WPS5220 Policy Research Working Paper 5220 Estimating the Fiscal Multiplier in Argentina Paloma Anós-Casero Diego Cerdeiro Riccardo Trezzi The World Bank Latin America and the Caribbean Economic Policy Sector February 2010 Policy Research Working Paper 5220 Abstract Argentina's government has resorted to fiscal policy as a factors, including the higher propensity of households to countercyclical tool to mitigate the negative impact of save during the economic downturn, the implementation the current economic downturn on aggregate demand. lag of public expenditures, particularly of capital Empirical results based on a vector error correction model expenditures, and the narrow tax base that limits the suggest, however, that the fiscal multiplier is relatively impact of countercyclical revenue measures on domestic small and short-lived. This could reflect a number of demand. This paper--a product of the Economic Policy Sector, Latin America and the Caribbean--is part of a larger effort in the department to estimate the fiscal multiplier in LAC countries. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at panoscasero@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team ESTIMATING THE FISCAL MULTIPLIER IN ARGENTINA Authors: Paloma Anós-Casero, Diego Cerdeiro and Riccardo Trezzi 1. INTRODUCTION The ongoing global downturn and financial turmoil have led many countries to resort to fiscal stimulus to cushion the impact on domestic economic activity. In turn, this has reignited the long-standing debate among economists about the effectiveness of fiscal policy as a counter-cyclical tool. Supporters of an active role for fiscal policy suggest that economies lack an efficient mechanism to return to full potential. Critics, on the other hand, argue that economic agents will offset the impact of fiscal policy on aggregate demand through changes in their saving behavior. A middle-of-the-road view holds that fiscal policy can be effective provided certain conditions hold, including sound macroeconomic fundamentals, nominal wage and price stickiness, and/or economic agents with finite horizons and liquidity constraints. Argentina's policy makers have also resorted to fiscal stimulus packages to mitigate the negative impact of the current economic slowdown on aggregate demand (see Table 1 for details). Table 1 - Response to the Financial Crisis ­ Countercyclical Measures Fiscal costs Countercyclical fiscal Details (US$) measures (-)=savings Credit lines US$2.7bn to the agricultural sector US$500mn to the manufacturing sector US$380mn to the SMEs US$880mn to those buying their first car US$910mn to those buying appliances US$1mn The newly created SIPA (Social Security fund) conducted an auction to allocate time deposits across Other subsidized credits for No immediate different banks. The winning bids will be those that consumption fiscal costs, charge less to finance these activities. The cost is paid by future retirees. Safety Nets US$400mn Lump-sum payment to ARS200 lump sum payment for retirees (for 2008 US$ 290mn retirees only) Other programs US$110mn PIT/exemptions/deductions US$1.2bn From 28 percent to 23 percent for wheat, 25 percent to Lower export taxes US$ 170mn 20 percent for corn The elimination of the tax brackets put in place in 2001 would benefit those who earn (after some Social Income tax reduction US$850mn Security taxes) more than ARS7000 (US$2060) per month Infrastructure Investment US$3.6bn 4-year infrastructure spending plan of about AR$71 bn (US$20bn). Fiscal impact will depend on the sources and details of financing. Most of the capital expenditures to be announced for 2009 are likely to be already included in the budget. It is assumed that 40 Program of public works US$1.5bn percent of the additional revenues from the nationalization of private pension funds will be used to finance additional infrastructure expenditures (about AR$6bn or about 1.5 bn US$) Other programs in US1.9bn Infrastructure Other Measures -US$760mn Capital repatriation Fiscal impact unclear Tax moratorium 2 percent