Fiscal Stabilisation Policy in a New Keynesian

0 downloads 0 Views 283KB Size Report
Page 1 ... More recently, some attention has also been given to the issue of monetary& .... conventional Keynesian effects, in that an increase in government spending causes a ... Smets and Wouters (2002) we assume that habits depend on past aggregate .... lapses to the purely forward looking IS curve (see Carroll, 2000).
Fiscal Stabilisation Policy in a New Keynesian Dynamic General Equilibrium Model with Liquidity Constraints: Evidence from the Euro-area V. Anton Muscatelli University of Glasgow and CESifo, Munich Tiziano Ropele Università Milano-Bicocca

Patrizio Tirelli Università Milano-Bicocca

First Draft: 15 October 2004 This Version: 29 April 2005

Abstract We show that …scal policy, as automatic stabilisers, can play a useful stabilisation role when monetary policy is controlled by an independent central bank. We use a New Keynesian DGE model with liquidity constrained consumers and sticky prices, estimated using Euro area data. It allows for various transmission channels for …scal policy. We extend our model to include consumers with …nite horizons, and in a two-country version analyse the interaction of monetary and …scal policies in a monetary union. We demonstrate that, surprisingly, the potential stabilisation role of …scal automatic stabilisers is reduced in the case of certain types of shocks. JEL Codes: E58, E62, E63

1

Introduction

There is a large and growing literature on the optimal design of monetary policy in the context of New Keynesian dynamic general equilibrium models. More recently, some attention has also been given to the issue of monetary…scal policy interactions1 . However, most of these models are calibrated rather than estimated on empirical data2 . In this paper we estimate a small econometric model for the Euro area over the sample period 1970-1998, and then consider the stabilization performance of …scal stabilizers in the presence of a monetary policy rule operated by an independent central bank. While the structural model used in this paper has many elements in common with other New Keynesian dynamic stochastic general equilibrium (DSGE) models, our analysis di¤ers in a number of respects. First, we extend some current DSGE models to include a wider range of …scal policy transmission channels, including "non-ricardian".e¤ects on consumption due to rule-of-thumb (RT henceforth) consumers (as in Galì et al. 2002; Amato and Laubach, 2003) and supply-side distortions. Second, our model is estimated, in contrast to some attempts to calibrate or numerically simulate these models.(Andres and Domenech, 2003) Third, we examine a basic two-country version of our Euro-area model, showing how automatic …scal stabilisers interact with monetary policy in a monetary union. The key question which we ask of this model is whether …scal policy, through automatic stabilisers (feedback rules on output) adds value to the stabilization role played by monetary policy in the Euro area. In this regard, our simulations investigate the role of …scal policies conditional to our estimate of the proportion of RT consumers. Earlier contributions (Andres and Domenech, 2003; Gordon and Leeper, 2003; Muscatelli et al., 2003a) found that countercyclical …scal policy can be welfare-reducing in the presence of optimizing consumers. Muscatelli et al. (2003b) estimate the proportion of RT consumers in the US and show that automatic stabilizers based on taxation improve the performance of the economy. Here we …nd that the proportion of RT consumers is far larger in the Euro area, possibly due to …nancial market imperfections. This result is in line with a relatively large body of empirical evidence. Asdrubali and Kim (2004) …nd that, following an out1

See for example Leith and Wren-Lewis (2000), Schmitt-Grohé and Uribe (2001), Benigno and Woodford (2003) for an analysis of …scal and monetary interactions in theoretical models. 2 With the exception of Muscatelli et al. (2003a,b)

1

put shock, EU capital markets enable a very limited degree of consumption smoothing relative to the US. Fair (2001) …nds that, unlike the US, in most EU countries there is little evidence of real interest rate e¤ects on aggregate consumption. This is easily reconciled with our result about the large share of RT consumers who do not directly react to interest rate shocks. Finally, our estimate is consistent with the cross country evidence about the share of current income consumers presented in Sarantis and Stewart (2003). Then we use our estimates to simulate a two-country version of the model with a single central bank and independent …scal authorities. We …nd that the stabilising e¤ects of …scal policy at the national level are strictly dependent on the existence of a home bias in the composition of national consumption bundles. Thus our results suggest a novel approach to the philosophy of EMU macroeconomic policymaking. At the Euroarea level, the action of automatic stabilisers should be regarded as a useful complement to the ECB actions. On the other hand the usual case for …scal stabilisation of withinEMU asymmetries is con…rmed only if the composition of national aggregate demand functions are su¢ ciently biased towards domestic production. This is in sharp contrast with the "Brussels consensus" based on the view that the ECB alone should stabilize the union-wide economy (Buti et al., 2001) In the next section we brie‡y survey the existing literature on …scalmonetary policy interactions in New Keynesian DGE models. In Section 3, we outline the structure of our estimated model and the empirical methodology. In Section 4, we report our econometric results and illustrate some of the dynamic simulation properties of our model. In Section 5 we illustrate interaction of the monetary and …scal policy rules, and examine the welfare e¤ects of including …scal policy. In section 6 we present the two country monetary union model, and consider the role of …scal policy in this context. Section 7 concludes.

