Fiscal Policy in Open Economies

Transcription

1 Fiscal Policy in Open Economies Harris Dellas Klaus Neusser Manuel Wälti This draft: February 2005 PLEASE DO NOT QUOTE Abstract We study the effects of fiscal policy in a small, open economy. Under sluggish prices, the model s fiscal implications are quite similar to those obtained in the standard IS-LM model. In particular, the effectiveness of fiscal policy depends negatively on: (mostly) the degree of flexibility of the exchange rate, the degree of openness and the degree of capital mobility. Under flexible prices, the latter two elements play a similar but smaller quantitatively role and the exchange rate regime does not matter at all. We then investigate the effects of government spending shocks in a large set of countries using a VAR identification scheme suggested by Fatás and Mihov, We find that none of these three factors is systematically related to the the effects of spending shocks. We interpret these results as being more in line with the predictions of the RBC rather than the Keynesian model. JEL classification: E2, E3, C32 Keywords: government expenditures, small open economy, VAR, Department of Economics, University of Bern, CEPR, IMOP. Address: VWI, Gesellschaftsstrasse 49, CH 3012 Bern, Switzerland. Tel: (41) , Fax: (41) , harris.dellas@vwi.unibe.ch, Homepage: Department of Economics, University of Bern, CEPR, IMOP. Address: VWI, Gesellschaftsstrasse 49, CH 3012 Bern, Switzerland. Tel: (41) , Fax: (41) , klaus.nuesser@vwi.unibe.ch, Homepage: Swiss National Bank, International Research. Address: P.O. Box, CH-8022 Zurich, Switzerland, manuel.waelti@snb.ch. 1

2 1 Introduction The IS-LM model has well known implications for the effects of fiscal policy on economic activity. A fiscal expansion (whether in the form of a budget deficit or balanced budget) increases output, employment and consumption, while crowding out private investment through the resulting increase in real interest rates. Moreover, the total effect on output is a multiple of the initial change in government spending (the fiscal multiplier is greater than one). The effects of fiscal actions become more ambiguous once the links with the rest of the world are taken into account. Under a flexible exchange rate regime, a fiscal expansion at home, by raising real interest rates, leads to a domestic currency appreciation. Given nominal price sluggishness, this translates into a domestic real exchange rate appreciation and a loss in international competitiveness. The trade balance deteriorates, and this restraints the expansionary effects of the higher government spending. If the economy is sufficiently small and capital is internationally mobile then the resulting reduction in exports completely offsets the fiscal expansion. In other words, the fiscal multiplier becomes zero. On the other hand, if the exchange rate were fixed, monetary policy would have to loosen up to prevent the real interest rate increase and hence the currency appreciation, amplifying the effects of the initial fiscal expansion. The multiplier becomes greater than what it would have been in a closed economy. The leading alternative to the Keynesian model, the Real Business Cycle henceforth, RBC model also has predictions about the effects of fiscal policy, which do not coincide with those of the Keynesian model. While both models agree that the short term effects of an increase in government spending on output, employment and real interest rates are positive while that on investment is negative, there is no multiplier in the RBC model. Furthermore, the RBC model predicts a decline in private consumption (see Baxter and King [2] for an in depth discussion). The predicted effects of fiscal policy remain unambiguous even in the open economy version of the RBC model. The main difference from the closed economy version is that changes in the current account replace changes in the real interest rate. It must be emphasized also that, unlike the Keynesian model, there is no relationship between the exchange rate regime and the magnitude of the fiscal impact (due to money neutrality). Given the diverging theoretical predictions of different models, it is not surprising that several attempts have been made in the literature to use the empirical evidence to evaluate the relative validity of these competing models. It is noteworthy that most of the 2

