WAS FISCAL STIMULUS IN USA IN RECENT CRISES REASONABLE? EMPIRICAL INVESTIGATION OF DISCRETIONAL FISCAL POLICY ON OUTPUT IN A SVEC FRAMEWORK

Transcription

1 WAS FISCAL STIMULUS IN USA IN RECENT CRISES REASONABLE? EMPIRICAL INVESTIGATION OF DISCRETIONAL FISCAL POLICY ON OUTPUT IN A SVEC FRAMEWORK Adrian Burda Uniwersytet Ekonomiczny w Krakowie Koło Naukowe Makroekonomii i Gospodarki Światowej

2 Fiscal stimulus package 1, which was enacted in 2009, and its effectiveness caused controversy among economists. Some, as Berstein and Romer (2009) found their impact on output and employment as very positive. However others, as Cogan and others (2009) were much more skeptical, stressing their negative impact on private consumption and investment. Today, in the time of weak recovery, question about effectiveness of this stimulus is one of the top. This paper examines the effects of fiscal policy shocks (such as the stimulus package) on the GDP in USA economy for which SVEC (structural vector error correction) approach is used. This method takes on account the cointegration between variables and differentiates between permanent and transitory shocks to identify structural shocks (Krusec, 2003). It can asses short- and long term impact of this shocks. The analysis is based on quarterly, seasonally adjusted data from first quarter 1983 to first quarter Both Chow s structural break and splitting tests, and previous analysis pointed at great structural change from pre and post Great Moderation's period (Krusec, 2003) (Ilzetzky and others, 2009). Thus this sample is choose in to analysis. Selected variables are as follows: Interest rate (y), Taxes Revenues (Tax), Output (GDP), Inflation Rate (i), Government Expenditures (Gov). Five-variables system allows to analyze relations between fiscal policy and financial market and seems to be more reliable than standard three-variable system (Krusec, 2003). The data are expressed in logs. All of them have one unit root, thus they could be included in model in levels. There are four variants of model. Details are shown in table 1. Table 1: Specifications of models. Gov and Tax Interest rate 1 'alla Blanchard & Perotti [2] 10 years treasury bond (10y) 2 'alla Blanchard & Perotti [2] Average 10y and corporate bonds 2 3 All tax. Revenues and gov. expenditures +transfers 4 All tax. revenues and gov. expenditures +transfers Source: own preparation 10 years treasury bond (10y) Average 10y and corporate bonds 1The American Recovery and Reinvestment Act (ARRA) 2With Moody's rating AAA&BAA. Average rate = (10y+AAA+BAA)/3 1

3 For cointegration analysis between variables, Johansen trace test is used (Lϋtkepohl). The Cointegrating Rank differs for various specifications and numbers of lags. Although some information criteria pointed one, or two lags as optimal, just for five lags (for model 4) residuals autocorrelations are insignificant on 0,05 level. For this lags, Johansen trace tests results are as follows: the cointegrating rank for models 1,3 and 4 is two, and for model 2 four. After estimation of unrestricted VEC, there were added some structural restrictions on residuals[6]. For all models there are that assumptions: there are no effects of Gov shocks on Tax and shocks in i on Tax and Gov within quarter. Shocks in Gov, i and GDP have no long run effects on Tax and Gov, and Tax shock have no long run effects on Gov. In model 2 were added two additional contemporaneous restrictions and there shocks in Gov are no effects within quarter on GDP and i. Impulse response from models are shown in tables 1 and 2. They are multiplied such a way, to interpreted change on GDP in % to one fiscal shock, as a 1% of GDP. Surprisingly government expenditures have negative impact in three models in first two years. Further, it is positive, but in the long run, in two models remain negative, and for neither model bigger than one. Negative tax shocks have positive both short and long-run impact on GDP, but in three models it is not significantly bigger than one. Table 2: Impuls response of GDP on government expenditures and taxes revenues shocks 3 1 quarter 8 quarters 20 quarters 40 quarters Max Min Tax Gov Tax Gov Tax Gov Tax Gov Tax Gov Tax Gov 1 0,7-0,3 0,79-0,27 0,81-0,09 0,76-0,12 0,95 (11q) -0,06 0,7 (1q) -0,36 (3q) 2 1,02-0,08 1,18-0,76 1,33 1,03 1,25-0,72 1,69 (12q) 1,03 (20q 0,56 (30q) -1,03 (4q) 3 0,47-0,18 0,6-0,73 1,31 0,92 0,98 0,55 1,36 1,19 0,39 (31q) -1,19 (30q) 4 0,73 0,87 0,29 0,42 0,57 0,54 0,86 0,42 0,86 (37q) 0,92 (2q) 0,29 (7q) 0,37 (5q) Source: own calculations Results from this investigation don't advocate the stimulus package, and overall discretional, spending-based fiscal policy. They give only moderate support for temporal cutting taxes. These 3Positive shocks of 1%GDP on government spending and negative shock (i.e. Cutting taxes) on taxes revenues 2

4 conclusions are quite consistent with another study, but they differs in detail (Berstein and Romer, 2009) (Ilzetzki and others, 2009). However this approach has some limitations, as possibly mistakes in identification of episodes of discretional fiscal policy, or assuming invariant structure of economy in all sample period. Furthermore some coefficient are insignificant. That is why putting constraint on coefficients, SVEC with time varying coefficients, or Bayesian approach is worth to be considered. These issue were left for further research. 3

5 REFERENCES: 1. Blanchard O., Perotti R., An empirical characterization of the dynamic effects of changes in government spendings and taxes on output., The Quarterly Journal of Economics, November 2002,pp , 2. Berstein J., Romer C., The job impact of the American Recovery and Reinvestment Plan, 2009, n.pdf 3. Cogan J.F., Tobias C., Taylor J. B.,Wieland W. New Keynesian versus old Keynesian Government Spending Multipliers, 2009, 4. Krusec. D, The effects of fiscal policy on output in a structural VEC model framework: The case of four EMU and four non-emu OECD countries, European University Institute, Florence, Ilzetzki E., Mendoza E.G., Vegh C.A., How big are fiscal multipliers,? CEPR Policy Insight no.39, Lϋtkepohl, K., New Introduction to multiple time series analysis, Springer Berlin,