Printer Friendly

Should we care about the composition of tax-based stimulus packages?

I. INTRODUCTION

In February 2009, the United States passed a fiscal stimulus package totaling US$787 billion in government spending and tax cuts over a period of 2 years. The stimulus package, the largest in U.S. history, amounts to an annual average of 2.5% of gross domestic product (GDP) in each of the 2 years, with approximately one-third of the package allocated to tax cuts. Amounting to US$288 billion, the tax cuts exceeded the magnitude of any previous tax cuts in the United States (Ahern 2004).

Despite the magnitude of the stimulus package, economists such as Krugman (2009) and Feldstein have criticized the package for being both too small and too cautious in light of the weakness of economic indicators, including the sharpest fall in employment since the great depression. Consequently, both authors predicted the need for a second stimulus package. Notwithstanding Krugman's and Feldstein's observations on the relative weights of expenditure increases and tax cuts in the package, the public debate over the stimulus package mainly focused on the size of the package, whereas the macroeconomic effects of the composition and, in particular, the composition of its tax component were somewhat ignored. By analyzing the macroeconomic impact of disaggregated tax shocks, we concentrate on the composition of fiscal stimuli.

The previous literature has suggested that different tax components have different effects on macroeconomic variables in the long run. (1) Yet, the effects of alternative tax instruments on short-run macroeconomic activity has only been indirectly addressed by Arin et al. (2009), who focus on the effects of tax policy on financial markets. While the latter authors also use quarterly data to study the United States in addition to Japan and Germany, we focus on the United States and differ in two dimensions: first, Arin et al. (2009) use a recursive partial identification scheme, whereas we choose a structural vector autoregressive (SVAR) framework building on the seminal work by Blanchard and Perotti (2002); second, while Arin et al. (2009) only include one tax variable at a time, we study the impact of indirect, labor, and corporate tax shocks accounting for the dynamics of the other sources of tax revenue; moreover, we study the impact of separating social security from income taxes. (2-3)

Our choice of an SVAR model has the recognized advantage of incorporating institutional information about the tax system to identify automatic responses of taxation and spending to output and other macroeconomic variables. Ultimately this knowledge allows for the identification of structural fiscal innovations. The effects of these fiscal innovations on output, inflation, and the interest rate may then be traced out. SVAR models based on the Blanchard and Perotti (2002) framework, with modifications to the selection of macroeconomic variables, have been applied to the United States and other Organization for Economic Cooperation and Development (OECD) economies. (4)

Our investigation of tax policies on economic activity contributes to the literature by focusing on the short-term dynamic effects of shocks to disaggregated tax policy instruments. Differentiating among corporate, labor, and indirect taxes, we empirically investigate whether policymakers should pay closer attention to the composition of their stimulus packages. Our results suggest that disaggregating tax shocks helps us to scrutinize the hidden fabric. The positive tax multipliers documented by the previous literature for total tax shocks are only found for indirect taxes when disaggregating tax revenues. Hence, the composition of the fiscal stimulus might help us to understand the cross-country variation in the success rate of fiscal stimulus packages. We also provide a simple theoretical model that can account for the seemingly puzzling positive multipliers for indirect tax innovations.

Among the different tax innovations considered, we find the largest output multipliers in the case of labor tax shocks. This finding suggests that income tax cuts are more likely to achieve success in stimulating the economy. The positive and negative responses of inflation following respectively corporate and labor tax shocks provide evidence of the former shocks working through aggregate supply, whereas the latter work predominantly through aggregate demand.

In the following, Section II briefly summarizes the related literature on the effects of government spending and taxation. Section III outlines a theoretical model with government sector. Section IV describes the data, before our baseline SVAR specification is presented in Section V. Section VI summarizes the findings, including impulse response diagrams and fiscal multipliers. Section VII concludes.

II. RELATED LITERATURE

The recent financial crises have seen a renewed interest in the effects of fiscal policy on macroeconomic activity. For instance, Agnello and Sousa (2011) point out significant multiplier effects in the face of severe housing busts, supporting the recent implementation of fiscal stimulus packages. Agnello and Sousa (2013) use a panel of industrialized countries and show that a positive fiscal shock has a negative (although temporary) impact on stock prices and a negative (persistent) effect on housing prices. Consequently, they argue that the attempts of fiscal policy to mitigate stock price developments may severely destabilize housing markets. Preceding these studies, Blanchard and Perotti (2002) pioneered the application of the SVAR model to fiscal policy questions starting with two fiscal variables, government spending and net taxes. Subsequently, the SVAR methodology has been applied in search of transmission channels of fiscal shocks with various model modifications and additional variables. (5) Vis-a-vis standard VAR models, the SVAR model has the advantage in its identification being less restrictive on the contemporaneous effects of the endogenous variables. Key contemporaneous parameters, which are assumed to equal zero under the Choleski ordering (recursive partial identification scheme), can be identified ex ante in the SVAR model. In incorporating information on, for instance, the contemporaneous effect of output on tax revenue and government spending, the SVAR approach may be seen as providing a more robust model for the characterization of the short-run effects of government spending and taxation shocks. (6)

Among U.S. studies, the qualitative results for the responses of major macroeconomic variables to shocks to aggregated government spending and net taxation are fairly consistent across VAR studies using different identification schemes (7): Positive shocks to net taxes (8) generally result in negative responses of output (mostly significant) and prices (often insignificant) in the short and long terms. Despite the fairly consistent qualitative results across the four major VAR identification approaches on output and prices, there is significant variation in the findings on the magnitude of the effects.