3 modern work has concerned itself with the patterns obtained in closed economies 1. For instance, Galí, Lopez-Salido and Valles [6], and Fatás and Mihov [5] have compared the two models in terms of their implications for private consumption and real wages. The verdict seems rather ambiguous at this point as some studies document an increase in private consumption (Galí, Lopez-Salido and Valles [6], Fatás and Mihov [5], Perotti [10]) while others find no effect (Mountford and Uhlig [9]). The goal of our study is threefold. First, to establish the nature of the relationship between government spending and macroeconomic activity within the context of the new, open economy, Keynesian model. Remarkably, this is an issue that has not been investigated in the literature. In particular, we investigate whether the three factors mentioned above (exchange rate regime, degree of capital mobility and degree of openness) play the same role as in the IS-LM model. Second, to examine the role of capital mobility and openness for the relationship between government spending and macroeconomic activity under flexible prices (in the RBC) model. Again, this is an issue that remains unexplored in the literature. And third, to use information from a large set of countries in order to investigate how the size of the fiscal impact is related to these open economy characteristics 2. The findings can then be used to evaluate the relative performance of the Keynesian and the RBC models. We study 22 countries over the last two decades employing the VAR scheme suggested by Fatás and Mihov to identify government spending shocks. We compute the fiscal effects in individual countries over various time horizons and we relate cross country differences in these effects to variation in the exchange rate regime, the degree of capital mobility and degree of openness. Our key findings can be summarized as follows. First, at the theoretical front, we find that the fiscal implications of the open economy, new Keynesian model are quite similar to those obtained in the traditional IS-LM analysis. The only noteworthy exception regards the prediction that, under flexible exchange rates, consumption decreases following a positive fiscal shock. Second, the fiscal implications of the RBC model are similar to those of the Keynesian model with two important differences. First, the exchange rate regime in place does not matter for the fiscal effects. And second, both capital mobility and degree of openness play a smaller quantitative role than in the Keynesian model. This is due to the fact that price rigidities amplify the effects of shocks on quantities. Turning to the empirical evidence now, we find that the effect of a positive fiscal shock 1 There is a huge literature dealing with the effects of fiscal shocks in open economies. But this literature has not yet incorporated the recent advances in the time series methods for identifying fiscal shocks. We return to this issue later in the literature review. 2 Tests that rely on output rather than consumption seem a preferable strategy given the difficulty of measuring consumption of non-durables. 3

4 on economic activity is positive in the majority of countries but tends to be close to or smaller than unity. And that these effects are not systematically related to the exchange rate regime in place, to the degree of capital mobility and trade openness. While some of these findings present problems for both models, they seem to be more consistent with the flexible price rather than the rigid price model. The rest of the paper is organized as follows. Section 1 presents the model. Section 2 derives the implications of the model for macroeconomic activity as a function of the exchange rate regime, the degree of capital mobility and degree of openness under both flexible and sluggish prices. Section 3 presents the econometric methodology and section 4 the main results. A detailed list of the data used can be found in the appendix. 2 The model The economy under consideration is a small, open one. In this economy there are two types of firms. The first type produces final goods and the second intermediate goods. 2.1 Final sector firms Following standard practice in the literature we assume that the domestic final good y is produced by perfectly competitive domestic firms by combining domestic (x d ) and imported (x m ) intermediate goods. Final good production is described by the following CES function y t = (ω ) 1 1 ρ x dρ t + (1 ω) 1 1 ρ x m ρ 1 ρ t where ω (0, 1) and ρ (, 1). Minimization of total expenditures, P xt x d t + P mt x m t, where P xt and P mt denote the price of the domestic and the foreign bundle of goods in domestic currency results in the standard demand equations: 3 x d t = x m t = ( Pxt P t ( Pmt where P t is the domestic CPI, given by P t ( ρ ρ ρ 1 P t = ωpxt + (1 ω)p mt (1) ) 1 ρ 1 ωyt (2) ) 1 ρ 1 (1 ω)yt (3) ρ 1 ) ρ 1 ρ. (4) 3 Note that in the rest of the world, the demand for the domestic good is given by x d t = ω )yt. ( P xt e t P t ) 1 ρ 1 (1 4

