital and provisioning process (Beltratti and Stultz, 2009). So the higher the capitalization and the reliance on deposits for the financing, the higher the performance is. Indeed, capital is a natural safeguard against uncertainty as testified also by the stress test for adverse economic and financial market conditions, applied to the 19 largest bank holding companies in the spring of 2009 (see Board of Governors 2009). More capital implies improving the ability to cope with worse-thanexpected economic conditions and losses. Besides, this result seems determined by the dimension of the banking institutions and by the measures of leverage used. While the Tier 1 ratio is significantly positive connected to the performance, independently from the banking institution dimension, the other two leverage measures, appeared in literature, ratios of equity to assets and tangible equity to liabilities, become not significantly connected with the performance when large banks are considered. A sort of similar cushion against adversity is supplied by deposit financing strategy, measured by the ratio of deposits to assets (Brunnermeier, 2009), compared to money market funding, measured by the ratio of money market funding to assets. Indeed during a recession, the lending channels can be dried up by the individual prudent action causing an unintentional systemic problem. The provisioning process has a relevant role even when a measure of short term funding, following on Demirgüç-Kunt and Huizinga (2009), is considered, whatever is the composition of the sample: depositary banks or merchant banks (Beltratti and Stulz, 2011). Again, during adverse conditions a positive connection with performance is expected for banking institutions which prefer loans to securities, measured by the ratio of loans to total assets. The last ones are the fruits of the “originate to distribute” banking model (Brunnermeier, 2009). Although securities were born with the purpose of transferring risk to the most able to bear, they ended with favoring poor lending standards and to make the banking system more unstable. In order to cope with financial turmoil, also the liquidity, defined as the ratio of liquid assets to total assets, is positively connected to a better performance. The last two determinants are then: the dimension of the banking institution, as Beltratti and Stultz (2011) showed that larger banks with high leverage performed worse, and a profitability measure, as this surely fell down over the period 2006-2008. This paper differs from the previous literature because it tries not only to look for factors determining a bad performance during the crisis but it also tries to identify the potential causes of the differences among bad performers. 4. Data The sample comprehends 543 banking institutions worldwide distributed (Table 3): 44 banks, out of 1053, come from the WDCI list published by Bloomberg, and the other 499 banks come from the Bloomberg database. The sample selection assures a worldwide coverage and a heterogeneous representation of the banking system, in particular in terms of best and worst performers during the crisis (see Table 4).

3

Data availability reasons has limited the number of banks within the WDCI list to be used in our study. 29

Journal of Advanced Studies in Finance Table 3. Geographical distribution of the sample Banks in the Banks outside the list list North 23 337 America

Europe 15 46 Asia 6 89 Worldwide 44 499* Sources: Bloomberg and Bloomberg WDCI Notes: * Worldwide includes also Central and South American Countries, Israel and Jamaica.

ROA Number of employees Total assets (Millions $) Interbank assets (Millions $) Tier 1 capital ratio

Table 4. Characteristics of the banks in the sample Banks in the list Banks outside the list Max. Min. Mean Max. Min. 3.4 -0.07 1.0 14.6 -3.2

Mean 1.34

313000

963

50098

91320

24

2563

1703162

9956

473874

399507

104

10768

291090

0

39418

33215

0

825

14.7

5.94

8.65

34.49

2.42

12.18

Sources: Bloomberg and Bloomberg WDCI; Notes: The values are the simple mean of the 2005-2006 quarterly data.

Indeed the worst performers come from the WDCI list, mainly composed by large banking institutions with assets around 473,874 US millions on average, high number of employees (over 50,000 employees on average) and a large interconnection in the interbank channel (on average 39,418 US millions). The best performers are instead represented by the banking institutions coming from the Bloomberg database. These are banks with an average number of employees equals to 2,563 and with total assets and exposure to interbank markets both equal to around 2% of the mean banking institution in WDCI list. Differently from the WDCI list banks, these banks have a higher level of capitalization (on average 12.18 vs 8.65). Besides the write-downs data of the 44 banks of the WDCI list used in this study refer to the following horizon: from the second quarter of 2007 to the end of May 2011. 5. Banks determinants Following Belratti and Stulz (2009, 2011), our model is based on a capital measure, tier 1 ratio, while the exposure to the risk of losses, differently from them, is captured by a profitability measure (Golin, 2001), return on assets (ROA), which appears inversely related to proportion of loan loss provisions, by following Staikouras and Wood (2003). As our aim is to find the existence of a correlation between risky assets and losses, related to subprime crisis, ROA is a perfect candidate because it expresses the gain of a bank related to its activities and normally the higher the risk the higher the return. Compared to other variables, for example the return on equity (ROE), ROA shows also the advantage to be less affected by the balance sheet rules used in the countries and by the moral suasion of the authorities. Indeed, equity is often influenced by the choices of managers and shareholders to raise or not to raise capital, by rules on capital requirements adopted in different markets and by the severity of the authorities. A special attention is given instead to the dimension factor, measured by two variables, total assets and number of employees, and to the contagion factor, measured by interbank assets. The size, represented by total assets, mostly used in banks performance literature4, represents the level of potential economies to be used. These potential economies can assume two forms: economies of scale, through the reduction of costs, and scope economies, through both the entrance into markets with size limits access and a better diversification of products and loans (Kosmidou et al., 2006, Barros et al., 2007). For our purpose, the size of a bank 4

The evidence is not certain as a strand of literature is in favor of a positive connection of bank size and economies of scale in case of large banks (European Commission, 1997; Berger and Humphrey, 1997; Altunbas et al., 2001), while another strand supports the idea that just small banks are positively influenced by the size while large one are negative influenced (Vander Vennet, 1998; Pallage, 1991). 30

Volume V Issue 1(9) Summer 2014 can be strictly connected to the higher probability of making negative business, of lending money to unreliable debtors and of a reduced control on the whole structure of the bank. The second measure of the size, the number of employees, has the role of taking into account the fact that the risk of losses can result even from inadequate internal processes, people and/or systems, as expressed by the Basel Committee on Banking Supervision (BCBS). Chernobai and Yu (2008) seem to confirm the relevance of the number of employees as one of the determinants of operational risk, by founding a significant and concave relation between operational risk and the human factor. We separately use these two measures of the size in the regressions because we found high correlation between these data. The interbank assets5 variable is, according to our view, another possible source of danger for the bank soundness. The literature is not unambiguous on the evaluation of the interbank assets, indeed, while Allen and Gale (2000) and Freixas et al. (2000) consider interbank assets as a source of contagion, Rochet and Tirole (1996) and Calomiris, (1998) consider it a source of prudence, as banks are better in evaluating other banks status and so this lending favors less risky behavior. We employ this variable both for the extended and restricted sample that includes only the banks in the WDCI list. Then, in order to inspect the role of the capital in ensuring the robustness of the banks’ financial position, the Tier1 ratio is considered. Indeed, not only it detects the ability to absorb losses but it also measures the soundness of a bank. Hence, a database comprehensive of the above mentioned variables is used based on Bloomberg data availability In particular all these variables, used as independent variables, enter our explicative models as the simple mean of the 2005-2006 quarterly data (the two years before the onset of the crisis). 6. Model Differently from Beltratti and Stulz (2011) our approach consists in searching for the role of certain determinants in increasing the probability to be part of bad performers (WDCI list). After having defined the variable y as a dummy variable coded “1” for banks in the list or “0” for the opposite case, and xi, with i=1-6, the set of explanatory variables, the probability function of y conditional on the regressors is specified by a probit model. To be in the WDCI list implies to suffer losses related to 2007-2009 financial crisis, which is a latent variable y* generated by a regression: y*= α + βxi + ε

(1)

where α is a constant, xi represents the covariates (total assets, ROA, number of employees, interbank assets, tier 1 capital ratio and a dummy) and ε is a normally distributed random term. So the probability of being in the WDCI list can be expressed as: y = 1 if y* > 0 i.e εi < α + βxi , y = 0 otherwise.

(2)

Pr(y=1|xi)=Pr(εi chi2 = 0.0000 Regressors Coefficients Standard error Z Prob. > |Z| Tier 1 -0.1958658 0.0714 -2.74 0.006 ROA 0.4296135 0.2050 2.09 0.036 Tot. Assets 0.0000118 1.71e-06 6.88 0.000 Constant -0.6418107 0.6362 -1.01 0.313 Marginal effect on regressors mean values: tier 1=11.8, ROA=1.23, tot assets=50037 y = Pr (WDCI1 = 1) = 0.0319 dy/dx Standard error Z Prob. > |Z| Tier 1 -0.0140404 0.00365 -3.85 0.000 ROA 0.0307963 0.0138 2.23 0.026 Tot assets 8.44e-07 0.0000 2.26 0.024 Marginal effect for the following regressors values: Tier 1= 5.9, ROA=2.5, Tot assets=100000 y = Pr (WDCI1 = 1) = 0.6752 Tier 1 -0.0704698 0.01597 -4.41 0.000 ROA 0.1545691 0.04932 3.13 0.002 Tot assets 4.24e-06 0.00000 4.00 0.000 Table 6. Probit model on the full sample with number of employees as regressor Dependent variable: WDCI1. Pseudo R2: 0.6230 Total number of banks: 434; number of WDCI banks: 39 LR chi2 = 163, Prob > chi2 = 0.0000 Regressors Coefficients Standard error Z Prob. > |Z| Tier 1 -0.2901069 0.072 -4.02 0.000 ROA 0.1759923 0.260 0.68 0.499 Num.employ. 0.0000713 9.83e-06 7.25 0.000 Constant 0.599045 0.582 1.03 0.304 Marginal effect on regressors mean values: Tier 1=11.8, ROA=1.21, Num employ=6303 y = Pr (WDCI1 = 1) = 0.0144 dy/dx Standard error Z Prob. > |Z| Tier 1 -0.0106 0.004 -2.76 0.006 ROA 0.0064 0.009 0.73 0.467 32

Volume V Issue 1(9) Summer 2014 Num.employ. 2.61e-06 0.000 1.93 0.053 Marginal effect for the following regressors values: Tier 1= 5.9, ROA=2.5 and Num employ.=50000. y = Pr (WDCI1 = 1) = 0.9980 Tier 1 -0.0017 0.003 -0.58 0.562 ROA 0.0010 0.001 0.73 0.468 Num.employ. 4.37e-07 0.000 0.55 0.580 Table 7. Probit model on the full sample with interbank asset as regressor Dependent variable: WDCI1. Pseudo R2: 0.6456 Total number of banks: 450; number of WDCI banks: 37. LR chi2 = 165.11, Prob > chi2 = 0.0000 Regressors Coefficients Standard error Z Prob. > |Z| Tier 1 -0.1594107 0.0744 -2.14 0.032 ROA 0.4389796 0.2087 2.10 0.035 Tot. Assets 0.0000165 3.74e-06 4.42 0.000 Interbank assets -0.0000438 0.0000 -1.44 0.150 Constant -1.106258 0.6570 -1.68 0.092 Marginal effect on regressors mean values: tier 1=11.8, ROA=1.22, tot assets=42843, interbank assets=4073. y = Pr (WDCI1 = 1) = 0.0264 dy/dx Standard error Z Prob. > |Z| Tier 1 -0.0097475 0.0032 -3.06 0.002 ROA 0.0268423 0.0112 2.40 0.017 Tot assets 1.01e-06 0.0000 1.99 0.047 Interbank assets -2.68e-06 0.0000 -1.16 0.245 Marginal effect for the following regressors values: Tier 1= 5.9, ROA=2.5, Tot assets=100000, interbank assets=8000. y = Pr (WDCI1 = 1) = 0.6381 Tier 1 -0.0597451 0.0199 -3.00 0.003 ROA 0.1645239 0.0559 2.94 0.003 Tot assets 6.20e-06 0.0000 3.32 0.001 Interbank assets -0.0000164 0.0000 -1.34 0.179 7.2 Bank’s structure and performance as determinants of intra-banks differences in losses While in the previous sections we have tried to study the causes of the different results of the banks affected by the same negative economic phenomenon, the sub-prime crisis, in this paragraph we inspect the causes of the ample heterogeneity in losses among the banks inside the WDCI list, the bad performers. Therefore, we consider a new sample composed of 44 banks in the WDCI list and we grouped these banks on the base of the severity of their losses through a dummy variable which assumes value “1” or “0”, if the losses are higher or lower than 9 million6 respectively. . Then, a new probit model is estimated by using the same regressors we have chosen in the previous steps, in order to study the effect of these regressors over the probability of losing more than 9 million dollars. After numerous attempts, the model showed in table 8 proved to be the best in terms of parsimony and efficiency. It includes three regressors, tier 1 ratio, ROA and interbank assets7, which are all significant and with the expected signs, positive for ROA and interbank assets and negative for Tier1 ratio. Interesting results emerge by looking at the marginal effects. Indeed, with the regressors at their mean values, the probability of a loss above 9 million dollars becomes equal to 83,3 percent. Then, repeating the exercise by contemporarily reducing the Tier1 ratio to 4.3 percent, increasing the ROA to 2 percent and leaving the interbank assets at its mean value the probability to be in the group with more than 9 million losses grows to 99.9 percent. This implies that large banks, exposed to interbank markets and characterized by a low tier 1 ratio and high profitability have been more exposed to severe losses. This partition divides the list in nearly two half: 22 are the banks with a loss over 9 million dollars while 23 show a loss lower than 9 million dollars. 7 We disregarded total assets because of the obvious correlation between the size and the amount of losses. 6

33

Journal of Advanced Studies in Finance Table 8. Probit model on WDCI banks list Dependent variable: WDCI2. Pseudo R2: 0.4140 WDCI banks: 37 LR chi2 = 20.95, Prob > chi2 = 0.0001 Regressors Coefficient Standard error Z Prob. > |Z| Tier 1 -0.6393536 0.2934 -2.18 0.029 ROA 1.81589 0.8523 2.13 0.033 Interbank assets 0.000064 0.0000 2.49 0.013 Constant 1.999483 1.6408 1.22 0.223 Marginal effect on regressors mean values : Tier 1=8.59, Roa=0.99, Interbank assets=41443 y = Pr (WDCI1 = 1) = 0.8338 dy/dx Standard error Z Prob. > |Z| Tier 1 -0.159447 0.0868 -1.84 0.066 Roa 0.452862 0.2263 2.00 0.045 Interbank assets 0.000016 0.0000 3.12 0.002 Marginal effect for the following regressors values: Tier 1= 4.3, Roa=2, Interbank assets=41443 y = Pr (WDCI1 = 1) = 0.9999 Tier 1 -7.83e-08 0.000 -0.07 0.941 ROA 2.22e-07 0.000 0.07 0.941 Interbank assets 7.84e-12 0.000 0.07 0.941 Conclusions In this study we try to explain the probability of banks to suffer severe or highly severe losses, represented by the inclusion in the WDCI list proposed by Bloomberg, analyzing their structure and performance. These characteristics are represented by 5 variables: the ROA, the tier1 ratio, the number of employees, the interbank assets and the total assets. Firstly, we compare banks inside and outside this list in order to understand the role of these variables. Then, we only focus on the banks in the WDCI list to study the heterogeneity of their losses. The results of the probit models used in the two analyses give us two identikits of banks: the first identifies the characteristics of banks included in the WDCI list and the second identifies the banks with highly severe losses. In both cases, big banks, in terms of number of employees or total assets, with a strong pressure for profitability and exposed to interbank markets have been vulnerable to losses or severe losses related to the sub-prime crisis. These findings confirm the direction of the new rules inspired by the Basel 3 agreement and by the Financial Stability Board, which seem generally in favor of a solid patrimonial structure, with an addiction of capitalization for the bigger banks, and a lower leverage ratio, in order to reduce the risk of losses and to create a safer financial structure. The focus on the size is absolutely relevant as large banks, in case of fragility and/or negative shocks, through their interconnections, can start a domino effect that can affect the whole financial system. These results also confirm what already appeared in literature: when a financial crisis occurs, we would expect banks with more capital and more stable financing to perform better (Beltratti and Stultz 2009). References [1] Acharya, V.V., Schnabl, P., Suarez, P. (2009). Securitization without risk transfer, mimeo. [2] Adrian, T., Shin, H. (2008). Liquidity and leverage, FRB of New York Staff Report 328. [3] Altunbas, Y., Gardener, E.P.M., Molyneux, P., Moore, B. (2001). Efficiency in European Banking, European Economic Review, 45: 1931-955. [4] Arpa, M.I. Giulini, A. Ittner, F. Pauer (2001). The influence of macroeoconomic developments on Austrian banks: implications for banking supervision’, BIS Papers, 1: 91-116. [5] Barros, C.P., Ferreira, C., Williams, J. (2007). Analyzing the determinants of performance of best and worst European banks: A mixed logit approach’, Journal of Banking and Finance, 31: 2189-2203. [6] Bebchuk, L., Spamann, H. (2010). Regulating banker’s pay, Georgetown Law Review, 98: 247-287. [7] Beck,T., Levine, R. (2004). Stock Markets, Banks and Growth: Panel Evidence, Journal of Banking and Finance, 28: 423-42.

34

Volume V Issue 1(9) Summer 2014 [8] Beltratti, R.M., Stultz, R.M. (2009). Why Did Some Banks Perform Better during the Credit Crisis? A CrossCountry Study of the Impact of Governance and Regulation, Finance Working Paper 254/2009. [9] Beltratti, N. Spear, Szabo M. D. (2010). The Value Relevance of Write-downs during the Subprime Financial Crisis, mimeo. [10] Beltratti, N., Stultz, R.M. (2011). The Credit Crisis Around the Globe: Why Did Some Banks Perform Better?, Fisher College of Business WP 2010-03-005. [11] Berger, A.N. (1995). The profit-structure relationship in banking: tests of market-power and efficient-structure hypotheses, Journal of Money, Credit and Banking 27, 2: 404-431. [12] Berger, A.N., Humphrey, D.B. (1997). Efficiency of financial institutions: international survey and directions for future research’, European Journal of Operational Research, 98: 175-212. [13] Bikker, J.A., Haaf, K. (2002). Competition, concentration and their relationship: An empirical analysis of the banking industry, Journal of Banking and Finance, 26: 2191-214. [14] Bikker, J.A., Hu, H. (2002). Cyclical patterns in profits, provisioning and lending of the new Basel capital requirements’, BNL. Quarterly Review, 221: 143-75. [15] Boyd, J., Runkle, D. (1993). Size and performance of banking firms: testing the predictions of theory, J. Monetary Econ., 31: 47–67. [16] Brunnermeier, M. K. (2009). Deciphering the liquidity and credit crunch 2007-2008, Journal of Economic Perspectives, 23: 77-100. [17] Chernobai, A., Jorion, P., Yu, F. (2008). The Determinants of Operational Losses, mimeo. [18] Demirgüç-Kunt, A. B. Karacaovali, L. Laeven (2005), ‘Deposit insurance around the world: A comprehensive database’, World Bank working paper. [19] Diamond, W. Douglas, R. G. Rajan (2009), ‘The credit crisis: Conjectures about causes and remedies’, American Economic Review, 99: 606-610. [20] D’hulster, K. (2009). The Leverage Ratio, A New Binding Limit on Banks, Note n.11, The World Bank. [21] Furfine, H. (2001). Banks as Monitors of Other Banks: Evidence from the Overnight Federal Funds Market, The Journal of Business, 74(1): 33-57. [22] Furlong , F. (2011). Stress Testing and Bank Capital Supervision, FRBSF Economic Letters, pp.2011-020. [23] Goddard, J.A., Molyneux, P., Wilson, J.O.S. (2001). European Banking: Efficiency, Technology and Growth. Chichester, UK: John Wiley & Sons. [24] Golin, J. (2001). The Bank Credit Analysis Handbook: A Guide for Analysts, Bankers and Investors, John Wiley & Sons (Asia) Pre Ltd. [25] Hellwig, M. (2009). Systemic Risk in the Financial Sector: An Analysis of the Subprime-Mortgage Financial Crisis’, De Economist, 157(2): 129-207. [26] Kirkpatrick, G. (2008). The corporate governance lessons from the financial crisis, OECD, Paris, France. [27] Kosmidou, K.,Tanna, S., Pasiouras, F. (2006). Determinants of profitability of domestic UK commercial banks: panel evidence from the period 1995-2002, Applied Research Working Papers Series, Coventry University Business School. [28] Levine, R. (2005). Finance and growth: Theory and evidence, in Aghion, P., and Durlauf, S. (ed.), Handbook of Economic Growth, Edition 1, Volume 1, Chapter 12, pp. 865-934, Elsevier. [29] Molyneux, P., Altunbaş, A., Gardener, E.P.M. (1996). Efficiency in European Banking, Chichester, UK: John Wiley and Sons. [30] Pallage, S.J. (1991). An econometric study of the Belgian banking sector in terms of scale and scope economies, Cahiers Economiques de Bruxelles, 130: 126-43 [31] Rochet, J.C., Tirole, J. (1996). Interbank Lending and Systemic Risk, Journal of Money, Credit and Banking, 28: 733-762. [32] Staikouras, G. Wood (2004). ‘The determinants of European bank profitability’, International Business & Economics Research Journal 3, pp. 57-68. [33] Stiglitz, J.E. (2010), Freefall: America, free markets, and the sinking of the world economy, W.W. Norton & Company, Inc. New York, NY.

35

Journal of Advanced Studies in Finance [34] Stulz, R., Williamson, R. (2003). Culture, openness, and finance, Journal of Financial Economics, 70: 313–49. [35] Vander Vennet, R. (2002). Cost and Profit Efficiency of Financial Conglomerates and Universal Banks in Europe’, Journal of Money, Credit and Banking, 34: 254-82. *** Board of governors of the federal reserve system (2009). The Supervisory Capital Assessment Program: Overview of Results, May 7, Washington DC. http://www.federalreserve.gov/newsevents/press/bcreg/ bcreg20090507a1.pdf *** European Commission (1997), ‘Impact on Services: credit institutions and banking’, Single Market Review, Subseries II, Vol 4., London: Office for Official Publications of the European Communities and Kogan Page Earthscan.

36

Volume V Issue 1(9) Summer 2014

DOI: http://dx.doi.org/10.14505/jasf.v5.1(9).03

DOES PAYMENT METHOD OF MERGERS AND ACQUISITIONS MATTER? AN EXAMINATION OF THE MEDICAL INDUSTRY IN CHINA Chengkui YE Cardiff University, Cardiff Business School, UK [email protected] Hao YUAN Hong Kong Baptist University, Hong Kong [email protected] Suggested Citation: Ye, C., Yuan, H. (2014). Does payment method of mergers and acquisitions matter? An examination of the medical industry in China, Journal of Advanced Studies in Finance, (Volume V, Summer), 1(9):37-47. DOI:10.14505/jasf.v5.1(9).03. Available from: http://www.asers.eu/journals/jasf/curent-issue. Article’s History: Received April, 2014; Revised May, 2014; Accepted July, 2014. 2014. ASERS Publishing. All rights reserved.

Abstract This study examines the impact of payment method and share of large shareholder on individual stocks after mergers and acquisitions. A research framework based on the Capital Asset Pricing Model (CAPM) is developed to identify factors of merger and acquisition that would affect stock return of the medical industry in China using transaction value, price-earnings ratio, debt asset ratio total asset as control variables. The results reveal that PE ratio, percentage of large shareholder ownership and pure cash payment method take positive effect on 30-day cumulative abnormal returns. Mergers and acquisition settled in cash has positive effect, so is share of large shareholders. Keywords: capital asset pricing model, mergers and acquisitions, event study. JEL Classification: C23, G34 1. Introduction 1.1. Statement of problem Merger and acquisitions (M&A) activities have been a common form, staple transaction activities in the mature financial markets. For instance, America has experienced five tides of the merging enterprises in the recent centuries. However, since China joined the World Trade Organization (WTO) in 2001, Chinese enterprises just started to merger or acquire other companies to raise their competitiveness. Recently, the Chinese capital market also experiences a merging tide in the last five years. According to Wind Info, there are 4,707 merging issues, including 311 issues in the medical industry, between 2008 and 2013. Dodd and Ruback (1977) demonstrate that the abnormal return of bidding companies and target firms around the first time public announcement of a takeover have positive return to their shareholders and result in substantial price change in the stock market. So far, most studies also have proved that there are abnormal returns after announcement of M&A information. Hence, it is necessary to research what factors lead to the abnormal return after the new M&A information in the China stock market. Nevertheless, due to the complex data and economic condition, this paper mainly concentrates on examining empirically stock return of the medical industry in China.

