Economic Modelling 34 (2013) 15–24
Contents lists available at ScienceDirect
Economic Modelling journal homepage: www.elsevier.com/locate/ecmod
How best to measure discretionary fiscal policy? Assessing its impact on private spending Luca Agnello a, b,⁎, Davide Furceri b, c, Ricardo M. Sousa d, e a
Banque de France, Service d'Etude des Politiques de Finances Publiques (FIPU), 31 Rue Croix des Petits Champs, 75001 Paris, France University of Palermo, Department of Economics, Business and Finance, Viale delle Scienze, 90128 Palermo, Italy International Monetary Fund, 700 19th Street NW, 20431 Washington DC, USA d University of Minho, Department of Economics and Economic Policies Research Unit (NIPE), Campus of Gualtar, 4710-057 - Braga, Portugal e London School of Economics, Department of Economics and Financial Markets Group (FMG), Houghton Street, London WC2 2AE, United Kingdom b c
a r t i c l e
i n f o
JEL classification: E0 E6 Keywords: Discretionary fiscal policy Private spending Crowding-in Crowding-out effects
a b s t r a c t We develop a novel empirical approach to assess the effect of discretionary fiscal policy on private spending consisting of three stages: 1) extract the discretionary component of fiscal policy by estimating a fiscal policy rule; 2) use the residuals of the first-stage regression to investigate the existence of crowding-in and/or crowding-out effects both in the short and the medium term; and 3) condition the response of private spending on a set of country characteristics. We find that an expansion in discretionary fiscal policy boosts growth in the short term, but is detrimental in the medium term. In addition, the empirical findings suggest that the effect of discretionary fiscal policy on private spending varies across regions and income groups, and depends on countries' economic characteristics such as the level of economic development, trade openness, government and country size. © 2012 Elsevier B.V. All rights reserved.
1. Introduction To counteract the impact of the Great Recession, governments have implemented significant fiscal stimulus packages. While before the crisis, there was broad consensus that fiscal policy should play little role beyond allowing automatic stabilizers to operate, during the crisis almost all major countries adopted discretionary fiscal policy measures. In the debate on the fiscal policy response to the economic downturn, the effectiveness of fiscal policy to support the economy has regained importance and renewed the interest on the topic among researchers (Agnello and Sousa, 2011; Agnello et al., 2012; Baldacci and Kumar, 2010; Baldacci et al., 2009). In the last years, the theoretical and empirical literature has provided extensive analysis on the effect of government spending on economic activity. Despite this, there is no consensus on the effects of government spending on private consumption and investment (both in the short and in the medium term) neither from a theoretical nor from an empirical point of view. Indeed, recent economic developments suggest that fiscal stimulus can lead to business cycle desynchronization (Mallick
⁎ Corresponding author at: University of Palermo, Department of Economics, Business and Finance, Viale delle Scienze, 90128 Palermo, Italy. E-mail addresses:
[email protected],
[email protected] (L. Agnello),
[email protected],
[email protected] (D. Furceri),
[email protected],
[email protected] (R.M. Sousa). 0264-9993/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.econmod.2012.10.020
and Mohsin, 2010; Rafiq and Mallick, 2008) and affect the relationship between monetary and financial stability (Granville and Mallick, 2009; Sousa, 2010, forthcoming). Moreover, from a theoretical perspective, the effect of an increase of government spending on private consumption and investment can be of both signs. On the one hand, Real Business Cycle (RBC) type-models predict a decline in private consumption and an increase in private investment in response to a rise in government spending (Aiyagari et al., 1990; Baxter and King, 1993; Christiano and Eichenbaum, 1992). On the other hand, Keynesian type-models predict that consumption should rise and investment should decline in response to a positive government spending shock (Blanchard, 2003). From this discussion it emerges that the predictions of the abovementioned theories are orthogonal to each other. These contrasting views gave rise to several empirical studies attempting to assess the impact of public expenditure on private consumption and private investment. Unfortunately, the existing empirical evidence is also quite mixed in support of one theory or the other (Furceri and Sousa, 2011a, 2011b). Similarly, although there is a consensual view on the need to withdraw such fiscal stimulus as the economic recovery takes place, there is also substantial uncertainty about the macroeconomic impact of fiscal retrenchments (Alesina and Ardagna, 2010; Giavazzi and Pagano, 1990). Moreover, while most of the previous studies focused on singlecountry evidence (typically, the United States) and/or a group of advanced economies, fewer studies have analyzed the heterogeneity of the response of private consumption and investment across countries and regions. Indeed, recent work shows that the correlation between
16
L. Agnello et al. / Economic Modelling 34 (2013) 15–24
private consumption (investment) growth and government spending growth typically varies between advanced and developing economies (Furceri and Sousa, 2011a, 2011b). In addition, the existing research has also suggested that the effect of government spending on private demand may differ across countries depending on the level of economic development, trade openness, initial fiscal position and size (Giavazzi et al., 2000; Ilzetki et al., 2010). The main goal of this work is to shed some light on the existence of crowding-in or crowding-out effects of government spending shocks for private consumption and investment using a new methodology to identify discretionary fiscal policy. For this purpose, we consider an empirical framework consisting of three steps. First, we extract, for each country, the discretionary component of fiscal policy by estimating a fiscal policy rule and, thereby, quantifying the unexpected variation in fiscal policy. The rule is very similar to the structure of fiscal policy Vector Auto-Regressive (VAR) approaches, making our method comparable to most empirical studies (Afonso et al., 2010; Fatás and Mihov, 2003, 2006). Second, we use the residuals of the first-stage regression to evaluate, in a panel framework, the impact of discretionary fiscal policy on private spending (i.e. private consumption and investment growth). This provides an assessment of the crowding-in and the crowding-out effects associated with fiscal policy and a comparison between their short-term and medium-term effects. Finally, we condition the response of private consumption and private investment to government spending shocks on a set of country characteristics such as, economic development, region, income level, trade openness, government size and country size. Relying on the full panel of countries and data for the period 1960– 2008, the first stage regression shows that government spending responds, on average, in a “procyclical” manner with respect to fluctuations in the economic activity (Alesina et al., 2008; Ilzetki and Vegh, 2008). This result is particularly relevant for developing countries and gives support to the role of imperfections in credit markets preventing borrowing in bad times (Gavin and Perotti, 1997) or rent-seeking behavior in good times (Tornell and Lane, 1999). In addition, the empirical findings corroborate previous evidence suggesting that government spending negatively responds to the initial level of debt, thus contributing to the public debt sustainability. The results of the second-step regression highlight that discretionary changes in government spending lead to: (i) crowding-in effects in the short-run; and (ii) crowding-out effects in the medium-run. Such effects are particularly amplified for the private investment component. However, at the aggregate demand level, the strength of the crowding-out impact is generally not enough to completely annihilate the expansionary effect of the discretionary component of government spending. Despite this, there are important differences across countries and regions, in particular, when we control for countries' characteristics. Indeed, the results of the third-step analysis suggest that that the crowding-in effects mildly persist in the medium term and are particularly large for OECD countries, while the medium-run crowding-out effects seem to be more damaging for non-OECD countries and partially offset the short-lived crowding-in effects. Turning to the effects of fiscal shock at regional levels, we find that the crowding-out effects over the medium-run impinge more seriously on European countries in line with the current macroeconomic developments observed for this group of countries. We also find that the magnitude of the initial crowding-in effects and the medium-run crowding-out effects is larger for countries with higher income level and is inversely related with the degree of trade openness. Finally, medium-run crowding-out effects are in larger countries with smaller government size and in smaller countries, which are less able to smooth out unexpected variation on public consumption. The rest of the paper is as follows. Section two provides a description of the empirical methodology, whereby we measure the discretionary