SVAR studies for the United States (Blanchard and Perotti 2002; Caldara and Kamps 2008; Perotti 2002, 2005, 2007) find that positive shocks to net taxes have a negative impact on output. Blanchard and Perotti (2002) find that a positive shock to government spending has a positive effect on output, while unexpected increases in taxation have a negative effect on output. They also report that a positive shock to either taxation or spending has a strong negative impact on investment, indicating substantial crowding out effects.

The SVAR studies conducted by Caldara and Kamps (2008) and Perotti (2007) are comparative in nature. Using SVAR, narrative, and sign restrictions approaches, the former authors confirm that the mixed results provided by different identification schemes used for tax policy shocks cannot be reconciled by merely controlling for differences in model specification across VAR studies. (9) The latter author compares the results from an SVAR and a narrative approach focusing on the effect of government spending shocks. He concludes in favor of the SVAR approach due to the issues arising with the identification of truly exogenous fiscal shocks when applying the narrative approach. We adopt the SVAR methodology in our investigation of the effects of disaggregated tax shocks on output, prices, and the interest rate. This choice is motivated by (1) the SVAR approach being more explicit than other methodologies in its assumptions, (2) the robust results found for both output and price as well as transmission effects of fiscal shocks across SVAR studies (Fry and Pagan 2007; Perotti 2007), and (3) the unavailability of dates which could be used to identify disaggregated tax shocks. (10)

III. A RAMSEY MODEL

We provide a simple theoretical model with government as a point of reference for our empirical analysis. Using the Ramsey model, the workhorse model in modern macroeconomics, we keep the presentation parsimonious. (11) There is neither technical change nor population growth. The production function for final output is

(1) y = [Ak.sup.[alpha]] 0 < [alpha] < 1

where k represents the capital stock per capita. Competitive factor markets imply that the rental rate of capital and the wage rate are given by, respectively,

(2a) r = [alpha][Ak.sup.[alpha]-1]]

and

(2b) w = (1 - [alpha]) [Ak.sup.[alpha]].

The utility function of the representative individual is of the constant relative risk aversion (CRRA) type

(3)

u(c) = ([c.sup.1-[gamma]] - 1) / (1 - [gamma]), with [gamma] > 0

with the coefficient of relative risk aversion given by [gamma]. The representative individual maximizes its lifetime utility subject to the capital accumulation constraint

(4) k = (1 - [[tau].sub.y]) [rk + w] - [[tau].sub.c] C

where [[tau].sub.y], the tax on income, and [[tau].sub.c], the tax on consumption are exogenously given, along with the returns to capital and labor, r and xv. The government sets taxes to finance government spending as a fraction, [s.sub.g], of total output, namely g = [s.sub.g] y. It maintains a balanced budget, which may be expressed as

(5) [[tau].sub.y] [rk + xv] + [[tau].sub.c]c = g = [s.sub.g]y.

With this specification, an increase in the tax on consumption can be balanced with a reduction in the tax on income. A competitive equilibrium of the model consists of paths of per capita consumption, the capital stock, wage rates, and rental rates of capital, such that factor prices w and r are given by Equations (2a) and (2b), and the representative individual maximizes Equation (3) subject to Equation (4) given initial per capita capital stock holdings k(0) > 0, factor prices and the government policy given by Equation (5).

To save space, we refrain from an analytical representation of the steady state of the model, which is rather standard, and focus on the numerical calibration of the model and the tax shock. In order to solve the model numerically, we set the parameter values as in Table I. These parameter values are standard in the literature. Owing to the findings of positive output multipliers in the previous empirical literature and in our own analysis, we apply a hypothetical indirect tax shock to the model economy by increasing the tax rate on consumption from 10% to 15%. This tax shock leads to a positive output response as depicted in Figure 1.

Intuition can be gathered from inspecting the dynamic responses of consumption and income tax revenue variables in the model as shown in Figure 2. The increase in the tax on consumption, in combination with the decline in distortionary income taxes, increases savings from the beginning. As a result, capital accumulates and produces the positive output effect observed in Figure 1. Driven by the specification of the budget constraint in Equation (5), the behavior of income tax revenue and output in the theoretical model are consistent with the empirical response of direct tax revenues and output in Figure 7.