5 We assume producer currency pricing and purchasing power parity for traded goods. That is, P mt = s t P t currency price. x d t where s t is the nominal exchange rate and a denotes foreign and x f t are themselves combinations of the domestic and foreign intermediate goods according to ( 1 x d t = 0 ) 1 x d t (i) θ θ di ( 1 and x m t = 0 ) 1 x m t (i) θ θ di where θ (, 1). Note that ρ determines the elasticity of substitution between the foreign and the domestic bundle of goods, while θ determines the elasticity of substitution between goods in the domestic and foreign bundles. (5) 2.2 The intermediate goods firms Each intermediate firm i [0, 1] produces an intermediate good x(i) using physical capital k(i) and labor h(i) according to a constant return-to-scale technology (α k, α h [0, 1], α k + α h = 1) represented by the production function x t (i) = A t k t (i) α k h t (i) α h (6) where A t is an exogenous stationary stochastic technological shock. Assuming that each firm i operates under perfect competition in the input markets, the production plan is determined by minimizing total cost, W t h t (i) + P t z t k t (i), where z t is the real rental of capital, and W t is the nominal wage, subject to the production function (6). The input demand functions are given by (dropping i) α k ψ t P t x t = P t z t k t (7) α h ψ t P t x t = W t h t (8) where the real marginal cost, ψ t is given by ψ t = zα k t (W t /P t ) α h A tς. Intermediate goods producers are monopolistically competitive. prices for the good they produce. Therefore, they set We introduce price stickiness by assuming that it is costly to change prices. In particular, firms face an adjustment cost when they change their prices relative to some benchmark rate of inflation. problem is given by max P xt (i) { E t n=0 D t,t+n Π xt+n (i) } Their profit maximization (9) where the discount factor is D t,t = 1 and D t,t+1 = β Λ t+1(j) Λ t(j) problem of the household. comes from the optimization 5

6 Profits are Π xt (i) = (P xt (i) P t ψ t )x t (i) ξ x 2 ( Pxt (i) P xt 1 (i) π x) 2 Pt y t. The last element represents the cost of changing prices relative to some benchmark rate of inflation, expressed in units of the final good. The first-order condition with regard to the choice of price, P xt (i), is θ θ 1 x t(i) P tψ t θ 1 +E t D t,t+1 P xt+1 (i) P xt (i) 2 ξ x 1 P xt (i) x t(i) ( Pxt+1 (i) P xt (i) ( ) 1 P xt 1 (i) ξ Pxt (i) x P xt 1 (i) π x P t y t π x ) P t+1 y t+1 = 0. (10) 2.3 The Household There exists a continuum of identical households. The preferences of the representative household are given by [ ] ν E t β τ ct c 1 σ c t+τ νh h 1+σ h t+τ (11) 1 σ c 1 + σ h τ=0 where 0 < β < 1 is a constant discount factor, c t denotes consumption and h t is the quantity of hours supplied by the household. ν h and ν ct are constant. In each period, the representative household faces a budget constraint Bt+1 d + s t Bt+1 f + M t + P t (1 + η(v t, ζ t ))c t + +P t (I t ) + T t ) + W t h t = R t 1 Bt d + Rt 1s f t Bt f +P t z t u t k t + W t h t + M t 1 + N t + Π t (12) where Bt d and Bt f are domestic and foreign currency bonds. The foreigners do not hold any domestic bonds so Bt d = 0. W t is the nominal wage; P t is the nominal price of the domestic final good; c t is consumption and I t is investment expenditure; k t is the amount of physical capital owned by the household and leased to the firms at the real rental rate z t. M t 1 is the amount of money that the household brings into period t, M t is the end of period t money and N t is a nominal lump sum transfer received from the monetary authority; T t is the lump-sum taxes paid to the government and used to finance government consumption. Π t denotes the profits distributed to the household by the firms. η(v t ; ζ t ) is a proportional transaction cost that depends on the household s money to nominal consumption ratio η(v t ; ζ) = ζ v t = P tc t M t ( Av t + B 2 ) AB v t 6