of annual tax collection (estimated) -US$760mn Employment formalization No fiscal effect expected Total fiscal costs for 2009 US$ 2bn 2 But the implementation of countercyclical fiscal policies can be challenging, particularly for emerging economies. For start, economic fluctuations tend to be more intense for these countries. A typical recession in Argentina reduces output by an average of 6.2 percent. In the OECD, output falls by 2 percent, on average. As the tax base is narrower and revenues are largely dependent on commodity exports, external shocks and fluctuations in economic activity often resulted in large changes in government revenues. Recessions can cost the Treasury an average of 9 percent of its revenues, compared with six percent for the OECD as a whole. Additionally, automatic stabilizers are weak when compared to those in developed economies, and discretionary government spending might take several months to be implemented. Another complicating factor common to most LAC economies is the increased cost of borrowing during recessions, as creditors are wary of buying emerging-market bonds during bad times 1. This, in turn, prevents emerging-country governments from smoothing the cycle by borrowing, as their rich counterparts can afford to do. This paper empirically analyzes the effectiveness of fiscal policy in stimulating economic activity in Argentina, drawing on a vector error correction model to derive the fiscal multiplier. 2. BRIEF LITERATURE REVIEW Empirical studies based on structural vector autoregression models conclude that fiscal multipliers have declined over time and, in some cases, may even have been negative (on this specific point see Perotti, 2005). According to Eskesen (2009), these results, which differ widely across countries, likely reflect: (i) increasing leakage through the trade channel due to higher openness of economies; (ii) a decline in the share of liquidity constrained households due to better access to credit; and (iii) a sharper focus of monetary policy on price stability. In his study for OECD countries, Perotti (2005) finds that the effects of fiscal policy on GDP are small. He finds that only for the U.S. prior to 1980 the fiscal multiplier for government expenditures was larger than one. The figures below summarize these findings. 1 During the 2001-02 crisis interest rate shifted from 15.4 percent in June to 132.6 percent in November 2001. 3 Effect on GDP of a shock to government spending of 1 p.p. of GDP cumulative response at 12 quarters (source: Perotti, 2005) 4 1960-1980 3 1980-2001 2 1 0 -1 -2 -3 Australia Canada Germany United Kingdom United States Effect on GDP of a cut in net taxes of 1 p.p. of GDP cumulative response at 12 quarters (source: Perotti, 2005) 3 1960-1980 2 1980-2001 1 0 -1 -2 -3 Australia Canada Germany United Kingdom United States Studies drawing on aggregate dynamic macro models conclude that fiscal multipliers are in the range of 0.3 to 1.2 percent, and that expenditure measures appear to have a larger effect than tax measures (Hemming et al 2002, Botman 2006). Other studies find that government investment has the largest impact on economic activity and inflation. However, the size of the estimated multipliers depends on assumptions about, among other things, the monetary regime, labor supply elasticities and liquidity constraints (IMF, 2008). 3. METHODOLOGY This paper focuses on two measures: a general government tax cut and a general government expenditure increase. The expenditure variable is defined as total purchases of goods and services, meaning government consumption plus government investment. The revenue variable is defined as total tax revenues minus transfers. The traditional approach in the literature consists in estimating impulse reaction functions using a vector autoregressive (VAR) model (see for example Blanchard and Perotti, 1999 or Bernanke and Mihov, 1998 for monetary policy). Fiscal multipliers are than computed summing the first four, eight or 12 periods of the reaction functions. The basic reduced VAR specification is as follows: Z t = ( L) Z t -1 + u t (1) where Z t is a three- dimensional vector ( Z t = [Yt , Et , Rt ]' ) in the logarithms of quarterly observations for government consumption, government revenues and private consumption. 