2

Our contribution to the analysis of …scal stabilization Policy in New Keynesian DSGE Models

Early vintages of New Keynesian DSGE models involved a limited role for …scal policy, by assuming lump-sum and representative agents with in…nite planning horizon. Another strand of literature has modelled more complex 2

supply-side e¤ects for …scal stabilization policies by allowing taxation distortions (see Andres and Domenech, 2003 and the references therein). Studies of the business cycle using VAR-type models do not provide empirical support for this simple description of demand-side e¤ects in the New Keynesian model. Giavazzi et al. (2000) show that both Keynesian and neoclassical (Ricardian) e¤ects are present. Fatas and Mihov (2001), Blanchard and Perotti (2002) and Muscatelli et al. (2004) show that …scal shocks have conventional Keynesian e¤ects, in that an increase in government spending causes a persistent rise in output3 and consumption. Galí et al. (2002) demonstrate that this problem can be addressed by adding non-optimising behavior to the conventional New Keynesian model. They assume that a proportion of consumers are constrained to consume out of current income and show that, under plausible parameterizations, this provides an explanation for the positive response of consumption to a temporary government spending shock. The increase in government spending generates an increase in the real wage (providing the substitution e¤ect between consumption and leisure dominates the wealth e¤ect), and causes an increase in aggregate consumption because ’RT’consumers spend out of current income.4 Introducing nonoptimising consumers also potentially allows for other transmission channels for …scal policy. Even if taxes are lump sum, they will impact on aggregate consumption behaviour through their e¤ect on the current nominal income of RT consumers. Furthermore, payroll taxes a¤ect marginal costs and in‡ation. This, in turn, has an obvious impact on aggregate consumption if wages are sticky. There are other ways, however, to model non-ricardian consumers. Debt…nanced …scal de…cits will have an impact on aggregate demand in versions of the New Keynesian model which depart from Ricardian equivalence because of the presence of …nite horizons, as in the classic Blanchard-Yaari model (Leith and Wren-Lewis, 2000). Other e¤ects of government debt on consumer behavior can also be considered, such as the impact that …nancial wealth has on household transactions costs, which can explain the observed positive correlation between public expenditure shocks and consumption (see Schabert, 2004). Unfortunately there is a trade-o¤ in estimating New Keynesian models 3

The implied …scal multiplier is close to or greater than unity. Whether consumption actually increases in such models depends crucially on the assumptions made about labour supply and price-stickiness, given the linkage between consumption and leisure (and hence the real wage) via the consumer’s optimization problem. 4

3

which contain all these channels …scal policy transmission and maintaining the model su¢ ciently parsimonious in terms of structural parameters to allow classical econometric techniques to be employed. A complex model will typically be underidenti…ed with respect to the structural parameters of interest or will result in poorly-de…ned (in a statistical sense) estimated parameters. In modelling the transmission of …scal policies, we have therefore chosen to restrict the demand side to the inclusion of non-optimising (RT) consumers. On the supply side, we allow for taxation e¤ects on …rms’ marginal costs (through payroll taxes). This, in contrast with early attempts to estimate structural New Keynesian models5 , allows for a richer range of transmission channels for …scal policy, whilst still maintaining a model where the structural parameters are obtained using econometrics.

3

The Model

We develop a small forward-looking New Keynesian DSGE model, comprising a dynamic IS equation for output and a “New Keynesian Phillips Curve” speci…cation for in‡ation. In addition to the channels for …scal policy transmission outlined above, we allow for habit formation in consumption, which is necessary to capture the observed persistence in the output-gap response to shocks.(Leith and Malley, 2002). We model a hybrid Phillips curve, allowing for partly backward-loking in‡ation expectations (Galí, Gertler and D. López-Salido, 2001, 2003).