7 ζ t is a constant. The function η is borrowed from Schmitt-Grohe and Uribe [14]. ( η(v t, ζ) = ζ Av t + B 2 ) AB. (13) v t Capital accumulates according to k t+1 = I t ϕ 2 ( ) 2 It κ k t + (1 δ)k t (14) k t where δ [0, 1] denotes the rate of depreciation. κ > 0 is a constant. The capital adjustment costs are assumed to be zero in the steady state. The household then determines consumption/saving and money holdings maximizing (11) subject to (12) and (14). 2.4 Financial Markets The nominal interest rate on foreign liabilities carries a risk premium Rt f ( π = R t B f ) π ϱ t+1 P t (15) where ϱ(.) is a strictly increasing function in the aggregate level of real foreign debt. R t is the world nominal interest rate, and π and P are the foreign inflation rate and price level respectively. 2.5 Monetary policy We study two international monetary arrangements: A flexible system and a unilateral peg. In the latter case, the monetary authorities in the small open economy keep the nominal exchange rate vis a vis the rest of the world perfectly constant. Under a flexible exchange rate system, monetary policy is assumed to be conducted according to the money rule M t M t 1 M t 1 = constant, (16) We plan to study the implications of using instead a standard interest rate policy rule in the near future. 7

8 3 Calibration We are mostly interested in investigating the effects of fiscal policy for a generic rather than for a particular, real world economy. Hence, we use parameter values that are commonly used in the open economy literature. The benchmark parameters are reported in table 1. Table 1: Benchmark Parameter Values Transactions cost A Transactions cost B Money demand ζ 1E 8 Production α k Production α h Utility σ h Utility σ c Capital adjustment ϕ SS capital target i k κ Trade elasticity ρ Mark up θ Openness ω Discount factor β Preferences ν c Preferences ν h Depreciation δ Steady state inflation π 1.00 Working time h Risk premium ϱ Fiscal shock: persistence ρ g 0.90 Fiscal shock: volatility σ g 0.01 The fiscal shock is the only one in the model. It is assumed to follow an AR(1) processes with an autoregressive coefficient of 0.9. In the experiments run, the standard deviation of the shock has been set to The theoretical results The model is solved under flexible (ξ = 0) and sluggish prices after taking a first-order log approximation around the deterministic steady state. Table 2 reports the contemporaneous response (elasticity) of four macroeconomic variables, y, c, h, π to a fiscal shock under price rigidity. Figures 1 and 2 give the corre- 8

9 sponding IRFs in the benchmark case under a flexible and a fixed regime respectively. Finally, the left panel of Table 4 gives the fiscal effect as a function of the exchange rate regime over different time horizons. As can be seen the patterns obtained in the new Keynesian model are fairly consistent with those obtained in the traditional IS-LM analysis. That is, the effect of a positive fiscal shock on output is always greater under a fixed exchange rate system and it is decreasing in the degree of openness (the lower ω), in the degree of capital mobility (the lower prem) and slightly decreasing in the degree of substitutability between domestic and foreign goods (higher ρ). The main difference from the IS-LM model is that the response of consumption under a flexible exchange rate is, unlike the IS-LM model, negative. Table 2: Contemporaneous Response to a Fiscal Shock: Sluggish Prices Flexible Peg y c h π y c h π benchmark ω = prem = ρ = Note: output, y, consumption, c, employment, h, inflation, π. ω is the degree of closeness, prem is the risk premium on domestic external debt and ρ determins the elsticity of substitution between domestic and foreign intermediate goods. Table 3 reports the same information when prices are perfectly flexible. Figure 3 gives the corresponding IRFs in the benchmark case under a flexible regime (they are identical under a fixed regime). Finally, the right panel of Table 4 gives the fiscal effect as a function of the exchange rate regime over different time horizons. As can be seen the patterns obtained bear similarities to those obtained above under rigid prices. Namely, the effect of a positive fiscal shock on output is decreasing in the degree of openness (the lower ω) and in the degree of capital mobility (the lower prem) and slightly decreasing in the degree of substitutability between domestic and foreign goods (higher ρ). There are three main difference from the fixed price case. First, the exchange rate regime does not matter at all. Second, the variation in the fiscal impact arising from variation in openness and capital mobility is smaller. And third, the model always delivers a negative effect on consumption. To summarize: The new Keynesian model predicts significant variation in the effects of fiscal policy, in particular as a function of the exchange rate system. The flexible price model, on the other hand, predicts limited if any variation. With these predictions in mind we now turn to the empirical analysis. 9