IV. DATA

The data are obtained primarily from the OECD Economic Outlook Database, and comprise U.S. quarterly observations from the third quarter of 1972 to the last quarter of 2008. (12) Details on the these data series used are contained in Table 2. The sample period covers all quarters after the fall of the Bretton-Woods system. All nominal values are deflated by the GDP deflator. All variables, with the exception of the interest rate and inflation rate are in natural logs. All series are detrended with a deterministic trend.

Following Perotti (2005), the short-term interest rate on government bonds is used as a proxy for monetary policy shocks. (13) Direct taxation components include income, social security, and corporate taxes, which Kneller et al. (1999) found to have long-run growth-retarding effects. Indirect taxation comprises sales taxes and tariffs, for which evidence suggests no or significantly less growth-retarding long-run effects (Kneller et al. 1999). We should note that the United States is the only OECD country without value added taxes. Yet, we believe that the relevance of our results extends to other OECD countries, because the value added tax is ultimately also a consumption tax and, as such, not expected to distort individuals' labor-leisure choice. Building a bridge to the studies investigating the longrun growth effects of taxation, we chose to use taxes gross of transfers in contrast to the net-tax measure used in other studies. We further disaggregate direct taxes into corporate and labor taxes. Labor taxation includes taxes on household income (including the capital tax) and social security taxes. (14)

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

V. METHODOLOGY

We use an SVAR framework and identification procedure following the seminal work by Blanchard and Perotti (2002) and extensions thereto which include additional variables (Perotti 2002, 2005). As an extension to the aforementioned papers, our benchmark SVAR model includes seven macroeconomic variables and may be represented as

(6) [x.sub.t] = c + A(L)[x.sub.t-1] + [u.sub.t]

where the vector [x.sub.t] collects the seven endogenous variables,

[x.sub.t] = (government spending, indirect taxation, labor taxation corporate taxation, output, inflation, interest rate),

and A(L) is a lag coefficient polynomial with a lag length of four periods. (15) c is a vector of constants. Finally, the vector [u.sub.t], = ([u.sub.gt], [u.sub.st], [u.sub.lt], [u.sub.bt], [u.sub.yt], [u.sub.pt], [u.sub.rt]) comprises the reduced form error terms, which are serially uncorrelated but generally contemporaneously correlated. As the reduced form residuals [u.sub.t] are merely linear combinations of the structural shocks [v.sub.t], = ([v.sub.gt]), [v.sub.st], [v.sub.lt], [v.sub.bt], [v.sub.yt], [v.sub.pt], [v.sub.rt]), (16) each reduced form error term uit can be decomposed into a linear combination of reduced form error terms [u.sub.j], (J [not equal to] i) of all other variables plus the structural error term of the variable itself denoted [v.sub.it],:

(7) [u.sub.t], = A[u.sub.t], + B[v.sub.t]

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

We are interested in investigating the effects of the structural fiscal shocks to government spending, as well as to indirect, labor, and corporate taxation ([v.sub.gt], [v.sub.st], [v.sub.lt], and [v.sub.bt]) on the remaining macroeconomic variables (output, prices, and interest rates), hence the choice of matrices A and B. To recover the structural errors, we can rewrite the system in Equation (8) to see that the reduced form errors are in fact only functions of the structural errors (17):

(8) ([I.sub.7] - A) [u.sub.t] = [Bv.sub.t],

which may be written as

(9) [u.sup.CA.sub.t] [equivalent to][u.sub.t] = ([I.sub.7] - A) [u.sub.t] = B[v.sub.t].

In general, if we have a system of N endogenous variables, then identification of all parameters in Equation (7) requires the imposition of [([N.sup.2])--N]/2 restrictions on the parameters in matrices A and B. With N = 7, we need to identify the values of at least 21 parameters in A and B. Based on empirical or theoretical ex ante information about structural relations between the variables, these parameters (elasticities) can be either zero or non-zero constants. (18) Equation (10) is replicated below imposing the restrictions on the parameters.

The structural fiscal error terms [v.sub.gt], [v.sub.st], [v.sub.lt], and [v.sub.bt] capture the discretionary, but random, spending and tax policy shocks, which we are interested in recovering. The parameters [[alpha].sup.i.sub.j] in matrix A capture a combination of two types of shocks: (a) systematic automatic changes in spending and taxation in response to changes in other macrovariables (output, price level, and interest rates) and (b) systematic discretionary policy responses of fiscal variables to other macrovariables. Owing to policy lags, the contemporaneous effects [[alpha].sup.i.sub.j] of non-fiscal macrovariables on fiscal variables are assumed to only capture the automatic responses of fiscal policy variables to other macrovariables. Therefore, they can be identified ex ante using structural information about the tax and transfer systems.