component of fiscal policy and analyze its macroeconomic effects. Section three discusses the evidence of the unconditional impact of discretionary government spending on output, private consumption and private investment growth. Section four investigates the potential heterogeneity of the response of private spending to discretionary fiscal policy. Finally, in Section five, we conclude. 2. Empirical methodology Most of the existing studies on the effect of fiscal policy shock has relied on VAR models. Following Blanchard and Perotti (2002), several studies have identified government spending shocks based on the assumption that government consumption does not simultaneously react to contemporaneous change in economic activity.1 An alternative approach in identifying government spending shocks in a VAR model is the one proposed by Mountford and Uhlig (2009), who consider an identification scheme based on sign restrictions. An alternative empirical strategy has been designed by Ramey and Shapiro (1998), who consider a number of narrative events to detect fiscal policy actions. 2 The abovementioned frameworks, however, are only feasible for a restricted number of countries for which quarterly data and fiscal policy events are available. In addition, the linear structure of standard VARs prevents the analysis of the heterogeneity of the response of private consumption and private investment to government spending shocks. Ilzetki et al. (2010) try to address this problem using quarterly data for subgroups of countries distinguished by macroeconomic characteristics (such as exchange rate regime, public debt or trade openness). However, there are also limitations associated with this methodology, as the classification of sub-groups is somewhat arbitrary, and it does not account for time-variation in countries' characteristics. To address these constraints, we use a three-step approach that is described in the next sub-sections. 2.1. Discretionary fiscal policy We start by following Fatás and Mihov (2003, 2006) and Afonso et al. (2010), to obtain the discretionary component of government spending by estimating, for each country i (with i=1,…,N), the following fiscal policy rule: Δg i;t ¼ θi þ λi Δg i;t−1 þ βi Δyi;t þ Γ i X i;t þ εi;t ;
ð1Þ
where g is the logarithm of real government consumption, y is the logarithm of real GDP, and X is a set of controls including inflation, inflation squared, a time trend and the logarithm of real public debt. The country-specific coefficients λi, βi and εi,t in specification (1) represent the measures of persistence, responsiveness and discretion of fiscal policy, respectively. We use the discretionary component of government spending as a measure that is not related to the business cycle and, consequently, reflects unexpected variation in fiscal policy. We also account for the endogeneity of output, by regressing Eq. (1) with an Instrumental Variables – Two-Stage Least Squares (IV-2SLS) estimator. 2.2. The macroeconomic effect In this stage, we evaluate the macroeconomic effect of discretionary fiscal policy. More specifically, we use the residuals of regression (1) obtained for each country i in the dynamic panel equations for GDP
1 This assumption has been criticized as, for instance, social expenditure indeed simultaneously reacts to the economic cycle (Darby and Melitz, 2008; Furceri, 2010). 2 See Burnside et al. (2004), Devries et al. (2011) and Ramey (2011).
L. Agnello et al. / Economic Modelling 34 (2013) 15–24
growth (Δyi,t), private consumption growth (Δci,t) and private investment growth (Δii,t), that is Δyi;t ¼ κ i þ ρΔyi;t−1 þ
J X
ϑj ε^ i;t−j þ ηΖ i;t þ ξi;t
ð2Þ
ϑj ε^i;t−j þ ηΖ i;t þ ξi;t
ð3Þ
j¼1
Δci;t ¼ κ i þ ρΔci;t−1 þ
J X j¼1
Δii;t ¼ κ i þ ρΔii;t−1 þ
J X
ϑj ε^i;t−j þ ηΖ i;t þ ξi;t
17
• Real government spending, proxied by public consumption which is provided by the WDI; • Population, used to compute per capita terms of the abovementioned variables and provided by the WDI; • GDP deflator, used to convert nominal in real constant terms and provided by the WDI; and • Public debt, provided by the Historical Public Debt Database and compiled by the Fiscal Affairs Department of the International Monetary Fund (Abbas et al., 2010). 3.2. Fiscal policy persistence, responsiveness, and discretion
ð4Þ
j¼1
where yi,t is the logarithm of real GDP, ci,t is the logarithm of real private consumption, ii,t is the logarithm of real private investment, ε^ i;t is the discretionary component of fiscal policy, Zi,t is a vector of control variables (namely, population growth, trade openness and inflation), κi denotes country-fixed effects, ξi,t is the disturbance term, Δ is the first-differences operator, and J = 4. Eqs. (2)–(4) are regressed using a Fixed-Effects (FE) estimator, although, for robustness, we also consider a two-step robust Generalized Method of Moments (GMM).3 Finally, Eqs. (3) and (4) allow us to explore the existence of crowding-in or crowind-out effects associated with discretionary fiscal policy. 2.3. The heterogeneous response In a similar vein to the work of Agnello and Sousa (2011), we explore the heterogeneity of the response of private consumption and private investment to discretionary fiscal policy across income groups and regions, by re-estimating Eqs. (2)–(4) for: (i) OECD and non-OECD countries; (ii) Europe, North America, South America and the West Indies; and (iii) Africa, Asia and the Pacific, and the Middle East. Then, we analyze whether the response of private consumption and investment to government spending shocks depends on the level of income, namely, by re-estimating Eqs. (2)–(4) for: (i) low-income; (ii) lower-middle income; (iii) developing; (iv) upper-middle income; and (v) high-income countries. Finally, we condition the impact of discretionary fiscal policy on three major countries' characteristics: (i) the degree of trade openness; (ii) the government size; and (iii) the size of the country. 3. Data and empirical results 3.1. Data We gather the annual data, from 1960 to 2008, for 132 countries. The set of variables included in the estimations are as follows: • Real GDP, provided by the World Bank's World Development Indicators (WDI); • Real private consumption, provided by the WDI; • Real private investment, proxied by gross fixed capital formation which is provided by the WDI; 3 The two-step estimator uses the estimated residuals in order to construct a consistent variance covariance matrix of the moment conditions. Although it is asymptotically more efficient than the one-step estimator and it relaxes the assumption of homoscedasticity, the efficiency gains are not that important even in the case of heteroscedastic errors (Arellano and Bond, 1991; Blundell and Bond, 1998; Blundell et al., 2000). This result is further supported by the empirical findings of Judson and Owen (1999), who perform Monte Carlo experiments for several cross sectional and time series dimensions and show that the one-step estimator outperforms the twostep estimator. Moreover, the two-step estimator imposes a downward (upward) bias in standard errors (t-statistics) due to its dependence on the estimated values (as it uses the estimated residuals from the one-step estimator) (Windmeijer, 2005), which may impact on the statistical inference (Bond, 2002; Bond and Windmeijer, 2002). This issue is particularly relevant in the case of data samples with relatively small crosssection dimension (Arellano and Bond, 1991; Blundell and Bond, 1998).