10 Table 3: Contemporaneous Response to a Fiscal Shock: Flexible Prices Flexible Peg y c h π y c h π benchmark ω = prem = ρ = Note: output, y, consumption, c, employment, h, inflation, π. ω is the degree of closeness, prem is the risk premium on domestic external debt and ρ determins the elsticity of substitution between domestic and foreigh intermediate goods. Table 4: Fiscal Multipliers: Benchmark Case Fixed Prices Flexible Prices Flexible Exchange Rate periods Fixed Exchange Rate Note: The entries give the multiplier (cumulative effect) associated with government spending at 4, 8, 12 and 20 periods respectively. 10

11 5 Methodology 5.1 Review of the literature A key difficulty in the analysis of the effects of fiscal policy is the identification of fiscal shocks. For instance, some of government spending is due to automatic stabilizers and cannot be used to in the determination of the causal effect of spending on macroeconomic activity. Some other represents an adjustment to past tax changes. And so on. Two alternative methods have been proposed in the literature. 4 The first is called the narrative or event-study approach. It identifies historical events which have led to unusually large military build-ups and then asks whether this increased military spending stimulates aggregate demand, as it should from a Keynesian perspective. The event-study approach hinges upon a reliable identification of country-specific historical episodes. Moreover, it assumes that these episodes are entirely unanticipated and do not coincide with different types of fiscal shocks. Two prominent examples of this approach are Ramey and Shapiro [12] and Burnside et al. [4]. The so-called structural VAR approach, on the other hand, rests on one of two assumptions. Either that current government expenditures do not automatically react within the period to changes in economic conditions - an institutional feature of the budget process which seems to be shared by a wide range of countries and political systems. It follows that shocks to government spending can be taken as predetermined with respect to all other structural shocks in a VAR. Two prominent examples of the approach are Blanchard and Perotti [3] and Fatás and Mihov [5]. Or that fiscal shocks can be identified by using sign restrictions on the impulse responses and by imposing orthogonality to business cycle shocks and monetary policy shocks (Mountford and Uhlig [9]). advantage of the latter approach is that it overcomes the difficulty that changes in fiscal policy may manifest themselves in variables other than fiscal variables first and that fiscal variables may also respond automatically to business cycle conditions. The narrative and the structural VAR approach both lead for the U.S.A. to the same conclusions regarding output and employment: the two variables increase in response to a positive government spending shock. An They produce, however, conflicting evidence regarding the response of private consumption and real wage: while Ramey and Shapiro [12] and Burnside et al. [4] report that real wage and private consumption fall, 4 A third method proposed in the literature is the use of deficit measures (such as the full-employment deficit or a cyclically-adjusted fiscal balance) as an indicator of fiscal stance. This method, however, does not allow to make the discrimination between spending increases and tax cuts, which makes it less useful for the purpose at hand. 11