4 U t [u ty , u te , u tr ] is the corresponding vector of reduced form residuals, which in general will have no-cross correlations, and can be represented as follows: (2)u ty = ey u te + ry u tr + vty (3)u te = y u ty + re vte + vte e (4)u tr = ry u ty + er vte + vtr However, as is well known in the literature, the reduced form residuals represented by equations (2), (3) and (4) have little or no economic significance. In fact, they are linear combination of the underlying "structural" tax, spending and private consumption shocks. Equation (2) states that unexpected movements in output can be due to unexpected movement in taxes and/or unexpected movements in government spending. Equation (4) states that unexpected movements in taxes within a quarter can be due to one of the following factors: the response to unexpected movements in private consumption (captured by ry u ty ), the response to structural shocks in spending (captured by er vte ), and to structural shocks to taxes, captured by vtr . Finally, vte and vtr are the mutually uncorrelated structural shocks to government expenditure and revenues that we seek to identify. In order to do so, we follow a four-step procedure. In the first step, the reduced form residuals are estimated from the VAR specified in equation (1). Given the fact that the reduced form residuals are correlated with the structural shocks, it is necessary to apply further restrictions in order to identify the coefficients y and y . However, since ordinary least squares in equation (2) and (4) e r would not generate consistent estimates y and y , external estimates are used. 2 r e With these estimates, in the second step, the cyclically adjusted fiscal shocks, u e ,adj and y u y ,adj can be calculated. Expressed in more detail: r (5)u te ,adj u te - y u ty = re vtr + vte e (6)u tr ,adj u tr - y u ty = er vte + vtr r In the third step, assuming that structural revenue shocks have no impact on structural spending shock (meaning that re = 0 ), the structural fiscal shocks are identified. Equation (5) reduces to u te ,adj = vte = u te . This simply implies that structurally adjusted expenditure shocks are effectively assumed equal to the cyclically adjusted expenditure 2 The elasticity of expenditure with respect to changes in economic activity is assumed to be close to zero within the quarter, as commonly assumed in many other empirical studies. The elasticity of revenues to economic activity is estimated to be 0.45 for Argentina. 5 shocks. Finally, equation (6) can be estimated using ordinary least squares to get consistent estimates of er . 3 In the fourth step, the coefficients ey and ry in equation (2) can be determined using the structural uncorrelated fiscal shocks vte and vtr as instruments for u te and u tr , respectively. Combined, the four steps allow us to estimate the A and B matrices of the standard relation between the reduced-form innovations and the structural shocks Au t = Bvt , where: 1 - ey - ry 1 0 0 A = - y e 1 0 and B = 0 1 0 which brings to: - y r 0 1 0 er 1 -1 1 - ey - ry 1 0 0 vty u t = - y e 1 0 0 1 0 vte - y r 0 1 0 er 1 vtr With this, impulse responses are computed using Monte Carlo simulations based on 10,000 replications. Given the non-stationary properties of the underlying series, we test for co-integrated relationships among variables. If no co-integrated relationships are found, a Vector Auto Regressive (VAR) model, specified in first log differences, is estimated. If there are co- integrated relationships, a vector error correction (VEC) model is estimated, without imposing additional restrictions in order to avoid over-identification. In its basic form, a VEC model can be represented as: y t = y t -1 + y t -1 + t Where is the long-run matrix which contains the equilibrium correction terms. As usual, the number of co-integrating vectors is given by the number of linearly independent rows of , namely the rank of . 