37

Journal of Advanced Studies in Finance Most of the literature is about how macroeconomic factors affect abnormal returns8. The objectives of this study are to examine empirically evidence of M&A abnormal returns after announcement and measure the factors leading to the result in the China A-shares market over the period from January 1, 2009 to December 31, 2013. The paper mainly concentrated on the factors, which result in the abnormal return in the stock market. Specifically, we intend to examine how the payment mechanism of M&A and acquiring firm's nature would affect cumulative abnormal returns. This study research will give some suggestions to A-shares investors for investing into the restructuring stocks, and guide the acquirers and targeted firms in corporate governance. 2. Literature review 2.1 Abnormal return for shareholders after announcements Abnormal returns are measured by the difference between actual and expected stock returns around the announcement date. Most of the studies show that target firms and bidders earn significant abnormal positive returns on the day of the announcement of the proposal which is associated with M&A and capital restructuring announcements. Dodd and Ruback (1977) report that target firm stockholders earned large and significant positive abnormal returns of 20.58% for successful offers and 18.96% for unsuccessful offers in the month of the first public announcement of the tender offer. Besides, Dodd (1980) uses a data set of 151 merger proposals from 1970-1977, to research abnormal returns in the M&A. The result showed large positive abnormal return earned by stockholders of the target firms on the day of the announcement of the proposal and the day before. The average abnormal return on day 0 is 4.30% and on day -1 is 8.74%, with t-statistics of 11.71 and 23.80, respectively. The mean cumulative abnormal return from the date of the first announcement of the proposal until the final approval by stockholders is 11.20%. Furthermore, Franks et al. (1991) survey 399 US takeovers from 1975 to 1984. For the entire sample, targets experienced substantial announcement gains averaging 28%. Moreover, Jensen and Ruback (1983) summarize thirteen studies up to that date. Their results indicate that targeted companies realize substantial and significant increases in their stock prices, ranging from 16.7% to 34.1%, in the announcement month. Jensen and Rubck (1983) demonstrate that bidding firms’ stockholders gained an average of 4% around tender offers around the merger. Besides, Ruback (1977), Kummer and Hoffmeister (1978) and Dodd (1980) also indicate that bidding firms gained from the acquisitions which derive from synergistic effect. The bidder firms performed better in expanding through cooperating with target firms after merging. However, Agrawal et al. (1992) find stockholders of acquiring firms suffer a statistical significant loss of about 10% over the five-year post-merging period. The Cumulative Average Abnormal Returns (CAARs) are significantly negative for holding periods of two to five years. For the five-year period, the CAAR is -10.26% (t = 2.37). In addition, the results of 10 years of post-merging data indicate that the abnormal returns level off after the fifth year. These authors argued that the merger cannot contribute to the fortune of shareholders, because of the increasing of fortune in target firms only comes from the transfer from the acquiring firms. For example, Aquith and Kim (1982), “investigating conglomerate mergers, find that only the target company shareholders gain from mergers. The acquiring firms’ shareholders and bondholders’ gain were shown to be insignificant”. Nevertheless, there are some limitations on this line of research. First, most of the samples of the acquiring companies are listed in developed countries, for instance, the New York Stock Exchange (NYSE). There are big differences between the China financial market and the American financial market, the US stock market is mature, more of an efficient-market and investors are more rational, but in the Chinese security market still have some loopholes and irrationalities, investors are keen on stock speculation. Second, many targeted firms are unlisted companies in China, so it is more difficult to measure their stock price performance after merging. 2.2 Abnormal return and method of payment

The method of payment used by the bidding company to the acquisition is one of the most important factors which affect the acquirers and target firms’ stock price change. Bidding companies typically offer the shareholders of target companies either cash or stock, or both ways. According to Rappaport and Sirower (1999), the main distinction between cash and stock transactions is that in cash transaction the acquiring company shareholders take the entire risk if the synergies of acquisition do not materialize, while in share transactions synergy risks are shared between bidding and target company shareholders. 8

One of the interesting recent development is the resurrection of Consumption-Based Capital Asset Pricing Model (CCAPM)( See Lettau and Ludvigson, 2001., Fung et. al., 2014a; Fung et al. 2014b), [34,35,36]. 38

Volume V Issue 1(9) Summer 2014 Hence, many empirical studies indicate that cash payments during acquisitions earn more returns for target shareholders than do share payments. One consideration is that share transactions are equivalent to public equity offerings, which typically show negative returns to shareholders around the equity issue announcement date. Travlos (1987) shows that, in pure share exchanges, bidding companies experienced significantly negative abnormal return during the takeover announcement, while for cash acquisitions, bidding companies’ returns were normal and cash-based transaction returns were significantly higher than those of share transactions. In addition, Draper and Paudyal (1999) illustrate that receiving payments in cash only, bidding firms’ excess return was higher than share payments, but these are significantly different from zero. Moreover, Wansley et al. (1983) report that target companies’ shareholders gain about 33.5% on average during cash payment acquisitions, which compared to 17.5% when it is paid by shares. Huang and Walkling (1987) also confirm that abnormal returns to target firms, which are related to cash offers, are significantly higher than those associated with share offers. In addition, Loughran and Vijh (1997) indicate that acquirers firms’ stock abnormal return perform higher than matching firms’ stock return in the tender in cash. While the bidding companies perform lower stock return than the common enterprises by merger in stock. As a result, tender in cash has a larger market reaction and higher market return than payment in cash. Besides, Draper and Paudyal (1999) also find that target company abnormal returns were highest when these companies received either cash or shares in comparison to pure share transactions or pure cash transactions. Consequently, the target firms earned significantly higher abnormal returns for cash takeovers than for share takeovers. 2.3 Abnormal return and size of acquirer The size of the acquirer is a key factor which determines the bidding companies announcement period excess return with normal firms. In general, the amount of transactions and market return by small firms is also relatively smaller than in the case of large companies. However, the small acquirers outperform the large ones and the small firms have positive abnormal returns, whereas the large acquirers tend to have negative gains as reported by (Moeller et al. 2005). Loderer and Martin (1990), Schwert (2000), and Moeller et al. (2004) point out that those bigger buyers tend to have lower announcement period return under the size effect. Zhu (2006) also contended that Chinese stock market prefers the smaller size of bidding companies which are easier to speculate and their volatility is also larger than others. 2.4 Abnormal return and P/E ratio

Petmezas (2009) examines bidder performance of firms that had larger P/E ratio than the P/E of the market one month prior to the announcement and those that had smaller P/E ratio than the P/E ratio of the market one month prior to the announcement. He finds that bidders generate significantly larger abnormal returns during high valuation periods than low valuation periods for all and private acquisitions in the positive P/E portfolio. Jovanovic and Rousseau (2001) find that merger waves coincide with periods of high price-earnings ratios on the stock markets, which is used as a proxy of market valuations. Li et al. (2004) draw a conclusion that P/E ratio has a positive correlation with the performance of M&A, which not only reflects the optimistic expectations from market to merger prospects, but also shows that the M&A which improves the company's performance can bring abnormal returns to shareholders. 2.5 Merger & aquisition performance and the shareholding proportion of the controlling shareholder Feng and Wu (2001), and Zhu and Wang (2002) illustrate the shares proportion of largest shareholders has influence on the performance of M&A. The largest shareholders’ share holdings have a positive correlation with the performance of M&A in the same year rather than the next few years. Li et al. (2001) find that the stake of the largest shareholder can well explain the M&A performance. The higher the share proportions of largest shareholders are, the better the performance of M&A will be. Liu and Ouyang (2010) claim that when the status of the biggest shareholder do not change, the higher the stock exchange ratio is, the more benefits to improve the performance of firms in the short term and the long run. 2.6 Abnormal returns and debt asset ratio Rosenfeld (1984) and Sicherman and Pettway (1992), find that abnormal returns are lower for selling firms in relatively poor financial health. To control the financial health of the firms, the regressions include the return on assets, interest coverage, and debt to assets ratios for both firms. Amira et al. (2009) analyzed the impact of debt size on stock returns. For debt level, the sample is divided into high debt and low debt asset buyers. From the mean cumulative abnormal returns, it is evident that asset buyers with higher total debt-to-total asset ratios have

39

Journal of Advanced Studies in Finance higher returns, confirming that announcement period returns increase along with buyer leverage. Zhu (2006) also proves that the high debt or high financial risk companies generate lower cumulative abnormal return, which means the low debt companies would be more popular with common investors. 2.7 Merger & aquisition a performance and the transaction value Feng and Wu (2001) investigate the operational cash flow and rate of return on total assets of acquiring firms in Taiwan. They find that mergers and acquisitions led to a slight dip in profits. In addition, the transaction price plays a role in the performance of acquiring firms after M&A. The study made by Li and Zeng (2004) show that there is a positive relationship between the transaction price and performance. Transaction price has a significant effect on the performance of M&A in the same year. Shi (2008) uses 24 Chinese enterprises which have multinational M&A during the 2001-2005 as the samples to evaluate the performance of China’s enterprises multinational M&A. She argues that the revenue growth of M&A firms whose transaction price is more than thirty million dollars outstrips some firms with ten million dollars. She thinks that the higher transaction price can lead to a better M&A performance. Finally, according to the Hubris Hypothesis of Corporate Acquisitions (Roll, 1986), decision makers in acquiring firms pay too much for their targets on average in the sample. Managers assess the outlook for M&A over-optimistically and take measures of over-paying for the target firm during the time of M&A, which brings poor performance of acquiring firms. The higher transaction price results in higher costs that acquiring firms need to pay and more resources that acquiring firms have to allocate, which are not conductive to the improvements of M&A performance. 3. Methodology and data The methodology adopted by Zhang (2003) using financial data of Chinese medical firms was used in this paper. Suppose the announcement date of mergers and acquisitions is defined as t(0). We assume the stock price will not be affected by the mergers and acquisitions news and the stock price goes up or down by the market fluctuation in the clean period. So we choose the period between t-180 and t- 90 days (90 trading days in total) as the clean period. We follow Zhang (2003) to set (T0,T+30) as the event period (30 trading days in total) which means that we focus on the abnormal return between event day and 30 trading days later. First, we use the CAPM to calculate each individual stock beta value in the clean period (market return is CSI300 index return, risk-free rate is 1 year deposited rate in mainland China). According to Horn and Wachowicz (2005), adjustment beta is an estimate of a security’s future beta that involves modifying the security's historical (measured) beta owing to the assumption that the security's beta has a tendency to move over time toward the average beta for the market or the company's industry. So each stock is adjusted by multiplying the raw beta by 0.67 and adding 0.33. In the event period, the adjusted beta value is used to estimate the abnormal return, which will be used as dependent variable for cross-sectional regression. Abnormal returns are computed as the difference between actual returns and estimated expected returns: ARj,t = Rj,t - (αj+ adjβj*Rm,t)

(1)

where ARj,t is the abnormal return, Rj,t is the individual stock return; Rm,t denotes the market return. Then, we calculate the cumulative abnormal return by adding all abnormal return in the 30 event trading days. In order to examine the cumulative effect of events, the cumulative abnormal returns are produced. CAR(T1,T2)=

(2)

where T1 is the announcement M&A information day, T2 is the last day of an event period (t+30). CAR is cumulated abnormal return between T1 and T2 The following model is used to test whether the method of payment, PE ratio of acquirer at the announcement day, debt asset ratio of bidding firm ahead 1 year of M&A announcement, size of acquirer, large shareholder ownership and transaction value would affect the abnormal return after the announcement of new information about M&A activities in the China stock market. CARy1y2 = α0 +β1PEi +β2DRi +β3LSOi +β4LogTAi +β5D1+β6LogVi +μi

40

(3)

Volume V Issue 1(9) Summer 2014 where: CARy1y2 = cumulative abnormal return from day y1 to day y2; Pei = PE ratio of bidding firm at the announcement M&A activities day; Dri = Debt to asset ratio ahead 1 year for the M&A announcement; LSOi = the percentage of large shareholder ownership at the M&A year; LogTAi = Log total asset of acquirer ahead 1 year of the M&A announcement; D1 = 1 if payment method is cash; = 0 otherwise; LogVi = Log transaction value between targeted firm and bidding firm; μi = error terms. This study covers all M&A issues dealing with the medical listed companies in the Shanghai and Shenzhen stock markets from 2009 to 2013, but the deals only paid by cash or shares and just adopt horizontal or vertical merging strategies. So the payment by cash plus share, debt plus shares and other combinations are not in the research list and the mixed merging strategy deals also have been excluded. Besides, we also exclude M&A cases that overlap between event period and another clean period for the same bidding companies. This means that if one company issue merging or acquisition news not less than 30 trading days, but the companies declared M&A news again and their clean period overlapped with former event period. Then the second time M&A case will not be counted in this research. The data are collected from the Wind information M&A database9, there are 55 listed medical companies which were involved in M&A activities and comply above conditions in the China stock market from January 2009 to December 2013. As for securities returns, we adopt CSI300 index as the benchmark of market return, the daily closing rate of return for individual securities. 4. Empirical results Table 1 and Table 2 report the descriptive statistics of CAR(0,30). It can be seen that the CAR of bidding companies contribute to average positive return at 3.89 % between the announcement day and 30 trading dates later. The minimum and maximum returns are -35.51% and 63.31%, respectively. However, comparing with other mature securities market studies, the distribution of CAR for individual securities are not quite efficient. We believe that it is mainly because some insiders have already known the M&A information before the announcement of new information and make inside trading, but it is not key areas in this study. Table 2 shows that 47.3 per cent firms recorded positive cumulative returns after mergers and acquisitions. Table 1. Descriptive Statistics of cumulative abnormal return CAR(0,30)

N

Minimum

Maximum

Mean

Std. Deviation

55

-35.51%

63.31%

3.90%

0.211

Source: Wind Financial investment bank database

Table 2. Summary of CAR of bidding companies involved in M&A activities Cumulative Abnormal Return (0,30) below -10%

16

0~ -10%

13

0~ 10%

6

upper 10%

20

Source: Wind Financial investment bank database This paper aims to examine how the nature of acquiring firm affects the returns of the target. The first variable of interest is the percentage of large shareholder ownership. The literatures suggest that large shareholder ownership has a positive relationship with abnormal return of acquirers (Feng and Wu, 2001). We follow the early studies using PE ratio, Debt asset ratio and transaction value as control variables. All standard errors are robust adjusting for heteroskedasticity.Table 3 shows that PE ratio, Debt asset ratio and transaction value are all significant. The equation can be written as: CAR (0, 30) = 0.66 +0.0006PE- 0.345DR -0.087LogV +0.38LSO +μi (0.359) (0.000135) (0.135) (0.0414) (0.1756) 9

The data are collected from the Wind Financial Terminal 41

(4)

Journal of Advanced Studies in Finance Table. 3: Summary of CAR of bidding firms with shareholder ownership only Variables Dependent Variables

Ordinary Least Square Estimated Result Coefficient T- statistic

Prob.

0.0006 -0.345

4.115*** -2.55**

0.0001 0.0138

-0.087

-2.12**

0.0413

0.38

2.163**

0.035

0.66

1.61

0.1137

CAR(0,30)

Control Variable PE ratio (t statistics) Debt Asset Ratio (t statistics) Log Transaction Value (t statistics) Independent Variable Percentage of Large shareholder ownership (t statistics) C (t statistics)

Notes: Adjusted- R-square: 0.218; F-statistic: 4.757; Prob(F-stat): 0.002 ***, **, * stand for statistical significance at the 1%, 5% and 10% level respectively t statistics adjusted by White heteroskedasticity correction method.

The signs of PE ratio and debt-asset ratio are all correct. As for the independent variable, the percentage of large shareholder ownership is statistically at 5 percent significance level. Besides, it also shows that the higher proportion of large shareholder ownership has higher CAR from the announcement day to 30 trading days later. The R-square is 21.8 percent. Our result indicates that when larger shareholder ownership increases by one per cent, abnormal return increases on average by 2.163 percent. Our second variable of interest is payment method. The literatures suggest that if payment method is by cash, the impact on stock returns is positive (Travols, 1987). The control variables are PE ratio, Debt asset ratio and transaction value. The dummy variable that payment method is independent variable and CAR (0, 30) is dependent variable. The regression is illustrated as below: CAR (0, 30) = 0.691 + 0.0005PE -0.29DR -0.093β3LogV +0.12D1+μi (0.304) (0.0001) (0.139) (0.038) (0.048)

(5)

Table. 4: Summary of CAR of bidding firms with payment method only Variables Dependent Variables


Prob.

CAR(0,30) Control Variable PE ratio (t statistics)

0.0005

3.857**

0.0273

Debt Asset Ratio (t statistics) -0.290 --2.08** 0.0427 Log Transaction Value -0.093 -2.44** 0.00181 (t statistics) Independent Variable Payment Method 0.152 3.183** 0.0025 (t statistics) C (t statistics) 0.69 2.27** 0.0273 Notes: Adjusted- R-square: 0.235; F-statistic: 5.14; Prob (F-stat): 0.0015 ***, **, * stand for statistical significance at the 1%, 5% and 10% level respectively t statistics adjusted by White heteroskedasticity correction method.

Table 4 shows that the coefficient of payment by cash is positive and significant at 1 per cent. On average, when the mergers and acquisitions are settled in cash, the individual stock return increases by 3.183 per cent. Comparing to Table 3, the sign of control variable coefficients remain the same. The R-square is slightly higher -

42

Volume V Issue 1(9) Summer 2014 23.5 percent. The coefficients of explanatory variables are jointly significant as indicated by the low p-value of Fstatistics. We estimate the full model and report the results in Table 5. The result shows that the coefficient of bidding firms’ PE ratio is statistically significant at 1 percent significant, the Debt ratio, payment method and payment method are statistically significant at 5 percent significance level, and the percentage of large shareholder ownership is statistically significant at 10%. These mean that the five variables have a significant effect to the CAR after announcement. Furthermore, these also demonstrate that the PE ratios of bidding companies, the percentage of large shareholder ownership and pure cash payment method have a positive relationship with CAR. Especially, the percentage of large shareholder ownership has the biggest positive effect to CAR and amount to 0.304 at 10% significance level, but the PE ratio is the smallest positive effect, just represents 0.00572. However, debt asset ratio and transaction value have negative effect to CAR which means the higher debt ratio and higher deal price would cause lower or negative CAR of bidding companies. Nevertheless, the transaction value and total asset do not affect CAR of bidding companies after the announcement period in this regression model. Table. 5: Summary of CAR of buyers in the period between January 2009 and December 2013. Variables Dependent Variables


Prob.

CAR(0,30) Independent Variable PE ratio (t statistics)

0.0006

4.463***

Debt Asset Ratio -0.36 -2.33** (t statistics) Log Transaction Value -0.087 -1.922* (t statistics) Percentage of Large shareholder ownership 0.304 1.72* (t statistics) Log Total Asset 0.031 0.461 (t statistics) Payment Method 0.121 2.269** (t statistics) C (t statistics) 0.303 0.581 Notes: Adjusted- R-square: 0.2411; F-statistic:3.859; Prob(F-stat): 0.003 ***, **, * stand for statistical significance at the 1%, 5%,10% level respectively t statistics adjusted heteroskedasticity correction method.

0.0000 0.0239 0.0605 0.0918 0.6466 0.0278 0.6047 by White

5. Discussion and limitation 5.1 Discussion From the regression outcome, it can be concluded that these five variables have a strong relationship with CAR between the announcement day and 30 trading days later, namely, PE ratio, debt asset ratio, transaction value, payment method and the proportion of large shareholder ownership, because the significance are all smaller than the 5 percent level in the various regression. In these variables, PE ratio, pure cash payment method and percentage of large shareholder ownership have a positive effect. Debt asset ratio and transaction value have a negative relationship with CAR. According to Table 1 and Table 2, we can find that the average of CAR is positive and amounts to 3.89% in the 55 M&A cases, but the distribution of CAR (0, 30) is not quite normal and concentrated. For instance, there are 16 medical firms’ stock return loss more than 10%, however, there are also 20 companies that can get CAR excess 10%. It is mainly because the Chinese securities market is inefficient; some investors can get the M&A information ahead of the announcement day in some securities, so that the market price had already reacted to the new information. Some common investors can not share the excess return after announcement day and undertake the risk of falling stock price. But the average cumulative abnormal return is still positive since announcement day and the CAR still exists in the Chinese stock market.

43

Journal of Advanced Studies in Finance In all cases, the PE ratio of bidding companies still has a positive relationship with CAR and significant, which is consistent with the results of previous studies (Jovanovic and Rousseau, 2001; Li et al., 2004). The higher PE ratio represents investors’ confidence about the market prosperity and companies performance, so the M&A announcement would be good news to common investors. But the coefficient is very small, just about 0.0005, which means the positive effect would be relatively smaller, since the too high PE ratio would generate a price bubble. Considering the debt asset ratio of bidding companies, the high debt ratio would result in low CAR, since it has a negative relationship with CAR (0, 30) at the 5% significance level in our regression model. This result is also similar to the previous research by Zhu and Tao (2006). The high debt ratio bidding firms are more risky to merger or acquire other companies because they should issue more bonds or stocks to paid targeted firms. It would increase the debt burden and operation pressure, and these would affect the investors’ decision and stock return. In terms of transaction value, it has a negative effect on CAR in the various regression models, which means that the higher price of M&A targeted firm leads to lower CAR. However, the regression result is different from some previous research result. For instance, Facio et al. (2004) found that there are illiquidity discounts in the non-listed targeted firms. In order to get the liquidity, the targeted firms shareholders are easy to accept lower transaction value to exchange liquidity of shares, so the listed bidding companies can merger or acquire more valuable non-listed firms than the book price; however, most of the M&A cases in our sample pay the fund to targeted firms by cash, too expensive deal price would affect companies’ cash flow and daily operation. So we still conclude that the transaction value still generates a negative effect on CAR in the medical industry in the period between January 2009 and December 2013. According to the Table 3 and Table 5, the percentage of large shareholder ownership causes a positive relationship with CAR, the significant level also decreases from 10 percent in the total variables regression model to 5 percent in the control variable regression model. These results also comply with the former studies ( Feng and Wu, 2001 and Zhu and Wang, 2002). The big shareholders hold more shares, which is beneficial for decision-making, reducing the dispute and promoting merging or acquisition success. Speaking of payment method in the M&A activities, we utilize a dummy variable to test the effect, where pure cash is 1 and pure stock is 0. From Table 4 and Table 5, the pure cash method has a significant positive effect to CAR (0,30). The result is consistent with previous studies (Travlos,1987; Draper and Paudyal, 1999; Wansley, Lane and Yang, 1983). On the one hand, the bidding firms pay the targeted firms pure cash, which can be able to enjoy all profit of M&A targeted. On the other hand, listed firms pay shares to targeted firms which are similar to issue new shares that would affect stock price and CAR. Lastly, based on Table 5, total asset of bidding firms do not affect the CAR. Although it does not comply with previous research (Moeller & Schlingemann and Stulz 2005), the market participants may not make a difference with big or small listed medical companies now. 5.2 Limitation and further research There are at least two limitations of this paper. Firstly, we adopt the Capital Asset Pricing Model (CAPM) to regress the beta and adjusted beta value at the clean period and use them to estimate expected return in the event period. Although CAPM quite commonly used in financial research, it still has some controversies and weaknesses. For example, it assumes no transaction cost and tax cost and can be able to lend money at the risk free rate which cannot be reached in the real applications. Besides, some listed medical companies belong to the growth enterprise market, but we utilized CSI 300 index return as market return overall in the CAPM. The estimated adjusted value in the clean period would be not very accurate. Secondly, although it took place more than 4000 M&A cases in Chinese stock market between 2009 and 2013, the selected M&A case must comply with medical industries, payment method and type of M&A limited conditions, so there are only 55 cases in this period, which may affect the regression result. Thirdly, we follow Zhang’s clean period and event period in this research, but we cannot guarantee the clean period which was not affected by the M&A announcement information. At last, M&A sometimes increases the risk exposure of the corporation (Amihud et al. 2002). We should include a control variable to measure risk associated with M&A. In order to complete the research and make further research, we should collect more M&A cases randomly in the mature industry or market, besides; combing the CAPM, arbitrage pricing model and market adjusted would be more accurate to estimate the expected return.