12 Blanchard and Perotti [3] and Fatás and Mihov [5] find that the same variables rise, while Mountford and Uhlig find no effect. 5.2 Econometric framework In this paper we follow the semi-structural approach proposed by Fatás and Mihov [5] and Blanchard and Perotti [3] to identify fiscal shocks. Consider for this purpose the multidimensional stochastic process {X t } which consists of X t = (G t, Y t, π t, T t, R t, T OT t ) where G t, Y t, π t, T t, R t, and T OT t denote the logarithm of real government consumption, the logarithm of real GDP, the inflation rate computed with respect to the GDP deflator, the logarithm of tax revenues deflated by the GDP deflator, the long term interest rate, and the logarithm of the terms of trade (defined as import prices over export prices), respectively. 5 This is, with the exception of the terms-of-trade, basically the set of variables proposed by Fatás and Mihov [5] and Blanchard and Perotti [3]. We assume that {X t } can be approximated by a VAR model of order p: X t = c 0 + c 1 t + Φ 1 X t Φ p X t p + Z t (17) where {Z t } WN(0, Σ) and where c 0 and c 1 denote vectors of constants. As in Fatás and Mihov we assume that the reduced form shock Z t is a linear combination of the structural shock V t = (v (g) t, v (y) t, v (π) t, v (T ) t, v (r) t, v (tot) t ) such that BZ t = CV t (18) where {V t } WN(0, Ω) and where B and C are 6 6 matrices. Assuming as usual that the structural shocks are contemporaneously uncorrelated, this implies that Ω is a diagonal matrix with the variances of the structural shocks on the diagonal. As we are only interested in the effect of shocks to real government consumption, v (g) t, we do not assign any interpretation to the other structural shocks nor do we make any attempt to identify them. The government consumption shock v (g) t is identified by restricting B to be equal to I 6 and by setting the diagonal elements of C equal to one. In addition, and most importantly, we assume that real government consumption does not react within the period to any other shock nor to any other variable. Although this assumption may criticized for being too simplistic and unrealistic, it has, compared to Blanchard and Perotti [3] and Perotti [10], the great advantage that it does not require the estimation of nuisance elasticities. Given the ordering of the variables in X t, the fiscal shock is then identified through the Cholesky decomposition of Σ. 5 For some countries we had to use a slightly different set of variables (see Appendix). 12

13 Although several arguments can be made in favor of a specification in growth rates, we stick to the level (log-level) specification for comparison purposes to the studies by Fatás and Mihov [5], Blanchard and Perotti [3] and Perotti [10]. 6 6 Results Table 5 reports the fiscal effects at period 1 (impact effect), and the cumulative effects up to periods 4 and 8. Two patterns stand out. First, while for the majority of countries these effects are positive, there are cases of a negative effect. The negative effects pose a theoretical problem as the standard versions of both the rigid and flexible price models predict a non-negative effect. And second, for most countries there is no multiplier, that is, the effects fall short of unity. Graphs 4a-c and Tables 6-8 relate the estimated fiscal effects over the various time horizons to the the exchange rate regime. One would expect two things if the Keynesian model were true: First a plot with a negative slope (the more flexible the exchange rate regime the smaller the fiscal impact). And second a line that becomes steeper as the horizon over which the fiscal effects are computed is increased. This is due to the fact that, due to lags involved in trade, one should expect that the fiscal effect would only gradually disappear over time under a flexible exchange rate system. Neither of these patterns is observed in the data. There is no systematic relationship between the size of the fiscal effects and the degree of exchange rate flexibility. And there is no systematic variation as a function of the time horizon. This finding constitutes prima facie evidence against the Keynesian model and in favor of the RBC model. Similarly insignificant results obtain regarding the relationship between the other two open economy characteristics discussed in the theoretical part of the paper (the degree of trade openness and the degree of international capital mobility) and the size of the fiscal effects (also compare Tables 6-8). Moreover, while qualitatively the effect of capital mobility is consistent with the prediction of the theories (the more restrictions on the capital account, the higher the fiscal multiplier), that of trade openness is not. Again based on Table 4 one can claim that this finding is more consistent with the flexible than with the the fixed price model. 6 See also the discussion in Ashley and Verbrugge [1], and, with respect to the confidence intervals of impulse response functions, in Kilian [7] and Pesavento and Rossi [11]). 13