3 er is estimated to be 0.7541 for Argentina. 6 4. ESTIMATING THE FISCAL MULTIPLIER Consistent with the visual inspection of the data, no evidence for co-integration was found among the variables (see annex). Figure 1 shows diverging patterns of government revenues and private consumption 4. Thus, we specify a VAR model 5 and estimate the impulse response functions of shocks to both government expenditures and revenues, specifying the variables in log differences to eliminate the non-stationary properties. Finally, in order to estimate the fiscal multiplier, we calculate the cumulative response of private consumption to each of these shocks by adding the first four, eight and twelve lagged quarter responses. Figure 1. Tax Revenues, Government Expenditure and Private Consumption Source: National Authorities. Series seasonally adjusted. In Argentina, the fiscal multiplier is small and short-lived. A (one standard deviation) shock to government expenditures, which represents additional AR$ 1.17 million in government expenditures, is estimated to increase private consumption by nearly 0.5% within the same year. Put it differently, each AR$ 1 million of additional expenditures, private consumption increases 0.39%. The multiplier of revenue measures is also small. The cumulative impact on real private demand of a (one standard deviation) shock in tax revenues, which represents AR$ 1.16 million, is close to zero. The fiscal multiplier evaporates quickly (lasting one year). It should be noted that the coefficients after the fourth quarter should be interpreted with caution since only within the first year coefficients are statistically significant. The limited effectiveness of revenue measures could be related to a narrow tax base since the informal sector is not negligible. The small and short-lived cumulative impact 6 of increases in expenditures may reflect a number of factors, including a high propensity to save among households (possibly, in part, reflecting the lack of a comprehensive social safety net) and the implementation lag of public investment programs. 4 During the period that spans from 1993Q1 through 2003Q1 (end of the most recent financial crisis), the seasonally-adjusted series for private consumption increased by 1.72 percent (from 157,762.6 to 160,482.1), while government tax revenues increased 38.6 percent (from 8,207.2 to 11,380.7). The substantial increase in both government revenues and expenditures in 1997Q1 was not associated with a similar movement in private consumption. Finally, an exceptional increase (80.1 percent) in government tax revenues during 1998 was associated with a negligible 0.007 percent increase in private consumption. 5 It includes two dummy variables, corresponding to 1998-crisis and to the crisis in 2001-2002. 6 In fact, only the first four quarters are associated to statistically significant coefficients. 7 That the fiscal multiplier is small is not a surprising result. The small size of the fiscal multiplier in Argentina is comparable to the size of fiscal multipliers in other LAC economies (Perry et al., 2008). In many industrialized countries, the fiscal multipliers can range from less than zero to at most three (Perotti and Blanchard, 2005). In the US, Barro (2009) estimates that the multiplier of the increase in government expenditures at the peak of World War II (1943-44) was 0.8. But when Barro estimates the fiscal multiplier associated with peacetime government purchases in the US, he obtains a number close to 8 zero. The evidence for Argentina and elsewhere points to one of the key challenges of fiscal policy as a counter-cyclical tool: the relatively small size and short-lived nature of the fiscal multiplier. Finally, we estimate the fiscal multiplier in Peru, to compare the results obtained for Argentina. The Johansen's procedure suggests one co-integrated relationship, consistent with the visual inspection of the data for Peru (see Figure 