Table 1 presents the estimates of fiscal policy persistence, responsiveness and discretion, which are constructed based on country-specific regressions summarized by Eq. (1). It is clear that the lower and less significant the coefficients associated with GDP growth (i.e. fiscal policy responsiveness, βi) and with lagged spending growth (i.e. fiscal policy persistence, λi) are, the larger the discretionary component of govern ment spending will be ε^ i;t . The positive sign of the coefficient associated with responsiveness is consistent with the empirical literature suggesting that, over the post-World War II period, fiscal policy in emerging and industrialized countries has been “procyclical” (Alesina et al., 2008; Ilzetki and Vegh, 2008). This characteristic is particularly relevant for developing countries (the coefficient associated with GDP growth is 0.509 for non-OECD countries, which compares with 0.275 for OECD countries) and gives support to the role of imperfections in credit markets preventing borrowing in bad times (Gavin and Perotti, 1997) or rent-seeking behavior in good times (Tornell and Lane, 1999). As for the persistence of fiscal policy, our results suggest that it is mainly relevant for Europe and North America, suggesting some inertia in the budgetary process in these countries and a degree of unstable or erratic fiscal policy behavior in the remaining ones (Agnello and Sousa, forthcoming-a). Interestingly, the results also show a significant and negative response of government consumption to changes in the debt level. This highlights the existence of an important stabilizing effect: when public debt increases, government consumption contracts, thereby, bringing the debt to a sustainable path. 3.3. Evidence for the full sample In this section, we discuss the results of our baseline models (2)–(4) using the full sample. Table 2 provides a summary of the findings and Columns (1)–(3) show the parameter estimates for the GDP growth, the private consumption growth and the private investment growth equations, respectively. Interestingly, in Column (1), we find that while the short-term impact of discretionary fiscal policy is positive and statistically significant (0.0451), the medium-term effect ends up being detrimental for growth (a negative coefficient of −0.0196). Moreover, although the effect is negative after one year, it becomes statistically significant after four years. In what concerns private consumption growth and private investment growth (i.e. Columns (2)–(3)), the empirical findings are similar, as they corroborate the existence of an expansionary impact of discretionary fiscal policy in the short-term and a contractionary effect in the medium-run. Putting it differently, there are “crowding-in” effects in the short term and “crowding-out” effects in the medium term. The magnitude of the “crowding-in” effects is larger for investment growth (0.1287) than for consumption growth (0.0645). However, it is also important to note that while medium term “crowding-out” effects are not statistically significant for consumption growth, they cancel out the short-term expansionary impact of discretionary fiscal policy on investment growth (as can be seen by the coefficient of −0.1303). Therefore, for investment, the net effect of discretionary fiscal policy is negative.
18
L. Agnello et al. / Economic Modelling 34 (2013) 15–24
Table 1 Fiscal policy responsiveness and persistence.
Real GDP growth Lagged spending growth Inflation Inflation squared Public debt Constant Total observations Number of countries R-square
Major sub-samples
Income level
OECD
Non-OECD
Low
Lower-middle
Upper-middle
Region Asia and Pacific
South America and West Indies
Middle East
Africa
Europe
North America
0.275*** [0.044] 0.102
0.509*** [0.092] −0.042
0.477*** [0.158] −0.069*
0.419*** [0.097] 0.003
0.625*** [0.089] −0.091
0.457*** [0.115] −0.092
0.443*** [0.116] 0.038
0.277* [0.153] 0.063
0.541*** [0.145] −0.087**
0.554*** [0.088] 0.223***
0.365*** [0.125] 0.382***
[0.070] −0.006 [0.033] 0.000 [0.001] −0.590*** [0.190] 17.562*** [5.261] 1114
[0.028] −0.015 [0.025] 0.000 [0.000] −1.458** [0.584] 40.785*** [13.404] 2817
[0.038] −0.061 [0.044] 0.000 [0.000] −0.934 [1.007] 33.132 [23.711] 1234
[0.045] −0.003 [0.027] 0.000 [0.000] −2.531*** [0.897] 61.740*** [20.192] 1074
[0.056] 0.038 [0.027] 0.000 [0.000] −1.195 [0.875] 37.626** [18.535] 600
[0.075] −0.214 [0.143] 0.001 [0.001] −0.124 [0.505] 12.85 [13.129] 641
[0.050] 0.034 [0.026] 0.000 [0.000] −3.003*** [1.129] 59.325** [23.548] 745
[0.087] 0.008 [0.131] −0.005 [0.007] −4.584** [2.075] 112.054** [48.563] 270
[0.038] −0.062 [0.055] 0.000 [0.000] −1.149 [1.059] 29.258 [25.383] 1130
[0.078] −0.004 [0.039] 0.000 [0.000] −1.066*** [0.327] 27.069*** [7.974] 1012
[0.095] −0.086 [0.093] 0.001 [0.001] −1.024 [0.957] 31.637 [25.264] 133
30
128
52
46
29
31
26
14
44
40
3
0.296
0.094
0.120
0.106
0.213
0.155
0.137
0.268
0.114
0.250
0.734
Note: Fixed-effects (FE) estimation. Robust standard errors in square brackets. *, **, *** statistically significant at 10%, 5% and 1% level, respectively. For brevity, time-effects estimates are not reported.
As for the additional control variables, we show that: (i) population growth has a positive and significant impact on GDP, private consumption and private investment growth; (ii) the degree of openness is positively related with all dependent variables under consideration; and (iii) inflation is detrimental for GDP growth and also investment growth, but its effect does not seem to be significant for consumption growth. Finally, the results point to a reasonable degree of persistence of the dependent variables, as can be seen by the the coefficient associated with the lagged dependent variable. While this highlights the importance of distinguishing between the short-term and the medium term effects of discretionary fiscal policy, it also suggests that the Eqs. (2)–(4) should be estimated by using an IV-GMM framework in the context of a dynamic panel. To investigate this issue, we report in Table 3, the results from the IV-GMM specification based on the work of Arellano and Bond (1991). The empirical findings are very close to the ones found for the
FE estimator, and provide support to the existence of an expansionary effect of discretionary fiscal policy in the short term, which contrasts with a contractionary impact over the medium term. In the case of private investment growth, these medium term “crowding-out” effects are particularly strong and lead to a net negative impact. The coefficients associated with the control variables (population, trade openness and inflation) are as expected and all dependent variables (GDP, private consumption and private investment growth) display persistence. Additionally, the validity of the moment conditions is confirmed by the Hansen (1982) J-test and the disturbance terms, ξi,t, do not exhibit higher order correlation. Given the similarity of results and the fact that the use of the IV-GMM framework would suffer from the problem of overidentification and proliferation of instruments in the third-step analysis (where we consider sub-samples and look at the heterogeneity of the response of private spending to discretionary fiscal policy), in the next sub-sections, we report the findings using the FE estimator. 3.4. OECD versus non-OECD countries
Table 2 The macroeconomic impact of discretionary fiscal policy. Full sample.
Lagged dependent variable ε^ i;t−1 ε^ i;t−2 ε^ i;t−3 ε^ i;t−4 Population growth Openness Inflation # Observations # Countries R-squared
GDP growth
Consumption growth
Investment growth
0.1324**
−0.0685*
0.0553
[0.0602] 0.0451* [0.0240] −0.0158 [0.0144] −0.0096 [0.0126] −0.0196** [0.0098] 0.7528** [0.3465] 0.0213*** [0.0056] −0.0205*** [0.0043] 3160 131 0.08
[0.0360] 0.0645** [0.0251] 0.0409 [0.0261] −0.0235 [0.0272] −0.0146 [0.0205] 0.4344* [0.2556] 0.0367*** [0.0092] −0.0112 [0.0121] 3091 130 0.03
[0.0451] 0.1287* [0.0774] −0.0909 [0.0568] −0.1303** [0.0499] −0.0085 [0.0460] 1.2564* [0.7329] 0.0695*** [0.0166] −0.0293*** [0.0111] 2908 129 0.02
Note: Fixed-effects (FE) estimation. Robust standard errors in square brackets. *, **, *** statistically significant at 10%, 5% and 1% level, respectively.