In regards to the fiscal policy variables, systematic discretionary spending decisions can effect within quarter discretionary taxation decisions, and vice versa, depending on whether taxation or spending decisions are made first. As the [[beta].sup.i.sub.j] do not only capture automatic responses, they must be estimated. In order to do so, we orthogonalize the structural residuals in the model by imposing restrictions on the [[beta].sup.i.sub.j] parameters. Following Perotti (2007), we assume that spending decisions are made prior to taxation decisions which means that spending effects taxes contemporaneously, but not the other way around. Hence, [[beta].sup.g.sub.j] = 0, j [member of] {s,l,b}. Also, indirect taxation decisions come before labor and corporate tax decisions. (19) So, [[beta].sup.s.sub.l] = [[beta].sup.s.sub.b] = 0. Finally, labor tax decisions come before corporate tax decisions implying [[beta].sup.l.sub.b] = 0. The imposition of these restrictions rentiers matrix B lower triangular. Identification requires the imposition of a further 15 restrictions on the parameters [[alpha].sup.i.sub.j] in matrix A. As we are not interested in recovering the structural shocks to the three non-fiscal variables, we simply impose a Cholesky ordering, that is, [[alpha].sup.y.sub.p] = [[alpha].sup.y.sub.r]. = [[alpha].sup.p.sub.r] = 0. To identify the SVAR, this leaves us with 12 parameters [[alpha].sup.i.sub.j] to be estimated ex ante. (20) In the computation of the 12 elasticities to be determined, we follow Blanchard and Perotti (2002) and Perotti (2005). Table 3 displays the elasticities of the tax revenue components with respect to output and prices. In determining the remaining six elasticities, we also build on Perotti (2005).

Incorporating all the ex ante identified and estimated parameters into Equation (10) yields--as reflected in matrices [TAU] and B--the following relation between the contemporaneous reduced form errors u and the structural error terms v (21):

(10) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

[FIGURE 3 OMITTED]

The identification of the contemporaneous coefficients, along with the lag coefficients (four lags) estimated from the reduced form of the VAR, allows us to trace the output, price, and interest rate effects in response to alternative tax shocks. The dynamic responses of these variables to a one-standard deviation positive shock to the respective fiscal variable at time zero (t = 0), are graphically depicted in the form of impulse response functions (IRFs). One-standard deviation Monte Carlo confidence bands, corresponding to a 10% level of significance, are drawn around the impulse response functions.

VI. RESULTS

A. Five Variable Model

The impulse response functions to a onestandard deviation shock to real government expenditures in the five variable model, which does not disaggregate taxes, are presented in Figure 3. Consistent with the previous literature (Blanchard and Perotti 2002; Burnside et al. 2004; Fatas and Mihov 2001 among others), we observe a positive response of output as well as inflation to a positive real government spending shock. The response of the interest rate is negative, which is suggestive of budget deficits being monetized. The impulse response functions of a similar shock to tax revenues are illustrated in Figure 4. Consistent with Blanchard and Perotti's (2002) seminal paper, a negative response of output is documented. The effect on inflation is small, yet positive, pointing to a supply side effect of aggregate tax shocks. The effect on the interest rate is positive.

[FIGURE 4 OMITTED]

B. Six Variable Model

In order to understand the importance of disaggregating taxes, we separate direct tax revenues (income, social security, and corporate taxes) and indirect tax revenues (tax on goods and services). The impulse response functions for a shock to government expenditures (Figure 5) and a shock to direct tax revenues (Figure 6) remain similar to those we obtained for the five variable model. Our findings with regard to general tax and direct tax revenue shocks are in line with Afonso and Sousa (2012), who report a negative response of output with respect to government revenue shocks in a Bayesian SVAR including asset prices. Yet, we document positive output multipliers (similar to Perotti 2002) and negative inflation multipliers for indirect taxes (Figure 7).

Given the fact that consumption taxes affect the economy through aggregate demand, the negative response of inflation is hardly surprising. The interest rate response to an indirect tax shock, diverging from our previous results, is negative. Positive output multipliers in the case of general government revenue shocks, as opposed to indirect tax revenue shocks, were also found by Afonso and Sousa (2011) and Agnello and Sousa (2013). The former authors find positive output multipliers in a Bayesian SVAR when they include debt feedback into their empirical model. With the effect being negative on impact, their findings suggest positive output multipliers after six quarters. While turning positive earlier in our study, output multipliers are also negative on impact. The latter authors find evidence of positive output multipliers, which are on impact negative, for fiscal shocks in a panel VAR. Both Agnello and Sousa (2013) as well as Afonso and Sousa (2011) include asset prices in their empirical models, with the latter incorporating debt feedback in addition and the former including credit conditions. Ascertaining the role these variables play in the documented positive output multipliers of the shocks studied deserves further study.