2). Thus, we estimate a VECM to estimate fiscal multipliers, that are calculated on a cumulative basis over the first four, eight and twelve quarters (see annex for details). Figure 2. Peru :Tax Revenues, Government Expenditure and Private Consumption Source: National Authorities. Series seasonally adjusted. The fiscal multiplier in Peru is also small, close to zero, and short-lived. The impact of one standard deviation shock in government expenditure on private domestic demand is close to zero. 9 ANNEX Johansen's Procedure ­ ARGENTINA Date: 05/20/09 Time: 14:56 Sample (adjusted): 1993Q4 2008Q3 Included observations: 60 after adjustments Trend assumption: Linear deterministic trend Series: PRIV_CONSSA GOV_EXPSA GOV_REVSA Lags interval (in first differences): 1 to 2 Unrestricted Cointegration Rank Test (Trace) Hypothesized Trace 0.05 No. of CE(s) Eigenvalue Statistic Critical Value Prob.** None 0.253503 26.17100 29.79707 0.1237 At most 1 0.129025 8.629179 15.49471 0.4008 At most 2 0.005661 0.340648 3.841466 0.5595 Trace test indicates no cointegration at the 0.05 level * denotes rejection of the hypothesis at the 0.05 level **MacKinnon-Haug-Michelis (1999) p-values Unrestricted Cointegration Rank Test (Maximum Eigenvalue) Hypothesized Max-Eigen 0.05 No. of CE(s) Eigenvalue Statistic Critical Value Prob.** None 0.253503 17.54182 21.13162 0.1479 At most 1 0.129025 8.288531 14.26460 0.3501 At most 2 0.005661 0.340648 3.841466 0.5595 Max-eigenvalue test indicates no cointegration at the 0.05 level * denotes rejection of the hypothesis at the 0.05 level **MacKinnon-Haug-Michelis (1999) p-values VAR results for ARG, variables specified in log differences. Vector Autoregression Estimates Date: 05/21/09 Time: 14:36 Sample (adjusted): 1993Q4 2008Q3 Included observations: 60 after adjustments Standard errors in ( ) & t-statistics in [ ] D(LOG(PRIV_C D(LOG(GOV_E D(LOG(GOV_R 10 ONSSA)) XPSA)) EVSA)) D(LOG(PRIV_CONSSA(-1))) 0.643741 1.757509 1.819987 (0.14466) (1.01856) (0.97272) [ 4.45018] [ 1.72548] [ 1.87102] D(LOG(PRIV_CONSSA(-2))) -0.152693 0.470141 0.747600 (0.15219) (1.07164) (1.02341) [-1.00329] [ 0.43871] [ 0.73050] D(LOG(GOV_EXPSA(-1))) -0.039537 -0.616201 -0.152425 (0.03644) (0.25658) (0.24504) [-1.08500] [-2.40157] [-0.62205] D(LOG(GOV_EXPSA(-2))) -0.014673 -0.255629 -0.250986 (0.03580) (0.25210) (0.24075) [-0.40983] [-1.01401] [-1.04251] D(LOG(GOV_REVSA(-1))) 0.052093 0.095097 -0.295705 (0.03730) (0.26263) (0.25081) [ 1.39665] [ 0.36210] [-1.17900] D(LOG(GOV_REVSA(-2))) 0.019451 0.024786 -0.023454 (0.03651) (0.25707) (0.24550) [ 0.53278] [ 0.09642] [-0.09554] C 0.005087 -0.000273 -0.002308 (0.00301) (0.02120) (0.02025) [ 1.68930] [-0.01289] [-0.11397] DUMMY98 -0.008476 0.098150 0.092226 (0.00956) (0.06730) (0.06427) [-0.88686] [ 1.45847] [ 1.43502] DUMMY01 -0.018267 -0.003741 0.017975 (0.01343) (0.09459) (0.09033) [-1.35977] [-0.03955] [ 0.19898] R-squared 0.518550 0.250246 0.230159 Adj. R-squared 0.443028 0.132637 0.109400 Sum sq. resids 0.014482 0.718021 0.654846 S.E. equation 0.016851 0.118654 0.113314 F-statistic 6.866250 2.127784 1.905932 Log likelihood 164.7394 47.63172 50.39469 Akaike AIC -5.191314 -1.287724 -1.379823 Schwarz SC -4.877163 -0.973572 -1.065671 Mean dependent 0.006997 0.012353 0.013774 S.D. dependent 0.022579 0.127404 0.120072 Determinant resid covariance (dof adj.) 1.57E-08 11 Determinant resid covariance 9.66E-09 Log likelihood 298.2462 Akaike information criterion -9.041539 Schwarz criterion -8.099084 Johansen's Procedure ­ PERU Date: 05/20/09 Time: 14:33 Sample (adjusted): 1981Q2 2008Q3 Included observations: 110 after adjustments Trend assumption: Linear deterministic trend Series: PRIV_CONSSA TAX_REVSA GOV_EXPSA Lags interval (in first differences): 1 to 4 Unrestricted Cointegration Rank Test (Trace) Hypothesized Trace 0.05 No. of CE(s) Eigenvalue Statistic Critical Value Prob.