14 Table 5: Fiscal Multipliers at Lag 1, 4, and 8 Country Lag Australia Belgium Canada Denmark Finland France Greece Hong Kong Italy Japan Mexico Netherland New Zealand Norway Portugal South Africa Spain Sweden Switzerland Turkey UK USA Note: The impulse responses to Cholesky (d.f. adjusted) one S.D. innovations are accumulated up to lag 1, 4, and 8, respectively. To make the findings comparable across countries, the multipliers are divided by the estimated country-specific standard deviation of the fiscal shock. Model specification: X t = (G t, Y t, π t, T t, R t, T OT t) (see Section 5 and Appendix), with lag length 2. Frequency and time period: quarterly, 1988:1-2004:3. 14

15 7 Summary and Conclusions We have derived the implications of the standard open economy models with flexible and rigid prices for the effects of fiscal policy. The former model (the new Keynesian) predicts that the macroeconomic effects of government spending shocks are greater, the lower the flexibility of the exchange rate and the smaller the degree of international capital mobility and the trade openness. The flexible price model agrees with the latter two predictions but implies a weaker quantitatively relationship and moreover, it leaves no room for the exchange rate system to matter for fiscal policy. We have used data from a large number of countries to investigate the existence of these patterns. Our main finding is that the size of the fiscal effect on economic activity does not seem to vary with these key open economy characteristics. While the evidence must be viewed with some caution as the estimated fiscal effects on output are negative for some countries, a feature that is not captured by either of these two models (at least in their standard versions) our interpretation is that the evidence seems to slightly favor the flexible price model. A fruitful path for future research could be the repetition of the analysis using alternative government spending identification schemes. This could serve to check the robustness of our findings. Moreover, the development of models that can account for negative fiscal effects should also be a high priority for future research. Table 6: Explaining the Fiscal Multiplier at Lag 1 Regressor Coefficient Exchange rate regime (0.007) (0.010) Capital mobility (0.126) (0.150) Openness to trade (0.001) (0.001) Note: The entries give the slope of the robust regression line for the fiscal multiplier at lag 1, with standard errors in parenthesis. The following proxies are used: Reinhart and Rogoff s [13] exchange rate regime classification (the higher the value, the more flexible the exchange rate), Miniane s [8] measures on capital account restrictions (the higher the value, the more restrictions are in place), and the variable openness in current prices from the Penn World Table (the higher the value, the higher the degree of openness). For details compare the Appendix. The first three columns contain the slope coefficients for the individual bivariate regressions, the last column those for the multiple regression model. 15

16 Table 7: Explaining the Fiscal Multiplier at Lag 4 Regressor Coefficient Exchange rate regime (0.036) (0.051) Capital mobility (0.644) (0.746) Openness to trade (0.003) (0.004) Note: The entries give the slope of the robust regression line for the fiscal multiplier at lag 4 (cumulated response), with standard errors in parenthesis. Variables are defined in notes to Table 6. The first three columns contain the slope coefficients for the individual bivariate regressions, the last column those for the multiple regression model. Table 8: Explaining the Fiscal Multiplier at Lag 8 Regressor Coefficient Exchange rate regime (0.090) (0.122) Capital mobility (1.624) (1.795) Openness to trade (0.006) (0.008) Note: The entries give the slope of the robust regression line for the fiscal multiplier at lag 8 (cumulated response), with standard errors in parenthesis. Variables are defined in notes to Table 6. The first three columns contain the slope coefficients for the individual bivariate regressions, the last column those for the multiple regression model. 16