44

Volume V Issue 1(9) Summer 2014 Conclusion This paper mainly examines what factors affect the abnormal return after M&A information announcement by the empirical data between January 2009 and December in 2013. Furthermore, although not every firm has positive abnormal return in the M&A cases, the average of CAR is still positive, which means Chinese securities investors still have a chance to get abnormal return. According to literature reviews, we collect six variables that are PE ratio, Debt asset ratio, transaction value, and payment method, percentage of large shareholders and total asset of bidding firms to test the effect with CAR after announcement. We find that PE ratio, percentage of large shareholder ownership and pure cash payment method take positive effect on CAR (0, 30), especially, PE ratio just takes a minute positive effect to CAR, which has been proven by regression models. However, Debt asset ratio and transaction value have a significant negative effect to bidding firms CAR in the period between the announcement day and 30 trading days later. The four regression models in this study indicate that the total asset of bidding firms does not have a significant relationship with cumulative abnormal return. These results highlight that common investors should pay attention to the higher PE ratio and percentage of large shareholder ownership, pure cash payment method in the M&A activities. These firms usually can generate higher abnormal return after the announcement of the new information. Nevertheless, the poor debt ratio and expense of targeted firms usually cannot give rise to higher abnormal return. Only by adopt the above five variables investment strategies, Chinese securities investors can be able to get greater cumulative abnormal return in the period from announcement the day to 30 trading days later. References [1] Agrawal, A., & Jaffe, J.F. (1992). The post-merger performance puzzle, Advances in Mergers & Acquisitions, 1: 7-41. [2] Amihud, Y., Delong G.L., & Saunders, A. (2002). The effects of cross-border bank mergers on bank risk and value. Journal of International Money and Finance, 21(6): 857-877. [3] Amira, K. (2009). Asset buyers and leverage (Doctoral dissertation, New York University) http://www.fma.org/ Reno/Papers/AssetBuyersandLeveragefma09.pdf [4] Asquith, P., Kim, E. (1982). The impact of merger bids on the participating firms' security holders, The Journal of Finance, 37(5): 1209-1228. [5] Dodd, P., Ruback, R. (1977). Tender offers and stockholder returns: An empirical analysis, Journal of Financial Economics, 5(3): 351-373. [6] Dodd, P. (1980). Merger proposals, management discretion and stockholder wealth, Journal of Financial Economics, 8(2): 105-137. [7] Draper, P., Paudyal, K. (1999). Corporate takeovers: mode of payment, returns and trading activity, Journal of Business Finance & Accounting, 26(56): 521-558. [8] Feng, F., Wu, L. (2001).The empirical research about the performance of merge and acquisition by Chinese listed companies, Economics Research, (9). [9] Faccio, M., McConnell, J.J., Stolin, D. (2004). When do bidders gain? The difference in returns to acquirers of listed and unlisted targets. The Difference in Returns to Acquirers of Listed and Unlisted Targets, February 9. [10] Franks, J., Harris, R., Titman, S. (1991). The postmerger share-price performance of acquiring firms, Journal of Financial Economics, 29(1): 81-96. [11] Fung, K.W.T., Demir, E., and Zhou L. (2014b). Capital Asset Pricing Model and Stochastic Volatility: A Case study of India. MPRA working paper no. 56180. [12] Fung, K.W.T., Lau, C.K.M., Chan, K.H., (2014a). The Conditional CAPM, Cross-Section Returns and Stochastic Volatility, Economic Modelling, 38: 316-327 [13] Horne, J. C., Wachowicz, J.M. (2005). Fundamentals of financial management (12th Ed., ), Chapter 5. Essex, UK: Pearson Education publishing. [14] Huang, Y.S., Walkling, R.A. (1987). Target abnormal returns associated with acquisition announcements: Payment, acquisition form, and managerial resistance, Journal of Financial Economics, 19(2): 329-349. [15] Jensen, M.C., Ruback, R.S. (1983). The market for corporate control: The scientific evidence, Journal of Financial economics, 11(1): 5-50.

45

Journal of Advanced Studies in Finance [16] Jovanovic, B., Rousseau, P. (2001).Mergers and technological change: 1885-2001. Working Paper, New York University. [17] Kummer, D.R., Hoffmeister, J.R. (1978). Valuation consequences of cash tender offers, The Journal of Finance, 33(2): 505-516. [18] Lettau, M., Ludvigson, S. (2001). Resurrecting (C)CAPM: A Cross-Sectional Test When Risk Premia are Time-Varying. Journal of Political Economy, 109: 1238-1287. [19] Liu, L., Ouyang, Z. (2010). The analysis of M&A performance of Chinese firms. Economic Vision. doi:10.3969／ j.issn.1672—3309(x)． http://www.cqvip.com/Read/Read.aspx?id=36205360 [20] Li, S., Zeng, Z., Wang, C., Zhu, T., Chen, Y. (2004). Analysis of factors affect the M&A performance World Economic, 60-67 [21] Loderer, C., Martin, K. (1990). Corporate acquisitions by listed firms: The experience of a comprehensive sample. Financial Management, 17-33. [22] Loughran, T., Vijh, A.M. (1997). Do long‐term shareholders benefit from corporate acquisitions?. The Journal of Finance, 52(5), 1765-1790. [23] Moeller, S.B., Schlingemann, F.P., Stulz, R.M. (2004). Firm size and the gains from acquisitions, Journal of Financial Economics, 73(2): 201-228. [24] Moeller, S.B., Schlingemann, F.P., Stulz, R.M. (2005). Wealth destruction on a massive scale? A study of acquiring‐firm returns in the recent merger wave, The Journal of Finance, 60(2): 757-782. [25] Petmezas, D. (2009). What drives acquisitions?: Market valuations and bidder performance. Journal of Multinational Financial Management, 19(1): 54-74. [26] Rappaport, A., Sirower, M.L. (1998). Stock or cash? The trade-offs for buyers and sellers in mergers and acquisitions, Harvard Business Review, 77(6): 147-58. [27] Richard, R. (1986). The hubris of corporate takeovers, The Journal of Business, 59(2): 197-216. [28] Rosenfeld, J.D. (1984). Additional evidence on the relation between divestiture announcements and shareholder wealth,The Journal of Finance, 39(5): 1437-1448. [29] Schwert, G.W. (2000). Hostility in takeovers: in the eyes of the beholder?.The Journal of Finance, 55(6): 2599-2640. [30] Shi, X. (2008). The factors of multinational merger and acquisition performance in China. Southwestern University of Finance and Economics, http://cdmd.cnki.com.cn/Article/CDMD-10651-2009061884.htm [31] Sicherman, N.W., Pettway, R.H. (1992). Wealth effects for buyers and sellers of the same divested assets. Financial Management, 119-128. [32] Travlos, N. G. (1987). Corporate takeover bids, methods of payment, and bidding firms' stock returns. The Journal of Finance, 42(4): 943-963. [33] Wansley, J.W., Lane, W.R., Yang, H.C. (1983). Abnormal returns to acquired firms by type of acquisition and method of payment. Financial Management, 16-22. [34] Zhang Xin (2003). The theoretical and empirical research whether merger and acquisition create value in China security market. Economic Research Journal. [35] Zhu, B., Wang, Y.(2002). The effect analysis about the performance of merging and acquisition by Chinese listed companies, Economics Research, 11: 20-26. [36] Zhu, T. (2006). The short-term and long-term stock performance of listed companies after merger and acquisition in 2006. Journal of Modern Economic Science 28(3): 31-39. http://www.cqvip.com/Read/ Read.aspx?id=21866274 (June 2003) http://222.73.229.10/cs/zhs/xxfw/research/plan/plan20030701f.pdf

46



THE GINI COEFFICIENT. DECOMPOSITION AND OVERLAPPING Augustine C. ARIZE College of Business and Entrepreneurship, Texas A&M University-Commerce Paraskevas BAKAREZOS Agricultural Bank of Greece, Republic of Greece Krishna M. KASIBHATLA Department of Economics, School of Busines s and Economics North Carolina Agricultural and Technical State University John MALINDRETOS Department of Economics, Finance and Global Business Cotsakos College of Business, William Paterson University Alex PANAYIDES Cotsakos College of Business, William Paterson University Suggested Citation: Arize, A.C., Bakarezos, P., Kasibhatla, K.M., Malindretos, J., Panayides, A. (2014). The GINI coefficient. Decomposition and overlapping, Journal of Advanced Studies in Finance, (Volume V, Summer), 1(9):47-55. DOI:10.14505/jasf.v5.1(9).04. Available from: http://www.asers.eu/journals/jasf/curent-issue. Article’s History: Received June, 2014; Revised June, 2014; Accepted July, 2014. 2014. ASERS Publishing. All rights reserved.

Abstract The purpose of this paper is to present the process of decomposition and its relation to the Gini coefficient with emphasis on the explanation of the nature of the overlapping component (or interaction effect) which arises during decomposition and it is usually treated as an awkward by-product of this process. It is shown that exact decomposition without the presence of the overlapping effect is possible even when group income distributions are overlapping under condition that data on individual incomes are available. The problem is purely mathematical and it arises when decomposition is accompanied by replacement individual income comparisons with differences in mean group income between pairs of groups. In this case, the overlapping component can be completely separated from the other two, the “within” and “between” inequality components. Furthermore, the case of grouped data is considered that does not allow exact separation of the component. Keywords: GINI coefficient, decomposition, overlapping. JEL Classification: C23, G34 1. Introduction The measurement of personal income inequality has been for long a matter of interest in economics. There exist several measures of inequality and Stark (1972) gives a comprehensive list. The most widely used measure of inequality is probably the Gini coefficient. The investigation of its various characteristics on both the theoretical and the empirical level has also been long with major contributions by Atkinson (1970) and Sen (1973). An interesting development occurred when Bhattacharya and Mahalanobis (1967) and later Rao (1969) introduced the decomposition approach. According to it the population of income-receiving units is divided into two groups and consequently inequality is distributed to disparities in income between groups. Pyatt (1976) analyzed the same disaggregation in matrix form and his method corresponds directly to that used here. Exact decomposition without the presence of the overlapping effect, in the case of exact data, is the subject matter of the first section. The next section deals with the same case of exact data, the condition under

47

Journal of Advanced Studies in Finance which the overlapping component arises and its mathematical isolation and estimation. The third section considers the more realistic case of non-availability of exact data that makes exact separation of the overlapping component impossible and a combination of estimating techniques is suggested to approximate it. The forth section consists of an illustration of all these results that are applied to data on income in Greece 1962-88. The application is methodologically similar is not to analyze the components and the trend of inequality. The fifth section outlines briefly the necessary extensions of this analysis. The sixth section is a short summary with a presentation of the conclusions of the paper. In conclusion, this paper was developed and written in the hope of contributing to further understanding of the process and nature of decomposition thus facilitating analysis of income inequality situation. 2. Exact decomposition of the Gini coefficient The most suitable formulation of the Gini coefficient for the purposes of the paper is: ∑

∑

|

|

(1)

̅

where: p is total number of income-receiving units, Xd is income from the dth unit, d,f = 1 … p, and is average income. This formulation is therefore based on inter-unit income comparisons and it requires data on individual incomes, called here exact data as opposed to grouped data.2 The numerator in (1) can be considered as a norm of the matrix of absolute differences A of dimension ( p*p).3 It is possible, however, to consider instead of p number of units k number of groups of populations p1, =p. In this case, A can be partitioned into k2 sub-matrices Ai,j , i,j= 1…k. The sub-matrices along the diagonal of A are symmetric and of dimension (pi*pi), i= 1…k. The rest of the sub-matrices are of dimension (pi*pi), i,j= 1…k. i≠j, and depending on the division into groups, none, some, or all can be non-square. All this means that A can be decomposed into k2 sub-matrices each having the following norm:4 ∑

∑

|

(2)

|

Where i,j = 1…k ∑

∑

∑

∑

|

|

The result of this procedure is the decomposition of A which can be interpreted as follows: The submatrices Ai,j, i=j, contain as elements comparisons between pairs of incomes within each group. The rest of the sub-matrices Ai,j, i≠j, consist of elements which are the comparisons of pairs of income between units of i th and jth groups. The decomposition of A can be used to obtain expressions analogous to (1). These, for the sub-matrices Ai,j, i=j, can be interpreted as “within” group Gini coefficients: ̅̅̅̅

,

(3)

i=j

For the sub-matrices Ai,j , i≠j, the following expressions can be defined which are interpreted as “between” group Gini coefficients measuring inequality between pairs of income groups5: ,i≠j (4) ̅̅̅̅ where: ̅̅̅̅

̅̅̅ ̅̅̅

Using the results so far, matrix G can be defined with elements Gij from expression (3) and (4) and of dimension (k*k). Weighting these Gij by group income and population proportions results in an alternative additively decomposable expression for G in (1) as follows: G = `Gm (5) Where is a row vector (1*k) of aggregate group income proportions and m is a column vector (k*1) of group population proportions.6 Expression (5) is an exact and symmetric decomposition of the Gini coefficient.7,8 `

48

Volume V Issue 1(9) Summer 2014 3. Decomposition and overlapping9 The result of the previous section is an exposition of the well-known fact that exact decomposition of the Gini coefficient is possible when exact data are available, and it is included to preserve continuity of the discussion. Exact decomposability, however, it not usually encountered in practice, or it might be computationally awkward even when it does. It would be of advantage to replace as many Aij in (2) with a summary measure. One possible replacement is to express Aij = Aji, i≠j in terms of differences in group mean incomes, i.e. ∑

∑

|

|

|̅̅̅

̅|

,i≠j

(6)

Where R is a remainder and the explanation of its nature is the subject matter of this section. The investigation of the relationship between the left hand (LHS) and right hand (RHS) side expressions in (6) requires again analysis based on exact data as in the last section but before taking absolute values. The ordering of individual incomes in ascending (or descending) order according to size of income results in skew symmetric sub-matrices, Ai,j, i=j, that is elements above the diagonal are the negatives of the elements below it. The elements of these diagonal sub-matrices are not affected by the RHS replacement in (6). The elements of the sub-matricesAi,j, i≠j, are also placed symmetrically along the diagonal of the partitioned A. The difference between, Ai,j, i=j, and Ai,j, i≠j, is that the elements of the latter might or might not be neatly separated with respect to their signs and that depends on overlapping between pairs of group income distributions. In the case of nonoverlapping distributions each Ai,j, i≠j, above the diagonal consists of elements that are all negative, while each Ai,j, i≠j, below it (from now on denoted by ) consists of the same elements placed symmetrically but all positive. 10 When the distribution are overlapping and Ai,j, i=j, contain both positive and negative elements not at random but in the following order: The pair-wise comparisons in between members of group i and members of group j for which Xir < Xjw, come out negative while the rest of the comparison turn out positive since Xir > Xjw (and this is the result of overlapping).The corresponding comparisons in pairs in yield the same elements but of opposite sign. 11 It is the existence of elements of both sign within each Ai,j, i=j, that creates the overlapping effect when LHS expression in (6) is replaced by the summary measure on the RHS combined with the existence of absolute values. The addition during the computation of the double sum on the LHS of (6) is performed after talking absolute values and that means that the outcome of each comparison (or element of Ai,j,) counts positively regardless of its sign. However, the RHS is computed differently since calculation of requires addition before taking absolute values and differences of opposite signs off-set against each other.12 In other words, the “between” group Gini coefficient can be computed using as numerator in (3) either the LHS or the first term in the RHS in (6). Since these two expressions are always equal, because they are not computationally equivalent, it is necessary to make them equal under all conditions. This can be done in 2 ways: firstly, the addition on the LHS expression can be performed before talking absolute signs thus allowing offsetting due to opposite signs on the LHS also and in this case R need not be include in the RHS13. Secondly, the order of operations during computation on the double sum in (6) remains unchanged while the RHS expression in (6) is ∑ | |, u = 1…nij, nji being the number of differences of opposite sign in Aij increased by the amount R and i = 1…k. This amount is, therefore, computed as the absolute value of the sum of all differences of opposite sign14 multiplied by two15 thus making all the differences on the RHS expression in (6) of the same sign (negative in our case). This can become clear by focusing temporarily on one comparison only, X ir – Xjw = L, Xir > Ejw, by adding to L the quantity of -2L. Performing this for all the differences within each and separating the sum , we allow addition to be performed in the RHS expression in (6) before taking absolute values, while at the same time, we compensate for offsetting, since that is the only way to estimate ̅̅̅ ̅ 16. In the case of non-overlapping distributions L=0. The first way of making the two expressions in (6) equal by taking absolute values after performing the addition (the equivalent of subtracting the above quantity from the LHS expression) is not correct taking expression (3) into consideration since it underestimates the Gini coefficient by losing the quantity from offsetting due to positive and negative elements. The second way, of adding separately the same quantity to the RHS expression in (6) is in accord with expression (3) since no difference is lost only they are separated. Consideration of all in A results in the following decomposition of the Gini coefficient: G = `[B + X + V]m

(7)

Where B is a (k*k) diagonal matrix, each diagonal element being a group Gini coefficient calculated according to expression (3). X is a (k*k) symmetric matrix with diagonal elements equal to zero and all other elements equal to 49

Journal of Advanced Studies in Finance Gini coefficients measuring in equality between all pairs of the k number of groups due to differences in mean group incomes its calculation based on the RHS of expression (6). That is: |̅

̅| ⁄ ̅ (

(8)

̅)

Where V is again a symmetric matrix with all diagonal elements equal to zero while all other elements are Gini coefficients their computation based on overlapping elements as follows: ∑

|

|

̅̅̅̅

(9)

These coefficients have no special meaning but they are used to make expression (7) equal to (5). In the case of non-overlapping distributions V contains zeros as its elements.17 4. Decomposition and grouped data It was shown in the previous section that the Gini coefficient can be neatly decomposed into three components as given in (7). However, a number of practical problems may arise with this analysis. The basic difficulty has to do with the fact that exact decomposition is feasible only when data on individual incomes are available. The first problem arises even when data of this kind are available, and it relates to the dimension of matrix A and the difficulty of handling it mathematically even with the aid of computers. One way to avoid this problem, on the empirical level, is to consider average sub-group incomes in the place of individual incomes. In this case the dimension of A is drastically reduced. This formulation creates another well-known problem however, since it underestimates the Gini coefficient. The reason for this is twofold: Firstly, it replaces submatrices along the diagonal of A with zeros thus destroying a number of differences and therefore making the numerator in (1) smaller than it would be in case of individual differences for the same data. Secondly, it further underestimates the numerator in (1) since it replaces blocks of elements in each off-diagonal sub-matrices. Aij by their means thus creating an overlapping effect if the income distributions of the sub-groups, which are now being replaced by means multiplied by sub-group populations, are overlapping. The second problem related to the search of methods to improve estimation in the case of grouped data. Although this is not the objective of this paper, an attempt is made to develop a method with this problem in mind. (See next section for its application). Kakwani and Podder (1973) suggest an efficient method that can be used to estimate Gini coefficients. (The elements of B of the previous section in our case). Then the elements of X can be estimated using mean group incomes. Following this the Gini coefficient for the total population (denoted by RG) can be computed independently using the method suggested by Kakwani and Podder. Then another approximate Gini coefficient for the total population can be calculated using a modified form of (7) that does not include V, that is: MG = `[B + X] m

(10)

Subtracting MG from RG we obtain an estimate of the total overlapping component as a residual.The final step is to allocate the overlapping component among elements of V in (7). The straightforward way to do this is to compute overlapping elements using mean sub-group incomes in the place of individual incomes and then normalize taking as base the overlapping component computed as a residual.18 This method results in an exact decomposition of the Gini coefficient but at the same time it “dumps” all other estimation inefficiencies onto the overlapping component since it is estimated as a residual. The practical problem is, therefore, to obtain some idea of magnitude of these inefficiencies relative to the magnitude of the overlapping component. A simple way is to compare the value of the component before and after normalization. This is done with the application in the sections. (See Figure 3)19 5. Decomposition of inequality of Greece 1962-88 All the results of the previous sections were applied to a data base20 on incomes in Greece for the years 1962-88. The data allowed estimation of mean incomes and populations of sub-groups. The sub-groups were allocated to three groups: The group of pensioner including seventeen sub-groups, the group of wage-salary earners with twenty-three sub-groups and finally the group of entrepreneurs containing thirty sub-groups. Taxes and transfer payments are accounted for in the estimation of incomes. Since these are grouped data, they are subject to all the problems described in the last section. The decomposition of the Gini coefficient is done

50

Volume V Issue 1(9) Summer 2014 following the two methods outlined in the previous section of this paper.21 The first method, of using sub-group mean differences, multiplied by the corresponding subgroup populations, in the place of individual incomes, is applied for two reasons: firstly, because it is exact and it serves the purpose of verifying empirically all the results of this paper and secondly, because it can serve as an empirical comparison to the more efficient method described in the last section. Table 1 shows the application of expressions (3) and (4) to the data. This is called “the method of comparison between sub-group means.” Table 2 shows the results from the application of expression (7) using again the method of comparison between sub-group means. The lower case letters denote the weight Gini coefficient i.e. g11, g22, g33. Columns (1), (2), (3) are the elements of in (7) where B is a (3*3) matrix with its diagonal elements equal to G11, G22, G33, columns (1), (2), (3) of Table 1. Columns (5), (6), (7) are again the elements of X in (7) weighted by group population proportion and income shares, while columns (9), (10), (11) show the weighted overlapping elements. Column (4), (8), (12) are the sums of column (1), (2), (3) respectively (the “within” inequality component), (5), (6), (7) (the “between” component), and (9), (10), (11) (the overlapping component). The sum of column (4), (8), (12), in column (13) equals the total (not decomposed) population Gini coefficient, denoted by G, that was calculated using expression (1). The same data are used to decompose the Gini coefficient following the procedure suggested in the last part of section III. This method is called “combined method of estimation, regression and comparison between subgroup means,” and all the coefficients relating to it are labeled by the prefix R. Table 3 is analogous to table 2 only the second method is used now. Column (1), (2), (3) show the weighted elements of the “within” component where the elements of B are calculated using the method suggested by Kakwani and Podder. It should be noted here that all regression coefficients used to calculate the elements of B are statistically significant. Column (5), (6), (7) include the same entries with the corresponding columns of table 2. Total population Gini coefficient, RG, column (14) is computed independently using the method suggested by Kakwani and Podder. Columns (12), (13) show that normalized and estimated overlapping components correspondingly. It is interesting that none of the normalized overlapping elements, column (9), (10), (11) comes out negative and that fact could be considered to be an empirical verification of the discussion in section III. Figure 1 is a graphic exposition of how the three components from the decomposition vary among one another and also between each other according to the two methods of estimation. Figure 2 shows how the total population Gini coefficients, estimated by two methods, very together. It is also observed empirically that the first method of estimation generally underestimates inequality (as suggested in section III), that is G < RG, except for year 1968, and it is more volatile. Figure 3 shows how the estimated and normalized overlapping elements vary together. 6. Possible extension22 The purpose of this work is, as already stated, the explanation of the nature of the overlapping component. This required in-depth analysis from the basic level which is the framework used by Bhattacharya and Mahalanobis (1967) as well as that used by Pyatt. The analysis, however, must be brought to higher levels to be in accord with its current status. To this end it can be extended in at least two respects. The first aspect relates to generalization of the Gini indices as presented in Donaldson and Weymark (1980). They present a class of relative inequality indices characterized by a single parameter and the class of social evaluation functions defined by this class. They also present a corresponding class of absolute indices. The framework of our analysis, in the case of exact decomposition, seems to correspond to their absolute inequality indices (it depends on income differentials). Accordingly, the introduction of “between” inequality, in the case of grouped data, corresponds to their relative indices (only income shares are considered). It seems, therefore, necessary to investigate how the process of generalization is affected by decomposition. This refers to the additional restrictions that need to be placed on the generalization form of the index, in conjunction with the restrictions necessary to satisfy usual conditions an inequality index must satisfy as the principle of population, transfer, and income homogeneity. The second aspect relates to the analysis introduced by Shorrocks (1980). The whole analysis there derives the class of additively decomposable inequality measures under the condition that population sub-groups are disjoint. Since our work explains the nature of the overlapping component it would be advantageous to investigate how the restriction imposed by additive decomposability on the form of the index and on satisfaction of the related principles are affected. The analysis in this respect could be further advanced along the lines of the analysis in Shorrocks (1984) to generalization to just decomposable Gini indices as supposed to only additively decomposable ones.