** None * 0.194352 35.72291 29.79707 0.0092 At most 1 0.080222 11.95094 15.49471 0.1593 At most 2 0.024711 2.752365 3.841466 0.0971 Trace test indicates 1 cointegrating eqn(s) at the 0.05 level * denotes rejection of the hypothesis at the 0.05 level **MacKinnon-Haug-Michelis (1999) p-values Unrestricted Cointegration Rank Test (Maximum Eigenvalue) Hypothesized Max-Eigen 0.05 No. of CE(s) Eigenvalue Statistic Critical Value Prob.** None * 0.194352 23.77196 21.13162 0.0207 At most 1 0.080222 9.198577 14.26460 0.2700 At most 2 0.024711 2.752365 3.841466 0.0971 Max-eigenvalue test indicates 1 cointegrating eqn(s) at the 0.05 level * denotes rejection of the hypothesis at the 0.05 level **MacKinnon-Haug-Michelis (1999) p-values VECM results for PERU, variables specified in log. Vector Error Correction Estimates Date: 05/20/09 Time: 14:40 Sample (adjusted): 1981Q1 2008Q3 12 Included observations: 111 after adjustments Standard errors in ( ) & t-statistics in [ ] Cointegrating Eq: CointEq1 LOG(PRIV_CONSSA(-1)) 1.000000 LOG(GOV_EXPSA(-1)) -5.539632 (1.08443) [-5.10835] LOG(TAX_REVSA(-1)) 4.116692 (1.03463) [ 3.97889] C 2.489465 D(LOG(PRIV_C D(LOG(GOV_E D(LOG(TAX_R Error Correction: ONSSA)) XPSA)) EVSA)) CointEq1 0.003664 0.075820 -0.017105 (0.00584) (0.01777) (0.01415) [ 0.62711] [ 4.26768] [-1.20873] D(LOG(PRIV_CONSSA(-1))) -0.003520 0.580140 -0.482135 (0.10315) (0.31366) (0.24984) [-0.03412] [ 1.84956] [-1.92976] D(LOG(PRIV_CONSSA(-2))) -0.142206 0.012405 -0.171618 (0.09339) (0.28398) (0.22620) [-1.52273] [ 0.04368] [-0.75871] D(LOG(PRIV_CONSSA(-3))) -0.223590 -0.085381 0.061868 (0.08841) (0.26883) (0.21413) [-2.52911] [-0.31760] [ 0.28893] D(LOG(GOV_EXPSA(-1))) 0.075170 -0.089658 0.024927 (0.03794) (0.11536) (0.09188) [ 1.98151] [-0.77723] [ 0.27129] D(LOG(GOV_EXPSA(-2))) 0.066604 -0.192547 -0.048395 (0.03599) (0.10943) (0.08716) [ 1.85083] [-1.75960] [-0.55523] D(LOG(GOV_EXPSA(-3))) 0.014277 0.138317 0.057988 (0.03345) (0.10170) (0.08101) [ 0.42687] [ 1.36003] [ 0.71583] D(LOG(TAX_REVSA(-1))) 0.013674 -0.229854 -0.155426 (0.04470) (0.13592) (0.10827) [ 0.30591] [-1.69107] [-1.43559] 13 D(LOG(TAX_REVSA(-2))) 0.136474 -0.072110 0.216053 (0.04604) (0.14001) (0.11152) [ 2.96396] [-0.51502] [ 1.93729] D(LOG(TAX_REVSA(-3))) -0.021874 -0.349052 0.237864 (0.04481) (0.13626) (0.10853) [-0.48816] [-2.56166] [ 2.19160] C 0.011261 0.014617 0.013869 (0.00376) (0.01143) (0.00910) [ 2.99672] [ 1.27925] [ 1.52385] DUMMY87 -0.028946 -0.029857 -0.080481 (0.01310) (0.03982) (0.03172) [-2.21016] [-0.74971] [-2.53711] DUMMY90 -0.021113 -0.094835 0.038780 (0.01708) (0.05193) (0.04136) [-1.23639] [-1.82632] [ 0.93759] R-squared 0.320937 0.382791 0.216285 Adj. R-squared 0.237786 0.307214 0.120320 Sum sq. resids 0.102162 0.944644 0.599334 S.E. equation 0.032287 0.098180 0.078203 F-statistic 3.859701 5.064932 2.253783 Log likelihood 230.4833 107.0373 132.2892 Akaike AIC -3.918618 -1.694366 -2.149355 Schwarz SC -3.601286 -1.377033 -1.832023 Mean dependent 0.006284 0.007022 0.006045 S.D. dependent 0.036982 0.117956 0.083379 REFERENCES Blanchard, Olivier and Roberto Perotti, 2002, "An Empirical Characterization of the Dynamic Effects of Changes in Government Spending and Taxes on Output," Quarterly Journal of Economics 117, 1329­1368. Hemming Richard, Michael Kell, and Selma Mahfouz, 2002, "The Effectiveness of Fiscal Policy in Stimulating Economic Activity--A Review of the Literature," IMF Working Paper 02/208 (Washington: International Monetary Fund). Perotti, Roberto, 2005, "Estimating the Effects of Fiscal Policy in OECD Countries," CEPR Discussion Paper 4842 (London: Centre for Economic Policy Research). 14 ­­­­­­­­, 2007, "In Search of the Transmission Mechanism of Fiscal Policy," NBER Working Paper 13143 (Cambridge, Massachusetts: National Bureau of Economic Research). Shapiro, Matthew D. and Joel Slemrod, 2003, "Consumer Response to Tax Rebates," American Economic Review, Vol. 93 (1), 381­396 ­­­­­­­­, 2002, "Did the 2001 Tax Rebate Stimulate Spending? Evidence from Taxpayer Surveys," NBER Working Paper 9308 (Cambridge, Massachusetts: National Bureau of Economic Research). wb340866 L:\ARGENTINA\sVAR\Argentina-Theroleoffiscalpolicyascounterciclicaltool-estimating the fiscal multiplier-PAC_V2.doc 05/21/2009 4:05:00 PM 15