3.1

Households

We follow Gali et al. (2002) in assuming two types of households: those in the …rst group, denoted by i, bene…t from full access to the capital markets and as such are free to optimize. The proportion of optimising consumers in the economy is given by (1 #): Each optimizing consumer is assumed to maximize an intertemporal utility function given by: Et

1 X s=0

5

1

s

1

oi i =Ht+s )1 (Ct+s

"lt+s (N oi )1+' 1 + ' t+s

See Gali et al. (2001), Leith and Malley (2002), Smets and Wouters (2002).

4

(1)

where Cto represents consumption of a basket of goods (to be de…ned below), Ht is an index of external habits and Nto is the level of employment. 2 (0; 1) represents the subjective rate of time preference, the coe¢ cient of relative risk aversion, ' the inverse of the elasticity of labour supply with respect to real wage and "lt is a shock to labour supply. Finally, Et denotes the expectation operator conditional on the time t information set. Following Smets and Wouters (2002) we assume that habits depend on past aggregate consumption, C T : T i = Ct+s Ht+s

1

.

(2)

Optimizing consumers maximize (1) subject to their intertemporal budget constraint. This is expressed in real terms as: (1=rt )ait+1 = ait

Ctoi +

Wt oi N + Dti + GTt Ri Pt t

Tti

(3)

Their …nancial wealth (at ) is held in the form of one-period state-contingent securities, which yield a return of rt . The optimizing consumer’s disposable income is de…ned as consisting of labour income wt Ntoi plus the dividends from the pro…ts of the imperfectly competitive …rms Dti , plus public transfers GTt Ri minus personal taxes Tti , which are lump-sum by assumption. The second group of households, a proportion # of the total, follows a rule of thumb behaviour: these households consume out of current disposable income, and supply a constant amount of labour6 , N RT . This admittedly ad hoc assumption may be justi…ed assuming myopia or limited participation to capital markets. Thus, the consumption function of the representative rule-of-thumb consumer is: Wt + GTt Rj Ttj . (4) Pt Consumers spread their total consumption over a set of imperfectly substitutable consumption goods, indexed by z 2 [0; 1] :Their total consumption is given by a standard CES function: CtRT j = N RT

6

Galì et al. (2002) show that supplying a constant amount of labour is optimal when net taxes, GTt R Tt , levied on rule-of-thumb consumers are always nil. This result would never obtain in our model, where taxes and transfers are explicitly modeled. Thus, for sake of simplicity we assume a constant labour supply. Since consumption cannot be negative, this implies that we impose a lower bound on GTt R Tt for any given level of the real wage.

5

21 Z i Ct = 4 Cti (z)

1

0

3

1

dz 5

(5)

where the constant elasticity of substitution, , between di¤erentiated goods is assumed greater than one. Solving the intratemporal optimal allocation across each variety of the consumption goods leads to the following demand for each good z: Pt (z) Cti (6) Cti (z) = Pt where Pt (z) is the price of good z, and Pt is the consumption price index given by the aggregator: 21 Z Pt = 4 (Pt (z))1 0

3.2

311

dz 5

.

(7)

Firms

In the model economy there is a continuum of …rms, producing the varieties of goods, and they act as monopolistic competitors. We assume a …xed capital stock, and …rms’output of each good variety z follows a Cobb-Douglas technology which depends on employment, N and a technology parameter At : Yt (z) = At [Nt (z)]1 (8) We introduce a channel for …scal policy by assuming that taxes on labour take the form of a uniform payroll tax7 . Firms’demand for labour is therefore de…ned as: (1

) At [Nt (z)]

=

Wt + tPt R Pt

where tPt R is the tax rate per unit of employed labour, i.e. tPt R = TtP R are the total revenues from the payroll tax. 7

(9) TtP R , Nt

where

This implies that the optimizing consumer’s choice between leisure and consumption is not a¤ected.

6

We turn next to …rms’pricing behavior. We consider a standard model of monopolistic competition with sticky prices, as set out in Galí, Gertler and López-Salido (2001), and Leith and Malley (2002)8 . More precisely, sticky prices are incorporated into this model by assuming a Calvo pricing mechanism, whereby only a given proportion of …rms, de…ned as (1 ), can adjust prices every period whereas the remainder supplies output on demand at a constant price. A share of the adjusting …rms is assumed to index prices to in‡ation in the previous period9 , while the remaininder, (1 ), set 10 their prices optimally to maximize expected discounted real pro…ts , with a discount factor .

3.3

The IS and the Phillips curve

By log-linearizing the model around the deterministic steady state we are then able to derive a hybrid dynamic equation for output and the New Keynesian Phillips curve (see the Appendix for a proof)11 . In what follows “hatted”lower-case variables denote percentage deviations from the steady state which, in turn, is denoted by “barred”variables. The log linearized equation for output is: 8