51

Journal of Advanced Studies in Finance It is obvious now that the final step could be a synthesis of the two lines of analysis referred to with the finding of this work, i.e. the derivation of a generalized additively or just decomposable Gini index allowing division of the populations considered into sub-groups in any possible way and not just disjoint. 7. Summary and conclusion A number of results have been derived in the previous sections of this paper, according to its purpose to analyze the nature of the overlapping component that arises during the decomposition of the Gini coefficient. The first step is to clarify the meaning of decomposition and this is done is section I. The conclusion there is that an exact decomposition of the Gini coefficient without the presence of overlapping is possible in the case of exact (non-group) data. The next step is to investigate the condition under which the overlapping component arises as well as its nature. The conclusion in section II is: firstly, the problem of overlapping appears when the “between” (groups) inequality is computed using mean group incomes in the place of interpersonal comparisons. Secondly, the overlapping component appears exactly because of this replacement and it constitutes a mathematical problem since the order of operations using the method of summing-up absolute values of interpersonal comparisons is not compatible with the order of operation during computation of means. The two methods can be made compatible by compensating for summing-up after taking absolute values in the first case and vise-versa for the second and this compensation is the overlapping component. The consideration of certain problems arising from the use of grouped data is the purpose of section III. The use of grouped data makes estimation of the overlapping component non-exact and two methods of decomposition are suggested in the view of this. The first method treats sub-group income means as individual incomes and it is introduced due to its exactness although it generally underestimates the Gini coefficient. The second method uses more efficient techniques and the overlapping component is treated as a normalized residual. All these results are applied to a data base on incomes in Greece for the period 1962-88. The results of this application are presented in section IV. In conclusion the Gini coefficient might well be the most widely used measure of disparities among incomes and furthermore its decomposition is an interesting and significant step towards better analyzing inequality in incomes. The problem of the non-well defined overlapping component is a disadvantage of this positive development and clarification of its nature strengthens the power of analysis that decomposition offers. References [1] Atkinson, A.B. (1970). On the Measurement of Inequality, Journal of Economic Theory, 2: 244-263. [2] Bakarezos, P. (1984). Personal non-Wealth Income Inequality and Tax Evasion in Greece 1962-1975, Spoudai, 34: 233-257. [3] Bhattacharya, N. and Mahalanobis, B. (1967). Regional Disparity in Household Consumption in India, Journal of the American Statistical Assiciation, 62: 143-161. [4] Debacker, J. Heim, B., Panoussi, V., Ramnath, S., and Vindagos, I. (2013). Rising Inequality: Transitory or Persistent? New Evidence from a Panel of US Tax Returns, Brookings Papers on Economic Activity, 67-142. [5] DeHaan, A. and Thorat, S. (2012). Addressing Group Inequalities: Social Policies in Emerging Economies’ Great Transformation, The European Journal of Development Research, 24(1): 105-124. [6] Donaldson, D. and Weymark, J.A. (1980). A Single-Parameter Generalization of the Gini Indices of Inequality, Journal of Economic Theory, 22: 68-86. [7] Elbers, C., Lanjouw, P., Mistiaen, J., and Ozler, B. (2008). Reinterpreting Between Group Inequality, Journal of Economic Inequality, 6(3): 231-245. [8] Frosini, B. (2012). Approximation and Decomposition of Gini, Pietra_Ricci and Theil Inequality, Empirical Economics, 43(1): 175-197. [9] Guvenen, F. (2011). Macroeconomics with Heterogeneity: A Practical Guide, Economic Quarterly Federal Reserve Bank of Richmond, 97(3): 255-327. [10] Kakwani, N.C., Podder, N. (1973). On the Estimation of Lorenz Curves from Grouped Observations, International Economic Review, 14: 278-292. [11] Kendall, M.G., Stuart, A. (1963). The Advanced Theory of Statistics, Volume I, 2nd ed. London: Griffin. [12] Lambert, P. and Aronson, J. (2000). Inequality, Decomposition analysis and the Gini Coefficient Revisited, The Economic Journal, 103(420): 1221 - 1231.

52

Volume V Issue 1(9) Summer 2014 [13] Mookherjee, D., Shorrocks, A. (1982). A Decomposition Analysis of the Trend in U.K. Income Inequality, Economic Journal, 92: 886-902. [14] Pyatt, G. (1976). On the Interpretation and Disaggregation of Gini Coefficients, Economic Journal, 86: 243255. [15] Rao, V.M. (1969). Two Decompositions of Concentration Ratio, Journal of the Royal Statistical Society, Series A, 132(3). [16] Rhode, P., Strumpf, K. (2003). Assessing the Importance of Tiebout Sorting: Local Heterogeneity from 18501900, American Economic Review, 93(5), 1648-1677. [17] Sen, A.K. (1973). On Economic Inequality. Oxford: Claredon Press. [18] Shorrocks, A.F. (1980). The Class of Additively Decomposable Inequality Measures, Econometrica, 48(3): 613-625.

53

Journal of Advanced Studies in Finance Endnotes See Kendall and Stuart (1963), pp 48-9 Grouped data convey information in the form of mean incomes for groups and sub-groups of the total income-receiving population along with the size of the population of these groups. “Exact data” refers to the situation that individual incomes that can be attributed to individual income-receiving units exist. 3 The elements of the first row of A consist of the absolute differences from the pair-wise comparisons between income of the first unit and all units, including own income and so on, for the second and subsequent rows up to the pth row. If the units are arranged in ascending (or descending) order according to size of income, A has zeros as diagonal elements and it is skewed symmetric. 4 Ai,j (boldface) denotes a sub-matrix while Ai,j its norm as defined in (2). 1

2

5

It is also possible to use for the calculation of

, since Aij = Aji, i≠j, instead of ̅̅̅̅,

and its symmetric counterpart

̅ , and ̅ correspondingly but then Gij ≠ Gji although Aij = Aji. The end result, however, will be the same as in (5) irrespective of whether ̅ and ̅ or ̅̅̅̅ are used. Only in the first case, it is possible for some of the Gji, i≠j, to turn out greater than unity and that means that they cannot be interpreted as Gini coefficients.

It should be noted that In the context of this section “exact decomposition” allows additive recomposition of the Gini coefficient from the “within” and “between” components to the original value, i.e. expression (5) equal expression (1). In the next section this definition is expanded to include the overlapping component as well, i.e. expression (7) equals expression (1). 8 Pyatt decomposes the Gini coefficient in a non-symmetric way similar to that suggested in footnote 5, while his method is to replace interpersonal comparisons with a statistical game. The end result is the same as in (5) only certain of his equivalent to “between” group Gini coefficients are greater than unity. 6 7

9

Overlapping between two group or sub-group distributions is defined as follows: Given ̅ , ̅ , the highest observed income(s) in group i is greater than the lowest income(s) observed in group j. Two distributions are non-overlapping when the highest observed income(s) in group i is less than the lowest income(s) in group j. Besides these two cases, it is also possible to have the highest income(s) in group i equal the lowest income(s) in group j. Then the result, in terms of exact decomposition, is the same as in the case of non-overlapping distributions. The final consideration refers to the extreme case which occurs when all individual incomes in group i are equal and also equal to all incomes in group j (see footnote 17).

can be interpreted as the result of comparisons made by members in the ith group “looking” at ̅ . The elements of the corresponding income of members in the jth group, ̅ are the differences of the opposite comparison. 11 The satiation described here sheds light on the relationship between the statistical game suggested by Pyatt and each pair of Aij, Aji. He seems to consider only positive elements of Aij and only positive elements of Aji, since it is possible to work only with the elements of A above or below the diagonal correcting expression (1) by eliminating division by two. The same result can be obtained by considering the opposite statistical game (elements with negative signs would be considered). In the case of non-overlapping distributions the game results in Aij with all elements equal to zero and all positive elements in Aji. In the case of the opposite statistical game situation reverses itself. 12 Computation if the group income means relates to calculation of the double sum in the way shown in (6). 13 That is: 10

The elements of each

|∑

∑(

|̅̅̅

)|

̅|

In our case (of ascending ordering in incomes) differences termed as of opposite sign are positive for each Aij and negative for each Aji. This definition would be reversed in the case of descending ordering. 15 In our case of ascending ordering of incomes the problem is to have: 14

∑

∑

|

|

|̅̅̅

̅|

|

∑

∑

|

| |, where Xir – Xjw = L, Xir > Ejw considering This can be done by adding to RHS expression the quantity R = ∑ elements of Aij (that is a difference of opposite sign), and nij is the number of differences of opposite signs in Aij (positive in this case). In other words, absolute signs are taken after addition is performed in ∑ and ∑ since that is the only ̅̅̅ ̅ order of operations that allows estimation of . 16

It is the case, when we have to use mean group income and not individual incomes, for one reason or another, that

54

Volume V Issue 1(9) Summer 2014 creates the whole problem. If we were free to use individual incomes, we would use decomposition as given in (5).

̅ , i≠j i.e. “within” inequality is When the distributions are overlapping in expression (8) will be zero when ̅̅̅ zero since group income means are equal. In this case, this LHS if expression (6) equals R (if we allowed addition in the LHS to be performed before taking signs then inclusion of R is not receive the same amount of income. 18 There is no warranty that the estimated value of the overlapping element will be positive. Nevertheless the procedure of normalization can still be carried out and negative overlapping elements have the same meaning as positive ones, that is they correct the elements of X due to consideration of mean instead of individual incomes. 19 The main result of this section which should be emphasized is that it is not possible to decompose the coefficient exactly in the case of grouped data and the best that can be done is to use some kind of approximation as done here. This problem is not a consequence of the process of decomposition but of the use of grouped data which affects estimation of the coefficient even in the form of expression (1). On the other hand, it is clear that there exists some relation, albeit approximate, between the cases of exact and grouped data since the elements of B in (7) are calculated using some efficient method (the one suggested by Kakwani and Podder in this case) and the elements of X in (7) are the same during ̅, . approximation, when ̅̅̅ 17

This data base was originally complied by author from a wide variety of sources for the year 1962-75, (1984). The base was re-calculated using an even wider variety of sources an expanded up to 1988 for the purposes of this study. 21 A Personal Computer was used for the calculations. No computer programs for the required work existed so they were developed by the author using LOTUS 123 and its macro commands (LOTUS 123 is a trademark of Lotus Development Corporation). 22 This is the content of my current research. 20

55



EFFECT OF INLATION ON NOMINAL AND REAL STOCK RETURNS: A BEHAVIORAL VIEW Andrey KUDRYAVTSEV The Economics and Management Department The Max Stern Yezreel Valley Academic College, Emek Yezreel 19300, Israel [email protected] Eyal LEVAV The Economics and Management Department The Max Stern Yezreel Valley Academic College, Emek Yezreel 19300, Israel Shosh SHAHRABANI The Economics and Management Department The Max Stern Yezreel Valley Academic College, Emek Yezreel 19300, Israel10 Suggested Citation: Kudryavtsev, A., Levav, E., Shahrabani, S. (2014). Effect of inflation on nominal and real stock returns: A behavioural View, Journal of Advanced Studies in Finance, (Volume V, Summer), 1(9):56-65. DOI:10.14505/jasf.v5.1(9).05. Available from: http://www.asers.eu/journals/jasf/curent-issue. Article’s History: Received March, 2014; Revised June, 2014; Accepted July, 2014. 2014. ASERS Publishing. All rights reserved.

Abstract: The main goal of this study is to shed light on different aspects of the correlation between stock market returns and inflation. Using historical monthly rates of inflation calculated by the Israel Central Bureau of Statistics and historical monthly returns on six leading stock indices of the Tel Aviv Stock Exchange, we document that nominal stock index returns do not compensate investors either for contemporaneous inflation (inflation rate registered during the same month as the index return) or for previous month's inflation (inflation rate officially announced during the month when the index return is observed). In addition, the results indicate that real stock index returns (calculated by deducting the rate of inflation from the nominal return rate) are negatively correlated with contemporaneous inflation rates. We suggest a behavioral explanation for these findings based on the assumption that investors perceive inflation itself as bad news and subsequently downwardly update their estimates of (at least real) expected cash-flows from holding the stocks and/or upwardly update the expected risk levels of the stocks. Keywords: heuristics; inflation; investors' expectations; nominal stock returns; real stock returns. JEL Classifications: D80, D84, E31, G12, G14. 1. Introduction The effect of inflation rates on nominal stock returns has been the subject of extensive research. Over the last decades, many studies have examined the linkage between inflation rate and nominal stock returns. Fisher (1930) suggests that nominal returns of stocks are a hedge against inflation, so that an increase in current and expected inflation should increase expected nominal dividend payments. This theory is also supported by Gordon (1959), who argues that inflation and stock returns should be highly and positively correlated since rational markets evaluate stocks based on their discounted expected nominal dividend cash flows. In contradiction with these classical economic theories, a number of more recent empirical studies have found very limited, if any, support for the hypothesis that nominal stock returns may serve as a hedge against inflation (e.g., Geske, and Roll 1983; Gallagher, and Taylor 2002; Kim 2003; Oxman 2012). This contradiction between classical 10

Authors' names appear in alphabetical order. 56

Volume V Issue 1(9) Summer 2014 economic theories and empirical results is known as the "inflation-stock returns puzzle" (Nelson 1976; Fama, and Schwert 1977). The current study examines the correlation between monthly inflation rates in Israel and monthly nominal and real index returns on the Tel-Aviv Stock Exchange in order to determine whether the inflation-stock returns puzzle is present in the Israeli market. Investigating the correlation between rates of inflation and nominal stock returns on the Tel-Aviv Stock Exchange is important for several reasons. First, the findings of this study are important not only for Israeli stockholders, but also for international investors and economic policy makers. Despite the complex geopolitical map of the Middle East, the Israeli market represents a long-term and reliable economy that attracts many investors and international companies 11 . If the inflation-stock returns puzzle is indeed present in the Israeli market, this study may provide valuable information for all kinds of investors in the Israeli market. Second, a number of empirical studies investigate the correlation between inflation rates and monthly nominal returns of stocks in many different markets (e.g., Solnik 1983; Kaul 1987; Ritter, and Warr 2002; Hagmann, and Lenz 2005). Yet, these previous studies concentrate mainly on the effect of inflation on nominal stock returns only. In this context, we amplify the analysis by testing for the effect of inflation on both nominal and real stock market returns. Third, this study offers an innovation by proposing that the inflation-stock returns puzzle is also present one month after the inflation itself actually took place. In other words, we expect that the effect of inflation on the nominal stock returns continues in the month following the one during which the inflation was actually registered. Thus, the study makes a new contribution to the current literature and extends the discussion on the long-term effect of the inflation-stock returns puzzle. Using historical monthly rates of inflation calculated by the Israel Central Bureau of Statistics (CBS) and historical monthly returns for six leading stock indices of the Tel Aviv Stock Exchange, we document that nominal stock index returns do not compensate investors either for contemporaneous inflation (inflation rate registered during the same month as the index return) or for previous month's inflation (inflation rate officially announced during the month when the index return is observed). We also find that real stock index returns (calculated by deducting the rate of inflation from the nominal return rate) are negatively correlated with contemporaneous inflation rates. We suggest a behavioral explanation for these findings based on the assumption that investors perceive inflation itself as bad news and subsequently downwardly update their estimates of (at least real) expected cash-flows from holding the stocks and/or upwardly update the expected risk levels of the stocks. The remainder of this study is structured as follows. Section 2 reviews the literature, Section 3 presents the research hypotheses, Section 4 describes the database employed in the study, Section 5 provides the results, and Section 6 concludes. 2. Literature review 2.1. Classical economic theories Fisher's model. Fisher (1930) argues that the nominal return of an asset for a certain period embodies the real yield of that asset and the market assessment of the expected inflation rate during that particular period. In other words, any real asset is fully hedged against inflation. Fisher believes that in an ideal world, price level has no impact on the real sector, which is affected exclusively by real factors such as productivity of capital, time preference and risk aversion. If the market is rational and efficient, the nominal yield of an asset at time t will optimally express the real return of the asset at time t and the inflation that occurred during the period preceding t. That is, nominal yields are positively correlated with expected inflation, so that a one percent increase in expected inflation rate is thought to increase an asset's nominal rates of return by one percent. Therefore, long-term investors should be fully compensated for the changes in the prices of real assets that occur as a result of inflation. Gordon's model. According to Gordon (1959), one can also assume that nominal stock returns and inflation rates should be positively and highly correlated. Gordon's model states that rational markets need to evaluate stocks by discounting their future nominal dividend payments. The respective discount rate should be determined by the rate of return that investors expect to gain as a dividend yield and/or a capital yield on the stock. Therefore, an increase in inflation expectations and actual inflation rates should increase the expected flow of future nominal dividend payments for the stocks, and thus create an upward revision of stock prices. 11

According to the Bank of Israel, at the end of 2013 foreign investments in Israeli stocks and bonds amounted to about USD 34 Billion.

57

Journal of Advanced Studies in Finance 2.2. Inflation-stock returns puzzle Contrary to Fisher's and Gordon's theories, over the last twenty years many empirical studies document that nominal stock returns are negatively correlated both with expected and with actual rates of inflation (e.g., Nelson 1976; Fama, and Schwert 1977; Geske, and Roll 1983; Solnik 1983; Kaul 1987; Balduzzi 1995; Hess, and Lee 1999; Oxman 2012; Kudryavtsev 2013). Fama (1981) suggests that the negative correlation between inflation and stock returns holds because inflation is negatively related to real economic activity and because there is a positive association between real activity and stock returns. Caporale, and Jung (1997) contend that stock returns are inversely related to inflation, even when controlling for economic output. Perez de Gracia, and Cunado (1999) analyze inflation rates and stock returns in Spain during the period from 1941 to 1999 and corroborate the existence of a Granger causality relationship between them. Also in the Spanish market, Ferrer (2000) uses cointegration techniques and VAR models and concludes that the relationship between inflation and stock returns is negative and permanent. Ritter, and Warr (2002) use a residual income model to show that valuation errors of leveraged stocks in the presence of inflation cause depressed stock prices. Therefore, decreasing inflation may have triggered the bull market in the United States from 1982 to 1999. Gallagher, and Taylor (2002) argue that the significant relation between stock returns and inflation is not purely due to demand innovations, but rather that stock returns are also strongly negatively correlated with inflation due to supply innovations. Hagmann, and Lenz (2005) employ a VAR approach and conclude that the negative correlation between inflation (expected inflation) and stock returns is explained by aggregate supply shocks being stronger than demand shocks in the US. With respect to monetary shocks, these researchers find a positive but not significant correlation between inflation and stock returns. Despite the above, the puzzle is still not fully explained. Indeed, the situation is particularly intriguing in the context of stocks which, if compared to real assets, should immediately compensate investors for inflation. In other words, the models of Fisher and Gordon fail when it comes to explaining the relationship between nominal stock returns and inflation (Fama, and Schwert 1977). The contradiction between the findings of these empirical studies and the predictions of classical economic theories is known as the inflation-stock returns puzzle, and may serve an indication of irrationality in capital markets. 2.3. Subjective probability heuristics One potential explanation for the inflation-stock returns puzzle is based on the notion of subjective probability heuristics. In general, heuristics are rules of thumb, based on common sense or intuition and offering a quick and easy way to make decisions. Tversky, and Kahneman (1974) argue that estimating the probability of an event is a complicated task for individuals since their time and cognitive resources are limited. As a result, people rely on heuristics while making subjective estimates of the likelihood of various events. Tversky and Kahneman mention three major types of subjective probability heuristics:  Representativeness heuristic: People estimate the probability that a certain event will occur according to its resemblance to another event that has happened or is going to happen. According to an example by Tversky, and Kahneman (1974), suppose an individual is presented with a description of a company that does not provide data that may be relevant for estimating the company's ability to generate profits. Subsequently, the individual is asked to predict the future earnings of that company. If the description of the company is positive, the individual tends to assume a higher probability that the company will generate profits, whereas if the description of the company is negative, the individual tends to predict a lower probability of profits. Although the description does not include variables relevant for forecasting the company's profits, the individual may see the similarity between a good company description and the company's future success and thus assume higher probability of profits, and vice versa.  Availability heuristic: People tend to determine the probability of certain events according to the ease of recalling similar events (in their own memory). That is, the easier an event can be recalled and imagined, the greater its predicted probability. For example, Tversky, and Kahneman (1974) conduct an experiment in which participants are ask to assess whether the English language contains more words beginning with the letter R or more words with R as the third letter. To answer this question, the participants probably attempt to recall as many words as they can that begin with the letter R, e.g., "road," and as many words that have R as the third letter, e.g., "car." Since it is easier to quickly recall words that begin with a given letter, most of the participants argue that these words are more prevalent in English, though in fact there are twice as many English words with R as the third letter.  Anchoring and Adjustment heuristic: People commonly use this heuristic to resolve numerical questions or problems. An individual who tries to assess a number that is not known precisely, such as price, cost or 58

Volume V Issue 1(9) Summer 2014 time, may tend to rely on a certain initial value that seems likely. This value is called the "anchor." The anchor may be explicitly given to the individual or may itself be the result of some partial calculation. The final estimation may rely on this anchor and be adjusted upwards or downwards as necessary, but in many cases it remains relatively close to the initial anchor. As a result, new relevant information may have a smaller impact on the individual, thus leading to a deviation in the individual's estimations. In other words, the adjustment offered by the additional information may be insufficient. For example, in an experiment conducted by Tversky and Kahneman, two groups of participants are asked to assess the percentage of African countries in the United Nations. One group is asked whether this percentage is greater or less than 10%, and then asked to actually evaluate it. The second group is asked whether this percentage is greater or less than 45%, and then asked to actually evaluate it. The participants appear to be strongly affected by the initial (and absolutely non-relevant) percentage. The median answer in the first group is 25%, while the median answer in the second group is 65%. Reliance on these heuristics may lead to systematic deviations. Many studies show that heuristics may have an effect on people's probability judgments and create behavioral patterns that do not match theoretical economic expectations (Chiodo et al. 2004). In the context of the current study, heuristics may be regarded as a potential explanation for the inflationstock returns puzzle. There is broad consensus that one of the responsibilities of the central bank is to take preventive actions in the presence of any concerns about the future of the economy, and that financial instabilities may be a result of the central bank's inability to maintain price stability. Furthermore, the prices of assets and stocks in particular are considered to serve as tools for predicting future economic changes, so that an increase in stock prices will eventually lead to inflation (e.g., Lee 1992). Consequently, the central bank will have to take preventive actions to decrease inflation expectations, including increasing the real interest rate, thus probably forcing investors to re-evaluate the financial assets they hold. This is because the value of assets is affected by the discount rate employed for the expected future cash flows on those assets, so that an increase in the real interest rate that increases the discount rate will reduce the value of the assets. The decrease in the value of assets will, in turn, lead to a decline in private consumption and reduce companies' investment plans. Ultimately, these changes will reduce economic activity and hopefully restrain inflation (Laopodis 2006)12. Now, let us suppose that inflation not only affects nominal dividend levels, but is also employed as a variable in people's information set. In addition, assume that investors believe that high inflation is usually followed by increases in real interest rates, thus depressing economic growth. Then, for the investors inflation may become a sign of future slowdown in economic growth, not only for objective reasons (see, for example, Barro (2013) for a review of the literature dealing with the negative effects of inflation on economic growth), but also because this is how investors think. In this case, a high level of inflation may be perceived as bad news. According to the availability heuristic, high current levels of inflation may cause people to more easily recall periods characterized by economic slowdown and, according to the representativeness heuristic, may create expectations that monetary policies intended to work against inflation should be implemented13. In practice, two "mental actions" take place simultaneously. On the one hand inflation raises expected nominal dividends, and on the other hand it creates a pessimistic outlook among investors, resulting in reduced expectations of future dividend payments. The direction of the overall effect is not clear, which is consistent with the fact that nominal stock returns do not increase as a result of inflation, and in many cases are even negatively affected by it. In other words, nominal stock returns usually do not compensate investors for inflation (Chiodo et al. 2004).

Nevertheless, modern history includes quite a few examples where high or even extremely high inflation rates persisted in spite of increased interest rates, just because people no longer relied on the respective currency's stability. In such cases, the effect of inflation on the economy is even worse, as it is accompanied by a general slowdown. 13 Note that sometimes high inflation occurs during periods of economic growth (and bull markets). Yet, when it occurs and starts to be an urgent macroeconomic problem, investors expect the central bank to react by increasing the interest rate, cooling down the economy, and thus shifting the stock market downward. 12

59

Journal of Advanced Studies in Finance 2.4. Investors' reaction to inflation Monthly inflation rates in Israel are routinely announced by the Israel Central Bureau of Statistics (CBS) fifteen days after the end of the sampling period, that is, the period (month) when the inflation actually took place. If investors receive information concerning inflation only from CBS inflation announcements, we may assume that the economy as a whole should make a complete adjustment for inflation on the day immediately following the announcement. But a study by Huberman, and Schwert (1985) examining the effect of inflation on bond prices in Israel finds that 85% of bond price reaction to inflation occurs during the sampling period (month of inflation) itself, that is, during the two- to six-week period prior to the inflation announcement. Only 15% of bond price reaction to inflation occurs on the day following the inflation announcement — clearly a low impact percentage compared to the theoretical assumption that the full adjustment should take place only on that day. In addition, according to Huberman and Schwert's findings, inflation has no effect on bond prices during the two weeks between the end of the sampling period and the inflation announcement. Huberman and Schwert argue that the information necessary for estimating the rate of inflation is freely accessible public information, available to all investors. Moreover, investors should have an economic incentive to correctly estimate the expected inflation rate, which may provide them with an important tool for making financial decisions. Consequently, many investors estimate the changes in the Consumer Price Index (CPI) and respond accordingly to inflation at the time the inflation occurs, that is, during the sampling period. However, due to the high costs involved in measuring inflation because measurement methods are only partially available to investors, and due to the lack of time, most investors fail to effectively measure inflation. Hence, the CBS inflation announcements still incorporate new, albeit minor, information about inflation, resulting in a market correction following the arrival of this new information. 3. Empirical models and research hypotheses In order to test the relationship between rates of inflation and nominal stock returns, the current study uses a model based on Fisher's (1930) hypothesis for stock returns, which argues that the nominal return of an asset during a certain period embodies the real yield of that asset and the market assessment of the expected inflation rate at that certain period. This hypothesis is empirically tested by running the following regression for each given stock market index i: N it   i   i t   it (1) where N it is the monthly nominal return of index i in month t, and  t is the monthly inflation rate in month t.