[FIGURE 5 OMITTED]

[FIGURE 6 OMITTED]

[FIGURE 7 OMITTED]

[FIGURE 8 OMITTED]

[FIGURE 9 OMITTED]

C. Seven Variable Model

Following Alesina et al. (2002), we further disaggregate taxes into labor and corporate taxes. Impulse response functions to a government spending shock are presented in Figure 8 and are fairly similar to our previous results in terms of sign, if not magnitude. With regards to tax revenue shocks, our results suggest that the macroeconomic effects of shocks to labor (Figure 9) and corporate (Figure 10) income are fairly similar to each other, with output and interest rate multipliers for labor taxes both being slightly larger in magnitude. The main difference between the two types of tax shocks lies in the response of inflation, which is positive for corporate and negative for labor tax shocks. (22) This result implies that corporate tax shocks work mainly through aggregate supply, whereas the transmission mechanism for labor tax shocks lies predominantly on the demand side. Given the widespread use of this classification of taxes in the wider literature, we treat this VAR specification as our benchmark model.

For the readers' convenience, the peak and cumulative responses of the three empirical models are presented in Table 4. In the next part, we present out-of-sample predictions of the six and seven variable models (Appendix SI, Supporting Information).

D. Out-of-Sample Predictions

We update the real GDP time series with realized data extended to include 16 more quarters from 2008:4 to 2012:3. Thereby, we conduct out-of-sample predictions 16 steps ahead (see Figures 11 and 12). The predicted average quarterly GDP growth rate over this period is 38.1 bps for the six variable model and 38.2 bps for the seven variable model, which are both very close to the actual average growth rate of 38.0 bps. The root-squared-mean-deviations (RSMDs) of 59 and 66 bps, respectively, however, are arguably not great. This is perhaps not surprising given that our sample period spans part of the great recession and recovery therefrom, a period of economic turmoil, which is unprecedented in the post-World War II era. VII. VII.

[FIGURE 10 OMITTED]

VII. CONCLUSION

This article contributes to the literature investigating the effects of fiscal policy on short-run economic performance by analyzing the effects of disaggregated tax shocks on output, inflation, and interest rates in an SVAR framework. Our results suggest that disaggregated tax shocks have rather different effects on the economy, which underlines the importance of giving careful consideration to the composition of tax-based fiscal stimulus packages.

The negative output multipliers found for corporate and labor tax shocks, where we consider labor tax revenues inclusive and exclusive of social security tax revenues, are in line with the sign observed for aggregate net tax shocks in the literature. Yet, we provide evidence of positive output multipliers for indirect taxes. We provide a simple theoretical model to account for the behavior of output. Nevertheless, this finding deserves further research, which is beyond the scope of this article. Given suitable data, we believe that an investigation into the shortterm effects of consumption taxes using the narrative approach should prove useful. At this point, we may merely view the finding as further support of consumption as a tax base instead of income. We conclude that in order to better understand the macroeconomic effects of tax shocks, more focus on the analysis of disaggregated tax shocks may not only help to improve our understanding of the consequences of alternative tax-based stimulus packages but may also help to explain the cross-country variation in tax multipliers. Further research might incorporate our research into different countries and different sample periods.

[FIGURE 11 OMITTED]

[FIGURE 12 OMITTED]

doi: 10.1111/coep.12131

REFERENCES

Acemoglu, D. introduction to Modern Growth. Princeton. NJ: Princeton University Press, 2009.

Afonso, A., and R. Sousa. "What Are the Effects of Fiscal Policy on Asset Markets." Economic Modelling. 28(4), 2011, 1871-90.

--. "The Macroeconomic Effects of Fiscal Policy." Applied Economics, 44(34), 2012, 4439-54.

Agnello, L., and R. Sousa. "Can Fiscal Policy Stimulus Boost Economic Recovery?" Revue economique, 62(6). 2011. 1045-66.

--. "Fiscal Policy and Asset Prices." Bulletin of Economic Research, 65(2), 2013, 154-77.

Ahem. W. Comparing the Kennedy, Reagan and Bush Tax Cuts. Tax Foundation: Washington, DC, 2004.

Alesina, A., S. Ardagna, R. Perotti. and F. Schiantarelli. "Fiscal Policy, Profits and Investment." American Economic Review, 92(3), 2002, 571-89.

Arin, K., A. Mamun, and N. Purushothman. "The Effect of Tax Policy on Financial Markets: G3 Evidence." Review of Financial Economics, 18(1), 2009, 33-46.

Beetsma, R., and M. Giuliodori. "The Effects of Governments Purchases Shocks: Review and Estimates for the EU." The Economic Journal, 121(555), 2010, F4--F32.

Blanchard, O., and R. Perotti. "An Empirical Characterization of the Dynamic Effects of Changes in Government Spending and Taxes on Output." Quarterly Journal of Economics, 117(4), 2002, 1329-68.

Burnside, C., M. Eichenbaum, and J. Fisher. "Fiscal Shocks and Their Consequences." Journal of Economic Theory, 115(1), 2004, 89-117.

Caldara, D., and C. Kamps. "What Are the Effects of Fiscal Shocks? A VAR-Based Comparative Analysis." ECB Working Paper No. 877. European Central Bank, 2008.