The analysis presented so far has shown evidence on the existence of crowding-in and crowding-out effects. But are these effects similar among the groups of countries? To answer this question, we start by replicating the estimation of Eqs. (2)–(4) for developed (OECD) and developing (Non-OECD) countries. Table 4 shows that the effect of discretionary spending varies substantially between the two groups of countries. With regards to the effects on consumption growth, we note that crowding-in effects differ in terms of timing and are strongly magnified in OECD countries: while, for non-OECD countries, the impact of discretionary fiscal policy on consumption growth is positive and statistically significant within the first year (0.0557), in the case of OECD countries, the effect only emerges after four years, but is almost two times stronger (0.1051). Turning to the impact on investment, we can see that crowding-out effects are only statistically significant for non-OECD countries: a 1% in discretionary spending typically reduces investment growth by 0.13 percentage points within the third year. Putting these two pieces of evidence together, it can be seen that discretionary fiscal policy ends up having some expansionary effects in the medium term in the case of OECD countries. In contrast, for non-OECD countries, this positive impact is confined to the short term, as important crowding-out effects emerge afterwards. This contrasting findings give support to the
L. Agnello et al. / Economic Modelling 34 (2013) 15–24 Table 3 The macroeconomic impact of discretionary fiscal policy. Full sample (IV-GMM framework).
Lagged dependent variable ε^ i;t−1 ε^ i;t−2 ε^ i;t−3 ε^ i;t−4 Population growth Openness Inflation # Observations # Countries AR(1) test statistic p-value AR(2) test statistic p-value Hansen J-Statistic p-value
GDP growth
Consumption growth
Investment growth
0.1423*
−0.0442
0.1162**
[0.0827] 0.0435 [0.0273] −0.0093 [0.0147] −0.0091 [0.0117] −0.0221* [0.0112] 0.6919* [0.3736] 0.0210*** [0.0062] −0.0192*** [0.0040] 3023 130 −3.55 0.00 −1.06 0.29 93.22 0.13
[0.0400] 0.0646** [0.0293] 0.0436 [0.0285] −0.0223 [0.0304] −0.012 [0.0249] 0.3978 [0.2624] 0.0339*** [0.0089] −0.0097 [0.0127] 2954 129 −4.98 0.00 −1.17 0.24 90.82 0.17
[0.0513] 0.0917 [0.0844] −0.0747 [0.0552] −0.0926* [0.0546] −0.0087 [0.0489] 0.941 [0.7853] 0.0570*** [0.0167] −0.0255** [0.0114] 2773 128 −4.24 0.00 −1 0.32 88.07 0.23
19
our findings confirm that, despite some expansionary effects associated with fiscal policy in the short term – albeit, not statistically significant –, crowding-out effects emerge in the medium term: a 1% increase in the discretionary component of government consumption reduces private consumption and private investment growth by 0.10 and 0.24 percentage points, respectively. In the case of North America, the discretionary component of fiscal policy does not have a significant effect on growth, although this result needs to be assessed with caution given that the model is estimated only for 3 countries. As for South America and the West Indies, we find important crowding-in effects in the short term and crowing-out effects in the medium term, which are particularly strong in the case of investment growth. In Table 6, we can see that discretionary fiscal policy: (i) does not play a statistically significant role for growth and private spending in African countries; (ii) produces some short-lived crowding-in effects in Asia and Pacific; and (iii) generates medium-run crowding-out effects in the Middle East. 3.6. Income level
Note: Dynamic panel-data estimation (IV-GMM). Standard errors in square brackets. *, **, *** statistically significant at 10%, 5% and 1% level, respectively.
argument that poor capacity of implementation and larger macroeconomic risks explain the limited scope for fiscal policy in developing countries (Botman and Kumar, 2006). 3.5. Regional effects In this sub-section, we present evidence on the impact of discretionary fiscal policy for different regions. Table 5 summarizes the findings for Europe, North America, and South America and the West Indies, while Table 6 reports the results for Africa, Asia and the Pacific, and the Middle East. Table 5 shows that, for Europe, the crowding-out effects associated with discretionary fiscal policy are quite strong: a 1% increase in discretionary spending leads to a fall in GDP growth of 0.04 percentage points within four years. When looking at the composition of private demand,
Tables 7 and 8 summarize the findings for different income groups. In Table 7, we consider low-income, lower-middle income and developing countries, while, in Table 8, we look at upper-middle income and high-income countries. The results suggest that, for lower-middle income and developing countries, discretionary fiscal policy has an expansionary effect in the short term. This, in turn, operates, mainly via boosting private consumption growth. In contrast, crowding-out effects emerge in the medium term and negatively impinge on private investment growth. For higher-middle income and high-income countries, we also find short-lived crowding-in effects associated with an increase in the discretionary component of government spending. Later, this impact reverts and crowding-out effects take place, leading to a fall in GDP growth and, in particular, in private investment growth. One should also highlight that, for these countries, both crowding-in and crowding-out effects are larger in magnitude than the ones associated with countries of lower income level, but the net effect of discretionary fiscal policy is still positive. 3.7. Trade openness We now condition the effects of discretionary fiscal policy on the degree of trade openness. More specifically, Eqs. (2)–(4) are estimated for
Table 4 The macroeconomic impact of discretionary fiscal policy OECD versus non-OECD countries. OECD
Lagged dependent variable ε^ i;t−1 ε^ i;t−2 ε^ i;t−3 ε^ i;t−4 Population growth Openness Inflation # Observations # Countries R-squared
Non-OECD
GDP growth
Consumption growth
Investment growth
GDP growth
Consumption growth
Investment growth
0.2926*** [0.0520] 0.0534 [0.0877] 0.0978 [0.0747] 0.0645* [0.0317] 0.0158 [0.0376] 0.0622 [0.3277] 0.0035 [0.0043] −0.0839*** [0.0173] 983 30 0.18
0.2860*** [0.0445] 0.1259 [0.0887] −0.0116 [0.0590] 0.0377 [0.0561] 0.1051** [0.0400] −0.2262 [0.3888] 0.0042 [0.0048] −0.0835*** [0.0113] 976 30 0.16
0.2133*** [0.0642] 0.112 [0.2298] 0.1964 [0.1902] −0.0484 [0.1343] −0.1007 [0.1255] 0.3617 [0.9563] 0.0128 [0.0120] −0.2167*** [0.0412] 908 30 0.11
0.1039 [0.0647] 0.0449* [0.0243] −0.0193 [0.0148] −0.0132 [0.0128] −0.0225** [0.0098] 0.8602** [0.3838] 0.0320*** [0.0087] −0.0200*** [0.0043] 2177 101 0.08
−0.0935** [0.0363] 0.0557** [0.0264] 0.0392 [0.0270] −0.0261 [0.0274] −0.0191 [0.0211] 0.5154* [0.2809] 0.0548*** [0.0138] −0.0105 [0.0119] 2115 100 0.03
0.0383 [0.0480] 0.1312 [0.0792] −0.0959 [0.0580] −0.1349*** [0.0507] −0.0085 [0.0470] 1.4162* [0.8041] 0.0927*** [0.0253] −0.0257** [0.0111] 2000 99 0.02
Note: Fixed-effects (FE) estimation. Robust standard errors in square brackets. *, **, *** statistically significant at 10%, 5% and 1% level, respectively.