According to Fisher's (1930) hypothesis for stock returns, the correlation between the inflation rate and the nominal stock returns should be equal to one (   1 ). The main goal of the present study is to estimate the actual correlations between the monthly inflation rates and the monthly nominal stock returns of TASE indices. In other words, our goal is to estimate the coefficients of β in regression (1). Other empirical studies document contradictions in Fisher's (1930) hypothesis, which claims that stocks should compensate investors for inflation. The empirical findings indicate that nominal stock returns do not compensate investors for inflation, and sometimes are even negatively correlated with inflation rates. The main objective of our study is to examine whether nominal index returns in the Israeli capital market embody a compensation for inflation. We hypothesize as follows: Hypothesis 1: There is no significant compensation for inflation in nominal TASE index returns during the month when the inflation takes place. This hypothesis emerges from a potential explanation for the stock return-inflation puzzle that is based on the concept of availability and representativeness heuristics. The heuristics explain the puzzle by suggesting that inflation generates pessimistic perceptions among investors, resulting in reduced expectations with respect to future dividend payments and therefore upsetting the increase in dividend payments following nominal inflation and creating a situation in which the nominal stock returns are not affected, and sometimes even negatively affected, by inflation. If the first hypothesis of this study is corroborated, we may expect regression (1) not to yield positive and significant coefficients of β. Moreover, we may expect to find significantly negative coefficients for some of the indices, which would represent negative effects of inflation on nominal stock return. Another matter of interest for our study is the correlation between inflation rates and nominal stock returns one month after the inflation actually occurred. Since investors have access to all information necessary for measuring inflation, many of them probably estimate the changes in the CPI and formulate a response to

60

Volume V Issue 1(9) Summer 2014 inflation during the period when inflation actually takes place, indicating that there should be no additional compensation for inflation in stock index nominal returns during the following period. In line with the study by Huberman and Schwert (1985) documenting the effect of inflation on nominal assets' returns during the same month (the month when the inflation took place) and the next one, we run the following regression: (2) N it   i   i  t 1   it and hypothesize that: Hypothesis 2: There is no significant compensation for inflation in nominal TASE index returns during the month following the one when the inflation took place. Finally, we are interested in analyzing the effect of inflation on real stock returns, that is, on stock returns after deducting the contemporaneous inflation rate. In continuation of Hypothesis 1, we expect that if nominal stock returns do not compensate investors for inflation, real stock returns should even decrease as a result of inflation. Therefore, for each of the indices, we run the following regression: (3) Rit   i   i t   it where

Rit is the monthly nominal return of index i in month t, calculated as: Rit  N it   t

(4)

and hypothesize that: Hypothesis 3: Real TASE index returns are negatively correlated with contemporaneous rates of inflation. 4. Data description 4.1. Rates of inflation The data used in this paper to determine the rate of inflation consist of monthly values of the Consumer Price Index (CPI) in Israel. The data were retrieved from the Bank of Israel historical database14. The sampling period is from January 2000 through July 2012. We employ natural logarithms of actual changes in the CPI to represent the rates of inflation. Table 1 reports basic statistics for the monthly inflation rates. Table 1. Monthly inflation rates – basic descriptive statistics Minimum

Maximum

-0.8%

1.5%

Standard deviation 0.4%

Median

Mean

0.1%

0.1%

No. of observations 150

Through

From

07/12

02/00

4.2 Nominal stock returns Natural logarithms of actual changes in the value of stock indices are employed to represent nominal stock returns. We employ monthly returns on the following stock indices routinely calculated on the Tel-Aviv Stock Exchange (TASE):  TA-25 – Index that tracks the share prices of the 25 companies with the highest market capitalization on the TASE.  TA-75 – Index that tracks the 75 companies with the highest market capitalization that are not included in the TA-25 index.  TA-100 – Index comprising the 100 shares with the highest market capitalization on the TASE that are included in the TA-25 and TA-75 indices.  TA MidCap-50 – Index consisting of the 50 shares with the highest market capitalization that are not included in the TA-100 index.  General Shares and Convertibles – Index that includes all the stocks and convertible securities traded on the TASE.  TA Blue Tech-50 – Index that comprises 50 shares with the highest market capitalization that are part of the TA-Technology or TA-Biomed sector indices. The data can be found on the official website of the Bank of Israel http://www.boi.org.il/en/DataAndStatistics/Pages/Default.aspx. 14

61

Journal of Advanced Studies in Finance The data were retrieved from the Tel-Aviv Stock Exchange historical database15. The sample period for the nominal stock indices is from January 2000 through July 2012. Table 2 reports basic statistics for the monthly nominal stock returns. Table 2. Monthly nominal stock returns – Basic descriptive statistics Standard No. of Median Mean Through From deviation obs.

Minimum

Maximum

Index

-18.1%

14.9%

6.1%

1.40%

0.70%

149

06/12

02/00

TA-25

-25.7%

29.1%

7.6%

0.60%

0.40%

149

06/12

02/00

TA-75

-19.6% -29.8%

15.0% 22.5%

6.1% 8.3%

1.40% 0.60%

0.60% 0.30%

149 109

06/12 07/12

02/00 07/03

TA-100 TA MidCap-50

-17.5%

17.1%

5.6%

1.30%

0.60%

150

07/12

02/00

General Shares and Convertibles

-27.2%

46.9%

9.0%

-0.04%

0.03%

150

07/12

02/00

TA Blue Tech-50

5. Results We begin by analyzing the effect of inflation on contemporaneous nominal stock returns, that is, on returns during the month when the respective rate of inflation was registered. We test Hypothesis 1 by running regression (1). Table 3 shows the basic regression results with respect to each of the indices. The results corroborate Hypothesis 1. None of the indices provides compensation for inflation in the form of contemporaneous nominal returns, with only TA MidCap-50 nominal returns showing a positive, though not significant, correlation with contemporaneous inflation rates. Moreover, TA Blue Tech-50 nominal returns are significantly and negatively affected by the inflation registered at the time the returns were measured. Thus, we may conclude that the inflation-stock returns puzzle is present in the Israeli stock market as well, probably indicating that Israeli investors, like most investors around the world, regard inflation itself as bad news and subsequently downwardly update their estimates of (at least real) expected cash-flows from holding the stocks and/or upwardly update the expected risk levels of the stocks. Table 3. Effect of inflation on contemporaneous nominal index returns DurbinWatson

R-squared

p-value of β

β

α

1.75 1.70 1.90 2.07

0.01 0.02 0.03 *0.09

0.31 *0.08 0.40 0.87

-1.05 -2.25 -0.91 0.27

0.009 0.008 0.007 0.003

-0.85 -3.45

0.006 0.005

1.88 0.03 0.38 2.01 **0.09 **0.01 Asterisks denote 2-tailed p-values: *p 35. 20 and Max-Eigen statistic: 281.60 > 22.30). In addition, we can also reject the null hypothesis that there is at most one cointegration and there are at most two co-integrations at a 95% confidence interval. In conclusion, there are two linear independent co-integration vectors in this system at a 95% confidence interval. 4.3. Granger causality/block exogeneity Wald tests According to Kozhan (2010), one of the main purposes of the VAR model is forecasting. If Y1 contains the useful information which can be utilized to forecast Y2 or a group of variables, then we can conclude that Y1 granger-causesY2. In accordance with the Granger Causality test, the Wald statistic is used to find out, whether the causality relationship is significant. If the granger causality effect from Y1 to Y2 is significant, we can say that Y1 can cause Y2 to move and the information is impounded into Y1 first. ∑

∑

∑

∑

4.4. Impulse – response function According to Diebold (2008), impulse-response is one of the methods of characterizing the dynamics in the VAR model. The impulse-response function can be utilized to measure the effect of one standard deviation innovation from the random disturbance term to the endogenous variable in the short-run and long-run. In the impulse-response formula, s denotes coefficient matrices, and the elements of coefficient matrices indicates the influence of shocks on Yt.

71

Journal of Advanced Studies in Finance ∑(

)(

)

∑(

)

In the impulse-response formula, s denotes coeffieient matrices, and the elements of coefficient matrices indicates the influence of shocks on Yt. Before we conduct the impulse-response function, we need to detect whether the VAR model is stable. The inverse roots test is employed to test the stability of the VAR models. If all the points are within the circle, we can conclude that the VAR model is stable and the subsequent impulse-response function is meaningful. Since all the VAR models we established in this research are stable, the usage of the impulse– response function to characterize the dynamics in the VAR models is feasible in the analysis. 5. Empirical results and discussion 5.1. Preliminary analysis: Unit root test and testing for optimal lag The augmented Dickey Fuller (1981) test indicates that all the price series are non-stationary. We transformed the market indexes and stock prices to stock returns which become stationary17. In addition, the Schwarz information criterion (SIC) points out that the optimal lag order of the VAR model is one for almost all of the price return series, indicating that for the same day return of one market, the one day lagged return of the other two markets should be used as explanatory variables. For simplicity purpose, we use one lag for all estimation below. The return of HNP in SSE 8

4

The return of HNP in SEHK

The return of HNP in NYSE

12

12

8

8

4

4

0

0

-4

-4

-8

-8

0

-4

-8

-12

-12

-12 I II III IV I II III IV I II III IV I II III IV I II III IV 2009 2010 2011 2012 2013

I II III IV I II III IV I II III IV I II III IV I II III IV 2009 2010 2011 2012 2013

I II III IV I II III IV I II III IV I II III IV I II III IV 2009 2010 2011 2012 2013

Figure 1: Line graph for the returns of the stocks triple-listed (HNP Period 4 as an example)

5.2. Market Co-integration Table 6: Trace statistic and max-Eigen statistic (for Hang Seng Index, Shanghai A-Share Index, S&P 500) Trace Test 0.05 Critical Max-Eigen 0.05 Critical Period r① Statistics Value Statistic Value 982.91② 35.20② r=0 431.35 22.30 Period 1 r1 551.56 20.26 296.08 15.89 2 255.48 9.16 255.48 9.16 r=0 1118.98 35.19 478.00 22.30 Period 2 r1 640.98 20.26 347.71 15.89 r2 293.27 9.16 293.27 9.16 r=0 630.92 35.19 308.04 22.30 Period 3 r1 322.89 20.26 180.20 15.89 r2 142.69 9.16 142.69 9.16 17

The detailed information of the ADF test is available upon request. 72

Volume V Issue 1(9) Summer 2014 Period

r①

Period 4

r=0 r1 2

Trace Test Statistics 1552.03 800.49 378.64

0.05 Critical Value 35.19 20.26 9.16

Max-Eigen Statistic 751.54 421.84 378.64

0.05 Critical Value 22.30 15.89 9.16

Note: r = 0 denotes the hypothesis that there is no co-integration, r 1 denotes that the hypothesis that there is at most one co-integration, and r 2 denotes the hypothesis that there is at most two co-integrations; If the Trace Test Statistics or the Max-Eigen Statistic is greater than the critical value, it denotes rejection of the hypothesis at a 95% confidence interval.

The first research question that needs to be answered is whether the price series of the SEHK, SSE and NYSE are co-integrated. The long-term co-integration relationships between the three markets are testes by the Johansen's co-integration analysis (Johansen & Juselius, 1990). As shown in Table 6, the results of Trace test and Max-Eigen test reveal that the hypothesis of no cointegration between the returns in the SEHK, SSE and NYSE for all stocks in all four periods is rejected at a 95% confidence interval. The same results were found in the testing for the co-integration of the return of the Hang Seng Index, Shanghai A Share Market Index and S&P 500 Index, indicating that there is a co-integrating relationship among the stock exchanges of these three markets in the time frame from January 1, 2000 to December 31, 2013. Furthermore, the hypothesis of at most 2 co-integration vectors is rejected in the Trace and Max-Eigen tests, indicating that there must be at least two independent linear co-integrating vectors between the three series. Hence, we can conclude that the stock returns in the three markets in Hong Kong, China and the U.S. have a long-term co-integrating relationship, which is consistent with the law of one price. 5.3. Granger causality effects (price discovery) Table 7 reports the results of the vector autoregressive Granger Causality Test. The results show mixed conclusions across the four sub-periods. According to the results in Tables. 7a and 7b, for almost all cross-listed stocks, the Chinese stock market seemed to be isolated and independent prior to financial crisis. The Chinese market does not exert granger causality over other markets (columns 4 and 5), nor was it affected by the other two markets (columns 1 and 2). Taking a more macroeconomic view, the same results were found among the returns of the indexes. In all cases, we cannot reject the null hypothesis that there is no causal relationship from China to the other markets and vice versa. This is consistent with the studies of Diekmann (2011) and Sheu and Cheng (2011) that the returns in the Shanghai Stock Exchange cannot be predicted by the returns in other global markets; nor can they predict the returns of other markets. Table 7: Results of Granger causality test a) Period 1 HK→China①

US→China

China→HK

US→HK

China→US

HK→US

INDEX

0.1012

0.5677

0.0140

0.0000

0.5106

0.7791

CEA CHU HNP SHI SNP YZC b) Period 2

0.7591 0.5288 0.3804 0.1141 0.3633 0.6809

0.8037 0.8730 0.8908 0.5795 0.5968 0.9875

0.1368 0.3763 0.9080 0.4106 0.6343 0.5709

0.0000 0.0052 0.0004 0.0002 0.0008 0.4506

0.1551 0.2461 0.7825 0.8367 0.2338 0.3874

0.0166 0.0001 0.4092 0.0000 0.0015 0.0000

HK→China

US→China

China→HK

US→HK

China→US

HK→US

INDEX

0.4608

0.7578

0.9277

0.0000

0.4269

0.1800

CEA CHU HNP

0.6199 0.3546 0.2754

0.2855 0.7902 0.5998

0.7418 0.0163 0.9215

0.0000 0.0000 0.0000

0.7974 0.2548 0.5026

0.0006 0.9251 0.5583

73

Journal of Advanced Studies in Finance SHI SNP YZC ZNH

0.0216 0.0100 0.0866 0.3455

0.7400 0.1252 0.5083 0.1758

0.4727 0.1420 0.8302 0.0090

0.0000 0.0000 0.0000 0.0000

0.4766 0.2826 0.7470 0.0211

0.7664 0.8415 0.7172 0.0001

c) Period 3

HK→China

US→China

China→HK

US→HK

China→US

HK→US

INDEX

0.5562

0.0002

0.0038

0.0000

0.3280

0.2620

CEA ACH CHU GSH HNP LFC PTR SHI SNP YZC ZNH d) Period 4

0.2210 0.3186 0.5432 0.1453 0.4272 0.0033 0.0588 0.4336 0.1471 0.3815 0.2735

0.0008 0.0002 0.0349 0.0134 0.0027 0.0000 0.0002 0.0424 0.0000 0.0276 0.0014

0.9789 0.0957 0.0280 0.1087 0.0710 0.4247 0.0380 0.0631 0.1029 0.4298 0.9999

0.0060 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

0.4714 0.4386 0.0752 0.3097 0.2061 0.2229 0.1561 0.5116 0.4016 0.6363 0.6597

0.0203 0.2186 0.9664 0.4644 0.9350 0.2503 0.3222 0.2880 0.9720 0.7334 0.2830

HK→China

US→China

China→HK

US→HK

China→US

HK→US

INDEX

0.0889

0.0000

0.2271

0.0000

0.6261

0.2110

CEA ACH CHU GSH HNP LFC PTR SHI SNP YZC ZNH

0.9289 0.0066 0.0055 0.4282 0.9169 0.0028 0.0127 0.0960 0.4462 0.0947 0.2295

0.0607 0.0000 0.0001 0.0830 0.0118 0.0000 0.0000 0.0007 0.0000 0.0000 0.0005

0.6946 0.9626 0.0342 0.0119 0.3686 0.6149 0.1122 0.9255 0.5747 0.6829 0.0014

0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

0.9029 0.9086 0.1570 0.0536 0.3402 0.8533 0.3757 0.8457 0.8246 0.3765 0.1031

0.0000 0.0954 0.0955 0.0167 0.4448 0.0072 0.0000 0.0003 0.0130 0.0874 0.0029

Notes: The order in the first row in each period, for example HK – China, indicates the return of stock listed in Hong Kong Granger cause the return of stock listed in China. ② The statistic denotes in the table denotes the p value of the Granger Causality at the 0.05 level. If p value is less than 0.05, it indicated Granger Causality exists.

74


Figure 2: Results of Granger causality test (The effects of the U.S. and Hong Kong on China) In the same time frame, a significant causal relationship was detected between the Hong Kong and the US. In all cases expect for the YZC stock in period I, the returns in the NYSE have a significant causality over returns in the SEHK. On the contrary, the SEHK only has a price influence over New York listed stocks in some cases (5 stocks in Period I and 2 stocks in Period II). By comparing Tables 7a and 7b, it appears that the QFII program did not cause the Chinese Stock Exchange to be more integrated with other global markets. The integration of China did not improve after 2003.07.09, when the first qualified foreign institutional investor started trading in China. This may be due to the tight restrictions of the QFII regulations (one foreign investor can only hold up to 10% of the shares of the Chinese company) and the rigorous and complex approving procedure (it usually takes more than one year to be approved as a qualified foreign investor). Furthermore, according to Sun & Lin (2006), up to September 7, 2005, the trading quota conducted by qualified foreign institutional investors reached 32.85 billion RMB, which only accounts for 0.93% of the total market value of the A share and B share exchange markets. Thus, the trading quota for those qualified foreign investors is limited. Hence, the QFII did not have any effect on the integration of the Chinese exchange market. However, during and posterior to the financial crisis, a new pattern seems to have been emerging. Starting from April 2007, a causal relationship from the NYSE to the SSE was discovered, as indicated by the small pvalues in the second column in the Tables 7c and 7d. For almost all stocks as well as the indexes, returns in the NYSE have a significant effect on the returns in the SSE, contradictory to the results of the data series prior to the financial crisis. Simultaneously, the SEHK has a significant causal effect on the SSE in the case of a few stocks. Yet, the number seems to increase with time. The sudden overwhelming effect of the NYSE and the increasing influence of the SEHK on the SSE in Period 3 and 4 suggest that the Chinese financial market has become more integrated with the other two global financial centers due to the financial crisis. Across the whole time series, China could never or only in rare cases influence the Hong Kong and US market. The returns in the NYSE always have a spillover effect on the returns in the SEHK, while the returns in the SEHK have only a partial influence on the returns in the NYSE. One implication derived from the Granger Causality test is that foreign markets play the dominant role in the price discovery process, which is conflicting with the home bias hypothesis that the home market is the dominant market in the price discovery process. According to our results, the information transmission seems to happen from the US market to the other markets. The second dominant market is the Hong Kong market, which in some cases is able to influence the Chinese and US market. The result is not consistent with the claim proposed by Qadan and Yagil (2010), Hauer, Tanchuma, and Yaari (1998), Eun and Sabherwal, Wong and Zurbrueg (1988), Melvin, Gramming and Schlag (2000), and Climent, Pascual and Pascual-Fuster(2001), who claimed that the home market is the dominant one by using the Granger causality test. In the case of Chinese triple-listed companies, the foreign markets seem to be the determining market, whose returns have a forecasting effect on the returns in the Chinese stock market. One possible explanation may be found in the market structure of the Chinese stock market. According to Yao and Yueh (2009), the financial markets in China were still underdeveloped compared to other global markets. This underdevelopment may cause the Chinese stock exchange to be less efficient in reacting to real economic implications (Wang, 2010). As a result, the relationship between the

75

Journal of Advanced Studies in Finance domestic (China) and foreign market (US) may be disrupted by the dominance of the US exchange market due to its superior market structure (Alaganar and Bhar, 2000).

6.4. Impulse – response function (price volatility) The three groups of graphs below display the results of the Impulse – Response Function of the SSE, SEHK, and NYSE for four distinctive sub-periods. Figure 3b: Response of return of SSE (Period 2)

Figure 3a: Response of return of SSE (Period 1) 1.6

1.4

Return of SSE Return of SEHK Return of NYSE

1.4 1.2


1.2 1.0

1.0 0.8 0.8 0.6 0.6 0.4

0.4 0.2

0.2

0.0

0.0

-0.2

-0.2 1

2

3

4

5

1

2

3

4

5

Figure 3d::Response of return of SSE (Period 4)

Figure 3c: Response of return of SSE (Period 3) 1.4

3.0


2.5 2.0


1.2 1.0 0.8

1.5

0.6 1.0

0.4 0.5

0.2

0.0

0.0

-0.5

-0.2 1

2

3

4

5

1

2

3

4

5

Figure 3: Impulse response of return of each stock exchanges to one standard deviation innovation from SSE First of all, the results reaffirm the results we generated from the Granger Causality Test. According to the figures 3a - 3d, (denoting the response of the return of the SSE to one standard deviation innovations from the SSE, NYSE, and SEHK during four distinctive sub-periods), we can conclude that the volatility of stock returns from the NYSE and SEHK has no significant effect on the return of SSE during Period 1 and Period 2. Subsequently, the one standard deviation innovation of the NYSE had increasingly greater influences on the return of the SSE. Thus, the volatility of stock return of the NYSE plays an increasingly important role in determining the return of the SSE. This finding is in accord with the result from the Granger Causality Test, which points out that the SSE seemed to be isolated and independent before the financial crisis and it has become more integrated with the other two global financial centers due to the financial crisis in Period 3 and Period 4. In addition, from the perspective of different stock exchanges, we can conclude that the volatility of stock return of the SSE in every period has the most significant effect on its own stock return in every period. The volatility of stock return of the NYSE does not affect the return of the SSE until the Period 3, and the effect is relatively small compared to that of the volatility of the SSE itself. The volatility of the SEHK has no influence on determining the return of the SSE.

76

Volume V Issue 1(9) Summer 2014 Figure 4a: Response of return of SEHK (Period 1) 1.6

Figure 4b: Response of return of SEHK (Period 2) 1.0


1.2


0.8

0.6 0.8 0.4 0.4 0.2 0.0

0.0

-0.4

-0.2 1

2

3

4

5

1

Figure 4c: Response of return of SEHK (Period 3) 2.5

2

3

4

5

Figure 4d: Response of return of SEHK (Period 4) 1.2


2.0 1.5


1.0 0.8

1.0

0.6

0.5

0.4

0.0

0.2

-0.5

0.0 -0.2

-1.0 1

2

3

4

1

5

2

3

4

5

Figure 4: Impulse response of return of each stock exchanges to one standard deviation innovation from SEHK Compared with the SEE, the SEHK is more integrated with the other two global financial centers as shown in Figures 4a-4d. The impact from NYSE is stronger especially in more recent period. For instance, in period 3, one standard deviation increase of NYSE leads to 1.1% increase in SEHK (Figure 4c). The volatility of stock return of the NYSE plays a significant role in determining the stock return of the SEHK in every period, indicating the strong relationship between the SEHK a\and the NYSE. As shown in Figure 5a-5d, while NYSE is affected by other markets, the impact is small. Finally, since the effects of one standard deviation innovations decay after two days, we can conclude that the effect of the volatility of stock return can only induce a sharp but short shock to the endogenous variable (the stock returns). In the other words, the effects of innovations from the stock exchange market are significant in the short-run (within 2 days), but they will dramatically decline in the long-run. The impulse responses of the 11 triple-listed companies are reported in the appendix. The dynamic pattern is similar to that of exchange market.