Canzoneri, M., R. Cumby. and B. Diba "Should the European Central Bank and the Federal Reserve Be Concerned About Fiscal Policy?" Working Paper. Federal Reserve Bank of Kansas City, 2002. Accessed September 30, 2015. https://www.kansascityfed.org/publicat/ sympos/2002/pdf/s02canzoneri.pdf.

de Castro, F. and P. Hernandez de Cos. "The Economic Effects of Exogenous Fiscal Shocks in Spain: A SVAR Approach." Banco de Espana Working Papers: 0604. Banco de Espana, Madrid. 2006.

Fatas, A. and I. Mihov. "The Effects of Fiscal Policy on Consumption and Employment: Theory and Evidence." CEPR Discussion Papers: 2760. C.E.P.R., 2001.

Favero, C., and F. Giavazzi. "Reconciling VAR-Based and Narrative Measures of the Tax Multiplier." American Economic Journal: Economic Policy, 4(2), 2012, 69-94.

Fry. R. and A. Pagan. "Some Issues in Using Sign Restrictions for Identifying Structural VARs." NCER Working Paper Series, No. 18. National Centre for Econometric Research, 2007.

Giordano, R., S. Momigliano, S. Neri, and R. Perotti. "The Effects of Fiscal Policy in Italy: Evidence from a VAR Model." European Journal of Political Economy, 23(3), 2007, 707-33.

Gordon, D., and E. Leeper. "The Price Level, the Quantity Theory of Money, and the Fiscal Theory of the Price Level." Scottish Journal of Political Economy, 53(1), 2006, 4-27.

Heppke-Falk, K., J. Tenhofen, and G. Wolff. "The Macroeconomic Effects of Exogenous Fiscal Policy Shocks in Germany: A Disaggregated SVAR Analysis." Journal of Economics and Statistics, 230(3), 2010, 328-55.

von Kalckreuth, U., and G. B. Wolff. "Identifying discretionary Fiscal Policy Reactions with Real-Time Data." Journal of Applied Econometrics, 43(6), 2011, 1271-85.

Kneller, R., M. Bleaney, and N. Gemmell. "Fiscal Policy and Growth: Evidence from OECD Countries." Journal of Public Economics, 74(2), 1999, 171-90.

Krugman, P. "Behind the Curve," New York Times, March 9, 2009, A23.

Lee, Y, and R. Gordon. "Tax Structure and Economic Growth." Journal of Public Economics, 89(5-6), 2005, 1027-43.

Leeper, E., T. Walker, and S. Yang. "Fiscal Foresight: Analytics and Econometrics." NBER Working Papers: 14028. National Bureau of Economic Research, Inc., 2008.

Mountford, A., and H. Uhlig. "What Are the Effects of Fiscal Policy Shocks?" Journal of Applied Econometrics, 24(6), 2009, 960-92.

Neri, S. "Assessing the Effects of Monetary and Fiscal Policy." Temi di discussione (Economic Working Papers), No. 425. Bank of Italy, Economic Research Department, 2001.

Perotti. R. "Estimating the Effects of Fiscal Policy in OECD Countries." European Central Bank. Working Papers Series: 168. European Central Bank, Frankfurt, 2002.

--. "Estimating the Effects of Fiscal Policy in OECD Countries." CEPR Discussion Papers: 4842. CEPR, 2005.

--. "In Search of the Transmission Mechanism of Fiscal Policy," in NBER Macroeconomics Annual. Chicago: University of Chicago Press, 2007, 169-249.

Ramey, V. "Identifying Government Spending Shocks: It's All in the Timing." Quarterly Journal of Economics, 126(1). 2011, 1-50.

Romer, C. D., and D. H. Romer. "The Macroeconomic Effects of Tax Changes: Estimates Based on a New Measure of Fiscal Shocks." American Economic Review, 100(3), 2010, 763-801.

SUPPORTING INFORMATION

Additional Supporting Information may be found in the online version of this article:

Appendix S1. Responses of shocks to corporate and labor taxes (1972:03-2008:04)

The Appendix includes two robustness checks: (1) we disaggregate labor taxes into income and social security taxes and present results for the case when social security taxes are excluded and (2) we replace the short-term interest rate with the long-term interest rate. In both cases our main results remained essentially the same.

K. PEREN ARIN, PETER H. HELLES, MURAT KOYUNCU and OTTO F. M. REICH*

* We would like to thank B. Hayo, F. Koray, W. Bosnian, and two anonymous referees, the seminar participants at Helmut Schmidt University, Philipps-University Marburg, Giessen University. RWTH Aachen. ESSEC Paris and University of Tromso, the conference participants at the NZESM 2010 in Auckland, the AMW 2010 in Wellington, and EcoMod 2010 in Istanbul for useful comments and suggestions, as well as R. Perotti for sharing his RATS code. Usual disclaimer applies. The views expressed in this article are not necessarily those of Bank of America Merrill Lynch.

