Bank Risk and Real Estate: An Asset Pricing Perspective - CiteSeerX

Journal of Real Estate Finance and Economics, 10: 199-224 (1995) 0 1995 Kluwer Academic Publishers

Bank Risk and Real Estate: An Asset Pricing Perspective JIANPING ME1 AND ANTHONY SAUNDERS Leonard N. Stern School of Business, New York University, 900 Tisch Hall, New York, NY 10003

Abstract While a number of papers have investigated the time-series behavior of expost bank stock returns and real estate returns, no study has comprehensively studied the relationship between ex ante risk premiums on both assets and the time-varying nature of such premiums in relationship to economic and real estate market conditions. In this study, we investigate how the changing nature of bank risk taking, especially in the real estate market, has affected the ex anre pricing of risk in the market for bank stocks. We tind that the time variation in bank risk premiums are partly determined by interest rate and real estate market conditions. We also discover that the real estate factor has been important for banks in the 1980s. Key Words: real estate factors, money center banks

The 1980s have posed many loan management problems for large U.S. banks. In the early 1980s these banks faced growing problems in both their LDC and oil and gas loan portfolios. In the mid-1980s farming and agricultural loans were added to this list, and more recently, especially since 1987, commercial real estate has become a new problem loan area. By contrast, the 1970s are often viewed as a period of relative growth and prosperity for U.S. banking. This unfolding of asset-exposure problems for major U.S. banks raises an important empirical question: To what extent do investors price such risks in the market? Previous studies of bank stock returns, such as Aharony, Saunders, and Swary (1986), Flannery (1983), Flannery and James (1984)) and James (1989) have concentrated on explaining the ex post behavior of bank stock returns, usually in an augmented two-factor return-generating model with a market and interest rate risk factor. However, no study has formally sought to identify whether such factors are prices ex ante, in the sense of the APT, or whether the influence of such factors changes over time as banks alter the nature of their loan exposures. Perhaps the paper closest to ours is by Sweeny and Warga (1986) who investigated the ex ante premiums in the pricing of public utility stocks in the presence of a market and interest rate risk factors. However, their methodology was too constraining in that their premiums were fixed (constant) through time. In this paper we use data for 1970-1989 on large bank stock returns to investigate the time-varying nature of three sources of ex ante risk: (i) the market factor, (ii) an interest rate factor, and (iii) a real estate factor. In light of the problems of many large banks in the real estate market, where at the end of 1990 ten major banks had commercial real estate loan portfolios exceeding $3.5 billion each, and their problem commercial real estate loannet worth ratios exceeded 30%) considerable attention is given to the exposure to and the

JIANPING ME1 AND ANTHONY SAUNDERS

200

pricing of real estate risk in bank stock returns. Moreover, since banking is a regulated industry, we try to identify regulatory induced impacts on these risk premiums. In particular, we seek to identify whether changes in the Federal Reserve’s monetary policy regime (e.g., October 1979 to October 1982) affected the relative pricing of risk. It is important for banks to understand the determinants of equity risk premium, since this premium not only affects their investment decision but also their financing decision. As is well known, the weighted average cost of capital (WACC) is a weighted average of the costs of debt and equity. The higher the equity risk premium, the higher the required rate of return on equity, and thus, the higher the WACC. The variation of risk premium is also of interest to regulators because it contains information about market perception of bank risk. Thus, if banks have increased their exposures to certain risks, regulators should consider actions, such as additional loss provisioning, additional capital infusion, as well as revising the required deposit insurance premium paid. In Section 1 we outline the methodology of our study. In particular we employ a multifactor latent-variable model on the lines of Campbell (1987), Campbell and Hamao (199 l), and Ferson (1989) to derive time-varying ex ante (or expected) risk premiums. In Section 2 the estimation procedure is described. This is followed by a description of data in Section 3. Section 4 presents the empirical results and Section 5 is a summary and conclusion, in which we discuss our major findings: the time variation in bank risk premiums are predictably affected by a real estate market condition variable, and the real estate factor has been important in the 1980s for banks.

1. The Asset Pricing Framework Following, Campbell and Hamao (199 1)) Ferson ( 1989)) and Liu and Mei ( 1993), we assume that asset returns are generated by the following K-factor model:

;i,t+l

=

Et[ri,t+ll

+k PikC,t+l + ti,t+l k=I

(1)

where ri,r+i is the return on asset i, held from time t to time t + 1, in excess of the riskfree (treasury bill) rate. Er [ ?,r+i] is the expected excess return on asset i, conditional on information known to investors at time t. The unexpected return on asset i equals the sum of K factor realizations f;c,$+t times their betas or factor loadings Pik, plus an idiosyncratic error term ti,i+i. We assume also that E,[f&+i] = 0, E,[E?,,r+,] = 0, and E[ K) economic or forecasting variables Xpt, p = 1 . . L, and that conditional expectations are linear in those variables! Then we can write Xkt as

Xkt

=k p=l

ekpXpt,

(3)

and substituting, we have

Et[~i,t+,l

=2 bik 2

k=l

p=l

OkpXpt =i aylpxpt. p=l

Equation 4 suggests that expected excess returns are time-varying and can be predicted by the economic variables, X,,, in the information set. It is easy to see from equation 4 that the model puts some restrictions on the coefficients of equation 4. These are

where fiik and are free parameters. The main differences between this model and the model employed by Sweeny and Warga (1986) are that we allow for time-varying factor premiums as a result of a changing economic environment, and we do not assume that the factors in the model can be observed a priori. Instead, we let the data tell us whether certain factors, such as the market risk, interest rate risk, and real estate market risk are systematic factors that affect bank stock returns. This added flexibility allows us to use equation 4 to examine the degree to which economic (or “forecasting”) variables, X,, explain the ex ante time-variation in expected excess returns on bank stocks. We can also use the model to examine the extent to which various economic risks drive the expost movements in bank stock returns. As noted above, previous studies have concentrated on the ex post pricing of risk in bank stock returns, as opposed to the ex ante pricing of risk. It is worth noting that part of equation 4 can be derived directly from linear projections without using the asset pricing framework of equations l-3. In other words, given that conditional expectations are linear in the forecasting variables, we will have 8kj

Eitii,i+iI =$ oipXpt p=l

(4’)

202


This is an important observation, since it implies that the unconstrained conditional risk premium estimated using equation 4 ‘, does not depend on the assumption that beta coefficients are constant through time, or the other restrictions imposed by the asset pricing model. Thus, the sensitivities of bank stocks toward economic changes can vary over time, as long as the product of beta and factor premiums are linear in these economic variables.

2. The Estimation Procedure A generalized method of moments (GMM) approach, similar to Campbell (1987)) Campbell and Hamao (1991), and Ferson (1989) is employed to estimate equation 4 ’ and to test the asset pricing restriction of (5). Equation 5 is a cross-equation restriction with unknown parameters; thus equation 4 must be estimated simultaneously across a number of assets to appropriately test the restriction. To ensure that the linear pricing condition holds for a wide range of assets, we use returns of five asset portfolios: the value-weighted index of NYSE and AMEX stocks, a long-term U.S. government bond portfolio, an equallyweighted index of real estate investment trust (REITs), a portfolio of money center banks, and a portfolio of nonmoney center banks. In other words, a wide range of asset portfolios are needed to test the asset pricing model restriction and study the risk premiums on stocks of money center banks and nonmoney center banks. The forecasting variables chosen reflect those widely used in previous stock return studies (see Campbell, 1987; Fama and French, 1989; Keim and Stambaugh, 1986; Ferson and Harvey, 1989; Liu and Mei, 1991, among others), and which can be expected to act as important variables in our study. These variables are a January dummy, the dividend yield on an equally weighted market portfolio, the level of interest rates, the spread between the yields on long-term AAA corporate bonds and the one-month treasury bill rate, and a proxy for the earnings-price ratio on a large well-diversified portfolio of real estate assets (the capitalization rate)? Although all these variables have been found to be useful in explaining the time-variation in expected returns on regular stocks, the last three variables may have particular relevance to the expected return on bank stocks. The treasury bill rate proxies for the level of interest rates. Changes in the level of interest rates will affect banks according to whether they mismatch the duration of their asset and liability portfolios and the direction of any such mismatch. Most studies of bank portfolios have concluded that on average banks have longer-term asset durations than liability durations (see Bernanke, 1990, for example)? As a result, given any positive duration gap, an expected rise in the level of rates, increases bank exposure to interest-rate risk. Thus in periods when interest rates are higher (or lower) than normal we might expect a change in the interest-rate risk premium to be impounded in bank stock returns. In particular, we will expect a negative relationship when rates are high if the market expects interest rates to revert to some normal level! The spread between the yield on long-term AAA bonds and the treasury bill rate proxies for the default risk exposure of banks. A widening of the corporate bond-bill rate spread reflects investors’ expectations of increased default risk in the corporate sector (see Friedman and Kuttner, 1991, for example) and thus may be interpreted as reflecting increased default risk exposure on bank loans.