77

Journal of Advanced Studies in Finance Fi gur e.5b: R e s p o ns e o f r e tur n o f N YS E ( P e r i o d 2 )

F i gur e 5 a : R e s p o ns e o f r e tur n o f N Y S E ( P e r i o d 1 ) .7

1.6

R e tu r n o f S S E R e tu r n o f S E H K R e tu r n o f N Y S E

1.4 1.2

R e tu r n o f S S E R e tu r n o f S E H K R e tu r n o f N Y S E

.6 .5

1.0

.4

0.8 .3 0.6 .2 0.4 .1

0.2

.0

0.0

-.1

-0.2 1

2

3

4

5

1

Figure 5c:: Response of return of NYSE (Period 3)

2

3

4

5

Figure 5d: Response of return of NYSE (Period 4) 1.2

2.0


1.6


1.0 0.8

1.2

0.6 0.8

0.4 0.4

0.2 0.0

0.0

-0.4

-0.2 1

2

3

4

5

1

2

3

4

5

Figure 5: Impulse response of return of each stock exchanges to one standard deviation innovation from NYSE 7. Limitations 1) Limited sample size We select all eleven China-based companies which are simultaneously listed on the NYSE, HKSE, and SSE. The group of these eleven companies represents the whole population of Chinese triple listed companies. However, the relationship among the three stock exchange markets cannot be totally reflected by our eleven samples, since each company has a relatively great influence on our research conclusion. The change of the sample size by one company can distort the whole results. 2) Artificial division of the time scope of events Our research artificially separates the data series into 4 distinctive sub-periods to reflect whether the markets have become more integrated throughout the time due to China's own policy efforts and the recent financial crisis in 2008 (As shown in Table 2). We select two important events from 2000 to 2013, namely the QFII program and financial crisis, based on the literature review. However, the time scope of those two events is subjectively determined according to 2 other papers. Therefore, we cannot assure that the changes between those four periods are solely determined by the events we stated. The changes could also be attributed to unexpected factors that have no relationship with the financial crisis and the QFII. 3) The disturbance during financial crisis In the third period during the financial crisis, we had to delete data on days, where trading was suspended either on the SSE, SEHK, and HYSE. This abnormal market behavior during the financial crisis had influenced the results of that period. Conclusion This paper examines Chinese triple-listed firms and the market indexes of Shanghai, New York and Hong Kong regarding their market co-integration and the theories around the issue price discovery by using several methodologies and a more completed data series set. One contribution to the discussion of price discovery is that this study separates the data series into four sub-periods so that the effects of the QFII program and the financial crisis are taken into consideration when examining the co-integration and price discovery.

78

Volume V Issue 1(9) Summer 2014 The findings in this report support the efficient market hypothesis and prior literature that the returns in the SSE, the SEHK and the NYSE are co-integrated across all examined periods. However, regarding the issue of price discovery, our results are inconsistent with the studies conducted about data series originated from other regions. Unlike their results, the home market (China) does not play the dominant role. In the case of triple-listed Chinese companies and the SSE, SEHK and S&P 500 indexes, the price adjustments occurs mainly in the US market. This outcome is consistent with the hypothesis that the market with superior market structure should play the major role in information transmission. Another interesting fact was found about the differences in results across the four sub-periods. According to the findings, China was relatively independent from the other two markets prior to the financial crisis. However, with the starting of the financial crisis, the returns in the US market have gained a significant price influence power over the returns in the Chinese market. Even the returns in the Hong Kong market have increased their impact on the returns in China. This finding suggests that the Chinese financial market, although isolated and independent in the first place, has become increasingly dependent on the global markets after the start of the financial crisis. Further research is required to examine the reasons for the increasing integration of the Chinese market. Furthermore, we can also conclude that the effect of the volatility of each stock return can only induce a short but strong shock to the stock returns. This result is consistent with the hypothesis that the price of an asset is the same regardless of its listing location in an efficient market in the long-run. In addition, we can also conclude that the volatility of stock return of the NYSE plays an increasingly important role in determining the return of the SSE. This finding points out that the Chinese market seemed to be isolated and independent before the financial crisis and it has become more integrated with the other two global financial centers after the financial crisis in 2008. We also notice that the volatility of stock return of the NYSE always plays a significant role in determining the stock return of the SEHK, indicating the strong relationship between the Hong Kong market and the New York market. The results shed more light on the relationship between the Hong Kong, China and US financial markets and have implications for both investors and policymakers. From the investor's point of view, fund managers and investors who manage the inclusion of cross-listed stocks in their portfolios should pay more attention to the US market because the information flow seems to happen mainly from the NYSE to the SEHK and the SSE. Due to its dominant role in price discovery, the NYSE would be a good pricing indicator, while stocks in the Chinese stock market, which have little co-movement with other markets, would be a good risk diversified investment. However, the value of the Chinese market being a good location for international diversification seems to diminish as the Chinese markets have become increasingly integrated with the other markets, especially with the US exchange market. From the policymaker's point of view, the QFII program did not contribute to the integration of the Chinese stock market with the other global markets, as the trading of the qualified foreign institutional investors still did not play a major role in the Chinese Stock Exchanges. It may be beneficial to strengthen the liberalization efforts of the QFII program by decreasing the barriers for foreign investors to conduct trading in the Chinese stock market. However, Chinese policymakers also should pay attention to the effects and aftereffects of the global financial crisis in 2008, since the Chinese stock market has became more vulnerable to the changes in the US market due to the crisis. However, we leave it for further research to find out why the US market has become more influential to the Chinese market in and after the financial crisis. References [1]. Alaganar, V. T., Bhar, R. (2002). Information and volatility linkage under external shocks: Evidence from dually listed Australian stocks. International Review of Financial Analysis, 11(1): 59-71. [2]. Bae, K.-H., Cha, B., Cheung, Y.-L.(1999). The Transmission of Pricing Information of Dually-Listed Stocks. Journal of Business Finance & Accounting, 26: 709–723. DOI: 10.1111/1468-5957.00272 [3]. Chen, K. C., Li, G., Wu, L. (2010). Price discovery for segmented us-listed Chinese stocks: Location or market quality?.Journal of Business Finance & Accounting, 37(1-2): 242–269. Retrieved from http://onlinelibrary.wiley.com/doi/10.1111/j.1468-5957.2009.02153.x/citedby [4]. Chen, N., Roll, R., Ross, S. (1986). Economic forces and the stock market. The Journal of Business, 59: 383-404. [5]. Choi, O. T. K., Wong, H., Yiu, C. K. F., Yu, M. (2013). In depth analysis of the dually listed companies in Hong Kong and china stock markets prior and posterior to the global financial turmoil, International Journal of

79

Journal of Advanced Studies in Finance Economics and Finance, 5(10): 100-110. Retrieved from http://search.proquest.com/ docview/1468524169?accountid=38789 [6]. Chui, A. and Kwok, C. (1998). Cross-Autocorrelation between A shares and B shares in the Chinese stock market. Journal of Financial Research, 21: 333 – 344. [7]. Climent, F., Pascual, R., Pascual-Fuster, B. (2001) Cross-Listing, Price Discovery and the Informativeness of the Trading Process. Universidad de las Islas Beleares. [8]. Coffee, J. (2002). Racing Towards the Top? The Impact of Cross-Listings and Stock Market Competition on International Corporate Governance, Columbia Law Review, 102(7). [9]. Dickey, D., Fuller, W. (1981). Likelihood ratio statistics for autoregressive time series with a unit root, Econometrica, pp. 83-106. [10]. Diebold, F.X. (2008). Unit Roots, Stochastic Trends, ARIMA Forecasting Models, and Smoothing: J. Calhoun, Elements of Forecasting, Forth Edition (pp 288-324).USA: South-Western. ISBN 13: 978-0-32432359-7. [11]. Diekmann, K. (2011). Are there spillover effects from Hong Kong and the united states to Chinese stock markets?.Empirical Economic Research, Retrieved from http://ideas.repec.org/p/iee/wpaper/wp0089.html [12]. Eun, C. S., Sabherwal, S. (2003). Cross-border listings and price discovery: Evidence from U.S.-listed Canadian stocks. The Journal of Finance, 58(2): 549-575. Retrieved from http://www.jstor.org/ stable/3094550 [13]. Fung, K.W.T and Wan, W, (2013). The impact of merger and acquisition on value at risk (VaR): A case study of China eastern airline. International Research Journal of Finance and Economics, 110: 121-127. [14]. Fung, K.W. T., Lau, C.K.M., Chan, K.H., (2014a). The Conditional CAPM, Cross-Section Returns and Stochastic Volatility, Economic Modelling. 38: 316-327. [15]. Fung, K.W.T., Demir, E., and Zhou L. (2014b). Capital Asset Pricing Model and Stochastic Volatility: A Case study of India. MPRA working paper #56180. [16]. Guo, L., Sun, Z., & Tong, Y. (2010). Why Do Chinese Companies Dual-List Their Stocks. Working paper. [17]. Hamilton, J. D. (1994). Covariance-stationary Vector Processes. Time Series Analysis (pp. 257-279). United States: Princeton University Press. ISBN: 0-691-04289-6 [18]. Haripriya, R., Thenmozhi, M., & Kumar, G. A. (2006).Impact of event in cross-listed market on the stock returns of ADR-listed companies, Journal of Management Research, 6(1): 35-47. Retrieved from http://search.proquest.com/docview/237227595?accountid=38789 [19]. Hansda, S., & Ray, P. (2003). Stock Market integration and dually listed stocks: Indian ADR and Domestic Stock Prices, Economic and Political Weekly, 38: 741-754. [20]. Hasbrouck, J. (1995). One security, many markets: Determining the contributions to price discovery. Journal of Finance. [21]. Hauser, S., Y. Tanchuma, and U. Yaari.(1998). International Transfer of Pricing Information Between Dually Listed Stocks. Journal of Financial Research, 21: 139–56. [22]. Hu, J.W., Yang, C., Huang, B., 2000, Causality and cointegration of stock markets among the United States, Japan and the South China Growth Triangle, International Review of Financial Analysis, 9: 281-297. [23]. Jian, Y., James, W.K., Insik, M. (2002). Stock market integration and financial crises: the case of Asia. Applied Financial Economics, 13(7): 477-486. Retrieved from http://www.tandfonline.com/doi/ abs/10.1080/09603100210161965 [24]. Johansen, S., Juselius, K. (1990). Maximum likelihood estimation and inference on cointegration with application to the demand for money, Oxford Bulletin of Economics and Statistics (pp. 169-209). [25]. Kadapakkam, P.R., Misra, L., Tse, Y. (2003). International price discovery for emerging market stocks: Evidence from Indian GDRs. Review of Quantitative Finance and Accounting, 21(2): 179-199. Retrieved from http://search.proquest.com/docview/210310720?accountid=38789 [26]. Koumkwa, S., Susmel, R. (2005). Arbitrage and convergence: Evidence from Mexican ADRs. Journal of Applied Economics, 11(2): 399-425. Retrieved from http://search.proquest.com/docview/ 233256547?accountid=38789

80

Volume V Issue 1(9) Summer 2014 [27]. Kozhan, R. (2010). Financial Econometrics – with EViews. Roman Kozhan & Wentus Publishing. ISBN: 97887-7681-427-4. [28]. Liu, L. (2012). Return & volatility disparity, slow adjustment process in Chinese triple-listed firms. GSTF Business Review, 2(2): 233-243. Retrieved from http://connection.ebscohost.com/c/articles/88010413/returnvolatility-disparity-slow-adjustment-process-chinese-triple-listed-firms [29]. Mak, B.C., Ngai, A.S. (2005).Market Linkage for Dual-Listed Chinese Stocks. Chinese Economy, 38(2): 88107. [30]. Melvin, M., Gramming, J., Schlag, C. (2000).Price Discovery in International Equity Trading. Arizona State University. [31]. Perron, P., Yabu, T. (2009).Testing for shifts in trend with an integrated or stationary noise component. Journal of Business & Economic Statistics, 27(3), 369–396. http://dx.doi.org/10.1198/jbes.2009.07268 [32]. Qadan, M., Yagil, J. (2012). Main or satellite? Testing causality-in-mean and variance for dually listed stocks. International Journal of Finance & Economics, 17(3): 279–289. Retrieved fromhttp://onlinelibrary. wiley.com/doi/10.1002/ijfe.458/abstract [33]. Romilly, P., Liu, X., Song, H. (1997). Are Chinese stock markets efficient? A cointegration and causality analysis, Applied Economics Letters, 4: 511-515. [34]. Sabherwal, S. (2000). Price discovery for dually traded securities: Evidence from the united states-listed Canadian stocks. (Order No. 9978423, Georgia Institute of Technology). ProQuest Dissertations and Theses, 85-85 p. Retrieved from http://search.proquest.com/docview/304592420?accountid=38789. [35]. Santos, C. (2011). The Euro sovereign debt crisis, the determinants of default probabilities and implied ratings in credit default swap market: an econometric analysis. Journal of Advanced Studies in Finance, 2(1): 53-61. [36]. Sheu, H.J., Cheng, C.L. (2011). A study of US and China’s Volatility Spillover Effects on Hong Kong and Taiwan. African Journal of Business Management, 5(13): 5232-5240. Retrieved from http://www. academicjournals.org/AJBM [37]. Spitzer, J. (2011). The Persistence of Pricing Differentials in Dual-listed Companies in Hong Kong and China. CMC Senior Theses. Paper 272, http://scholarship.claremont.edu/cmc_theses/272 [38]. Sun, L., Lin, L. (2006). QFII投资中国内地证券市场的实证分析. (The Empirical Analysis about the Investment of QFII on the Mainland Securities Market). Financial Research (pp. 123-133). Retrieved from Weipu database, http://www.cqvip.com/QK/97926X/200607/22334018.html [39]. Wang, X. (2010). The Relationship between Stock Market Volatility and Macroeconomic Volatility: Evidence from China, Int. Res. J. Finance. Econ., 49: 156-167. [40]. Wang, Z., Hsiao, C., Li, Q., Yang, J. (2006). The emerging market crisis and stock market linkages: further evidence. Journal of Applied Econometrics, 727-744. [41]. Yao, Y, Yueh, L. (2009). Law, Finance, and Economic Growth in China: An Introduction. World Dev., 37: 753-762. [42]. Wong, T., Zurbruegg, R. (1998). Pricing behavior of Asian dually listed stocks, UNSW. *** Working paper Series 1998-01, The School of Banking and Finance, The University of New. South Wales, Australia.

81


APPENDICES The result of trace statistic and Max-Eigen statistic Company Name

Period r① Trace Test Statistics

CEA Period 1

Period 2

Period 3

Period 4

ACH Period 3

Period 4

CHU Period 1

Period 2

Period 3

Period 4

GSH Period 3

Period 4

HNP Period 1

0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2

1250.683② 642.0251 301.9186 1329.226 676.6676 285.5262 595.9912 283.6277 128.2260 1940.399 965.1805 432.7962 801.0132 335.3923 142.2562 2014.589 993.7174 469.9278 295.1208 143.1899 56.20389 1687.378 854.7804 379.7177 912.7478 413.2430 193.5817 1977.948 974.4697 473.5439 843.2244 389.9506 173.8644 1937.009 1043.445 482.8089 622.8744 341.2756 144.5378 82

0.05 Max-Eigen Critical Value Statistic

0.05 Critical Value

608.6584 340.1064 301.9186 652.5581 391.1413 285.5262 312.3635 155.4016 128.2260 975.2188 532.3843 432.7962 465.6209 193.1361 142.2562 1020.872 523.7895 469.9278 151.9309 86.98604 56.20389 832.5980 475.0627 379.7177 499.5048 219.6613 193.5817 1003.478 500.9258 473.5439 453.2738 216.0862 173.8644 893.5642 560.6359 482.8089 281.5988 196.7378 144.5378

22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546

35.19275② 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546

Volume V Issue 1(9) Summer 2014 Period 2

Period 3

Period 4

LFC Period 3

Period 4

PTR Period 3

Period 4

SHI Period 1

Period 2

Period 3

Period 4

SNP Period 1

Period 2

Period 3

Period 4

0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1

1267.197 777.0491 359.2966 826.4225 352.8686 169.1307 1819.420 938.1704 415.4748 699.6052 325.6917 152.3033 1987.381 941.7948 438.1831 569.1857 245.3458 118.6052 2002.843 879.0836 423.4528 1047.511 555.9340 274.0093 1315.299 667.2051 301.8275 661.8033 289.5213 134.9617 1834.928 884.8798 403.4396 627.6224 296.2910 135.9409 1576.876 740.4039 338.4135 683.7067 318.7531 150.4488 2006.342 919.1800

83

35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184 9.164546 35.19275 20.26184

490.1481 417.7525 359.2966 473.5539 183.7379 169.1307 881.2493 522.6956 415.4748 373.9135 173.3884 152.3033 1045.586 503.6117 438.1831 323.8398 126.7406 118.6052 1123.759 455.6308 423.4528 491.5766 281.9248 274.0093 648.0939 365.3776 301.8275 372.2820 154.5596 134.9617 950.0485 481.4402 403.4396 331.3315 160.3501 135.9409 836.4716 401.9904 338.4135 364.9536 168.3043 150.4488 1087.162 493.0854

22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210 9.164546 22.29962 15.89210

Journal of Advanced Studies in Finance 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2 0 1 2

426.0946 9.164546 426.0946 9.164546 1150.321 35.19275 525.0798 22.29962 625.2416 20.26184 345.5266 15.89210 279.7150 9.164546 279.7150 9.164546 Period 2 1179.838 35.19275 461.6485 22.29962 718.1898 20.26184 415.1969 15.89210 302.9929 9.164546 302.9929 9.164546 Period 3 659.5517 35.19275 367.7091 22.29962 291.8427 20.26184 171.6921 15.89210 120.1505 9.164546 120.1505 9.164546 Period 4 1940.918 35.19275 1018.567 22.29962 922.3514 20.26184 485.9279 15.89210 436.4235 9.164546 436.4235 9.164546 ZNH Period 2 1495.357 35.19275 766.4975 22.29962 728.8594 20.26184 428.8373 15.89210 300.0221 9.164546 300.0221 9.164546 Period 3 703.1926 35.19275 318.3850 22.29962 384.8076 20.26184 226.2627 15.89210 158.5450 9.164546 158.5450 9.164546 Period 4 1929.915 35.19275 959.2427 22.29962 970.6722 20.26184 523.3116 15.89210 447.3606 9.164546 447.3606 9.164546 ① r = 0 denotes the hypothesis that there is non co-integration, r 1 denotes that the hypothesis that there is at most one co-integration, and r 2 denotes the hypothesis that there is at most two cointegrations; ② If the Trace Test Statistics or the Max-Eigen Statistic is greater than the critical value, it denotes rejection of the hypothesis at a 95% confidence interval. YZC Period 1