20
L. Agnello et al. / Economic Modelling 34 (2013) 15–24
Table 5 The macroeconomic impact of discretionary fiscal policy. Regional effects (Europe, North America, and South America and West Indies). Europe
Lagged dependent variable ε^ i;t−1 ε^ i;t−2 ε^ i;t−3 ε^ i;t−4 Population growth Openness Inflation # Observations # Countries R-squared
North America
South America and West Indies
GDP growth
Consumption growth
Investment growth
GDP growth
Consumption growth
Investment growth
GDP growth
Consumption growth
Investment growth
0.3419*** [0.0585] 0.0186 [0.0419] 0.0461 [0.0376] 0.0182 [0.0302] −0.0399** [0.0169] −0.1344 [0.2617] 0.0099* [0.0050] −0.0435** [0.0208] 830 37 0.21
0.1456* [0.0761] 0.1085 [0.0947] −0.0138 [0.0452] 0.0704 [0.0451] −0.0952* [0.0487] −0.3408 [0.4268] 0.0088 [0.0067] −0.0454*** [0.0106] 826 37 0.11
0.1602*** [0.0557] 0.0978 [0.1272] −0.0605 [0.1549] 0.0902 [0.0780] −0.2386*** [0.0771] −0.4853 [0.9077] 0.0321*** [0.0091] −0.1568*** [0.0316] 769 37 0.13
0.1752 [0.0911] 0.0086 [0.2461] −0.0475 [0.0830] 0.0429 [0.0532] 0.0083 [0.0515] −0.2281 [0.3834] −0.009 [0.0106] −0.1489 [0.0513] 123 3 0.28
−0.0751 [0.1826] 0.2466 [0.0871] −0.2787 [0.2016] −0.241 [0.2376] 0.1964 [0.3024] 1.0766 [0.7252] −0.0189 [0.0251] −0.1451 [0.0914] 121 3 0.11
0.0183 [0.1569] −0.2742 [0.7567] −0.281 [0.2828] −0.2147 [0.0811] 0.1608 [0.3124] 2.2538 [1.3476] −0.0361 [0.1025] −0.585 [0.3210] 121 3 0.21
0.3153*** [0.0677] −0.0145 [0.0298] −0.0399 [0.0312] −0.0474*** [0.0153] −0.0039 [0.0119] 0.8324 [0.6475] 0.0416*** [0.0073] −0.0136*** [0.0037] 624 23 0.18
0.0241 [0.0391] 0.1008** [0.0432] 0.0743 [0.0530] −0.2130*** [0.0745] 0.0153 [0.0414] 1.0174 [1.4018] 0.0927** [0.0346] 0.001 [0.0111] 614 23 0.09
0.1120** [0.0520] 0.3944*** [0.1113] −0.2093 [0.1299] −0.3372*** [0.0919] −0.0713 [0.0518] 2.7104 [1.8856] 0.1288*** [0.0311] −0.0128 [0.0084] 595 23 0.10
Note: Fixed-effects (FE) estimation. Robust standard errors in square brackets. *, **, *** statistically significant at 10%, 5% and 1% level, respectively.
two major groups of countries: (i) high degree of openness; and (ii) low degree of openness. We use the sample mean of the degree of openness as the threshold for distinguishing the two groups. There are two main reasons why trade openness may affect the impact of discretionary fiscal policy on GDP growth and private spending. First, while, in a flexible exchange rate regime, an increase in the discretionary component of government consumption can generate crowding-out effects through the exchange rate channel, in countries where the external sector is relatively small, expansionary fiscal policies can lead to important Keynesian (thereby, crowding-in) effects. Second, countries with a high degree of trade openness are also more vulnerable to shocks to the external sector and, consequently, countercyclical fiscal policies may be needed to promote private consumption smoothing (Rodrick, 1998). However, if the integration of financial markets is relatively high, there will be less scope for fiscal
policy stimulus, as trade shocks can be counterbalanced with risk sharing. The results are summarized in Table 9 and suggest that, for countries with high degree of trade openness, there are some short-term crowding-in effects associated with discretionary fiscal policy, which are mainly beneficial for private consumption growth, and medium term crowding-out effects that are particularly detrimental for private investment growth. In contrast, for countries with low degree of trade openness, the expansionary effects associated with discretionary fiscal policy are typically short-lived and operate via its positive impact on private investment growth. On balance, these empirical findings suggest that the scope for an expansion of discretionary fiscal policy is inversely related with the degree of trade openness, a result that is in line with the work of Corsetti et al. (2010).
Table 6 The macroeconomic impact of discretionary fiscal policy. Regional effects (Africa, Asia and Pacific and Middle East). Africa
Lagged dependent variable ε^ i;t−1 ε^ i;t−2 ε^ i;t−3 ε^ i;t−4 Population growth Openness Inflation # Observations # Countries R-squared
Asia and Pacific
Middle East
GDP growth
Consumption growth
Investment growth
GDP growth
Consumption growth
Investment growth
GDP growth
Consumption growth
Investment growth
0.0152 [0.0883] 0.0317 [0.0318] −0.0234 [0.0213] −0.0122 [0.0219] −0.0194 [0.0153] 1.3827*** [0.3493] 0.0437** [0.0165] −0.0346*** [0.0076] 904 38 0.09
−0.1521*** [0.0491] 0.0191 [0.0411] 0.0249 [0.0423] 0.0022 [0.0304] −0.0143 [0.0283] 0.8368*** [0.2028] 0.0649** [0.0252] −0.0489** [0.0184] 883 38 0.05
−0.0671 [0.0702] −0.0414 [0.0851] −0.1101 [0.0779] −0.0989 [0.0698] 0.0692 [0.0782] 2.6692*** [0.6409] 0.1530*** [0.0404] −0.0252 [0.0498] 803 38 0.03
0.1645* [0.0804] 0.1405*** [0.0438] 0.0319 [0.0333] 0.0338 [0.0262] −0.0184 [0.0200] 0.3051 [0.1998] −0.0042 [0.0068] −0.1194* [0.0586] 479 21 0.14
−0.006 [0.0960] 0.0758 [0.0611] 0.1127*** [0.0391] 0.0658 [0.0479] 0.0037 [0.0450] 0.4966** [0.2329] 0.011 [0.0082] −0.0759 [0.0583] 450 20 0.05
0.2667*** [0.0402] 0.1772 [0.1345] 0.1113 [0.1183] 0.0229 [0.1109] −0.0474 [0.0853] 0.2621 [0.4784] −0.0247 [0.0211] −0.2155 [0.1863] 432 20 0.10
−0.0472 [0.1684] 0.0689 [0.0586] 0.045 [0.0974] −0.0079 [0.0685] −0.0814* [0.0373] −0.1936 [0.4172] 0.0043 [0.0101] −0.0758 [0.0473] 200 9 0.04
−0.1844** [0.0551] 0.0125 [0.0652] −0.0069 [0.1255] 0.0598 [0.0483] 0.0233 [0.0793] −1.3769* [0.6501] 0.0066 [0.0210] −0.0863 [0.1116] 197 9 0.08
0.1436* [0.0736] 0.2681 [0.3149] −0.2198 [0.1410] −0.075 [0.2314] −0.0543 [0.2633] −1.6672 [1.3492] 0.0227 [0.0477] −0.4262 [0.3064] 188 8 0.12
Note: Fixed-effects (FE) estimation. Robust standard errors in square brackets. *, **, *** statistically significant at 10%, 5% and 1% level, respectively.
L. Agnello et al. / Economic Modelling 34 (2013) 15–24
21
Table 7 The macroeconomic impact of discretionary fiscal policy. Income level (low-income, lower-middle income and developing countries). Low-income countries
Lagged dependent variable ε^ i;t−1 ε^ i;t−2 ε^ i;t−3 ε^ i;t−4 Population growth Openness Inflation # Observations # Countries R-squared
Lower-middle income countries
Developing countries
GDP growth
Consumption growth
Investment growth
GDP growth
Consumption growth
Investment growth
GDP growth
Consumption growth
Investment growth
0.0054 [0.0855] 0.026 [0.0302] −0.0023 [0.0146] 0.0102 [0.0139] −0.0191 [0.0146] 1.3399*** [0.3956] 0.0421** [0.0161] −0.0244*** [0.0067] 982 45 0.09
−0.1195*** [0.0404] 0.0131 [0.0441] 0.0066 [0.0393] −0.0144 [0.0378] −0.0417 [0.0259] 0.8943*** [0.2298] 0.0623** [0.0240] −0.0004 [0.0166] 958 45 0.04
−0.0496 [0.0702] 0.044 [0.0849] −0.1015 [0.0711] −0.1154* [0.0677] 0.0739 [0.0753] 2.5879*** [0.6394] 0.1438*** [0.0411] −0.0414*** [0.0146] 876 45 0.04
0.1848** [0.0846] 0.0667** [0.0316] −0.0146 [0.0325] −0.0275 [0.0196] −0.0244 [0.0159] 0.2061 [0.1785] 0.0248*** [0.0078] −0.0226*** [0.0063] 843 38 0.14
−0.0175 [0.0545] 0.1267*** [0.0364] 0.0531 [0.0460] −0.0313 [0.0546] −0.0263 [0.0446] 0.1863 [0.4045] 0.0570** [0.0246] −0.0174 [0.0105] 813 37 0.06
0.1551*** [0.0498] 0.3516*** [0.1046] −0.2326*** [0.0752] −0.1021 [0.0742] −0.1023* [0.0547] −0.3011 [0.5071] 0.0437 [0.0272] −0.017 [0.0169] 800 37 0.08
0.1074* [0.0634] 0.0430* [0.0242] −0.0186 [0.0147] −0.0135 [0.0127] −0.0227** [0.0098] 0.8790** [0.3712] 0.0292*** [0.0077] −0.0209*** [0.0045] 2281 106 0.08
−0.0881** [0.0360] 0.0557** [0.0260] 0.0372 [0.0268] −0.0264 [0.0273] −0.018 [0.0209] 0.5266* [0.2737] 0.0487*** [0.0122] −0.0118 [0.0121] 2221 105 0.03
0.0432 [0.0472] 0.124 [0.0787] −0.0969* [0.0576] −0.1338*** [0.0506] −0.0119 [0.0466] 1.4268* [0.7813] 0.0794*** [0.0217] −0.0284** [0.0112] 2106 104 0.02
Note: Fixed-effects (FE) estimation. Robust standard errors in square brackets. *, **, *** statistically significant at 10%, 5% and 1% level, respectively.
3.8. Government size In this sub-section, we investigate the heterogeneity of the response of GDP, private consumption and private investment growth to an increase in the discretionary component of government consumption for two groups of countries: (i) large government size; and (ii) small government size. We proxy the government size by the share of public consumption in GDP and the sample mean is used to build the two sub-samples. Table 10 provides a summary of the empirical findings. It can be seen that, for countries with a large government size, an increase of 1% in the discretionary component of government spending leads to a rise of GDP growth of 0.26 percentage points within one year. For countries with a small government size, such fiscal stimulus does not lead to significant crowding-in effects. Indeed, discretionary fiscal policy appears to be responsible for some medium term crowding-out effects: a 1% rise in discretionary fiscal policy reduces private investment growth by 0.17 percentage point over the period of four years. This result suggests
that, for countries with a smaller government size and, therefore, a larger share of private spending, crowding-out effects tend to be larger than in countries with a larger share of public spending.
3.9. Country size We also condition the macroeconomic response to variation in discretionary fiscal policy on the size of the country. This is explained by the fact that size helps countries spreading the costs of government spending, while insuring them against idiosyncratic shocks. As a result, the impact of unexpected variation in fiscal policy on GDP, private consumption and private investment growth may be different depending on the country size. Consequently, we estimate Eqs. (2)–(4) for two sub-samples (i) large countries and (ii) small countries, and average population is used as the threshold. As is common in the literature, the country size is proxied by population (Agnello and Sousa, forthcoming-a).
Table 8 The macroeconomic impact of discretionary fiscal policy. Income level (upper-middle income and high-income countries). Upper-middle income countries
Lagged dependent variable ε^ i;t−1 ε^ i;t−2 ε^ i;t−3 ε^ i;t−4 Population growth Openness Inflation # Observations # Countries R-squared
High-income countries
GDP growth
Consumption growth
Investment growth
GDP growth
Consumption growth
Investment growth
0.0054 [0.0855] 0.3188*** [0.0760] −0.0533 [0.0393] −0.0761** [0.0304] −0.0940*** [0.0180] −0.017 [0.0196] 1.1486* [0.5659] 0.0137* [0.0079] 456 23 0.14
−0.1195*** [0.0404] −0.0968 [0.0990] 0.1026* [0.0515] 0.1139** [0.0502] −0.0842** [0.0395] 0.0938* [0.0523] −0.6855 [0.8167] 0.0217* [0.0110] 450 23 0.05
−0.0496 [0.0702] 0.1258** [0.0598] −0.1452 [0.2929] 0.0275 [0.1955] −0.3388** [0.1551] −0.1288 [0.1164] 2.5121* [1.2948] 0.0293 [0.0275] 430 22 0.04
0.3139*** [0.0549] 0.1498* [0.0815] 0.0154 [0.0624] 0.1100*** [0.0336] 0.033 [0.0461] −0.1737 [0.2493] 0.0012 [0.0048] −0.1044*** [0.0304] 879 25 0.21
0.3306*** [0.0477] 0.1825 [0.1201] 0.0264 [0.1302] 0.0544 [0.0474] 0.0544 [0.0807] −0.4613 [0.3277] 0.0015 [0.0046] −0.0933*** [0.0124] 870 25 0.19
0.2105*** [0.0680] 0.4325* [0.2480] 0.1798 [0.1913] −0.1796 [0.1643] 0.1259 [0.1461] −0.6384 [0.7919] 0.0243* [0.0125] −0.2772*** [0.0623] 802 25 0.12
Note: Fixed-effects (FE) estimation. Robust standard errors in square brackets. *, **, *** statistically significant at 10%, 5% and 1% level, respectively.
22
L. Agnello et al. / Economic Modelling 34 (2013) 15–24
Table 9 The macroeconomic impact of discretionary fiscal policy. Degree of openness. High degree of openness
Lagged dependent variable ε^
i;t−1
ε^ i;t−2 ε^ i;t−3 ε^ i;t−4 Population growth Openness Inflation # Observations # Countries R-squared
Low degree of openness
GDP growth
Consumption growth
Investment growth
GDP growth
Consumption growth
Investment growth
0.0702 [0.1222] 0.0655 [0.0472] −0.0299 [0.0292] −0.0316 [0.0217] −0.0194 [0.0169] 0.8612 [0.5584] 0.0267*** [0.0068] −0.0385*** [0.0107] 1180 56 0.09
−0.1195*** [0.0404] −0.0408 [0.0462] 0.0818* [0.0424] 0.038 [0.0448] −0.0145 [0.0348] 0.0026 [0.0327] 0.3601 [0.3018] 0.0406*** [0.0113] 1145 55 0.06
−0.0496 [0.0702] −0.0133 [0.0695] 0.1506 [0.1218] −0.0281 [0.0939] −0.2407** [0.0984] 0.0249 [0.0761] 1.4616 [1.5627] 0.0693*** [0.0267] 1079 55 0.03
0.3139*** [0.0549] 0.1815*** [0.0399] 0.0311 [0.0223] −0.0021 [0.0178] 0.0061 [0.0146] −0.0244 [0.0152] 0.4435 [0.3418] 0.0126* [0.0071] 1980 75 0.08
0.3306*** [0.0477] −0.0962* [0.0497] 0.0409 [0.0342] 0.0398 [0.0350] −0.0347 [0.0319] −0.0427 [0.0268] 0.4017 [0.5057] 0.0271** [0.0118] 1946 75 0.02
0.2105*** [0.0680] 0.0903 [0.0605] 0.1197** [0.1010] −0.1288 [0.0857] −0.0373 [0.0826] −0.0526 [0.0852] 0.4579 [1.2838] 0.0648*** [0.0245] 1829 74 0.03
Note: Fixed-effects (FE) estimation. Robust standard errors in square brackets. *, **, *** statistically significant at 10%, 5% and 1% level, respectively.