17 A Data The empirical part of our study involves the 22 OECD and Non-OECD countries listed in Table 5. We use two categories of data. The first one is utilized to estimate the fiscal multipliers at various lags. The following quarterly series over the period 1988:1-2004:3 are taken from either the Main Economics Indicators (MEI) or the International Financial Statistics (IFS) database: real GDP, implicit GDP deflator, real government consumption, nominal tax revenue, long-term interest rate, import price and export price deflator. If the variable tax revenue is contained in neither the MEI nor the IFS database, we use the series government revenue instead, taken from the Oxford Economic Forecasting (OEF) database. And when the import price and export price deflators are missing, then we proxy the terms of trade by the real effective exchange rate. Most series are seasonally adjusted by the original source; if not, we apply the Census X12 filter. The second category of data are used to get a measure for the three open economy characteristics: exchange rate regime in place, degree of capital mobility, and openness to trade. To classify a country s exchange rate regime, we utilize the Reinhart and Rogoff [13] classification. 7 From their annual classification we compute for each country the arithmetic mean over the period 1988 to To measure a country s capital account restrictions we use the dummy provided by Miniane [8], which is based on information from the IMF s Annual Report on Exchange Arrangements and Exchange Restrictions. 8 The proxy openness to trade, finally, is taken from the Penn World Table (the sum of nominal imports and nominal exports over nominal GDP). Miniane s measure and the openness to trade proxy are both at an annual frequency. For both series we compute the arithmetic mean over 1988 to the year of the latest available observation (2001 in the case of Miniane, 2000 in the case of the Penn World Table). References [1] R. A. Ashley and R. J. Verbrugge, To Difference or Not To Difference: A Monte Carlo Investigation of Inference in Vector Autoregressive Models. VPI Economics Department Working Paper E99-15, [2] Baxter and King, Fiscal policy in general equilibrium, AER, 83 (1993), pp The Reinhart and Rogoff classification allows for 14 categories of exchange rate regimes, ranging from no separate legal tender to a hyperfloat. 8 Miniane s dummy ranges from 0 to 1, where 0 means no restrictions. 17

18 [3] O. Blanchard and R. Perotti, An Empirical Characterization of the Dynamic Effects of Changes in Government Spending and Taxes on Output, Quarterly Journal of Economics, 117 (2002), pp [4] M. E. Burnside, Craig and J. D. Fisher, Assessing the Effects of Fiscal Shocks. NBER Working Paper 7459, [5] A. Fatás and I. Mihov, The Effects of Fiscal Policy on Consumption and Employment: Theory and Evidence. Mimeo, [6] J. D. L.-S. Gali, Jordi and J. Valles, Understanding the Effects of Government Spending on Consumption. ECB Working Paper Series No. 339 (April), [7] L. Kilian, Small-Sample Confidence Intervals for Impulse Response Functions, Review of Economics and Statistics, 80 (1998), pp [8] J. Miniane, A New Set of Measures on Capital Account Restrictions. IMF Staff Papers Volume 51, Number 2, [9] A. Mountford and H. Uhlig, What are the Effects of Fiscal Policy Shocks? CEPR, [10] R. Perotti, Estimating the Effects of Fiscal Policy in OECD Countries. CEPR Discussion Paper no. 4842, [11] E. Pesavento and B. Rossi, Small Sample Confidence Intervals for Multivariate Impulse Response Functions at Long Horizons. Mimeo, [12] V. A. Ramey and M. D. Shapiro, Costly Capital Reallocation and the Effects of Government Spending, Carnegie-Rochester Coference Series on Public Policy, 48 (1998), pp [13] C. M. Reinhart and K. S. Rogoff, The Modern History of Exchange Rate Arrangements: A Reinterpretation, Quarterly Journal of Economics, 119 (2004), pp [14] S. Schmnitt-Grohé and M. Uribe, Optimal Fiscal and Monatery Policy Under Sticky Prices, Journal of Economic Theory, 114 (2004), pp

19 Figure 1: Impulse Response Functions: Rigid Prices, Flexible Exchange Rate 19

20 Figure 2: Impulse Response Functions: Rigid Prices, Fixed Exchange Rate 20

21 Figure 3: Impulse Response Functions: Flexible Prices, Flexible Exchange Rate 21

22 Figure 4a: Fiscal Multiplier at Lag 1 vs. Exchange Rate Regime Figure 4b: Fiscal Multiplier at Lag 4 vs. Exchange Rate Regime 22

23 Figure 4c: Fiscal Multiplier at Lag 8 vs. Exchange Rate Regime 23