Arm: Professor, College of Business, Zayed University, Abu Dhabi. P.O. Box 144534, UAE; CAMA. Canberra, Australia. Fax (+971) 4 402 1015, E-mail kerim.arin@zu.ac.ae

Helles: Commodity Strategist, Global Research. Bank of America Merrill Lynch, London, EC 1A 1HQ, UK. Phone 0044 79 3140 3185, E-mail peter.helles@baml.com

Koyuncu: Assistant Professor, Department of Economics, Bogazici University, Istanbul. 34342. Turkey. Phone +902123597640, Fax +902122872453, E-mail mkoyuncu@boun.edu.tr

Reich: Assistant Professor, College of Business, Zayed University, Dubai, P.O. Box 19282, UAE. Phone (+971) 56 9500991, Fax (+971) 4 402 1015, E-mail Otto.Reich@zu.ac.ae

ABBREVIATIONS

CRRA: Constant Relative Risk Aversion

GDP: Gross Domestic Product

IRFs: Impulse Response Functions

OECD: Organization for Economic Cooperation and Development

RSMDs: Root-Squared-Mean-Deviations

SVAR: Structural Vector Autoregressive

VAR: Vector Autoregressive

(1.) See, for instance, Knelleret al. (1999), Lee and Gordon (2005), as well as Gordon and Leeper (2006).

(2.) We thank an anonymous referee for the suggestion of disaggregating labor taxes.

(3.) While Arin et al. (2009) use data from 1973 to 2005. we study the 1972 to 2008 period.

(4.) Perotti (2005) and Perotti (2007) study the United States, while Australia, Germany, Italy, Spain, and the United Kingdom have respectively been studied in Perotti (2005), Heppke-Falk et al. (2010), Giordano et al. (2007), and de Castro and Hernandez de Cos (2006). Beetsma and Giuliodori (2010) provide estimates for fiscal multipliers in 14 European Union (EU) countries.

(5.) See, for instance. Perotti (2005), Perotti (2007), Caldara and Kamps (2008), Afonso and Sousa (2011). as well as Perotti (2005), Perotti (2007), Caldara and Kamps (2008), Afonso and Sousa (2011).

(6.) Of course, the SVAR approach relies on the ability to ex ante identify the automatic responses of tax revenues and spending, some of which are derived from theory and some of which are estimated.

(7.) See, for instance, Blanchard and Perotti (2002), Perotti (2005), Neri (2001) using the SVAR methodology, Canzoneri et al. (2002) as well as Mountford and Uhlig (2009) employing the agnostic approach, and Romer and Romer (2010) follow the narrative approach.

(8.) Net taxes are defined as the sum of indirect and direct taxes, less transfer payments.

(9.) Favero and Giavazzi (2012) argue that the difference in estimated tax multipliers obtained from the narrative approach and in fiscal VAR models is due to different estimation approaches as opposed to being driven by the nature of the shocks.

(10.) Leeper et al. (2008) question the validity of the identification strategies of SVAR and narrative approaches in the context of a discussion of the consequences of fiscal foresight. Although the narrative approach is unaffected by fiscal foresight, it has been criticized for achieving "identification through a variety of heroic--and often implicit--identifying assumptions" (Leeper et al. (2008)). More recent criticisms of the SVAR approach include Ramey (2011) and von Kalckreuth and Wolff (2011); the former providing further insights into the question of the anticipation of fiscal policy and the latter questioning the speed of reaction of fiscal policy to events.

(11.) See Acemoglu (2009) among other textbooks for the details of the model.

(12.) In Section VI.D, we use real GDP data up to 2012:3 obtained from the same database.

(13.) We also use the long-term interest rate in a robustness check.

(14.) In a second robustness check, we exclude social security tax revenues from labor tax revenues.

(15.) The choice of lag length follows Blanchard and Perotti (2002) and is confirmed by a Lagrange multiplier test. We also use the Akaike information criterion which suggests six lags for the six variable model (direct and indirect taxes), and three lags for the seven variable model (labor, corporate, and indirect taxes). Given the varying results for optimal lag length, we stick to the choice of lag length 4, due to this being the convention in SVAR studies using quarterly data (Blanchard and Perotti 2002; Perotti 2002, 2005, 2007) and the fact that four lags are statistically significant at the 5% level in our model specifications.

(16.) The vector v, collects the structural errors which are assumed to be contemporaneously and serially uncorrelated.

(17.) We require matrix B to be invertible such that a solution [v.sub.t] = [B.sup.-1] ([I.sub.7] - A) [u.sub.t], exists. Moreover, for a unique solution, we require that [B.sup.-1] ([I.sub.7] - A) be of full rank.

(18.) Given the definition of the variables, the coefficients have the interpretation of elasticities or semi-elasticities.

(19.) In the six variable model, which includes government spending and two tax variables (direct and indirect tax revenues), we similarly rank indirect taxation before direct taxation. In both models, the results are robust to different orderings of the tax revenue decisions.

(20.) Note that the parameters [[gamma].sup.i.sub.j], capturing the contemporaneous effect of the system variables on non-fiscal variables, are estimated after imposing the relevant restrictions on the parameters [[alpha].sup.i.sub.j] and [[beta].sup.i.sub.j].