The results are reported in Table 11 and show that, for large countries, discretionary fiscal policy has a statistically significant expansionary effect, albeit confined to the first two years. In the case of small countries, similar (albeit weaker) crowding-in effects emerge in the short-run. However, they are completely annihilated by crowding-out effects in the medium term, which are particularly detrimental for private investment growth. 4. Conclusions In this paper, we develop a novel empirical approach to quantify the impact of discretionary fiscal policy on GDP and private spending growth. This framework consists of three major stages. First, we estimate a fiscal policy rule and use it to extract the discretionary component of fiscal policy. Second, we use residuals of the first-stage regression to assess the existence of crowding-in and/or crowding-out effects and to investigate the trade-off between such effects in the short and medium term. Third, we analyze the heterogeneity of the macroeconomic response to discretionary fiscal policy conditioning it on a set of
country characteristics (such as, economic development, region, income level, trade openness, government size and country size). Building on a panel dataset of 132 countries for the period 1960– 2008, we show that an increase in the discretionary component of fiscal policy has an expansionary effect in the short-term, but leads to crowding-out effects in the medium term. Considering the countries' economic development, we find that crowding-in effects are particularly large in OECD countries, while medium-run crowding-out emerge more importantly in non-OECD countries. While looking at regional differences, the empirical findings suggest that crowding-out effects over the medium term are larger for European countries. In what concerns the response of GDP growth and private spending conditional on the income level, the results show that short-lived crowding-in effects and medium-run crowding-out effects are larger in magnitude for higher-middle income and high-income countries than for countries with lower income level. In the case of low-income and lower-middle income countries, the crowding-in effects operate mainly via stimulating private consumption growth,
Table 10 The macroeconomic impact of discretionary fiscal policy. Government size. Large government size
Lagged dependent variable ε^ i;t−1 ε^ i;t−2 ε^ i;t−3 ε^ i;t−4 Population growth Openness Inflation # Observations # Countries R-squared
Small government size
GDP growth
Consumption growth
Investment growth
GDP growth
Consumption growth
Investment growth
0.0702 [0.1222] 0.2576*** [0.0428] 0.0653*** [0.0222] 0.0007 [0.0187] −0.0292 [0.0185] −0.0143 [0.0137] 0.3144 [0.2525] 0.0156*** [0.0060] 1363 56 0.11
−0.1195*** [0.0404] −0.1385** [0.0662] 0.0994* [0.0517] 0.1207** [0.0475] −0.0051 [0.0462] −0.0183 [0.0418] −0.5188 [0.5232] 0.0263* [0.0137] 1341 56 0.05
−0.0496 [0.0702] 0.1182** [0.0485] 0.1497 [0.1169] −0.1059 [0.1012] −0.0667 [0.1015] −0.0197 [0.0760] 0.1326 [1.2195] 0.0574* [0.0310] 1249 55 0.03
0.3139*** [0.0549] 0.0436 [0.0846] 0.0799 [0.0350] −0.0285 [0.0229] −0.0026 [0.0161] −0.0211 [0.0158] 0.9409* [0.5626] 0.0249*** [0.0076] 1797 75 0.08
0.3306*** [0.0477] −0.0273 [0.0385] 0.0496 [0.0332] 0.0057 [0.0352] −0.0251 [0.0282] −0.0085 [0.0262] 0.7515*** [0.2670] 0.0422*** [0.0103] 1750 74 0.03
0.2105*** [0.0680] 0.0236 [0.0668] 0.13 [0.1085] −0.0841 [0.0803] −0.1660** [0.0783] 0.0018 [0.0796] 1.6285 [1.5564] 0.0761*** [0.0248] 1659 74 0.03
Note: Fixed-effects (FE) estimation. Robust standard errors in square brackets. *, **, *** statistically significant at 10%, 5% and 1% level, respectively.
L. Agnello et al. / Economic Modelling 34 (2013) 15–24
23
Table 11 The macroeconomic impact of discretionary fiscal policy. Country size. Large country size
Lagged dependent variable ε^
i;t−1
ε^ i;t−2 ε^ i;t−3 ε^ i;t−4 Population growth Openness Inflation # Observations # Countries R-squared
Small country size
GDP growth
Consumption growth
Investment growth
GDP growth
Consumption Growth
Investment growth
0.0702 [0.1222] 0.2585*** [0.0539] 0.0135 [0.0301] 0.0153 [0.0288] 0.0125 [0.0210] −0.0380** [0.0188] −0.4173 [0.5323] −0.0038 [0.0117] 823 29 0.12
−0.1195*** [0.0404] 0.0611 [0.0653] 0.0902* [0.0471] −0.0119 [0.0486] 0.0451 [0.0365] −0.0361 [0.0313] −0.3836 [0.5187] 0.008 [0.0128] 812 29 0.03
−0.0496 [0.0702] 0.1830** [0.0869] −0.034 [0.1374] −0.1806 [0.1981] 0.0296 [0.1692] −0.0779 [0.1781] −1.9067 [1.2279] 0.0005 [0.0336] 780 29 0.05
0.3139*** [0.0549] 0.1105 [0.0724] 0.0485* [0.0277] −0.021 [0.0192] −0.0163 [0.0148] −0.0166 [0.0129] 0.8624* [0.4598] 0.0273*** [0.0056] 2337 102 0.08
0.3306*** [0.0477] 0.0586* [0.0388] −0.0843** [0.0311] 0.0484 [0.0321] −0.0401 [0.0284] −0.0114 [0.0260] 0.4992* [0.2801] 0.0437*** [0.0098] 2279 101 0.03
0.2105*** [0.0680] 0.0355 [0.0514] 0.1479* [0.0873] −0.066 [0.0636] −0.1569** [0.0664] −0.0013 [0.0590] 1.5488 [1.3401] 0.0847*** [0.0222] 2128 100 0.03
Note: Fixed-effects (FE) estimation. Robust standard errors in square brackets. *, **, *** statistically significant at 10%, 5% and 1% level, respectively.
while the crowding-out effects are due to the negative impact on private investment growth. We also show that the strength of the stimulus to growth generated by discretionary fiscal policy is inversely related to the degree of trade openness. Controlling for the government size, we find that the medium-run impact of discretionary fiscal policy is more detrimental for private investment growth in countries with a smaller government size. Finally, the medium-run crowding-out effects associated with an expansion of the discretionary component of fiscal policy are particularly damaging for small countries, making them more vulnerable to unexpected variation on public consumption. The current work opens new avenues for understanding the impact of discretionary fiscal policy on private spending. In particular and in light of the severity of the most recent financial turmoil, it would be interesting to condition the effectiveness of fiscal policy on the occurrence of crisis episodes. For instance, Agnello and Sousa (2011, forthcoming-b) suggest the existence of important multiplier effects during periods of severe housing busts. We aim at pursuing such investigation in the near future.
References Abbas, S.A., Belhocine, N., ElGanainy, A., Horton, M., 2010. A historical public debt database. International Monetary Fund, IMF Working Paper No. 245. Afonso, A., Agnello, L., Furceri, D., 2010. Fiscal policy responsiveness, persistence and discretion. Public Choice 145 (3), 503–530. Agnello, L., Sousa, R.M., 2011. Can fiscal policy stimulus boost economic recovery? Revue Économique 62 (6), 1045–1066. Agnello, L., Sousa, R.M., forthcoming-a. Political, Institutional and economic factors underlying public deficit volatility. Review of International Economics. Agnello, L., Sousa, R.M., forthcoming-b. Fiscal policy and asset prices. Bulletin of Economic Research. Agnello, L., Castro, V., Sousa, R.M., 2012. How does fiscal policy react to wealth composition and asset prices? Journal of Macroeconomics 34 (3), 874–890. Aiyagari, R., Christiano, L., Eichenbaum, M., 1990. Output, employment and interest rate effects of government consumption. Journal of Monetary Economics 30, 73–86. Alesina, A., Ardagna, S., 2010. Large changes in fiscal policy: taxes versus spending. In: Brown, J.R. (Ed.), Tax Policy and the Economy 24, 35–68. Alesina, A., Tabellini, G., Campante, F.R., 2008. Why is fiscal policy often procyclical? Journal of the European Economic Association 6 (5), 1006–1036. Arellano, M., Bond, S., 1991. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. The Review of Economic Studies 58, 277–297. Baldacci, E., Kumar, M.S., 2010. Fiscal deficits, public debt, and sovereign bond yields. International Monetary Fund, IMF Working Paper No. 184.
Baldacci, E., Gupta, S., Mulas-Granados, C., 2009. How effective is fiscal policy response in systemic banking crises? International Monetary Fund, IMF Working Paper No. 160. Baxter, M., King, R., 1993. Fiscal policy in general equilibrium. The American Economic Review 83, 315–334. Blanchard, O., 2003. Macroeconomics, 3rd ed. Prentice Hall. Blanchard, O., Perotti, R., 2002. An empirical characterization of the dynamic effects of changes in government spending and taxes on output. Quarterly Journal of Economics 117 (4), 1329–1368. Blundell, R., Bond, S., 1998. Initial conditions and moment conditions in dynamic panel data models. Journal of Econometrics 87, 115–143. Blundell, R., Bond, S., Windmeijer, F., 2000. Estimation in dynamic panel data models: improving on the performance of the standard GMM estimator. In: Baltagi, B. (Ed.), Non-stationary panels, panel cointegration and dynamic panels. : Advances in Econometrics, vol. 15. JAI Elsevier Science. Bond, S., 2002. Dynamic panel data models: a guide to micro data methods and practice. Portuguese Economic Journal 1 (2), 141–162. Bond, S., Windmeijer, F., 2002. Finite sample inference for GMM estimators in linear panel data models: A comparison of alternative tests. Institute for Fiscal Studies, mimeo. Botman, D., Kumar, M.S., 2006. Fundamental determinants of the effects of fiscal policy. International Monetary Fund, IMF Working Paper No. 72. Burnside, C., Eichenbaum, M., Fisher, J., 2004. Fiscal shocks and their consequences. Journal of Economic Theory 115 (1), 89–117. Christiano, L., Eichenbaum, M., 1992. Current real business cycles theories and aggregate labor market fluctuations. The American Economic Review 82, 430–450. Corsetti, G., Meier, A., Müller, G.J., 2010. What determines government spending multipliers? European University Institute . (manuscript). Darby, J., Melitz, J., 2008. Social spending and automatic stabilizers in the OECD. Economic Policy 23, 715–756. Devries, P., Guajardo, J., Leigh, D., Pescatori, A., 2011. A new action based dataset of fiscal consolidation. International Monetary Fund, IMF Working Paper No. 128. Fatás, A., Mihov, I., 2003. The case for restricting discretionary fiscal policy. Quarterly Journal of Economics 118, 1419–1447. Fatás, A., Mihov, I., 2006. The macroeconomics effects of fiscal rules in the US states. Journal of Public Economics 90, 101–117. Furceri, D., 2010. Stabilization effects of social spending: empirical evidence from a panel of OECD countries. The North American Journal of Economics and Finance 21 (1), 34–48. Furceri, D., Sousa, R.M., 2011a. The impact of government spending on the private sector: crowding-out versus crowding-in effects. Kyklos 64 (4), 516–533. Furceri, D., Sousa, R.M., 2011b. Does government spending crowd-out private consumption and investment? Theory and some empirical evidence. World Economics 12 (4), 153–170. Gavin, M., Perotti, R., 1997. Fiscal policy in Latin America. In: Bernanke, B., Rotemberg, J. (Eds.), NBER Macroeconomics Annual. MIT Press. Giavazzi, F., Pagano, M., 1990. Can severe fiscal contractions be expansionary? Tales of two small European countries. In: Blanchard, O., Fischer, S. (Eds.), NBER Macroeconomics Annual. MIT Press. Giavazzi, F., Jappelli, T., Pagano, M., 2000. Searching for non-linear effects of fiscal policy: evidence from industrial and developing countries. European Economic Review 44, 1259–1289. Granville, B., Mallick, S.K., 2009. Monetary and financial stability in the euro area: procyclicality versus trade-off. Journal of International Financial Markets, Institutions and Money 19, 662–674. Hansen, L., 1982. Large sample properties of Generalized Method of Moments estimators. Econometrica 50, 1029–1054.
24
L. Agnello et al. / Economic Modelling 34 (2013) 15–24
Ilzetki, E., Vegh, C.A., 2008. Procyclical fiscal policy in developing countries: truth or fiction? NBER Working Paper No. 14191. National Bureau of Economic Research. Ilzetki, E., Mendoza, E.G., Végh, C.A., 2010. How big (small?) are fiscal multipliers? NBER Working Paper No. 16479. National Bureau of Economic Research. Judson, R.A., Owen, L.A., 1999. Estimating dynamic panel data models: a guide for macroeconomists. Economics Letters 65, 9–15. Mallick, S.K., Mohsin, M., 2010. On the real effects of inflation in open economies: theory and empirics. Empirical Economics 39, 643–673. Mountford, A., Uhlig, H., 2009. What are the effects of fiscal policy shocks? Journal of Applied Econometrics 24, 960–992. Rafiq, M.S., Mallick, S.K., 2008. The effect of monetary policy on output in EMU3: a sign restriction approach. Journal of Macroeconomics 30, 1756–1791. Ramey, V.A., 2011. Identifying government spending shocks: it's all in the timing. Quarterly Journal of Economics 126 (1), 1–50.
Ramey, V., Shapiro, M., 1998. Costly capital reallocation and the effects of government spending. Carnegie Rochester Conference on Public Policy 48, 145–194. Rodrick, D., 1998. Why do more open economies have bigger governments? Journal of Political Economy 106, 997–1032. Sousa, R.M., 2010. Housing wealth, financial wealth, money demand and policy rule: evidence from the Euro area. The North American Journal of Economics and Finance 21 (1), 88–105. Sousa, R.M., forthcoming. Wealth, asset portfolio, money demand and policy rule. Bulletin of Economic Research. Tornell, A., Lane, P., 1999. Voracity and growth. The American Economic Review 89, 22–46. Windmeijer, F., 2005. A finite sample correction for the variance of linear efficient twostep GMM estimators. Journal of Econometrics 126 (1), 25–51.