84

Volume V Issue 1(9) Summer 2014 Results of Impulse – Response Function 1) ACH

Period 1

Period 2

Res pons e of CHINA_RETURN to Choles ky One S.D. Innovations


5

3.0

4

2.5 2.0

3

1.5

2 1.0

1

0.5

0

0.0

-1

-0.5

1

2

3

4

5

6

7

8

9

10

1

2

3

CHINA_RET URN HONGKONG_RET URN US_RET URN

4

5

6

7

8

9

10


Res pons e of HONGKONG_RETURN to Choles ky One S.D. Innovations


5

2.4

4

2.0 1.6

3

1.2

2 0.8

1

0.4

0

0.0

-1

-0.4

1

2

3

4

5

6

7

8

9

10

1

2

3


4

5

6

7

8

9

10


Res pons e of US_RETURN to Choles ky One S.D. Innovations


5

2.5

4

2.0

3

1.5

2

1.0

1

0.5

0

0.0

-1

-0.5

1

2

3

4

5

6

7

8

9

10

1


2

3

4

5

6

7

8


85

9

10

Journal of Advanced Studies in Finance 2) CEA Period 1

Period 2



2.5

2.5

2.0

2.0

1.5

1.5

1.0

1.0

0.5

0.5

0.0

0.0 -0.5

-0.5 1

2

3

4

5

6

7

8

9

1

10

2

3

4

5

6

7

8

9

10

CHINA_RET URN HONGKONG_RETURN US_RET URN


Res pons e of HONGKONG_RETURN to Choles ky One S.D. Innovations 4

Res pons e of HONGKONG_RETURN to Choles ky One S.D. Innovations 3.0 2.5

3

2.0

2

1.5

1

1.0 0.5

0 0.0

-1

-0.5

1

2

3

4

5

6

7

8

9

10

1

2

3


4

5

6

7

8

9

10




2.5

2.5

2.0

2.0

1.5

1.5

1.0

1.0

0.5

0.5

0.0

0.0 -0.5

-0.5 1

2

3

4

5

6

7

8

9

1

10

2

3

4

5

6

7

8



86

9

10


Period 4 Response of CHINA_RETURN to Cholesky One S.D. Innovations

Res pons e of CHINA_RETURN to Choles ky One S.D. Innovations 5

2.5

4

2.0

3

1.5

2

1.0

1

0.5

0

0.0 -0.5

-1 1

2

3

4

5

6

7

8

9

1

10

2

3

4

5

6

7

8

9

10



Response of HONGKONG_RETURN to Cholesky One S.D. Innovations


3.0

4

2.5 2.0

3

1.5 2

1.0 1

0.5

0

0.0

-1

-0.5 1

2

3

4

5

6

7

8

9

1

10

2

3


4

5

6

7

8

9

10



Response of US_RETURN to Cholesky One S.D. Innovations

5

2.5

4

2.0

3

1.5

2

1.0

1

0.5

0

0.0

-1

-0.5 1

2

3

4

5

6

7

8

9

10

1


2

3

4

5

6

7

8


87

9

10

Journal of Advanced Studies in Finance 3) CHU Period 1

Period 2


Response of CHINA_RETURN to Cholesky One S.D. Innovations

2.0

2.5 2.0

1.5

1.5 1.0 1.0 0.5 0.5 0.0

0.0

-0.5

-0.5 1

2

3

4

5

6

7

8

9

1

10

2

3


4

5

6

7

8

9

10




3.0

2.5

2.5

2.0

2.0

1.5

1.5 1.0 1.0 0.5

0.5

0.0

0.0 -0.5

-0.5 1

2

3

4

5

6

7

8

9

1

10

2

3


4

5

6

7

8

9

10




2.5

2.0

2.0 1.5 1.5 1.0

1.0

0.5

0.5

0.0 0.0 -0.5 -1.0

-0.5 1

2

3

4

5

6

7

8

9

1

10


2

3

4

5

6

7

8


88

9

10


Period 4 Res pons e of CHINA_RETURN to Choles ky One S.D. Innovations

Res pons e of CHINA_RETURN to Choles ky One S.D. Innovations 4

2.0

3

1.5

2

1.0

1

0.5

0

0.0 -0.5

-1 1

2

3

4

5

6

7

8

9

1

10

2

3

4

5

6

7

8

9

10





2.0

3

1.5

2

1.0

1

0.5

0

0.0

-1

-0.5 1

2

3

4

5

6

7

8

9

1

10

2

3


4

5

6

7

8

9

10




4

2.0

3

1.5

2

1.0

1

0.5

0

0.0

-1

-0.5 1

2

3

4

5

6

7

8

9

10

1


2

3

4

5

6

7

8


89

9

10

Journal of Advanced Studies in Finance 4) GSH Period 1

Period 2



4

2.0

3

1.5

2

1.0

1

0.5

0

0.0

-1

-0.5

1

2

3

4

5

6

7

8

9

10

1

2

3


4

5

6

7

8

9

10



Res pons e of HONGKONG_RETURN to Choles ky One S.D. Innovations 2.0 1.5

2

1.0

1 0.5

0 0.0

-1

-0.5

1

2

3

4

5

6

7

8

9

10

1

2

3


4

5

6

7

8

9

10




3

1.6 1.2

2

0.8

1 0.4

0

0.0

-1

-0.4

1

2

3

4

5

6

7

8

9

10

1


2

3

4

5

6

7

8


90

9

10

\

Volume V Issue 1(9) Summer 2014 5) HNP Period 1

Period 2 Res pons e of CHINA_RETURN to Choles ky One S.D. Innovations

Res pons e of CHINA_RETURN to Choles ky One S.D. Innovations 2.5

2.0

2.0

1.5

1.5

1.0 1.0

0.5 0.5

0.0

0.0 -0.5

-0.5 1

2

3

4

5

6

7

8

9

1

10

2

3

4

5

6

7

8

9

10



Res pons e of HONGKONG_RETURN to Choles ky One S.D. Innovations 2.5


2.0 2

1.5 1.0

1

0.5 0

0.0 -0.5

-1

1

2

3

4

5

6

7

8

9

10

1

2

3


4

5

6

7

8

9

10




2.0

2.0

1.5

1.5

1.0

1.0

0.5

0.5

0.0

0.0

-0.5

-0.5

1

2

3

4

5

6

7

8

9

1

10


2

3

4

5

6

7

8


91

9

10

Journal of Advanced Studies in Finance Period 3

Period 4



4

2.5 2.0

3

1.5

2 1.0

1 0.5

0

0.0 -0.5

-1 1

2

3

4

5

6

7

8

9

1

10

2

3

4

5

6

7

8

9

10

US_RET URN HONGKONG_RETURN CHINA_RET URN



Response of HONGKONG_RETURN to Cholesky One S.D. Innovations 2.0

3

1.5

2

1.0

1

0.5

0

0.0

-1 1

2

3

4

5

6

7

8

9

10

-0.5 1

2

3


4

5

6

7

8

9

10




4

2.0

3

1.5

2

1.0

1

0.5

0

0.0

-1

-0.5

1

2

3

4

5

6

7

8

9

10

1


2

3

4

5

6

7

8


92

9

10

Volume V Issue 1(9) Summer 2014 6) LFC Period 1

Period 2



4

2.0

3

1.5

2

1.0

1

0.5

0

0.0

-1

-0.5

1

2

3

4

5

6

7

8

9

10

1

2

3

4


5

6

7

8

9

10



Res pons e of HONGKONG_RETURN to Choles ky One S.D. Innovations 1.6

3

1.2

2 0.8

1 0.4

0

0.0 -0.4

-1 1

2

3

4

5

6

7

8

9

1

10

2

3

4

5

6

7

8

9

10





4

1.6

3

1.2

2

0.8

1

0.4

0

0.0

-1

-0.4

1

2

3

4

5

6

7

8

9

10

1


2

3

4

5

6

7

8


93

9

10

Journal of Advanced Studies in Finance 7) PTR Period 1


Response of CHINA_RETURN to Cholesky One S.D. Innovations 2.0

4

1.6

3

1.2 2 0.8 1 0.4 0

0.0 -0.4

-1 1

2

3

4

5

6

7

8

9

1

10

2

3

4

5

6

7

8

9

10

CHINA_RETURN HONGKONG_RETURN US_RETURN




4

2.0

3

1.5

2

1.0

1

0.5

0

0.0

-1

-0.5 1

2

3

4

5

6

7

8

9

1

10

2

3


4

5

6

7

8

9

10




4

1.6

3

1.2

2

0.8

1

0.4

0

0.0

-1

-0.4 1

2

3

4

5

6

7

8

9

10

1


2

3

4

5

6

7

8


94

9

10

Volume V Issue 1(9) Summer 2014 8) SHI Period 1

Period 2



2.5

2.5

2.0

2.0

1.5

1.5

1.0

1.0

0.5

0.5

0.0

0.0

-0.5

-0.5

1

2

3

4

5

6

7

8

9

10

1

2

3


4

5

6

7

8

9

10


Response of HONGKONG_RETURN to Cholesky One S.D. Innovations 4


3 2

2 1

1 0

0 -1

-1

1

2

3

4

5

6

7

8

9

10

1

2

3


4

5

6

7

8

9

10




3

2.5 2.0

2 1.5

1

1.0 0.5

0 0.0

-1

-0.5

1

2

3

4

5

6

7

8

9

1

10


2

3

4

5

6

7

8


95

9

10




4

2.5 3 2.0 2

1.5

1

1.0 0.5

0

0.0 -0.5

-1 1

2

3

4

5

6

7

8

9

1

10

2

3

4

5

6

7

8

9

10





4

2.5 3 2.0 2

1.5

1

1.0 0.5

0 0.0 -1

-0.5 1

2

3

4

5

6

7

8

9

1

10

2

3


4

5

6

7

8

9

10




4

2.5 2.0

3

1.5 2 1.0 1 0.5 0

0.0

-1

-0.5 1

2

3

4

5

6

7

8

9

10

1


2

3

4

5

6

7

8


96

9

10

Volume V Issue 1(9) Summer 2014 9) SNP Period

Period 2



2.0

2.5

1.5

2.0 1.5

1.0

1.0

0.5 0.5

0.0

0.0

-0.5

-0.5

1

2

3

4

5

6

7

8

9

10

1

2

3


4

5

6

7

8

9

10




2.0 2

1.5 1.0

1

0.5 0

0.0 -0.5

-1

1

2

3

4

5

6

7

8

9

10

1

2

3


4

5

6

7

8

9

10




2.0

2.0

1.5

1.5

1.0 1.0

0.5 0.5

0.0 0.0

-0.5 -1.0

-0.5

1

2

3

4

5

6

7

8

9

1

10


2

3

4

5

6

7

8


97

9

10




4

1.6

3

1.2

2 0.8

1 0.4

0

0.0 -0.4

-1 1

2

3

4

5

6

7

8

9

1

10

2

3

4

5

6

7

8

9

10





2.0

3

1.5

2

1.0

1

0.5

0

0.0 -0.5

-1 1

2

3

4

5

6

7

8

9

1

10

2

3

4

5

6

7

8

9

10




Response of US_RETURN to Cholesky One S.D. Innovations 4

1.6

3

1.2

2

0.8

1

0.4

0

0.0 -0.4

-1 1

2

3

4

5

6

7

8

9

1

10

2

3

4

5

6

7

8



98

9

10

Volume V Issue 1(9) Summer 2014 10) YZC Period 1

Period 2



2.5

2.5

2.0

2.0

1.5

1.5

1.0

1.0

0.5

0.5

0.0

0.0

-0.5

-0.5

1

2

3

4

5

6

7

8

9

10

1

2

3


4

5

6

7

8

9

10




4

4

3

3

2

2

1

1

0

0

-1

-1

1

2

3

4

5

6

7

8

9

10

1

2

3


4

5

6

7

8

9

10




3

2.5 2.0

2 1.5

1

1.0 0.5

0 0.0

-1

-0.5

1

2

3

4

5

6

7

8

9

1

10


2

3

4

5

6

7

8


99

9

10



Response of CHINA_RETURN to Cholesky One S.D. Innovations 3

5 4

2 3 1

2 1

0 0 -1

-1 1

2

3

4

5

6

7

8

9

1

10

2

3

4

5

6

7

8

9

10





5

2.4

4

2.0 1.6

3

1.2 2 0.8 1

0.4

0

0.0

-1

-0.4 1

2

3

4

5

6

7

8

9

1

10

2

3


4

5

6

7

8

9

10




4

2.5 2.0

3

1.5 2 1.0 1 0.5 0

0.0

-1

-0.5 1

2

3

4

5

6

7

8

9

10

1


2

3

4

5

6

7

8


100

9

10

Volume V Issue 1(9) Summer 2014 11) ZNH Period 2

Period 3



2.5

5

2.0

4

1.5

3

1.0

2

0.5

1

0.0

0

-0.5

-1 1

2

3

4

5

6

7

8

9

1

10

2

3


4

5

6

7

8

9

10




2.5 4

2.0 1.5

2 1.0 0.5

0

0.0 -2

-0.5 1

2

3

4

5

6

7

8

9

1

10

2

3

4

5

6

7

8

9

10




Response of US_RETURN to Cholesky One S.D. Innovations 2.5

5

2.0

4

1.5

3

1.0

2

0.5

1

0.0

0 -1

-0.5 1

2

3

4

5

6

7

8

9

1

10

2

3

4

5

6

7

8



101

9

10

Journal of Advanced Studies in Finance 1) Markets Impulse Responses Period 4 Response of CHINA_RETURN to Cholesky One S.D. Innovations 2.5 2.0 1.5 1.0 0.5 0.0 -0.5 1

2

3

4

5

6

7

8

9

10



2

1

0

-1 1

2

3

4

5

6

7

8

9

10


Response of US_RETURN to Cholesky One S.D. Innovations 2.5 2.0 1.5 1.0 0.5 0.0 -0.5 1

2

3

4

5

6

7

8

9

10


102



THE CARRY TRADE ON THE EURO AND THE EUROPEAN STOCK MARKET Fabio PARLAPIANO

Tasmanian School of Business and Economics, Australia [email protected] Sapienza University, Management Department, Italy [email protected] Suggested Citation: Parlapiano F. (2014). The carry trade on the Euro and the European stock market, Journal of Advanced Studies in Finance, (Volume V, Summer), 1(9):103-114. Doi:10.14505/jasf.v5.1(9).07. Available from: http://www.asers.eu/journals/ jasf/curent-issue. Article’s History: Received July, 2014; Revised July, 2014; Accepted August, 2014. 2014. ASERS Publishing. All rights reserved.

Abstract: We analyze the influence of carry trades involving the Euro on the Eurozone stock market, modeling returns as dependent on macroeconomic risk factors and speculative positions. The latter are measured by the net open futures positions recorded on the CME. While correlated with the standard carry-to-risk and forward discount proxies, net open positions have the considerable advantage of being directly observable. The results show that long future positions on the Euro rise contemporaneously with the positive returns in the stock market, supporting the hypothesis that speculative capital flows may inflate or stress the market price of financial assets. Keywords: carry trade, speculative net open future positions on the Euro, Eurozone stock market, carry-to-risk ratio. JEL Classification: F3, F32, G15 1. Introduction18 The appreciation of high yielding currencies and depreciation of low yielding currencies represents a puzzling phenomenon. According to uncovered interest rate parity (UIP), the interest rate relates to the expected inflation rate within a currency area; high interest rate currencies are expected to depreciate against low interest rate currencies which are more valuable. Therefore, investing capital across countries with different interest rates should not provide risk-free profits since the gain from interest rate differentials is offset by the changes in exchange rates. However, this well-known forward premium puzzle (Fama, 1984) makes it potentially profitable to trade currencies against the arbitrage-free condition of financial markets (La Marca, 2007, 2008; Cavallo, 2006). Investors that engage in long positions in high yield currencies and a simultaneous short position in low yield currencies obtain a pay-off that, on average, exceeds the interest rate differential. The carry trade is a financial operation based on this trading strategy. This paper investigates whether carry trades involving the Euro relate to the European stock market. We test the hypothesis that capital flows driven by speculative expectations on the Euro may partially inflate or stress the price of European stocks, particularly when large positions are built (sold). A majority of recent studies in this 18

This work was made possible by the facilities of the University of Cambridge Judge Business School - Centre for Financial Analysis & Policy. I would like to thank Prof. Mardi Dungey for many helpful comments and suggestions, Dr. Vitali Alexeev, Prof. Giovanni Palomba, and seminar participants at UTAS School of Economics and Finance.

103

Journal of Advanced Studies in Finance area analyse the returns of mimicking carry trade portfolios and compare these returns with benchmark investment strategies, such as stock, corporate and government bond portfolios. Carry trade displays prominent features, returns are positive and less volatile than other investment strategies; according to Burnside et al. (2011) and Das et al. (2013) carry trade returns exceeds stock market returns after accounting for different levels of risk. Another strand of studies explore risk factors of carry trades. Corcoran (2009) argues that changes in stock market volatility may lead to portfolio re-balancing among countries and this international flow of funds can contribute to adverse exchange rate fluctuations that hit carry trade returns. Lustig et al. (2011) argue that carry trade returns are exposed to changes in global equity market volatility. These studies assume that carry trade portfolios do not exert influence on equity market evaluations. We stretch this assumption and consider the scenario where the transfer of large amounts of currencies related to carry trades may temporarily influence stock market prices through an excess demand or supply for financial assets. International financial flows driven by factors such as the so called “hunt for yield” can lead to a mispricing of financial assets, and a potential impact on the real economy (Forster et al., 2011). In this regard, the Yen carry trade is a paradigmatic case (Hattori and Shin, 2007, 2009). Capital in-flows in the US sustained the credit expansion and the associated bubble of house prices until 2007. Carry traders unwound their positions in USD when the real estate bubble burst and the volatility of US market increased. As a result, the US stock market declined even further and the Yen appreciated suddenly. There is also evidence that the Euro was involved in carry trade activity. The role of Euro as a funding (investing) currency can be inferred from the carry-to-risk ratio (CTR), an indicator frequently used to estimate the ex-ante risk adjusted return of carry trades. Figure 1 shows the weekly CTR for 8 currencies against the Euro for the period 1999 to 2011, suggesting that the Euro may have played a role as a stable target currency for carry trades financed in JPY and CHF; in fact the CTR is always positive during the time period analysed. On the other hand, the Euro may have financed carry trades against the AUD, GBP and BRL given the negative trend of the CTR. The Euro might have had an alternating financing and investment currency role against the CAD, USD and SEK. EUR/BRL

EUR/AUD

0.2

60

0

30

0

50 -0.1

25

-0.2

40 -0.2

20

-0.4

30 -0.3

15

-0.6

20 -0.4

10

-0.8

10 -0.5

5

-1 Jan-99

Oct-01

Jun-04

Mar-07

Dec-09

0 -0.6 Jan-99

0 Oct-01

EUR/CAD

Mar-07

Dec-09

EUR/CHF

0.2

25 0.9

0.15

0.8

25

20 0.7

0.1

0.05

15

0

10

-0.05 -0.1

20

0.6

15

0.5 0.4

10

0.3

5 0.2 0.1

-0.15 -0.2 Jan-99

Jun-04

0 Oct-01

Jun-04

Mar-07

Dec-09

0 Jan-99

104

5 0 Oct-01

Jun-04

Mar-07

Dec-09

Volume V Issue 1(9) Summer 2014 EUR/GBP

EUR/JPY

0.2

25 0.7

40

0.1

0.6

35

20

0

30

0.5

-0.1

25

15 0.4

-0.2

20

-0.3

10 0.3

-0.4

0.2 5

-0.5 -0.6 Jan-99

0 Oct-01

Jun-04

Mar-07

Dec-09

15 10

0.1

5

0 Jan-99

0 Oct-01

EUR/SEK 0.25 0.2 0.15 0.1 0.05 0 -0.05 -0.1 -0.15 -0.2 -0.25 Jan-99

Jun-04

Mar-07

Dec-09

EUR/USD 20 0.3

30

0.2

25

15

0.1 10

20

0

15

5 -0.1

10

-0.2 0

Oct-01

Jun-04

Mar-07

Dec-09

5

-0.3

-5 -0.4 Jan-99

0 Oct-01

Jun-04

Mar-07

Dec-09

Source: Author’s calculation on Bloomberg data.

Figure 1. The attractiveness of carry trade involving the Euro. In Figure 1 the carry to risk (first axis) and the option-implied volatility (second axis) series are computed on a weekly basis assuming the Euro as the target currency. The series on AUD, CAD, and BRL are not complete due to the unavailability of data on option-implied volatility. We estimate returns of three carry trade mimicking portfolios: a comprehensive equally weighted portfolio (EWP), a portfolio composed by high absolute interest rate spread currencies (HIRS) and a portfolio of low absolute interest rate spread currencies (LIRS), where the Euro is the reference currency; Table 1 reports descriptive statistics and Figure 1A in Appendix 1 visualize the cumulative returns associated with these portfolios. Table 1. Carry trades portfolios involving the Euro.

Portfolios

Average

EWP HIRS LIRS Euro Stoxx TMI

0.03% 0.03% 0.05% 0.04%

St. Deviation Reward to risk Sharpe ratio Panel A. Full sample period 1999 - 2011 0.64% 4.18% 0.003 1.16% 2.41% 0.003 0.74% 6.29% 0.029 2.77% 1.26% 0.004

Skew

Kurt

-0.745 -0.668 -0.345 -0.325

2.277 3.012 1.597 3.178

Note: Constituents of HIRS are JPY, AUD and BRL and constituents of LIRS portfolio are CAD, SEK, and USD against the Euro. Source: Authors’ calculations based on Bloomberg data. In each case we observe that the returns of these portfolios outperform the Eurozone stock market both in terms of cumulative returns (see Figure 1A) and reward to risk. Large and volatile cross-border financial flows are associated with price volatility and bubbles, unexpected exchange rate fluctuations, credit expansion and unsustainable drops in risk premia, (Blanchard et al., 2012). In this paper we explore the potential contribution of carry trades to changes in stock market prices by providing evidence of joint movements of speculative positions on the Euro and stock market returns. The excess (or lack) of liquidity in the target currency area provides the channel of influence on stock prices. We model the excess market return as a linear function of both macroeconomic fundamentals and speculative activity on the foreign exchange market. We show that speculative future positions on the Euro track carry trade activity, hence this variable is employed as a proxy for currency speculative activity. While correlated with the standard carry-to-risk and forward discount proxies, net open positions have the considerable advantage of being directly observable. Our main finding is evidence of a pro-cyclical dynamic of carry trade activity for the Eurozone stock market. Long positions on the Euro are found to simultaneously rise with positive returns on the

105

Journal of Advanced Studies in Finance stock market, corroborating the hypothesis that speculative capital flows may affect the market price of financial assets. The remainder of this paper proceeds as follows: section 2 sets up the framework for modeling the potential impact of currency speculation on the stock market. Data are described in section 3 and findings are discussed in section 4. Section 5 concludes. 2. Modelling framework We investigate the relationship between capital flows associated with carry trades and stock market returns with a multiple risk factors model. The model takes a form where market portfolio excess return, ( ), and the speculative ) is modelled as a function of the expected macroeconomic risk factors, ( activity on the FX market, , as in (1): (

)

∑

(

)

(1)

where: , is the expected excess return of zero beta assets, not exposed to systematic risks, , is the risk exposure of market portfolio to each macroeconomic source of systematic risk, , is the sensitivity of market return to speculative activity on the FX market and , is a serially uncorrelated error term. The choice of macroeconomic variables affecting the stock market relies on the economic forces identified in previous studies: industrial production, term premium, default spread, inflation rate, oil price, exchange rate and trade balance. We expect that increasing speculative activity is related to a general increase in the returns recorded by the stock market. We approximate the level of carry trade activity by speculative net open future positions on the Euro (SNPEur); following previous studies of Klitgaard and Weir (2004), Galati et al. (2007), Nishigaki (2007) and Anzuini and Fornari (2012). Net open future positions of speculators are defined as long minus short future ) are stochastic contracts in foreign currency; this proxy is discussed in section 3. The risk factors, ( variables non-observable at the time market prices are settled. The expectations of systematic risk factors during ) are conditioned by the informative set available at time t, and we follow the relevant evaluation horizon ( convention that these unobservable expectations are provided by realized values ; Muth (1961) and Forsythe et al. (1982). However, this study is not conclusive on the causality between SNPEur and excess market return. Speculative positions may rise (long positions on the Euro) when simultaneous excess returns are recorded on Eurozone stock market. If so, an endogeneity issue in modelling this relationship may arise, since, independent and dependent variable are simultaneously determined19. Macroeconomic state variables may partially reflect non-systematic risks that should not have effects on asset prices. In fact, Ferson and Harvey (1991) observe that the state variables usually employed to test multiple risk factor models, may capture not only the relevant economic risks, but may instead jointly proxy for a set of latent variables that determine security returns. For example, by the diversification argument, an investor may not be exposed to the overall variation of industrial production. We estimate model (1) accounting for this observation by filtering specific risks from the set of macroeconomic variables. Common components that capture the underlying structure of macroeconomic data are extracted and employed as risk factors in a principal component regression model20.

3. Data 3.1. Macroeconomic fundamentals Macroeconomic variables employed as risk factors in (1) are identified according to Chen et al. (1986), Panetta (2002) and Angelini et al. (2008). These economic forces are either related to the stream of a firm’s future cash flows or its discount rate. In particular, nominal levels of cash flows are related to the inflation rate (I t), and one of its main determinant the oil price (OILt); while real levels are affected by the economic activity, as approximated by the growth rate of industrial production (IPt), the competitive potential on international markets as recorded by the real effective exchange rate index return (rREERt), and the trade balance (TBt). Computational Formal tests reject the possibility of biased OLS estimation in (1). In particular insignificant correlation is found between SNPEur and the error term in (1) and, the null hypothesis of no Granger causality between SNP Eur and the excess market return cannot be rejected. 20 As a sensitivity analysis, equation (1) is estimated using both extracted principal components and the original macroeconomic data series. The first approach obtains better results in terms of variance explained (R 2). 19

106

Volume V Issue 1(9) Summer 2014 details for each of the variables employed in the empirical analysis are summarized in Table 2. TP and TB are expressed in first differences to obtain stationary variables. Monthly data on macroeconomic variables for Eurozone countries are obtained from DataStream and the European Central Bank Data Warehouse. Table 2. Macroeconomic variables. Symbol

Variable and calculation

Source

OIL

Inflation ∆% Eurozone harmonized consumer price index Oil Price ∆%West Texas Intermediate index

dTP

Term premium ∆ (Euro area 10 years – 3 months government benchmark bond yield) ECB

DS

Default spread Iboxx Eur corp all maturities - 10 years Gov EMU

DataStream

IP

Industrial production ∆% Eurozone industrial production

ECB

rREER

Real effective exchange rate ∆% (Eurozone real effective exchange rate index)

ECB

dTB

Trade balance ∆[ln(exports)-ln(imports)]

ECB

I

ECB DataStream

Source: Authors’ calculations based on Bloomberg data. Macroeconomic variables are generally weakly correlated; see Table 2A in Appendix 2. Higher values are reasonably recorded between oil price changes, one of the main drivers of price dynamics in Europe, and inflation (0.59) and between the term premium and the risk premium (0.24) because of the use of the 10-year government bond yields in both series; for the full sample period the above cases are far from perfect collinearity. We also compute autocorrelations of the macroeconomic series at different lags (table available from the author upon request) and industrial production shows high serial dependence until the third lag, consistent with Chen et al. (1986) and Panetta (2002). This feature is common with OIL, changes in the term premium and default spread series while for the other variables the auto-correlation drops after the first lag. 3.2 Speculative positions on the Euro Open positions in major currencies held against the dollar are published by the US Commodity Futures Trading Commission (CFTC) in its weekly commitments of traders’ report. This report tracks positions of speculators, identified as “non-commercial” positions, on the Chicago Mercantile Exchange. However, the identification of non-commercial traders is based on what investors tend to do, and this classification does not change when a speculative transaction takes place by a commercial investor. We perform two groups of tests on SNPEur to assess whether its behaviour is consistent with the profitability and risks underlying carry trades. Firstly, we consider the extent of a significant relationship between SNPEur, the CTR and the forward discount (FWD). Both the CTR and FWD are assumed to approximate the attractiveness of carry trade operations. The forward discount (or premium) is defined as the difference between forward rate, , and spot rate, , as follows: , where and are expressed as units of foreign currency per Euro. Speculators hold deliberate positions against the UIP expectations, selling forward (short positions) currencies that are expected to appreciate (forward discount), which are typically low interest rate currencies. On the contrary, speculators buy forward (long positions) currencies that are expected to depreciate (forward premium). Hence we expect that positive SNPEur (the prevalence of long positions on the Euro) are negatively (positively) related to currencies quoted at forward discount (premium) at time t. Table 3 reports the association between SNPEur, the CTR and the FWD.

107

Journal of Advanced Studies in Finance Table 3. Speculative net positions on the Euro and ex ante profitability of carry trades. Pearson ρ (SNP; CTR) t statistic Pearson ρ (SNP; FWP) t statistic

JPY

CHF

AUD

SEK

CAD

USD

GBP

0.54

0.39

0.03

0.31

0.16

-0.18

-0.32

16.71***

10.76***

0.75

8.29***

4.06***

-4.74***

-8.77***

-0.13

-0.13

0.02

-0.07

0.00

0.02

0.02

-3.33***

-3.47***

0.43

-1.78*

-0.1

0.59

0.40

Note: *, **, *** denote significance at 10%, 5% and 1% respectively. Source: Authors’ calculations based on Bloomberg data Positive values of CTR are consistently associated with the prevalence of long positions on the Euro. The correlation coefficients are positive for the JPY (0.54) and CHF (0.39), the currencies that potentially financed carry trades toward the Eurozone and negative for the GBP (-0.32), a potential target currency of carry trades financed in Euro. The interpretation of the correlation coefficients for the SEK, CAD and USD is complicated for the overall sample period, since the Euro may have alternated between financing and investment roles (as Figure 1 shows). The evidence from the SNPEur and the FWD suggests that speculators had short positions on the JPY and CHF when these currencies were quoted at a forward discount, consistent with our expectations. Secondly, we consider whether carry trade risk factors identified in previous studies impact on the level of speculator positions in the future currency market. In particular, we test whether increasing levels of volatility on the stock market and the FX market are associated with the closure of future open positions on the Euro. Table 4 shows the correlations between the absolute value of the SNPEur and the carry trade risk factors: the Eurozone stock market portfolio returns as approximated by the Euro Stoxx TMI Index, rStoxxTMI , and volatility, σStoxxTMI, the global stock market volatility as approximated by the VIX, the realized volatility of the portfolio of currencies considered, σFX, and innovations in FX market volatility, ∆σFX, computed according to Menkhoff et al. (2012). Table 4. Carry trade risk factors and their correlations with speculative positions. σStoxx rStoxxTMI VIX σFX ∆σFX TMI

Pearson ρ

0.41

-0.09

-0.14

-0.25

-0.29

t statistic

11.59***

-2.37**

-3.59***

-6.58***

-7.82***

Note: *, **, *** denote significance at 10%, 5% and 1% respectively. Source: Authors’ calculations based on Bloomberg data

According to Brunnermeier et al. (2008), Christiansen et al. (2011), and Menkhoff et al. (2012) increasing levels of uncertainty in the stock market and the FX market, as captured by the VIX, realized and innovations in FX volatility have led to a decreasing level of speculative activity on the Euro. These findings provide us with support for the use of SNPEur as proxy of carry trade activity on the Euro. However, it should be borne in mind that this is a conservative measure of the actual level of speculative activity; in fact, not all of the speculative positions are held through future contracts. The model potentially suffers from endogeneity if speculative positions may rise when simultaneous excess returns are recorded on the Eurozone stock market. However, we pre-tested such that the null hypothesis of no Granger causality between SNPEur and the excess market return cannot be rejected, providing support for our specification.