(21.) The estimated coefficients in the matrices are displayed here as rounded to one. or in some cases two, decimal places. Whenever the estimated contemporaneous parameters ([[gamma].sup.i.sub.y] and [[beta].sup.i.sub.j]) are found to be statistically insignificant at the 10% level, we set their values to zero for the purpose of obtaining results and drawing impulse response functions.

(22.) Given the respective sizes of the interest rate and inflation multipliers, the corporate tax shock appears to imply an increase in the real interest rate in contrast to labor tax shocks. The same observation holds for the robustness check, where we use the long-term interest rate.
TABLE 1

Parameter Values for the Model Economy

Technology Parameters
 Scale parameter, A                                          1
 Share of capital, [alpha]                                 .33

Preference Parameters
 Coefficient of relative risk aversion, [gamma]             -2
 Rate of time preference, [beta]                           .04

Fiscal Policy Parameters
 Share of government spending in output, [s.sub.g] (%)      20
 Tax rate on income, [[tau].sub.y] (%)                      12
 Tax rate on consumption, [[tau].sub.c] (%)                 10

TABLE 2

Data--Definitions and Sources

Variable               Names According to OECD Outlook Database

GDP                    Gross domestic product at market prices
Consumption            Private final consumption expenditure
Government spending    Government final consumption expenditure
                          + government fixed capital formation
Interest rate--long    Long-term interest rate on government bonds
Interest rate--short   Short-term interest rate
Price level            GDP deflator (market prices)
Net taxes              Indirect taxes + direct taxes + social
                          security contributions by households +
                          (total current transfers paid by
                          households--social security contributions
                          paid by households) + capital tax and
                          transfers receipts--(current transfers
                          received by households + other current
                          transfers received by households)
Net taxes--direct       Total direct taxes + social security
                          contributions by households + (total
                          current transfers paid by households--
                          social security contributions by
                          households) + capital tax and transfers
                          receipts--(current transfers received by
                          households + other current transfers
                          received by households)
Net taxes--indirect     Indirect taxes
Net taxes--direct,      Direct taxes on business + capital tax and
   corporate              transfers receipts value
Net Taxes--direct,      Direct taxes on households + social security
   labor                  contributions by households + (total
                          current transfers paid by households--
                          social security contributions by
                          households)--(current transfers received
                          by households + other current transfers
                          received by households)
Taxes--direct           Total direct taxes + social security
                          contributions by households + capital tax
                          and transfers receipts
Taxes--direct,          Direct taxes on business + capital tax and
   corporate              transfers receipts
Taxes--direct, labor    Direct taxes on households + social security
                          contributions by households
Taxes--indirect         Indirect taxes

Notes: All data are sourced from the OECD. We exclude interest
payments from government spending and exclude property income and
interest receipts from tax revenues in order to make our results
comparable with Perotti (2005). Direct taxes on household income
include other capital taxes and transfers.

TABLE 3

Output and Price Elasticities of Tax Components

Elasticity of Output y and Price p with Respect to

S VAR-7
   Indirect tax    [[alpha].sup.s.sub.y]    [[alpha].sup.s.sub.p]
                   = 1.00 *                 = 0.00 *

   Labor tax       [[alpha].sup.l.sub.y]    [[alpha].sup.l.sub.p]
                   = 0.26                   = 0.83

   Corporate tax   [[alpha].sup.b.sub.y]    [[alpha].sup.b.sub.p]
                   = 4.95                   = 0.00 *

SVAR-6             [[alpha].sup.s.sub.y]    [[alpha].sup.s.sub.p]
   Indirect tax    = 1.00 *                 = 0.00 *

   Direct tax      [[alpha].sup.d.sub.y]    [[alpha].sup.d.sub.p]
                   = 0.89                   = 1.03

Note: An asterisk * denotes those elasticities that were not estimated
but assumed. Having determined the [a.sup.j.sub.i], we can

TABLE 4

Responses to Shocks

                       Five Variable Model   Six Variable Model

                        Peak    Cumulative    Peak    Cumulative

Government spending    0.152      0.228      0.113      0.182
Total taxes            -0.186     0.732
Direct taxes                                 -0.148     -0.798
Indirect taxes                               0.264      1.363
Labor taxes
Corporate taxes

                       Seven Variable Model

                        Peak    Cumulative

Government spending    0.160      0.314
Total taxes
Direct taxes
Indirect taxes
Labor taxes            -0.129     -0.768
Corporate taxes        -0.071     -0.542
COPYRIGHT 2016 Western Economic Association International
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2016 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Arin, K. Peren; Helles, Peter H.; Koyuncu, Murat; Reich, Otto F.M.
Publication:Contemporary Economic Policy
Date:Jul 1, 2016
Words:6613
Previous Article:Do minimum legal tobacco purchase age laws work?
Next Article:Willingness-to-pay for sporting success of football Bundesliga teams.
Topics:

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters