* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project
Download The Government Spending Multiplier: Evidence from County Level
Survey
Document related concepts
Transcript
The Government Spending Multiplier: Evidence from County Level Data Christopher A. Erickson, Ph.D. Department of Economic, Applied Statistics, and International Business College of Business New Mexico State University Box 30001, MSC 3CQ Las Cruces, NM 88003 [email protected] August 27, 2012 Acknowledgements I would like to acknowledge the able assistance of Eduardo Saucedo and Fred Owensby. The Government Spending Multiplier: Evidence from county level data By Christopher A. Erickson, Ph.D. Executive Summary The passage of the American Recovery and Reinvestment Act (ARRA) of 2009 has brought fiscal policy to the forefront once again. The size of the “multiplier” of government spending becomes of critical importance for determining the effect of ARRA. Yet there is considerable controversy on this issue with estimates varying from .8 to 1.2. This study seeks to estimate the impact of ARRA on local economic development by exploiting county level data. In particular, this study relies on data obtained from the Census to estimate the government spending multiplier. The spending associated with ARRA represents a relatively unique circumstance in that the expenditure occurred at a time during which the economy was in a sharp contraction and monetary policy was especially accommodating. Thus provides a particular pure test of the use of activist fiscal policy. The main results are presented in Table 1, which appears at the end of this paper. Two different versions of the model were estimated. In each model, lagged log difference of personal income, logged personal income (not differenced), log difference of state personal income and the change in the unemployment rate are included as control variables. In Column 1 of Table 1 is reported the ordinary least square results with each source of government earnings is entered separately, allowing estimation of the elasticity for each of local, state, federal military and federal civilian earnings.1 Column 2 is also estimated using ordinary least squares, but in this case, total government earnings are used with no distinction made as to the level of government. 1 Government sector earnings are calculated using revenue received from the governmental unit. So federal civilian government earnings is the budget allocated to the county to support federal civilian activity in that county. Page 1 of 12 There is reasonable consistency between the models, which gives confidence that the estimates are reasonable. The coefficient on ARRA spending is most troubling. The coefficient is marginally significant and positive in the first model (Column 1), but insignificant in the second model (Column 2), although the coefficients are similar in magnitude between the two models. Table 2 presents the implied government spending multipliers (GSMs) based on the estimated elasticities calculated at the mean value of county personal income and of the respective measures of government earnings. The results are reasonable and fall within the range found by other authors, as indicated in the literature review. The multiplier on ARRA is small and economically insignificant. At least within the one-year sample we have here, it appears that ARRA was not very effective. Page 2 of 12 The Government Spending Multiplier: Evidence from county level data By Christopher A. Erickson, Ph.D. Introduction and Literature Review The passage of the American Recovery and Reinvestment Act (ARRA) of 2009 has brought fiscal policy to the forefront once again. The size of the “multiplier” of government spending becomes of critical importance for determining the effect of ARRA.2 This study estimates the impact of ARRA on local economic development by exploiting county level data. In particular, data from Recovery.gov will be used to estimate the short run government spending multiplier. Because ARRA expenditures are determined exogenous to local economic conditions, the stimulus plan provides a natural experiment to study fiscal policy impact. Because of the detailed information provided by recovery.gov, it is possible to isolate the impact of government spending on the local economy. The spending associated with ARRA represents a relatively unique circumstance in that the expenditure occurred at a time during which the economy was in a sharp contraction and monetary policy was especially accommodating. We would expect the government spending to have maximum impact under these circumstances, thus, providing a particularly favorable environment to test fiscal policy for those who believe that fiscal policy can be used to stabilize the economy. Should the government spending multiplier be found to be small, it would be a strong argument against activist fiscal policy. Either way, the results will be interesting. There is considerable controversy as to the multiplier. Estimates vary from Barro (1981) who found a multiplier of .8 to Ramey (2010) who found a multiplier of 1.2. Hall (1986), using annual data back to 1920, found a slightly negative effect of government purchases on The multiplier determines the effect of a dollar of spending on income. Specifically, 𝑑𝑌 = 𝑚 𝑑𝐺, then m is the multiplier. 2 Page 3 of 12 consumption. An important determinate of the empirical multiplier is the treatment of the timing of the effect of government spending. Traditionally, the assumption has been that the effect of spending occurs in the quarter in which the spending occurs. Others, especially Ramey (2010) and Romer and Romer (2010), argue that the impact of government spending occurs at the time of announcement, which they study by analyzing news reports to determine when positive congressional action is expected. Those studies that assume that the effects of government spending are simultaneous with the actual expenditure of spending typically find that a positive government spending results in increased GDP, hours worked, consumption and real wages (Rotemberg and Woodford, 1992; Blanchard and Perotti, 2002; Fatas and Mihov, 2001; Monacelli and Uhlig, 2002; Perotti, 2005; Caldara and Kamps, 2006; Gali, Lopez-Salido, and Valles, 2007). By contract, studies that assume that government spending affects the economy via an announcement effect, find that while government spending raises GDP and hours, it lowers consumption and the real wage (Ramey and Shapiro, 1998; Edelberg, Eichenbaum and Fisher, 1999; Burnside, Eichenbaum, and Fisher, 2004; Cavallo, 2005; Romer and Romer, 2010). Papers investigating an announcement effect using event study methodology generally show negative effect of government spending on private consumptions (Giavazzi and Paganom, 1990; and Cullen and Fishback, 2006). The ambiguous results found in the empirical literature are reflected in published theoretical models. These models fall into two general categories: Neoclassical models that assume markets clear, and new Keynesian models that assume sticky prices. With Neoclassical models such as those of Aiyagari, Christiano, and Eichenbaum (1992) and Baxter and King (1993), a permanent increase in government spending financed by nondistortionary means (such as a lump sum tax) creates a negative wealth effect. Households respond by decreasing Page 4 of 12 consumption and increasing labor supply, causing output to rise. Increased labor supply causes the real wage to fall and a rise in the real return to capital in the short run. A higher real return causes capital accumulation, which ultimately causes the real wage to return to its original value. At the new steady state, consumption is lower and hours worked are higher. The neoclassical model is at odds with the empirical results of studies using contemporaneous timing. The new Keynesian approach, which is characterized by sticky prices in the short run, seeks to explain a rise in consumption, the real wage, and productivity found in most empirical analyses. For example, Drautzburg and Uhlig (2010), Rotemberg and Woodford (1992) and Devereux, Head and Lapham (1996) use models with imperfect competition and increasing returns to explain the rise in real wages and productivity. Gali, Lopez-Salido, and Valles (2006) showed that only an “ultra-Keynesian” can explain how consumption and real wages can rise when government-spending increases, thereby highlighting the many special features required to explain a positive correlation between consumption and government spending. Typically, New Keynesian models are neoclassical in nature in the long run; therefore, long-run issues tend to be of second order in importance raised here, are surely not a matter of principle, but a matter of degree. The Model Clearly, from the discussion above, economists have no consensus model for evaluating the effects of macroeconomic effects of government spending. To access the empirical effects of the ARRA, following Barro and Redlick (2011), the following model was estimated: (1) (y(i,t) – y(t-1,i))/y(i,t-1) = β0 + β1 ((g(i,t) – g(i,t))/g(i,t-1)] + β2z(i,t) Page 5 of 12 In the equation, y(i,t) is per capita real personal income for county i in year t, g(i,t) is per capital real ARRA expenditures for county i in year t, and z(i,t) is a vector of county level controls for year t. The above would be estimated using data on government expenditure from available through Recovery.gov, which includes ARRA expenditures at the county level, by zip code. Data on county level for control variables will be obtained from Bureau of Economic Analysis (www.bea.gov), and the Bureau of the Census (www.census.gov). In Eq. (1), the main variable of interest is β1, which gives the elasticity of personal income with respect to government spending. Methodology and Results The Eq (1) was estimated using cross sectional techniques with a potential n= 3143—the number of counties in the United States. After accounting for missing data, the sample size was reduced to 2766. Personal income and earnings are available for multiple years, so growth rates could be calculated, but data for ARRA expenditures by county was only available for 2009, so eq (1) could be estimated only for that one year. The main results are presented in Table 1, which appears at the end of this paper. Two different versions of the model were estimated. In each model, lagged log difference of personal income, logged personal income (not differenced), log difference of state personal income and the change in the unemployment rate are included as control variables. In Column 1 of Table 1 is reported the ordinary least square results with each source of government earnings is entered separately, allowing estimation of the elasticity for each of local, state, federal military and Page 6 of 12 federal civilian earnings.3 Column 2 is also estimated using ordinary least squares, but in this case, total government earnings are used with no distinction made as to the level of government. There is reasonable consistency between the three models estimated, which gives confidence that the estimates are reasonable. The coefficient on ARRA spending is most troubling. The coefficient is marginally significant and positive in the first model (Column 1), but insignificant in the second model (Column 2), although the coefficients are similar in magnitude between the two models. . Table 2 presents the implied government spending multipliers (GSMs) based on the estimated elasticities calculated at the mean value of county personal income and of the respective measures of government earnings.4 The results from ordinary least squares (Column 1 and Column 2) are reasonable and fall within the range found by other authors, as indicated in the literature review. The multiplier on ARRA is small and economically insignificant. At least within the one-year sample we have here, it appears that ARRA was not very effective. The problem here may be that ARRA data reflects more projects approved rather than projects actually undertaken. This, of course, is precisely the criticism of ARRA put forward by opponents of the program—that actually expenditure on projects were delayed, mitigating the impact of ARRA on the economy. 3 Government sector earnings are calculated using revenue received from the governmental unit. So federal civilian government earnings is the budget allocated to the county to support federal civilian activity in that county. 𝑌𝑃 4 The formula for calculating the government spending multiplier is 𝐺𝑆𝑀 = 𝛽̂𝑖 where 𝛽̂𝑖 is the estimated 𝐺𝑖 coefficient on government spending of type i, Gi is government spending of type i, and YP is personal income. Page 7 of 12 References Aiyagari, S. Rao, Lawrence Christiano, and Martin Eichenbaum, “The output, employment, and interest rate effects of government consumption,” Journal of Monetary Economics, vol. 30(1), pages 73-86, 1992. Barro, Robert, “Output effects of government purchases,” Journal of Political Economy, 89(6): 1115, 1981. Barro, Robert and Charles J. Redlick, “Macroeconomic effects from government Purchases and Taxes,” The Quarterly Journal of Economics (2011) 126, 51–102. Baxter, Marianne and Robert G. King, “Fiscal policy in general equilibrium,” The American Economic Review, 83, 3: 315-334, Jun., 1993. Blanchard, Olivier, and Roberto Perotti, “An Empirical Characterization of the Dynamic Effects of Changes in Government Spending and Taxes on Output,” Quarterly Journal of Economics, 117 (2002), 1329–1368. Burnside, Craig, Martin Eichenbaum, and Jonas Fisher, “Fiscal shocks and their consequences,” Journal of Economic Theory 115: 89-117, 2004. Caldara, Dario and Christophe Kamps, “What are the Effects of Fiscal Policy Shocks? A VAR-based Comparative Analysis,” Working Paper, September 2006.. Cavallo, Michele, “Government Employment Expenditure and the Effects of Fiscal Policy Shocks,” Federal Reserve Bank of San Francisco Working Paper 2005-16, 2005. Christiano, Lawrence, Martin Eichenbaum, and Charles Evans, “Nominal rigidities and the dynamic effects of a shock to monetary policy,” Journal of Political Economy, 113: 1—45, 2005. Cullen, Joseph, and Price V. Fishback, “Did Big Government Largesse Help the Locals: The Implications of WWII Spending for Local Economic Activity, 1939-1958.” NBER Working Paper No. 12801, 2006. Drautzburg, Thorsten and Harald Uhlig, “Fiscal Stimulus and Distortionary Taxation,” Draft, University of Chicago, 2010. http://www.frbatlanta.org/documents/news/conferences/10fiscal policy uhlig.pdf Edelberg, Wendy, Martin Eichenbaum, and Jonas D. M. Fisher, “Understanding the Effects of a Shock to Government Purchases,” Review of Economic Dynamics, 2 (1999), 166–206. Eggertsson, Gauti and Michael Woodford, “The zero interest-rate bound and optimal monetary policy,” Brookings Panel on Economic Activity, 2003. Fatas, Antonio, and Ilian Mihov, “The Effects of Fiscal Policy on Consumption and Employment: Theory and Evidence,” CEPR Discussion Paper No. 2760, 2001. Page 8 of 12 Gali, Jordi, J.David Lopez-Salido and Javier Valles , “Understanding the Effects of Government Spending on Consumption,” Journal of the European Economics Association, 5:1, 227-270, 2007. Giavazzi, Francesco and Marco Pagano, “Can Severe Fiscal Contractions be Expansionary?” NBER Macroeconomics Annual, 1990. Hall, Robert E. “The Role of Consumption in Economic Fluctuations,” in The American Business Cycle: Continuity and Change, Robert J. Gordon, ed. Chicago: NBER and The University of Chicago Press, 1986. Monacelli, Tommaso and Roberto Perotti , “Fiscal policy, wealth effects and markups,” National Bureau of Economic Research Working Paper 14584, 2008. Perotti, Roberto, “Estimating the Effects of Fiscal Policy in OECD Countries,” CEPR Discussion Paper 4842, January 2005. Ramey, Valerie A. and Matthew D. Shapiro, “Costly capital reallocation and the effects of government spending,” Carnegie-Rochester Conference Series on Public Policy, 48: 145194, June 1998. Ramey, Valerie A. “Identifying government spending shocks: it’s all in the timing,” forthcoming, Quarterly Journal of Economics, 2011. Romer, Christina D., and David H. Romer, “The Macroeconomic Effects of Tax Changes: Estimates Based on a New Measure of Fiscal Shocks,” American Economic Review 100 (2010), 763–801. Rotemberg, Julio, and Woodford, Michael, “Oligopolistic Pricing and the Effects of Aggregate Demand on Economic Activity,” Journal of Political Economy, 100, 1153–1297, 1992. Woodford, Michael (2009), “Convergence in Macroeconomics: Elements of the New Synthesis,” American Economic Journal: Macroeconomics, Vol. 1, No. 1, 267-279. Page 9 of 12 Table 1: Impact of Government Spending on County Personal Income Dependent Variable: Personal Income Ordinary Least Squares Variablea Model 1 Constant Model 2 0.05167 6.87998 ARRA Spending 0.05474 *** 0.00199 1.82463 Federal Gov't Contracts Federal Gov't Civilian Earnings *** 0.00137 * 0.00666 2.06922 7.55892 1.18032 0.01369 ** 3.95971 *** 0.02789 2.27044 Federal Gov't Military Earnings * 0.04502 2.17896 State Gov't Earnings * 0.01847 2.09557 Local Gov't Earnings * 0.05039 2.33593 * Total Gov't Earnings 0.16983 6.53053 Lagged Personal Income -0.34014 -20.73302 Log Personal Income -0.43869 *** -0.00556 -4.83478 State Personal Income Change in the Unemployment Rate *** Adjust R-Bar Durbin-Watson N Df t-statistic given in bold 0.22861 1.8387 2766 2755 * Significance level between 5% and 10% **Significance level between 1% and 5% Page 10 of 12 *** -3.65614 *** 0.52769 *** -0.05841 -11.27912 -26.91167 -0.0045 0.53559 12.56123 *** 11.28338 *** -0.06062 *** -11.00317 0.26068 1.8657 2766 2755 *** *** Significance level between less than 1% a. Variables are expressed as log differences unless indicated otherwise. All variables are by county except for state personal income. Table 2: Implied Government Spending Multiplier (GSM) Variablea Model 1 Model 2 ARRA Spending 0.0872 0.0081 Federal Gov't Contracts 0.0269 0.0036 Federal Gov't Civilian Earnings 1.1244 Fedearl Gov't Military Earnings 3.1265 State Gov't Spending 0.6989 Local Gov't Earnings 0.6888 Total Gov't Earnings 1.2612 Page 11 of 12