AN ASSET PRICING PERSPECTIVE

203

The capitalization rate (cap-rate) on real estate assets seeks to capture changing expectations regarding future expected returns in the residential and commercial real-estate markets? The capitalization rate is equal to the earnings-price ratio on direct real estate investments. An increase in the risk of real estate investment, such as a glut of new commercial properties, will increase the required rate of return on real estate and thus lower the market value of real estate assets. This will result in an increase in the cap-rate. The cap-rate is constructed as the ratio of net stabilized earnings to the market-value (market price) of a well diversified property portfolio. The stabilized earnings factor adjusts actual earnings so that each holding in the property portfolio reflects a vacancy rate no higher than that exists for other buildings in the same (local) market (see Liu and Mei, 1992). These real estate data are reported by the American Council of Life Insurance (ACLI) publications (various issues).

3. Data The forecasting (economic) variables, discussed above were derived from a number of sources. Real-estate capitalization rates were taken from the ACLI quarterly publication Investment Bulletin (various issues). Yields on one-month bills and data for the AAA corporate bond-treasury bill spread were derived from the Federal Reserve Bulletin and Ibbotson and Associates (1989). The dividend yield variable, defined as the dividend paid during the last twelve months divided by current market price, was derived by using divided and price information on the CRSP file.

3.1. Bank Stock Returns We used Compustat to identify the appropriate banks and bank-holding companies in our sample for the 1971:2-1989:4 period. The bank group was divided into a money center bank (MCBs) group and a nonmoney center bank (NMCBs) group. This was done to examine whether there was any differential sensitivity to real estate (and other factors) between the largest banks and the more regionally specialized large nonmoney center banks. The money center bank group contained a total of eleven banks6 while the nonmoney center group contained a total of 180 banks. We used all banking firms listed on Compustat over the period, not just those with a continuous trading history, so as to avoid a selection bias. As a result approximately 90 banks are contained in the nonmoney center bank sample at any given time. Based on the money center and nonmoney center classifications, two montly return series on equally-weighted bank stock portfolios are derived from the CRSP (daily) tapes.

3.2. Other Portfolio Returns The market portfolio returns and the returns on long-term U.S. bonds were also derived from the CRSP tapes. The government bond return series was employed below to reflect the degree to which bank stock returns (which are claims on underlying portfolios of fixed

204


and variable rate, riskless, and risky bonds) mimicked the return behavior of bond assets rather than stocks. The government bond portfolio was formed from a portfolio of noncallable twenty-year Treasury bonds. The (stock) market portfolio comprised of the valueweighted index of NYSE and AMEX stocks. In addition to a money center, a nonmoney center, a stock return portfolio, and a bond return portfolio, we also analyze the comparative predictability of returns on a REIT portfolio. We construct an equally-weighted equity REIT return series using all equity REITs on the CRSP from January 1971 to December 1989. All equity REITs are included, not just those having a continuous price history over the period in question to avoid the problem of survivorship bias. The REIT portfolio consists of 50 equity REITs on average. A REIT is deemed to be an equity REIT if it is listed as such on at least two of the following three sources: (1) REZT Sourcebook published by the National Association of Real Estate Investment Trusts, Inc., (2) 7Fze Realty Stock Review published by Audit Investments, or (3) Moody’s Bank and Finance Manual, Volume 2.

4. Empirical Results Table 1 provides summary statistics for the variables used in this study. Panel A provides data on the monthly means, standard deviations (SDS) and first-order autocorrelations on five portfolios: (i) the market portfolio, (ii) the government bond portfolio, (iii) the REIT portfolio, (iv) the nonmoney-center bank (NMCB) portfolio, and (v) the money-center bank (MCB) portfolio. As can be seen, both MCBs and NMCBs had higher excess returns on their portfolios than both the market and the bond portfolio. However, their excess returns were lower than those for the portfolio of REITs. Interestingly, the excess mean return on the MCBs portfolio (0.381%) was lower than that for the NMCBs portfolio (0.401%) even though MCB banks had a higher standard deviation of returns (6.415% vs. 4.783%). Bank stock returns in general also appeared to exhibit a higher degree of first-order autocorrelation than the nonbank portfolios (assets). In Table 1, panel B, the sample is divided into three subsamples approximating the three monetary policy regimes that existed over this period. Specifically the mid subsample (1979: 10-1982: 10) is viewed as a period of relatively high interest rate volatility compared to the other two subsample periods. As can be seen, while the mean excess returns for NMCBs increase over three subperiods, those for MCBs were actually highest in 1979: lo1982: 10 even though the underlying volatility of returns was highest in 1982: 1 l-1989:3. Thus the subperiod results suggest that MCB and NMCB stocks have diverged in their return performance during the 1980s with the MCB stocks generating lower mean excess returns than NMCB stocks despite increasing relative return risk. Such evidence is consistent with the existence of implicit deposit insurance and safety-net guarantees for large money center bank investors in the 1980s which divorced the required returns on bank stocks from the underlying riskiness of bank portfolios (see O’Hara and Shaw, 1990). In Table 1, panel D, the correlation among excess returns of the market, bond, REIT, NMCBs, and MCBs portfolio are shown for the whole period. As can be seen both NMCBs and MCBs are more highly correlated with the REIT portfolio than with the bond portfolio.

205


Table 1. A. Summary statistics for 1971:2-1989:4 Mean (%)

S.D. (%)

PI

0.282 0.038 0.679 0.401 0.381

4.822 3.287 4.887 4.783 6.415

0.055 0.050 0.115 0.230 0.151

Dependent Variables: Excess return on the market portfolio Excess return on government bond portfolio Excess return on RElTs portfolio Excess return on NMCBs portfolio Excess return on MCBs portfolio B. Summary statistics for subperiods Mean (%) Subperiods Market Bonds REITs NMCBs MCBs

SD. (%)

1971:2-1979:9 0.018 4.627 -0.088 1.953 0.573 5.889 -0.052 4.616 0.058 6.090

Mean (%)

S.D. (%)

1979:10-1982:10 0.043 5.373 -0.344 5.440 0.863 5.276 0.469 5.137 0.834 6.475

Mean (%)

S.D. (%)

1982:l l-1989:3 0.748 4.831 0.382 3.382 0.732 2.860 0.974 4.827 0.598 6.856

C. Correlations among excess returns of different assets

Market Bonds REITs NMCBs MCBs

Market

Bonds

REITs

NMCBs

MCBs

1 .ooo

0.317 1.000

0.639 0.186 1.000

0.818 0.345 0.734 1.000

0.670 0.355 0.606 0.845 1.000

Notes: MCBs stands for money center banks, and NMCBs represents nonmoney center banks, The sample period for this table is 1971:2-1989:4, with 219 observations. Units on excess returns are percentage points per month. Units on one-month T-bill rate, term spread, dividend yield, and cap rate are percentage per annum. p, is the first-order autocorrelation coefficient of the series,

This suggests that bank exposure to real-estate risk may be as important if not more important than their exposure to interest rate risk (as reflected in bond returns)! The relationship between the excess returns on MCB stocks and REITs is also shown in Figure 1. While MCB stock returns appear to be less volatile than REITs in the 1970s they become more volatile in the 1980s. This is consistent with real estate factors having a relatively greater influence on bank stocks in the most recent decade as large banks expanded their loan portfolios in the real estate area. Table 2 examines the extent to which the forecasting variables (the January dummy, the dividend yield, the treasury bill yield, the corporate-treasury bill spread, and the cap-rate on real estate assets) explain the time-variation in ex ante excess returns on our five asset portfolios-and in particular, the ex ante excess returns on MCB and NMCB stocks. The t-statistic has been adjusted for heteroskedasticity and serial correlation in the regression using the GMM approach. The most interesting finding here is the significant contribution (at 5 % level) to the predictability of NMCB returns by the real estate variable (the cap rate). As we can see from

206

JIANPING

ME1 AND ANTHONY SAUNDERS

-

I

72

71.

76

78

Time Period:

80

82

84

t36

88

90

1 9 7 1 .2- 1 9 8 9 . 4

Figure 1. Excess returns on REITs and MCBs

the regression, the market will demand a high risk premium (expected excess returns) when the real estate market cap rate is high, and a lower risk premium will suffice if the cap rate is low. Since the cap rate is an indicator of real estate market condition, we can conclude that the risk premium on large nonmoney center banks is highly influenced by real estate market conditions. The interest rate variable also affects NMCB premiums in the expected direction. To make sure that our results on NMCB are not driven by the small firm effect or the January effect, we here also provide the regression result for a small firm portfolio! Although all coefficients are of same signs, it is worth noting that the January effect is much smaller for NMCB stocks while the same effect is much stronger in REIT. We can also see that short-term interest rates (T-bill) have a smaller effect on NMCB stocks comparing to that on REIT and small stocks. The term spread has a positive (insignificant) effect on the two bank portfolios but a negative effect on REIT and small stocks. Nevertheless, while interest rate and real estate variables have the expected signs they are not statistically significant in the MCB regression. Overall, approximately 8.2% of the variation in monthly excess returns on NMCBs (compared to 3 % on MCBs) is accounted for by our five forecasting variables after adjusting for degrees of freedom. Thus, the simple unconstrained version of the latent variable model implies greater interest rate and real estate exposure for large nonmoney center banks relative to money center banks!

207


Table 2. Difference between MCB and NMCB in unrestricted conditional risk premiums. rl.t+l = Cons. + p, Jandum + &T-Bill, + &Spread, + flqDivYld, + &CapRate, + i, Constant

Jandum

T-bill

Spread

DivYld

CapR

ti2

Market

-7.408** (-2.41)

1.815 (1.56)

-0.597** (-3.25)

0.007 (0.03)

1.429** (2.27)

0.724* (1.76)

,065 1 . 8 7

Equity REITs

-9.335** (-2.80)

5.316** (4.47)

-0.603** (-3.40)

-0.095 (-0.45)

1.110* (1.82)

1.040** (2.62)

.I52 1 . 7 7

Govt. Bonds

- I.652 (-0.69)

MCB Stocks

-9.494** (-2.03)

3.027* (1.92)

NMCB Stocks

-9.732** (-2.86)

Small Stocks

~ 12.562** (-2.58)

Asset Class

-0.747 (-0.92)

0.077 (0.61)

0.362** (2.41)

1.239** (2.84)

-0.322 (-1.29)

0.275 (0.94)

3.690** (3.23)

-0.402** (-2.22)

0.176 (0.82)

5.604** (3.42)

-0.841** (-3.30)

-0.104 (-0.34)

DW

-0.332 (-1.16)

,033 1 . 9 0

1.497* (1.76)

0.649 (1.17)

,030 1 . 6 7

0.934 (1.51)

0.921** (2.26)

,082 1 . 5 4

2.014** (2.3X)

1.114** (2.00)

,133 1 . 7 6

Regression of the returns on each asset class at time t + 1 on a January dummy, the yield on Treasury hills, the spread between the yield on AAA corporate bonds and the yield on T bills. the dividend yield for the overall stock market, and the cap rate on real estate all at time I. Regression coefticients are given by the first line of each TOW, while the f-statistics are given in parentheses in the second row. MCBs stands for money center banks and NMCBs represents nonmoney center banks. * indicates significance level at 10%. ** indicates significance level at 5%.

With respect to the other forecasting variables, the constant is significant for both bank stock return portfolios as is the January dummy (at the 10% level or better). The spread and dividend yield variables have the expected positive signs but are generally insignificant at the 10% level with one exception-the dividend yield variable in the MCB stocks regression. The time-varying forecasting variables in Table 2 along with their estimated coefficients can be used to generate expected excess returns E,[r,,,+,], or conditional risk premiums, for each portfolio. These can be compared to the actual excess returns, i,i+i , on each portfolio. Estimated expected excess returns relative to actual excess return for MCBs are shown in Figure 2. As can be seen, the monthly predictable or forecastable risk premiums on MCBs can be as high as 6% for some months, while actual monthly excess returns vary from about -28 % to 20%. Figure 3 plots the expected excess return (risk premiums) for MCBs relative to REITs over 1971:2-1989:4?O Overall, both MCBs and REITs expected excess returns move closely in tandem with a high correlation coefficient of 0.868. The risk premium for the NMCB group exhibits an even closer co-movement with that of the REITs. Figure 3 also provides a supply side explanation to real estate market cycles in the last twenty years. As we can see from the figure, the gradual reduction of risk premium for banks and real estate certainly is consistent with the increase in real estate lending and real estate development in the late 1970s while the high cost of leading in the recessions of 1980 and 1981-1982 also offer an explanation as to why larger numbers of U.S. banks adopted a relatively easy credit policy for real estate financing in the later 1980s.

209


l~~~l~~~l~~~I,,~I,,,l,,,i,,,,,,,I,,,,,,,,

72

74

76

78

80

,,,/,I

82

J

,,,,,,,,,,,,,,,,,,,,,,,,,,,

8LT

86

88

90

Time Period: 197 1 .2- 1989.4 Figure 3. Conditional risk premium on MCBs and REITs without the January effects.

to 1 for the market portfolio, the betas for both the MCBs and NMCBs are higher than unity (1.068 and 1.162 respectively) but smaller than that for REITs (pii = 1.525). As expected, the beta for the bond portfolio is low. Moreover, the chi-square test of the linear pricing restriction implied by equation 5 could not be rejected by the data at 18.1% level. The model was then reestimated assuming K = 2 (panel B) and K = 3 (panel C). That is, we also estimate a two-factor and a three-factor model. To gain insights into the effects of the interest-rate factor and the real-estate factor on bank stock returns, we undertook a normalization procedure. Specifically, we normalize the bond portfolio to have a beta of 1 on the first factor and a beta of 0 on the second factor, and we normalize REITs to have a beta of 0 on the first factor and a beta of 1 on the second factor. Using this normalization procedure, the first factor in panel B of Table 3 can be called the bond or interest rate factor, and the second factor the real estate factor, since a given change in these factors will result in a corresponding change in the relevant portfolios’ asset returns. From panel B, it can be seen that while the MCBs have a greater sensitivity to the interest rate factor than the NMCBs (0.734 vs. 0.323) over 1971-1989, they are slightly less sensitivity to the real estate factor (0.665 vs. 0.732). Nevertheless, in this model specification, the realestate factor is strongly significant for both MCBs and NMCBs. Further, the linear pricing restriction imposed by equation 5 on the two-factor model cannot be rejected by the data using a Chi-square test (p = 46.2%).

210

JIANPINC MEI AND ANTHONY SAUNDERS

7?1h/e 3. Estimation of the latent variable model (4) with the rank restriction of equation 5 imposed

E,[C,,+,l =i: &k i: k,x,,, =; Gxnt

k-l

“=I

(4)

“=I

A. The number of systematic factors in the economy equals one (K = 1). S.D.

P,I Estimated beta coefficient for the following assets: Excess return on the market portfolio Excess return on government bond portfolio Excess return on REITs portfolio Excess return on NMCBs portfolio Excess return on MCBs portfolio

1 .ooo* 0.083 1.525 I.162 1.068

0.148 0.318 0.161 0.167

&statistic of the rank restriction (5): 25.54 (DF = 20). Significance level: P = 0.181. B. The number of systematic factors in the economy equals two (K = 2).

Estimated beta coefficient for the following assets: Excess return on market portfolio Excess return on government bond portfolio Excess return on REITs portfolio Excess return on NMCBs portfolio Excess return on MCBs portfolio X2-statistic of the rank restriction (5): Significance level: P = 0.462.

P,l

S.D.

P,,

S.D.

1.068 I .OOo* o.OOo* 0.323 0.734

0.442

0.603 o.ooo* 1 .OOo* 0.732 0.665

0.160

0.243 0.464

0.089 0.141

I I .79 (DF = 12)

C. The number of systematic factors in the economy equals two (K = 3).

4, Estimated beta coefficient for the following assets: Excess return on market Excess return on government bond portfolio Excess return on REITs portfolio Excess return on NMCBs portfolio Excess return on MCBs portfolio

1.000* o.oOO* o.oOo* -0.461 -0.360

S.D.

P,*

o.OOo* 1.000* o.OOo* 0.649 0.832 0.985 1.099

S.D.

P,3

S.D.

o.ooo* o.OOo* 1.000* 0.624 1.006 0.400 0.841 0.871 0.587

x*-statistic of the rank restriction (5): 3.16 (DF = 6). Significance level: P = 0.788. Notes: MCBs stands for money center banks, and NMCBs represents nonmoney center banks. S.D. stands for standard error for the corresponding parameter estimates. Asterisks indicate these numbers are normalized to be I or 0. The sample period for this table is 1971:2-1989:4, with 219 observations. The standard errors reported here have been corrected for heteroskedasticity using the general method of moments (GMM) of Hansen (1982).


211

Finally in Table 3, panel C, we present the results of a three-factor model. As in the twofactor case, we normalize the betas on various asset portfolios. Specifically in the threefactor case, we normalize the betas on the stock market, bond, and REIT portfolios so that we have a stock factor, a bond factor, and a real estate factor. This allows us to identify the effects of these three factors on MCB and NMCB portfolio returns. As can be seen the sensitivity of bank stocks to the stock factor is both negative and insignificantly different from zero. The exposure of bank stocks to the interest rate factor is again stronger for the MCBs, although the significance level is lower than in the two-factor model. The strongest factor, in terms of explanatory power, appears to be the real estate factor. This factor is positive for both portfolios of banks, but it is larger and more significant for the NMCB group.‘* These results appear to confirm our earlier findings in Table 2 with the unconstrained model, that the real estate factor is more important for NMCBs than for MCBs.

4.2. Subsample Estimates As a test of robustness, we also split the sample into two subsamples: 1971:2-1979:9 and 1979: IO-1989:4. The latter period incorporates a more volatile interest rate environment as well as a period in which many banks were expanding their real estate portfolios. By splitting the sample in this manner, we allow for structural changes in factor sensitivitiesI For MCBs, the results in Table 4, panel A for the one-factor model suggest an increase in their sensitivity to the market factor in the more recent subperiod (0.975 vs. 0.394). The two-factor model (panel B) implies that while MCBs’ interest rate sensitivity increased marginally in the more recent subperiod, their sensitivity to the real estate factor increased more dramatically. This probably reflects the fact that the market is increasingly aware of the real estate risk exposure of large U.S. banks. The results for the three-factor model (panel C), generally support the two-factor model conclusions regarding changes in MCBs’ real estate factor sensitivity over the two periods. The results for NMCBs generally suggest an increase in sensitivity to all factors in the more recent subperiod. These results are supported by recent changes in bank real estate holdings. According to Standard and Poor’s Industry Surveys (Industry Survey, Banking Section 1991), U.S. commercial banks has $285.7 billion of real estate loans in 198 1. By 1990, the real estate loans almost tripled to $836.5 billion. In 1991, the ratio of real estate loans to all loans was 29%. By 1992, the same ratio became 40%!

5. Summary and Conclusions This paper has sought to examine the ex ante pricing of risk implicit in money center and nonmoney center bank returns. Unlike previous studies that have concentrated on the ex post return generating processes of bank stock returns, or have assumed risk premiums to hold constant over time, we investigate an asset pricing model from which ex ante timevarying risk premiums can be derived. We find that the time variation in bank risk premiums have been partly determined by interest rate and real estate market conditions. We also discover that the real estate factor

212

JIANPING MEI AND ANTHONY SAUNDERS

Table 4. Estimation of the latent variable model (4): Two subperiods

A. The number of systematic factors in the economy equals one (K = 1).

PII Estimated beta: Market Bonds REITs NMCBs MCBs

S.D.

S.D.

@,I 1979: IO-1989:3

1971:2-1979:9

1 .ooo*

I .ooo*

0.123 0.729 0.848 0.394

0.158 1.151 0.998 0.975

0.120 0.362 0.144 0.289

&statistic of the rank restriction (5): 19.40 Significance level: P = 0.495

0.207 0.182 0.142 0.183

&statistic of the rank restriction (5): 25.42 Significance level: P = 0.185

B. The number of systematic factors in the economy equals two (K = 2)

P,I Estimated beta: Market Bond REITs NMCBs MCBs

2.743 1.000* o.OOO* 1.428 0.637

S.D.

PI2

1971:2-1979:9 1.755 0.561 o.OOO* 1 .OOO* 0.986 0.690 1.088 0.390

&statistic of the rank restriction (5): Significance level: P = 0.471

S.D.

&I

0.236

0.211 1.000* o.OOO* 0.213 0.737

0.123 0.187

11.68

S.D.

PI2

1979: IO-1989:3 0.181 0.805 o.OOO* 1 .OOO* 0.147 0.852 0.313 0.931

X2-statistic of the rank restriction (5): Significance level: P = 0.507

S.D. 0.135

0.095 0.177

11.25

C. The number of systematic factors in the economy equals two (K = 3) S.D.

0,~

S.D.

Pi3

S.D.

-

o.OOo* 1.000* o.OOO*

-

o.OOo* o.OOo* 1.000*

-

1.224 1.234

3.267 2.931

3.193 3.029

1.038 0.813

0.653 0.631

1.58 (P = 0.953) -0.514 0.583 -1.041 1.210

0.378 1.097

0.310 0.643

1.218 1.663

0.519 1.021

P,I Estimated beta coefficient for the following assets: Excess return on market Excess return on government bond portfolio Excess return on REITs portfolio

1.000* o.OOO* o.OOO*

Subperiod 1: 1971:2-1979:9 &statistic: 1.76 (P = 0.940) Excess return on NMCBs portfolio -0.593 Excess return on MCBs portfolio -0.714 S u b p e r i o d 2 : 1979:10-1989:3 x2-statistic: Excess return on NMCBs portfolio Excess return on MCBs portfolio

Notes: MCBs stands for money center banks, and NMCBs represents nonmoney center banks. SD. stands for standard error for the corresponding parameter estimates. Asterisks indicate these numbers are normalized to be 1 or 0. The sample period for this table is 1971:2-1989:4, with 219 observations. The standard errors reported here have been corrected for heteroskedasticity using the general method of moments (GMM) of Hansen (1982).


213

has been most important for large nonmoney center banksi Moreover, we find that the time variation of risk premiums is consistent with changes in bank real estate lending during the sample period.

Acknowledgments We wish to thank John Campbell for letting us use his latent variable model algorithm and Doug Herold, Wayne Ferson, and Cracker Liu for providing data on real estate cap rates, REITs, and business condition variables to us. We are also grateful to Bin Gao for able research assistance. We have benefited from helpful discussion with Mitchell Berlin, Silverio Foresi, and Cracker Liu. We acknowledge financial support from the Salomon Center at New York University.

Appendix Money-Center Bank Date Begin

Name Bank of Boston Corp Bank New York Inc Bankers Trust NY Corp Chase Manhattan Corp Chemical Banking Corp Chemical Banking Corp Citicorp First Chicago Corp Manufacturers Hanover Corp Mellon Bank Corp Morgan JP & Co Inc Republic New York Cot-p

Name

710107 691204 6905 16 6503 15 690305 87052 1 681101 711230 690428 810710 69040 1 720112

Date End

Value (in $1000)

901231 901231 901231 901231 901231 901231 901231 901231 901231 901231 901231 901231

0.469908E+06 O.l23442E+07 0.350817E+07 O.l34632E+O7 0.967274E+06 0.337990E+05 0.419593E+07 O.l08930E+07 O.l52318E+07 O.l04650E+07 0.8217lOE+07 O.l71384E+07

Nonmoney Center Bank (NYSE & AMEX) Date Begin Date End

Amsouth Bancorporation Arizona Commerce Bank Tucson BBI Inc BSD Bancorp Inc Bathe Group Inc Baltimore Bancorp Bane One Corp Banctexas Group Inc Bank New England Corp Bank of San Francisco Hldg Co

8 10520 861223 620702 85 1010 72053 1 880204 830929 811023 880412 880712

901231 90083 1 730508 901231 810611 901231 901231 901231 901231 901231

Value (in $1000) 0.456554E+06 O.l54250E+03 0.989500E+03 0.242276E + 05 0.348742E+06 0.541875E+05 0.439665E+07 0.395719E+O4 0.478074E +05 O.l07540E+05

214


Nonmoney Center Bank (NYSE & AMEX) (continued) Name Date Begin Date End Value ($1000) Bankamerica Corp Barnett Banks Inc CVB Financial Corp Chase Convertible Fd Boston Inc Citizens & Southern Corp GA Citizens First Bancorp City National Corp Citytrust Bancorp Inc Continential Bank Corp Eldorado Bancorp CA Equimark Corp First of America Bank Corp First Bank Systems Inc First City Bancorporation TX Inc First City Bancorporation TX First Empire State Corp First Fidelity Bancorporation NE First Interstate Bancorp First Interstate Bancorporation First National Corp CA First Pennsylvania Corp First Union Corp First Virginia Banks Inc Firstar Corp New Guarantee Bancorp Inc Guardian Bancorporation LA HUBCO Inc Hibernia Cot-p James Madison Ltd Keycorp LITCO Bancorporation NY Inc La Jolla Bancorp Landmark Bancshares Corp MNC Financial Inc NBD Bancorp Inc National City Corp Norwest Corp PNC Financial Corp Pacific Western Bancshares Security Pacific Corp Shawmut National Corp Signet Banking Corp

760628 791217 891214 7207 17 890313 781227 9005 17 8807 11 730910 870924 710720 900629 840507 880720 761201 880317 710517 620702 8802 16 871109 720110 880801 710419 710913 791227 880209 84030 1 891026 861007 830527 780417 860604 841121 890112 73022 1 881025 621210 871021 890413 7903 14 881221 710106

90123 1 90123 1 901231 8203 19 90083 1 90123 1 901231 90123 1 901231 901231 901231 901231 901231 901231 880419 90123 1 901231 901231 901231 901231 900305 901231 901231 901231 840117 901231 901231 901231 901231 901231 820609 900823 901231 901231 901231 901231 901231 901231 901231 901231 901231 901231

0.565060E+07 O.l18465E+07 0.647932E+05 0.717080E+05 O.l29093E+07 0.504474E+05 0.3886068+06 0.796500E+04 0.440972E306 0.253036E+05 0.26577OE+O5 0.761685E+06 O.l00340E+07 0.845542E+O5 O.l89433E+05 0.361898E+06 0.1045898+07 O.l45206E+07 0.149504E +05 0.416981E+05 0.585003E+06 O.l67834E+07 0.485440E+06 0.657378E+06 0.3645348+05 0.429944E +05 0.2191928+05 O.l81803E+06 0.9735008+04 0.983475E+06 0.957960E+05 0.725355E+05 0.390601E+05 0.280836E+06 0.240659E+O7 O.l89056E+07 0.209434E+07 0.2057478+07 0.983854E+05 0.251307E+07 0.3737378+06 0.275560E+06

215


Nonmoney Center Bank (NYSE & AMEX) (continued) Value ($1000) Date End Date Begin Name Southeast Banking Corp Southwest Bancorp Southwest Bancshares Inc Sovran Financial Corp Sterling Bancorp Suntrust Banks Inc Texas American Bancshares Inc UJB Financial Corp Union Planters Corp Wells Fargo & Co New Westamerica Bancorporation Worthen Banking Corp

Name

720807 850716 760728 8809 13 620702 850701 820623 701123 890207 700209 870108 830815

901231 891019 841010 90083 1 90123 1 901231 890725 901231 90123 1 901231 901231 90123 1

Nonmoney Center Bank (NASDAQ) Date Begin Date End

Affiliated Bankshares Co10 Inc Ameritrust Corp Amsouth Bancorporation Arizona Commerce Bank Tuscan Baltimore Bancorp Bane One Corp Banco Popular de PR Bancorp Hawaii Inc Banctexas Group Inc Bank New England Corp Bank of San Francisco Hldg Co Bank South Corp Bankamerica Corp Banks Iowa Inc Barnett Banks Inc Baybanks Inc Boatmens Bancshares Inc CVB Financial Corp Central Bancshares South Inc Central Fidelity Banks Inc Citizens & Southern Corp GA City National Corp Citytrust Bancorp Inc Colorado National Bankshares Inc Comerica Inc Commerce Bancshares Inc Corestates Financial Corp

721214 721214 721214 861002 841025 721214 721214 721214 7704 18 721214 860627 721214 721214 721214 721214 721214 721214 8303 16 721214 721214 721214 721214 721214 721214 721214 721214 721214

901231 901231 8105 19 861222 880203 830928 901231 901231 811022 880411 8807 11 90123 1 760625 90123 1 791214 901231 90123 1 891213 90123 1 901231 890310 9005 16 880708 901231 901231 901231 901231

O.l64244E+06 0.382326E+05 0.373744E+O6 0.1583488+07 0.4505 14E +05 0.288424E+07 O.l38875E+04 0.319236E+06 0.117205E +06 0.297744E+07 0.828516E+05 O.l25706E+06

Value (in $1000) O.l09897E+06 0.303639E+06 O.l71887E+06 O.l52932E+05 O.l64866E+06 0.865059E+06 0.321568E+06 0.787539E+06 0.603595E+05 0.189932E +07 O.l07278E+05 0.214331E+06 0,388828E+07 0.205829E+06 O.l75601E+06 0.204204E+06 O.l10144E+07 O.l27820E+06 0.295730E+06 0.3798788+06 O.l64851E+O7 0.676099E+06 0.200450E+06 O.l04377E+06 0.702182E+06 0.430171E+06 O.l72402E+07


216

Nonmoney Center Bank (NASDAQ) (continued) Date End Date Begin Name Crestar Financial Corp Cullen Frost Bankers Inc Dauphin Deposit Corp Deposit Guaranty Corp Dominion Bankshares Corp Eldorado Bancorp CA Equitable Bancorporation Fifth Third Bancorp First Alabama Bancshares Inc First Bancorporation Ohio Inc First American Corp TN First of America Bank Corp First Bank Systems Inc First City Bancorporation TX First Commerce Corp New Orleans First Empire State Corp First Florida Banks Inc First Hawaiian Inc First Maryland Bancorp First National Corp CA First Security Corp DE First Tennessee Nat1 Corp First Union Corp Florida National Banks FL Inc Fourth Financial Corp Guarantee Bancorp Inc HUBCO Inc Hibernia Cot-p Huntington Bancshares Inc INB Financial Corp Integra Financial Corp Keycorp LITCO Bancorporation NY Inc Landmark Bancshares Corp MNC Financial Inc Manufacturers National Corp Marshall & Ilsley Corp Mellon Bank Corp Mercantile Bancorporation Inc Mercantile Bankshares Corp Merchants National Cot-p Meridian Bancorp Inc

721214 770708 740405 721214 721214 810318 721214 750423 721214 820104 721214 780504 721214 721214 721214 721214 721214 721214 721214 821101 721214 721214 721214 721214 721214 721214 721214 721214 721214 721214 810115 721214 721214 760312 721214 721214 721214 721214 721214 721214 730924 721214

901231 901231 901231 901231 901231 870923 900117 901231 901231 901231 901231 900628 840504 761130 901231 880316 901231 901231 89032 1 871106 901231 901231 880729 900126 901231 791226 840229 89 1025 901231 901231 901231 830526 780414 841120 890111 901231 901231 810709 901231 901231 901231 901231

Value ($1000) 0.430512E+06 0.565144E+05 0.336757E+06 O.l56025E+06 0.2625088+06 0.229560E-tO5 0.437543E+06 O.l30119E+07 0.581685E+06 0.301296E+06 0.148608E +06 0.841650E+06 0.750868E+06 0.2979908+06 O.l61084E+06 0.266084E+06 O.l41975E+06 0.6335608+06 0.703596E+06 0.2363388+05 0.355113E+06 0.355439E+06 0.253192E+07 0.675635E+06 0.316931E+06 0.3360348+05 O.l81125E+05 0.687808E+06 0.651491E+06 0.340879E+06 0.386272E+06 O.l94985E+06 0.457710E+05 0.3388708+05 O.l29424E+07 0.688882E+06 0.600012E+06 0.753109E+06 0.341397E+06 0.601708E+06 0.216706E+06 0.4158848+06

217


Nonmoney Center Bank (NASDAQ) (continued) Name Date Begin Date End Michigan National Corp Midlantic Corp Multibank Financial Corp NBD Bancorp Inc National City Corp Northeast Bancorp Inc Northern Trust Corp Old Kent Financial Cot-p PNC Financial Corp Pacific Western Bancshares Premier Bancorp Inc Puget Sound Bancorp Riggs National Corp Wash DC Security Pacific Corp Shawmut National Corp Society Cot-p South Carolina National Corp Southwest Bancorp Southwest Bancshares Inc Sovran Financial Corp Star Bane Corp State Street Boston Cot-p Sunwest Financial Services Inc Texas American Bancshares Inc Trustcorp Inc Trustmark Corp Union Planters Corp United Banks Colorado Inc United Missouri Bancshares Inc United States Bancorp OR United States Trust Cot-p Valley National Corp AZ West One Bancorp Westamerica Bancorporation Zions Bancorp

Name

721214 721214 8006 11 721214 73050 1 721214 721214 770802 721214 831102 850111 780705 721214 721214 721214 721214 721214 780802 721214 721214 721214 721214 770803 721214 730131 721214 721214 721214 721214 721214 721214 721214 741028 761015 721214

901231 901231 901231 730220 881024 901231 901231 901231 871020 890412 901231 901231 901231 790313 881220 901231 901231 850715 760727 880912 901231 901231 90123 1 820622 900105 901231 890206 90123 1 901231 901231 901231 901231 901231 870107 901231

REIT (NYSE & AMEX) Date Begin Date End

American Fletcher Mtg Invs American Health Pptys Inc American Realty Trust Inc Americana Hotels & Realty Corp

701106 870212 691002 830113

77 1202 901231 901231 901231

Value ($1000) 0.244902E+06 O.l90470E+06 0.355221E+05 0.292500Ef06 O.l34692E+07 0.375660E+05 0.100111E+07 0.615426E+06 0.255371E+07 0.818235E+05 0.676858E+05 0.299763E+06 0.1205498+06 0.761228E+06 O.l69942E+07 O.l0568OE+O7 0.354126E+06 O.l14538E+05 O.l50975E+O6 0.194216Ef07 0.483874E+06 0.1274658+07 0.1275628+06 0.257430E+06 0.428558E+06 O.l71938E+06 0.1973988+06 0.350532E+06 0.338229E+06 O.l22172E+07 0.283309E+06 0.215553E+06 0.306672E+06 O.l47537E+06 O.l90310E+06

Value (in $1000) O.l52lOOE+04 0.365804Et06 O.l16490E+05 O.l53750E+05


218

Name

REIT (NYSE & AMEX) (continued) Date Begin Date End

Angeles Mortgage Investment Tr Angeles Participating Mtg Trust Arizona Land Income Corp Asset Investors Corp Associated Mortgage Investors BB Real Estate Invt Corp BRT Realty Trust BRE Properties Inc BT Mortgage Investors Boddie Noel1 Restaurnt Bradley Real Estate Trust Burnham Pacific Properties Inc CR1 Insured Mtg Invs II Inc CR1 Insured Mortgage Assn Inc CR1 Liquidating REIT Inc CV REIT Inc California Real Estate Invt Tr Capital Housing Mtg Partners Inc Capstead Mortgage Corp Citizens Mortgage Investment Tr Columbia Real Estate Invts Inc Continental Mortgage Investors Copley Property Inc Countrywide Mortgage Invts Inc Delaware Valley Fin1 Corp Derwood Investment Trust Dial REIT Inc Duke Realty Investments Inc Duke Realty Investments Inc EQK Realty Investors 1 Eastgroup Properties Equitable Life Mtg & Rlty Invs FGI Investors Federal Realty Investment Trust First Union Real Est Eq & Mg Inv General Growth Pptys Guild Mortgage Investments Inc HMG Courtland Properties Ltd HRE Properties Harris Teeter Pptys Inc Health Care Ppty Invs Inc Healthvest

880104 891115 880607 861219 700303 85 1025 730523 8008 14 710318 870519 891005 870604 860630 891207 891207 73032 1 801229 890609 850905 710817 86022 1 650614 850719 850909 820714 700128 900518 860317 88083 1 850305 730117 710122 7 10430 750609 70052 1 730305 860624 720926 700518 860808 850523 860529

901231 901231 901231 901231 73 1026 890717 901231 901231 820618 901231 901231 901231 891127 901231 901231 901231 901231 901231 901231 760120 901231 750723 901231 901231 901231 860115 901231 880815 90123 1 90123 1 901231 830105 831208 901231 901231 850925 880629 90123 1 901231 871230 901231 901231

Value (in $1000) 0.292530E+05 0.289800E+05 0.966OOOE-tO4 0.111160E+06 0.740675E+04 0.200375E+05 O.l74491E+05 O.l82479E+06 0.3438508+04 0.270750E+05 0.208835E+05 0.686250E+05 O.l06700E+06 O.l35972E+06 0.368915E+06 0.216810E+05 0.215840E+05 0.305148E+05 O.l21800E+06 0.248675E+04 0.367579E+05 0.2344288+05 0.345395E+05 0.614025E+05 0.777600E+04 O.l32559E+05 0.565356E+05 0.61 lOOOE+O4 0.247078E+05 0.189725E +05 0.306862E+05 0.838058E-tO5 O.l02878E+05 0,240249E+06 O.l28598E+06 0.712598E+05 O.l89875E+05 0.562400E+04 0.5868508+05 0.219188E+05 0.38034lE+06 O.l88528E+05

219


Name

REIT (NYSE & AMEX) (continued) Date Begin Date End

Health & Rehabilitation Pptys Tr Health Care REIT Inc Health Equity Properties Inc Highlands National Inc Homeplex Mortgage Investments Hotel Investors Trust Hotel Properties Inc ICM Property Investors Inc IDS Realty Trust IRT Property Co International Income Ppty Inc Kavanau Real Est Tr Koger Equity Inc Landsing Pacific Fund Lincoln North Carol Rlty Fd Inc Linpro Specified Pptys MGI Properties Inc MIP Properties Inc MSA Realty Corp Massmutual Mortgage & Rlty Invs Medical Properties Inc Meditrust Metropolitan Realty Corp Mortgage & Realty Trust Nationwide Health Properties Inc New Plan Rlty Tr North American Mortgage Invs Northwestern Mutual Lf Mtg & Rlty Nova Corp Novus Properties Co One Liberty Properties Inc Pacific Realty Trust Pennsylvania Real Est Invt Tr Pittsburgh & West Virginia RR Plaza Realty Invs Presidential Realty Corp New Presidential Realty Corp New Property Capital Trust Property Trust Amer Prudential Realty Trust REIT America Inc RPS Realty Tr

861217 840223 861021 731128 88072 1 720503 840106 850125 73052 1 711011 840607 620702 880818 881213 851216 860723 720302 85070 1 841001 710113 8703 10 870623 881117 7 10503 851219 790307 690905 710722 731031 720428 861008 721219 700610 620702 730524 620702 620702 720906 890627 851001 710416 881228

901231 901231 901231 750609 901231 901231 860915 901231 761216 901231 900626 76052 1 901231 901231 901231 901231 901231 901231 901231 850625 901231 901231 90123 1 901231 901231 901231 810519 821115 750930 7404 18 901231 830215 901231 90123 1 760616 901231 901231 901231 90123 1 901231 841213 901231

Value ($1000) O.l33983E+06 0.8248368+05 0.204960E+05 0.382362E+04 0,423281E+05 O.l36485E+05 0.7584OOE+05 0.216630E+05 0,451688E+04 0.912102E+05 0.309983E-tO6 O.l45622E+O4 O.l12620E+06 0.4754628+05 O.l04895E+05 O.l04400E+04 0.693029E-tO5 O.l46575E+05 0.291836E+05 O.l21420E+06 O.l12528E+05 0.388482E+06 O.l86945E+05 0.263055E+O5 O.l95912E+06 0.601961E+06 O.l63899E+05 0.654225E+05 0.241600E+04 0.1766758+05 0.887088E+O4 0.382500E+O5 O.l24373E+06 0.887125E+04 0.1531758+04 O.l97588E+04 0.856800E+04 0.627491E+05 0.354970E+05 O.l73984E+04 0.923449E+05 O.l54096E+06

220


Name

REIT Bank (NYSE & AMEX) (continued) Date Begin Date End

Rampac Real Estate Investment Tr Amer Real Estate Investment Trust CA Realty Refund Tr Residential Resources Mg Invt Cp Residential Mortgage Invts Inc Resort Income Invs Inc Rockefeller Center Pptys Inc Rymac Mortgage Investment Corp Saul BF Real Estate Invt Tr Sierra Capital Realty Trust VIII Sizeler Property Investors Inc Storage Properties Inc Storage Equities Inc Strategic Mortgage Invts Inc TIS Mortgage Investment Co Trammel1 Crow Real Estate Invs Transcontinental Realty Invstrs Transamerica Realty Investors TRI South Investors Inc Turner Equity Invs Inc USP Real Estate Investmt Trust United Dominion Realty Tr Inc United Realty Investors Inc United States Realty Invts Universal Health Rlty Incm Tr Wachovia Realty Investments Washington Real Est Invt Tr Weingarten Realty Investors Wedgestone Financial Wells Fargo Mortgage & Equity Tr Western Investment Real Est Tr Wincorp Realty Invts

Name API Trust SBI American Equity Invt Tr Americana Hotels & Realty Corp Arlington Realty Investors Bradley Real Estate Trust Burnham Pacific Properties Inc

800417 620702 870819 721121 880624 860616 881025 850912 880916 7307 10 900727 870130 901001 821001 841212 880819 851121 860910 710719 720530 850719 880810 900507 7206 19 680226 870120 710128 710507 850816 851001 720425 840613 77 1004

REIT (NASDAQ) Date Begin 721214 740524 821104 721214 850517 870105

Value ($1000)

840123 831004 901231 901231 890316 901231 901231 901231 901231 880817 901231 901231 90123 1 901231 891106 901231 901231 901231 860930 850415 881221 901231 901231 831026 821223 901231 820318 901231 901231 901231 891222 901231 841207

O.l18184E+06 0.640952E+05 0.9236258+05 O.l34006E+05 0.2799388+04 0.896750E +04 0.327285E+05 0.712690E+06 0.3604568+05 0.147436E +06 O.l02064E+05 0.290568E+05 0.1850068+05 0.760351E+05 0.6967888+05 0.546750E+05 0.226875E+O5 O.l22760E+05 0.383952E+05 0.3693808+05 0.304020E+05 O.l11550E+05 O.l73041E+06 0.5433008+05 0.344158E+05 0.766361E+05 0.250125E+05 0.2593408+06 0.408202E+06 0.362188E+04 0.925925E+O4 0.2333208+06 0.581030E+05

Date End

Value (in $1000)

830414 8409 13 830112 880503 891004 870603

0.417000E+O4 0.596159E+05 O.l23714E+06 0.998000E+03 0.3822OOE+05 0.238560E+05


Name

221

REIT (NASDAQ) (continued) Date Begin Date End

CPL Real Estate lnvesmtne Tr CV REIT Inc Cedar Income Fund Ltd Cedar Income Fund 2 Ltd Central Realty Investors Inc Centrevest Corp Chicago Dock and Canal Trust Citizens Growth Pptys Clevetrust Realty Investors Commonwealth Fin Gp Rl Es Inv Tr Commonwealth Realty Tr Continental Mortgage & Equty Tr Cousins Properties Inc Delaware Valley Fin1 Corp Denver Real Estate Invt Assn Dial REIT Inc Dominion Mortgage & Realty Trust Eastover Cot-p First Fidelity Investment Trust Flatley Realty Investors Florida Gulf Rlty Tr General Real Estate Shs Grubb & Ellis Rlty Income Tr Health Care REIT Inc Highlands National Inc Hotel Properties Inc Independence Mortgage Trust International Income Ppty Inc JMB Realty Trust Landsing Institutional Ppts Tr V Lincoln Investors MYM Liquidating Trust Maxxus Inc Meditrust Mellon Part Mtg Tr Corn1 Pptys Merry Land & Investment Co Inc Miller Henry S Rlty Tr Monetary Realty Trust Monmouth Real Estate Invt Cot-p Murray Mortgage Investors Nationwide Real Estate Investors Nooney Realty Trust Inc

851220 721214 861217 881021 721214 861107 861021 721214 721214 721214 7403 13 821018 721214 730209 721214 861219 721214 730302 721214 810615 730504 721214 8504 19 780217 721214 760430 721214 791218 801128 850422 721214 901204 731218 85 1009 8502 13 8 10706 721214 721214 721214 73062 1 721214 851015

880205 730320 90123 1 890929 901231 880715 901231 901231 901231 90123 1 8803 11 901231 901231 820713 801222 900517 8602 13 901231 781109 810902 851210 901231 901231 840222 860430 840105 770808 840606 90092 8 881125 820308 901231 850116 870622 901231 901231 821223 821130 901231 840608 8 10220 901231

Value ($1000) O.l23761E+05 0.000000E+00 O.l04805E+05 0.6237OOE304 0.341031E+03 0.433838E-tO4 0.6507OOE+05 0.259875E+O4 0.3946OOE+04 0.5355948+03 0.227540E+05 0.1645588+05 0.1733708+06 0.203712E+05 0.401865E+05 0.771690E+05 O.l97635E+05 0.629200E+04 0.941775E+04 0.9625OOE+O4 0.600064E+05 0.423438E+02 O.l16OOOE+05 0.257118E+05 O.l20478E+05 O.l48046E+05 0.703125E+03 0.859275E+05 0.7115OOE+04 O.l21560E+05 0.228312E+04 0.2593268+05 O.l00462E+05 0.217100E+O6 0.389025E+05 0.358554E+05 O.l40OOOE+05 0.780000E+03 0.898625E+O4 0.100600E+04 O.l68829E+05 0.780300E+O4

A

222


Name

REIT (NASDAQ) (continued) Date Begin Date End

Nova Corp Novus Properties Co One Liberty Properties Inc Pacific Southern Mortgage Trust Plaza Realty Invs Property Trust Amer Rainier Realty Investors Real Estate Investment Trust CA Resources Pension Shs 3 Resources Pension Shs 1 Resources Pension Shs 2 Tr Riverside Properties IL Saul BF Real Estate Invt Tr Sierra Real Estate Equity Tr 84 Storage Equities Inc Terrydale Realty Trust Towermarc Travelers Real Estate Invt Tr Travelers Realty Income Invs USP Real Estate Investmt Trust Unicorp Realty Investors Inc United Dominion Realty Tr Inc United States Equity & Mtg Tr United States Mutual Fin1 Corp Vanguard Real Estate Fund II Vinland Property Trust Virginia Real Estate Invt Tr WMI Equity Invs Wedgestone Financial Wespac Investors Trust II Wespac Investors Trust III

721214 770628 830425 730409 771104 721214 850412 840823 850606 821126 831223 721214 721214 850513 801118 730403 721214 840426 85032 1 780425 721214 8002 11 721214 800307 900807 7606 11 721214 721214 821013 831212 85 1003

830630 861103 861007 820701 840918 890626 8707156 870818 88 1227 881227 881227 90123 1 730709 901231 820930 820111 841031 890330 890330 901231 811102 900504 770404 86082 1 901231 901231 810202 811013 850930 880622 880412

Value ($1000) O.l37918E+05 0.457875E+05 0.243971E+O5 0.735OOOE+04 0.922281E+04 0.5124268+05 0.3209758+05 O.l16960E+06 0.546411E+05 0.435705E+05 0.593060E +05 0.547250E+03 0.000000E+00 O.l03849E+05 0.294548E+05 O.l85288E+O4 0.110775E +05 0.207359E+05 0.237185E+05 O.OOOOOOE+00 0.750375E +04 O.l79816E+06 0.571050E+O4 O.l4623lE+04 0.416250E+05 0.745750E+03 0.215680E+05 0.414150E+04 O.l62049E+05 0.170888E-tO4 0.545912E+04

Notes 1. It is possible that the conditional expectation could be nonlinear in the forecasting variables. In that case, equation 3 could be thought of as a Taylor first-order approximation to the nonlinear relationship. 2. A constant is also included. 3. Although Flannery and James (1984) model the portfolio mismatch of banks using the repricing structure of assets and liabilities, their results are consistent with such a duration mismatch. 4. See Campbell (1987), Keim and Stambaugh (1986) for mean reversion. 5. Liu and Mei (1992) found that the cap rate is useful in explaining the time-variation in real estate expected returns. Mei and Liu (1994) also discovered that the cap rate can be used in market timing of investment in real estate portfolios. Because the large holdings of real estate loans in bank portfolios, it would be interesting to see how the cap rate would explain the risk premium in bank stock returns.


223

6. They are Bank of New York, Bankers Trust, Chase Manhattan, Chemical Bank, Citicorp, Manufacturers Hanover, J.P. Morgan, Republic New York, Bank of Boston, First Chicago, Continental Bank. The definition follows that of Salomon Brothers Commercial Bank Stock Research (various issues). 7. The positive correlation between bond returns and bank stock returns is consistent with a negative correlation between interest rates and bank stock returns. (See also, Flannery and James, 1984). 8. The small stock returns series are takn from Ibbotson and Associates’ Stocks, Bonds, Bills, and Inf7ation (SBBI) series on CRSP. 9. This result should not be surprising given the problems of regional banks (such as the Bank of New England) in the Northeast and California. 10. These results are wirhour the January effects. II. See Campbell (1987), Ferson (1989, 1990). and Liu and Mei (1991). 12. It is worth noting that the three factors we choose may not be orthogonal. Thus, it is possible that some of the fundamental forces that move the stock factor also move the real estate factor. Therefore, it is not contradictory to see bank stocks having exposure to market risk in the one-factor model but little such exposure in a three-factor model, since some of the variations in stock returns originally explained by the stock factor are now explained by the real estate factor. 13. The subperiod 1979:10-1982: 10 (the regime of nonborrowed reserve targets) was too short to include as a separate subperiod in these tests. 14. To test the robustness of our results, we also conducted similar tests on excess returns for a portfolio of ten banks with the largest commercial real estate loans holdings (as defined by Barrens as of December 1990). The results are quite similar to those of MCBs.

References Aharony, Joseph, Anthony Saunders, and Itzhak Swary. (1986). “The Effects of a Shift in Monetary Policy Regime on the Profitability and Risk of Commercial Banks,” Journal of Monetary Economics 17, 493-506. Bernanke, Ben S. (1990). “On the Predictive Power of Interest Rates and Interest Rate Spreads,” New England Economic Review November, December, 5 l-68. Campbell, John Y. (1987). “Stock Returns and the Term Structure,” Journal ofFinancial Economics 18, 373-399. Campbell, John Y., and Yasushi Hamao. (1991). “Predictable Stock Returns in the United States and Japan: A Study of Long-Term Capital Market Integration,” Working Paper. Princeton University. Chan, KC., Patric Hendershott, and Anthony Sanders. (1990). “Risk and Return on Real Estate: Evidence from Equity REITs,” AREUEA Journal 18, 431-452. Fama, E., and K. French. (1988). “Dividend Yields and Expected Stock Returns,” Journal of Financial Economics 22, 3-25. Fama, E., and K. French. (1989). “Business Conditions and Expected Return on Stocks and Bonds,” Journal of Financial Economics 25, 23-49. Ferson, W. (1989). “Changes in Expected Security Returns, Risk and Level of Interest Rates,” Journal ofFinance 44, 1191-1217. Ferson, W. (1990). “Are the Latent Variables in Time-Varying Expected Returns Compensation for Consumption Risk?” Journal of Finance 45, 397-430. Ferson, W., and C. Harvey. (1990). “The Variation of Economic Risk Premia,” Journal of Political Economy, forthcoming. Ferson, Wayne, Shmuel Kandel, and Robert Stambaugh. (1987). “Test of Asset Pricing with Time-Varying Expected Risk Premia and Market Betas,” Journal ofFinance 42, 201-219. Flannery, Mark J. (1983). “Interest Rates and Bank Profitability: Additional Evidence,” Journal of Money, Credit and Banking 15, 355-362. Flannery, Mark J., and Christopher M. James. (1984). “The Effect of Interest Rate Changes on the Common Stock Returns of Financial Institutions,” Journal of Finance 39, 114 I- 1153. Friedman, Ben M., and Kenneth Kuttner. (1991). “Why Does the Paper-Bill Spread Predict Real Economic Activity?” Working Paper, Federal Reserve Bank of Chicago.

224


Harvey, Campbell R. (1989). “Time-Varying Conditional Covariances in Tests of Asset Pricing Models,” Journal of Financial Economics 24, 289-317. James, Christopher. (1989). “Empirical Evidence on Implicit Government Guarantees of Bank Foreign Loan Exposure,” Carnegie-Rochester Conference on Public Policy, 30, 129-162. Keim, D., and R. Stambaugh. (1986). “Predicting Returns in the Stock and Bond Markets,” Journal ofFinancial Economics 17, 357-390. Liu, Cracker, and Jianping Mei. (1992). “Predictability of Returns on Equity REITs and Their Co-movement with Other Assets,” Journal of Real Estate Finance and Economics, forthcoming. Mei, Jianping, and Cracker Liu. (1994). “Predictability of Real Estate Returns and Market Timing,” Journal of Real Estate Finance and Economics, forthcoming. Mei, Jianping, and Ahyee Lee. (1994). “Is There a Real Estate Factor Premium?” Journal of Real Estate Finance and Economics, forthcoming. Miles, M., B. Webb, and D. Guilkey. (1991). “On the Nature of Systematic Risk in Commercial Real Estate,” Graduate School of Business, Indiana University, Working Paper #467. Nourse, Hugh 0. (1987). “The ‘Cap Rate’ 1966-1984: A Test of the Impact of Income Tax Changes on Income Property,” Land Economics 63, 147- 152. O’Hara, M., and W. Shaw. (1990). “Deposit Insurance and Wealth Effects: The Value of Being ‘Too Big to Fail,’ ” Journal of Finance 45, 1587-1600. Roll, Richard, and Stephen A. Ross. (1980). “An Empirical Investigation of the Arbitrage Pricing Theory,” Journal of Finance 35, 1073-l 103. Ross, Stephen. (1976). “The Arbitrage Theory of Capital Asset Pricing,” Journal of Economic Theory 13, 341-360. Saunders, Anthony, and Thomas Urich. (1988). “The Effect of Shifts in Monetary Policy and Reserve Management Behavior in the Federal Funds Market,” Journal of Banking and Finance 12, 523-535. Sweeny, Richard, and Arthur Warga. (1986). “The Pricing of Interest Rate Risk: Evidence from the Stock Market,” Journal of Finance 41, 393-410. White, Halbert. (1980). “A Heteroskedasticcity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity,” Econometrica 48, 817-838.