4. Results We analyze the impact of carry trades on the Eurozone stock market via estimation of model in (1) by a two-stage procedure. First, factors are extracted from the Eurozone macroeconomic fundamentals using principal components. The extracted factors and the speculative positions on the Euro are employed in model (1) for the Eurozone stock market returns. The number of principal components is determined according to the Kaiser criterion (Kaiser, 1961) and, all components with eigenvalues under unity are removed from the analysis leaving us with four components accounting for 75.77% of the overall variance of the original data set; Table 5 reports component loadings.

108


I OIL dTP DS IP rREER dTB

Table 5. Component loadings. Components F1 F2 F3 0.872 0.067 0.015 0.817 0.099 0.241 -0.003 0.719 -0.500 0.107 0.789 0.228 0.132 0.081 0.883 0.093 -0.010 -0.136 -0.503 0.363 0.171

F4 0.008 0.136 -0.236 0.133 -0.15 0.911 0.403

Source: Authors’ calculations based on Bloomberg, ECB and DataStream data.

The economic interpretation of the extracted components is straightforward. The first component (F1) shows a strong association with the inflation rate and its main driver, the oil price index variation. Hence, it can be interpreted as an inflation risk factor. The second component (F2) is highly correlated with the term premium and the default spread, and thus it can be interpreted as a discount rate risk factor. The third component (F3) captures only industrial production dynamics, therefore, it may be considered as an economic activity risk factor (or business cycle). Lastly, the forth component (F4) represents the international risk factor due to its association with the real exchange rate and the trade balance. The second step requires the estimation of the excess market return model in (1) on the lead values, , of identified macroeconomic risk factors and the currency speculation risk factor. The speculative net future position on the Euro series has been transformed into first differenced series following Klitgaard and Weir (2004) who found evidence that changes in speculative positions affect the spot exchange rate. In this analysis the choice of numbers of leads is based on the auto-correlation structure of factors. As in the original set of macroeconomic variables, significant auto-correlations are found until the third lag. That is, investors’ current evaluations of financial assets are based on the expected value of macroeconomic fundamentals for the quarter ahead. The results of the estimation of model (1) during the sample period 1999 – 2011 are shown in Table 6. Here two specifications of the model are compared, the first (model 1), which employs three macroeconomic risk factor leads, and the second (model 2), where non-significant risk factors are left out of the model. The variance explained in both models ( = 0.43 and 0.40 for model 1 and 2 respectively) is greater than reported in previous findings; Remolona (1991) found of 0.11 for Japanese market and 0.24 for UK market. The intercept term , is not significantly different from zero. Hence, there is no significant excess return for risky assets not exposed to systematic sources of risk. This evidence provides preliminarily support for the selection of explanatory variables in the model. On the macroeconomic side of the model, the signs of the coefficients for significant risk factors are aligned with the financial literature. In particular, the inflation risk factor (F1) is negatively associated with simultaneous stock market returns. Fama (1981) first documented this puzzling result and argued that the current inflation rate is related to increasing levels of uncertainty of future economic activity; see also Geske and Roll (1983) and Flannery and Protopapadakis (2002) for later evidence. Stock market prices are penalized for this uncertainty by decreasing market values of stocks. The discount risk factor (F2) has an ambiguous effect on the stock market. The simultaneous and first leads are positively related to market returns. On the other hand, the third lag is negatively related to current stock market returns reflecting the view that increasing future level of uncertainty penalizes stock market prices. Lastly, the business cycle risk factor (F3), from the first to the third leads, are positively related with the stock market value. Future levels of economic growth are consistently priced with increasing stock returns. It is surprising that despite the higher degree of openness of the Eurozone economy, the international risk factor component (F4) does not exert significant impact on stock market returns. Table 6. Eurozone excess market return model. Model 1

Variable

Coefficient

Intercept

-0.0007

Model 2

Std. Error

Coefficient

Std. Error

-0.0009

0.0036

F1

-0.0138***

0.0037

-0.0149***

0.0039

F2

0.0237***

0.0041

0.0216***

0.0039

109

Journal of Advanced Studies in Finance Model 1

Model 2

F3

0.0046

0.0044

F4

0.0017

0.0042

F1(1)

-0.0035

0.0039

F2(1)

0.0064

0.0044

0.0089**

0.0038

F3(1)

0.0072*

0.0041

0.0085**

0.004

F4(1)

0.0011

0.0042

F1(2)

0.0048

0.0041

F2(2)

0.0019

0.0043

F3(2)

0.0098**

0.0041

0.0102***

0.0038

F4(2)

-0.0053

0.0042

F1(3)

0.0011

0.0039

F2(3)

-0.0163***

0.0042

-0.0158***

0.0039

F3(3)

0.0108*

0.0041

0.0128***

0.0039

F4(3)

-0.0025

0.0043

dSNP

0.3984*

0.0038

0.4203**

0.2032

R2

0.43

0.4

Adjusted - R2

0.36

0.37

5.79***

11.64***

Akaike info criterion

-3.27

-3.34

Schwarz criterion

-2.91

-3.16

Hannan-Quinn criterion

-3.12

-3.27

Durbin-Watson statistic

2.01

2.02

F statistic

Note: Table 6 reports estimations of model in (1) on monthly data from February 1999 to June 2011. Model 1 is the complete model while in model 2 nonsignificant risk factors are left out of the model. dSNP series is scaled in thousands of Euro. *, **, *** denote significance at 10%, 5% and 1% respectively. Source: Authors’ calculations based on Bloomberg data

The results suggest significant sensitivity of Eurozone stock market returns to the simultaneous dynamics of speculative positions, even after controlling for macroeconomic fundamentals changes. In particular, during the sample period the Eurozone stock market expressed valuations in line with speculators’ expectations about the future value of the Euro. Positive returns are recorded when speculative positions on the Euro increased, and market slowdowns are associated with the prevalence of sell future positions on the Euro. The statistical significance of the estimated coefficients in model 2 is higher than model 1 by increasing from 10% to 5%. The sign of the coefficients is positive, hence positive net positions change (an increase of long positions on the Euro) are related to increasing stock market returns during the full sample period. Klitgaard and Weir (2004) first analysed the relationship of speculative position on the USD and the simultaneous bilateral changes of spot exchange rate of different currencies against the USD. They found that positions built by speculators on the future currency market, predict the correct dynamic of underlying spot exchange rate 75% of the time. The prevalence of long future positions on the USD, i.e. an optimistic perspective about the future currency value, is related to the appreciation of the USD on the spot market and vice versa when short positions prevail. The findings of this study extend the evidence regarding the effects associated with currency speculation. In particular, the analysis support the hypothesis that asset prices may be partially inflated by an excess demand attracted by investment (and financing) conditions. Concluding remarks This paper analyses the potential influence of carry trades involving the Euro on the Eurozone stock market, through a multiple risk factor model. The model consists of two facets: one relates the stock market value to expected macroeconomic fundamentals, while the other relates the stock market value to changes in the level

110

Volume V Issue 1(9) Summer 2014 of currency speculation. Four macroeconomic risk factors are identified: inflation, discount rate, business cycle and international risk factor. As speculative future positions on the Euro are shown to trace the actual level of carry trade activity this variable is employed to approximate the currency speculative activity. We investigate the potential role of the Euro as an investing or financing currency during the period 1999 – 2011. We estimate carry trade returns of portfolio strategies and, consistent with previous findings for other markets, we find a relationship with stock market performance. Carry trade activity on the Euro, as approximated by speculative future positions on the Euro, relates with the Eurozone stock returns. In particular, long positions on the Euro rise when the stock market experiences positive returns, while sell positions prevail when the stock market performs poorly. These results corroborate the hypothesis that cross-border speculative capital flows may partially inflate or stress the price of financial assets, supporting the view that the effects of speculative activity on the currency market may have important implications for other segments of financial markets. References [1]

Angelini, E., Camba-Mendez, G., Giannone, D., Runstler, G., Reichlin, L. (2008). Short-term forecasts of Euro area GDP growth. European Central Bank, Working Paper Series, 949: 1–29.

[2] Anzuini, A., Fornari, F. (2012). Macroeconomic determinants of carry trade activity. Review of International Economics 20: 468–488, http://dx.doi.org/10.1111/j.1467-9396.2012.01034.x [3] Blanchard, O., Hagan, S., Tiwari, S Siddharth, Vinals, J. (2012). The liberalization and management of capital flows: an institutional view. IMF Report. [4] Brunnermeier, Markus K., Nagel, S., Pedersen, Lasse H. (2008). Carry trades and currency crashes. NBER Working Paper 14473: 1–35. [5] Burnside, C., Eichenbaum, M., Rebelo, S. (2011). Carry trade and momentum in currency markets, Annual Review of Financial Economics, 3: 511–535, http://dx.doi.org/10.1146/annurev-financial-102710-144913 [6] Cavallo, M. (2006). Interest rates, carry trades, and exchange rate movements. FRBSF Economic Letter, 31: 1–4. [7] Chen, Nai-Fu, Roll, Richard and Ross Stephen A. 1986. Economic forces and the stock market. The Journal of Business, 59: 383–403. [8] Christiansen, C., Ranaldo, A., Söderlind, C. P. (2011). The time-varying systematic risk of carry trade strategies. Journal of Financial and Quantitative Analysis, 46: 1107–1125, http://dx.doi.org/10.1017/ S0022109011000263 [9] Corcoran, A. (2009). The determinants of carry trade risk premia. Institute for international integration studies, 287: 1–33. [10] Sougata Das, Palani-Rajan Kadapakkam and Yiuman, T. (2013). Is carry trade a viable asset class? Journal of International Financial Markets, Institutions & Money, 24: 247– 257. [11] Fama, F. E.. (1981). Stock returns, real activity, inflation, and money, The American Economic Review, 71: 545–565. [12] Fama, F. E. (1984). Forward and spot exchange rates, Journal of Monetary Economics, 14: 319–338, http://dx.doi.org/10.1016/0304-3932(84)90046-1 [13] Ferson, E. Wayne, and Harvey, Campbell R. (1991). The variation of economic risk premiums, Journal of Political Economy, 99: 385–415. [14] Flannery, Mark J., and Protopapadakis, Aris A. (2002). Macroeconomic factors do influence aggregate stock returns, Review of Financial Studies, 15: 751–782, http://dx.doi.org/10.1093/rfs/15.3.751 [15] Forster, Katrin, Vasardani, Melina, and Cà Zorzi, M. (2011). Euro area cross-border financial flows and the global financial crisis. ECB Occasional paper series, 126: 1–42. [16] Forsythe, Robert, Palfrey, Thomas R. and Plott, Charles R. (1982). Asset valuation in an experimental market. Econometrica, 50: 537–567. [17] Galati, G., Healt, A., and Guire, P. (2007). Evidence of carry trade activity. BIS Quarterly review 2007: 27–41.

111

Journal of Advanced Studies in Finance [18] Geske, R., Roll, R. (1983). The fiscal and monetary linkage between stock returns and inflation. The Journal of Finance 38: 1–33, http://dx.doi.org/10.1111/j.1540-6261.1983.tb03623.x [19] Hattori, Masazumi, and Shin, Hyun Song. (2007). The broad Yen carry trade. IMES discussion paper series 19: 1–26. [20] Hattori, Masazumi, and Shin, Hyun Song. (2009). Yen carry trade and the subprime crisis. IMF Staff Paper 56: 384–409. [21] Kaiser, Henry F. (1961). A note on guttman’s lower bound for the number of common factors, British Journal of Statistical Psychology, 14: 1–2, http://dx.doi.org/10.1111/j.2044-8317.1961.tb00061.x [22] Klitgaard, T., Weir, L. (2004). Exchange rate changes and net positions of speculators in the futures market. Economic Policy Review, 10: 17–28. [23] La Marca, M. (2007). Carry trade and financial fragility. UNCTAD Working paper series. [24] La Marca, M. (2008). Capital flow paradox, speculation and external adjustment in emerging market economies. The Berkeley Electronic Press, 24: 1–28. [25] Lustig, Hanno, Roussanov, Nikolai, and Verdelhan, A. (2011). Common Risk Factors in Currency Markets. Review of Financial Studies, 24: 3731–3777. http://dx.doi.org/10.1093/rfs/hhr068 [26] Menkhoff, L., Sarno, L., Schmeling, M., Schrimpf, A. (2012). Carry trades and global foreign exchange volatility. The Journal of Finance, 67: 681–718, http://dx.doi.org/10.1111/j.1540-6261.2012.01728.x [27] Muth, John. F. (1961). Rational expectations and the theory of price movements, Econometrica, 29: 315– 335. [28] Nishigaki, H. (2007). Relationship between the Yen carry trade and the related financial variables. Economics Bulletin, 13: 1–7. [29] Panetta, F. (2002). The stability of relation between the stock market and macroeconomic forces. Economic Notes 31: 417–450. [30] Remolona, Eli M. (1991). Do international reactions of stock and bond markets reflect macroeconomic fundamentals? FRBNY Quarterly review, 16: 1–13.

112

Volume V Issue 1(9) Summer 2014 Annex 1 80.00%

60.00%

40.00%

20.00%

0.00%

-20.00%

-40.00% Jan-99 May-00 Oct-01 Feb-03 Jun-04 Nov-05 Mar-07 Aug-08 Dec-09 May-11 EWP Cr

HIRS Cr

LIRS Cr

Stoxx TMI

Figure 1A. Carry trade portfolios involving the Euro. Note: Figure 1A reports estimates of cumulative carry trade returns of three portfolios: the equally weighted portfolio (EWP), the high interest rate spread portfolio (HIRS), and the low interest rate spread portfolio (LIRS). Returns are estimated considering the buy and hold strategy. Constituents of HIRS are JPY, AUD and BRL and constituents of LIRS portfolio are CAD, SEK, and USD against the Euro. The average absolute interest rate spread during the sample period 1999 – 2011 is calculated against the 3 months Eurozone bank deposit interest rate. The dotted line represents the weekly cumulative return of Euro Stoxx Total Market Index. Source: Author’s calculation on Bloomberg data.

113


Appendix 2 Table 2A. Correlation matrix for macroeconomic variables. Symbol I OIL dTP DS IP rREER dTB I OIL dTP DS IP rREER dTB I OIL dTP DS IP rREER dTB

I OIL dTP DS IP Panel A. Full sample period 1999 - 2011 1.000 0.597 1.000 0.015 -0.068 1.000 0.082 0.149 0.241 1.000 0.113 0.239 -0.192 0.151 1.000 0.030 0.073 -0.062 0.098 -0.120 -0.227 -0.146 0.089 0.097 -0.034 Panel. B Sub sample 1999 - 2007 1.000 0.544 1.000 0.127 -0.014 1.000 0.090 -0.048 0.481 1.000 -0.062 0.104 -0.061 -0.054 1.000 -0.037 -0.025 -0.034 -0.213 -0.111 -0.180 -0.176 0.058 0.135 0.063 Panel C. Sub sample 2007 - 2011 1.000 0.670 1.000 -0.090 -0.139 1.000 0.078 0.277 0.169 1.000 0.243 0.384 -0.318 0.213 1.000 0.094 0.205 -0.093 0.266 -0.150 -0.286 -0.097 0.131 0.103 -0.139

Source: Author’s calculation on Bloomberg data.

114

rREER

dTB

1.000 0.105

1.000

1.000 0.102

1.000

1.000 0.113

1.000


ASERS Publishing Publications  Journals…  Conferences Proceedings…  Books Collections….

115


Journals … Journal of Advanced Research in Entrepreneurship and New Venture Creation

Editor in Chief PhD Professor Domenico NICOLÒ Mediterranean University of Reggio Calabria, Italy

The Journal of Advanced Research in Entrepreneurship and New Venture Creation (JARE-NVC) seeks to promote further ideas and debates on the international, cross-cultural and comparative academic research about entrepreneurs and entrepreneurship. Manuscripts that are suitable for publication in JARE-NVC cover domains such as, but are not limited to these: Entrepreneurship and growth; Entrepreneurship support and policy implications; New firms and regional context; Academic entrepreneurship and technology transfer; International entrepreneurship; Innovation activities in SME; Entrepreneurial personalities, human capital and teams; Technology entrepreneurship; Corporate entrepreneurship; Start-ups development (or growth) strategies; Startups competitive strategies; Risk factors for start-ups; Business planning models; Family business; Entrepreneurial values and business success; Smart cities policies and start-ups; Venture capital and Private Equity (or Private equity investors); Venture creation models (or patterns); Start-ups Ecosystems. Journal of Advanced Research in Entrepreneurship and New Venture Creation, starting with its first issue, will be indexed in RePEC, CEEOL, ProQuest, and EBSCO databases. Web:

http://www.asers.eu/journals/jare_nvc

E-mail: [email protected] [email protected]

116


Journal of Advanced Research in Economics and International Business

Editor in Chief: PhD Mădălina CONSTANTINESCU Spiru Haret University, Romania And Association for Sustainable Research and Science

Education,

Co-Editor:

PhD Daniele SCHILIRÒ Messina University, Italy

Journal of Advanced Research in Economics and International Business provides a forum where academics and professionals can share the latest developments and advances in the knowledge and practice of Economics and International Business. It aims to foster the exchange of ideas on a range of important international subjects, to provide stimulus for research and the further development of international perspectives, and to publish empirical and applied research on issues relating to Economics and International Business. Journal of Advanced Research in Economics and International Business, starting with its first issue, will be indexed in RePEC, CEEOL, ProQuest, and EBSCO databases. Web: http://www.asers.eu/journals/jareib E-mail: [email protected]

117


Editor in Chief: PhD Andy Ştefănescu University of Craiova, Faculty of Economics and Business Administration, Romania Co-Editor: PhD Rajesh K. Pillania Management Development Institute, India The Journal aims to serve researchers, scholars through prompt publications of significant advances in any branch of management science, and to provide a forum for the reporting and discussion of news and issues concerning management science. Journal of Advanced Research in Management is indexed in RePEC, IndexCopernicus, ProQuest, CEEOL, CABELL'S Directories and EBSCO databases. Web: http://www.asers.eu/journals/jarm Email: [email protected]

118


Journal of Advanced Research in Law and Economics

Editor in Chief: PhD Mădălina CONSTANTINESCU Spiru Haret University, Romania and Association for Sustainable Education, Research and Science Co-Editors: PhD Russell PITTMAN International Technical Assistance Economic Analysis Group Antitrust Division, SUA PhD Eric LANGLAIS EconomiX CNRS and Université Paris Ouest-Nanterre, France

Journal of Advanced Research in Law and Economics provides readers with high quality and empirical research in law and economics. The Journal publishes analytical studies on the impact of legal interventions into economic processes by legislators, courts and regulatory agencies. Finally, important developments and topics in the analysis of law and economics will be documented and examined in special issues dedicated to that subject. The Journal is edited for readability; lawyers and economists, scholars and specialized practitioners count among its readers. Journal of Advanced Research in Law and Economics, starting with its first issue, will be indexed in SCOPUS, RePEC, EconLit, IndexCopernicus, CEEOL, ProQuest, Cabell Directories and EBSCO databases. Web: http://www.asers.eu/Journals/jarle E-mail: [email protected]

119


Journal of Advanced Research in Organizational Psychology

Editor in Chief: PhD Pompiliu CONSTANTINESCU Academy of Economic Studies, Romania and Association for Sustainable Education, Research and Science

Co-Editor: Andra Mihaela PALOȘ Association for Sustainable Education, Research and Science

The aims and scope of the Journal of Advanced Research in Organizational Psychology is to provide its readers with up-to-date studies and current trends in the field of organizational psychology. The Journal will host articles dedicated to the study of inner-group psychology and the social dynamics existing in the organization today. The contents of the Journal can be useful to students and practitioners alike, as they will provide insight to new theories, thoughts and perspectives with regards to the field of organizational psychology. Journal of Advanced Research in Organizational Psychology, starting with its first issue, will be indexed in RePEC, and CEEOL databases. Web: http://www.asers.eu/Journals/jarop E-mail: [email protected]

120



Editor in Chief: PhD. Laura ŞTEFĂNESCU Spiru Haret University, Romania and Association for Sustainable Education, Research and Science Co-Editor: PhD Rajmund MIRDALA Technical University of Kosice, Faculty of Economics, Slovak Republic

The Journal aims to publish empirical and theoretical articles which make significant contributions in all areas of finance, such as: asset pricing, corporate finance, banking and market microstructure, but also newly developing fields such as law and finance, behavioral finance, and experimental finance. The Journal serves as a focal point for communication and debates for its contributors for the better dissemination of information and knowledge on a global scale. Journal of Advanced Studies in Finance is indexed in EconLit, IndexCopernicus, RePEC, CEEOL, ProQuest, Cabell’s Directories and EBSCO databases. Web: http://www.asers.eu/Journals/jasf E-mail: [email protected]

121


Journal of Environmental Management and Tourism

Editor in Chief: PhD Ramona PÎRVU University of Craiova, Faculty of Economics and Business Administration

Journal of Environmental Management and Tourism will publish original research and seeks to cover a wide range of topics regarding environmental management and engineering, environmental management and health, environmental chemistry, environmental protection technologies (water, air, soil), at-source pollution reduction and waste minimization, energy and environment, modelling, simulation and optimization for environmental protection; environmental biotechnology, environmental education and sustainable development, environmental strategies and policies, etc. Journal of Environmental Management and Tourism is indexed in RePEC, IndexCopernicus, CEEOL, ProQuest, EBSCO and Cabell’s Directories databases. Web: http://www.asers.eu/Journals/jemt E-mail:[email protected]

122


Journal of Research in Educational Sciences

Editor in Chief: PhD Laura UNGUREANU Spiru Haret University, Romania and Association for Sustainable Education, Research and Science

Assistant Editor: Andra-Mihaela PALOŞ Association for Sustainable Education, Research and Science

The Journal is designed to promote scholarly thought in the field of education with the clearly mission to provide an interdisciplinary forum for discussions and debates about education’s most vital issues. We intend to publish papers that contribute to the expanding boundaries of knowledge in education and are focusing on research, theory, current issues and applied practice in this area. Journal of Research in Educational Sciences is indexed in RePEC, IndexCopernicus, CEEOL, ProQuest, Cabell’s Directories and EBSCO databases. Web: http://www.asers.eu/Journals/jres E-mail: [email protected]

123


Theoretical and Practical Research in Economic Fields

Editor in Chief: PhD Laura UNGUREANU Spiru Haret University, Romania and Association for Sustainable Education, Research and Science Co-Editor: PhD Ivan KITOV Russian Academy of Sciences, Russia

Theoretical and Practical Research in Economic Fields publishes original articles in all branches of economics - theoretical and empirical, abstract and applied, providing wide-ranging coverage across the subject area. The Journal promotes research that aims to unify the theoretical-quantitative and the empirical-quantitative approach to the economic problems that can be solved through constructive and rigorous thinking. Theoretical and Practical Research in Economic Fields is indexed in RePEC, EconLit, IndexCopernicus, CEEOL, ProQuest, Cabell’s Directories and EBSCO databases. Web: http://www.asers.eu/Journals/tpref E-mail: [email protected]

124


Active Calls for Book Chapters Contributions!

STARTUPS AND STARTUP ECOSYSTEMS. THEORIES, MODELS AND SUCCESSFUL CASES IN THE MEDITERRANEAN AREA A book edited by PhD Professor Domenico NICOLÒ, Mediterranean University of Reggio Calabria, Italy [email protected] To be published by ASERS Publishing in CD-ROM format with ISBN and DOI. Deadline for submission 29th October, 2014 Download Call for Book Chapters: http://www.asers.eu/asers-publishing/books.html

125


Conferences Proceedings …

Proceedings of the ASERS First on-line Conference on Global Trends in Finance Coordinator: PhD. Rajmund MIRDALA Format: 17cm x 24cm ISBN: 978-606-92386-8-4

Proceedings of the ASERS First on-line Conference on World’s Economies in and after Crisis: Challenges, Threats and Opportunities Coordinator: PhD. Laura ŞTEFĂNESCU Format: 17cm x 24cm ISBN: 978-606-92386-0-8

Proceedings of the ASERS First on-line Conference on Competitiveness and Economic Development: Challenges, Goals and Means in a Knowledge based Society Coordinator: PhD. Andy ŞTEFĂNESCU Format: 17cm x 24cm ISBN: 978-606-92386-4-6

126

ASERS


Web: www.asers.eu URL: http://www.asers.eu/asers-publishing

E-mail: [email protected] [email protected] ISSN 2068 – 8393 Journal DOI: http://dx.doi.org/10.14505/jasf Journal’s Issue DOI: http://dx.doi.org/10.14505/jasf.v5.1(9).00

127