WPS5982 Policy Research Working Paper 5982 Fiscal Multipliers over the Growth Cycle Evidence from Malaysia Sohrab Rafiq Albert Zeufack The World Bank East Asia and Pacific Region Poverty Reduction and Economic Management Unit March 2012 Policy Research Working Paper 5982 Abstract This paper explores the stabilisation properties of spending leads to a maximum output multiplier of fiscal policy in Malaysia using a model incorporating around 2.7 during growth recessions, and around 2 nonlinearities into the dynamic relationship between in normal times. The returns to government spending fiscal policy and real economic activity over the growth in Malaysia are greater when the focus is on public cycle. The paper also investigates how output multipliers investment, as opposed to consumption. Changes in tax for government purchases may alter for different policy are less effective in stimulating economic activity components of government spending. The authors find than direct government spending. These results provide that fiscal policy in Malaysia has become increasingly pro- empirical backing to conjectures in the recent literature cyclical over the last 25 years and establish that the size implying that procyclicality in fiscal policy reduces the of fiscal multipliers tend to change over the growth cycle. effectiveness of fiscal actions in emerging markets. A 1 Malaysian Ringgit rise in government (investment) This paper is a product of the Poverty Reduction and Economic Management Unit, East Asia and Pacific Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at azeufack@worldbank.org; Sohrab.Rafiq@khazanah.com.my. 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 Fiscal Multipliers Over the Growth Cycle: Evidence From Malaysia Sohrab Ra…q & Albert Zeufack Khazanah Research and Investment Strategy Abstract This paper explores the stabilisation properties of …scal policy in Malaysia using a model incorporating nonlinearities into the dynamic relationship be- tween …scal policy and real economic activity over the growth cycle. The paper also investigates how output multipliers for government purchases may alter for di¤erent components of government spending. We …nd that …scal policy in Malaysia has become increasingly pro-cyclical over the last 25 years and establish that the size of …scal multipliers tend to change over the growth cycle. A 1 Malaysian Ringgit rise in government (investment) spending leads to a maximum output multiplier of around 2.7 during growth recessions, and around 2 in normal times. The returns to government spending in Malaysia are greater when the focus is on public investment, as opposed to consumption. Changes in tax policy are less e¤ective in stim- ulating economic activity than direct government spending. These results provide empirical backing to conjectures in the recent literature implying that procyclicality in …scal policy reduces the e¤ectiveness of …scal actions in emerging markets. JEL codes: E32, E62, E65 Keywords: …scal multiplier, investment, consumption, nonlinearity 1 Introduction After a couple of decades when the primary focus was on monetary policy, the 2008 …nancial crisis refocused the attention of economists onto discretionary …scal policy as a potentially potent economic stabilisation tool. Despite the emergence ect The views in this paper solely re‡ those of the authors, and do not necessarily correspond to the views of Khazanah National Berhad. 1 of a large body of literature since then, surprisingly very little is known about the e¤ects of …scal policy on economic activity and on the short-versus long-run destabilising behaviour of …scal policy in emerging market countries. This is notwithstanding the fact that during the height of the crisis the largest …scal stimulus packages were enacted in emerging market economies.1 Malaysia, for instance, passed a stimulus package of around 8.6 percent of GDP in 2009. Emerging market economic cycles tend to be more volatile than developed countries’ whilst being more susceptible to large shocks such as …nancial crises …scal policy could play a central role in dampening economic ‡uctuations and promoting long-run growth.2 Aghion, Angeletos, Banerjee and Manova (2010), for example, show that the ability of …rms to …nance investment falls during an economic downturn due to a fall in pro…ts, which acts as a constraint on their borrowings. Indeed, the …nancial accelerator principle implies the credit wor- thiness of the borrower may change depending upon the stage of the economic cycle. In the presence of …nancial frictions the government, through undertaking counter-cyclical …scal policy and pushing out the demand for goods and services for these …rms, …scal policy could have a positive impact on long-term investment and productivity. As well as the habitual undulations in the economic cycle, it is particularly the role of …scal policy during crisis times that should be of relevance to emerging market economies. As noted in Perotti (2005b), in the aftermath of the …nancial crises of the late 1990s, many emerging markets were perceived as sacri…cing long- run growth in order to show signs of …scal discipline. Since the crisis, a¤ected economies - including Malaysia - have endured a downward shift in their long-run growth trajectory. 1 See Federal Reserve Bank of Dallas (2009). 2 See Aguiar and Gopinath (2007), which discuss the characteristics of economic ‡uctuations in emerging market countries. 2 1000 900 800 700 MYR mn 600 500 400 300 200 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Shaded area represents crisis period Trend output growth pre- and post-Asian crisis If there is a negative shock to economic activity, policymakers need to know how the economy would respond so as to decide whether to make a discretionary change in, say, government consumption. In investigating these questions, much of the literature has focused on the advanced countries. From an analytical standpoint documenting empirical similarities and observing whether they are the same across di¤erent levels of income provide an empirical basis for devising …scal policies that incorporate features and relationships that are particularly important for developing countries.3 If …scal policy mainly has demand e¤ects, and shifts out the demand for goods without crowding out private consumption, then clearly there is a role for a coun- tercyclical …scal policy during recessions, when individuals or …rms are more likely to be credit constrained. On the other hand, if …scal policy mainly has a negative wealth e¤ect on labor supply, and a crowding out e¤ect on private investment, then 3 Much of the literature measuring the size of …scal multipliers has continued to focus on the advanced economies. 3 there is a role for procyclical …scal policy. There is also currently little agreement on the e¤ects of …scal policy on private consumption.4 This latter disagreement forms part of the more broader issue of whether consumers are Ricardian or not.5 Recent evidence has highlighted the asymmetric response of the economy to …s- cal policy shifts as a result of whether the country is in boom or depression times.6 Tagkalakis (2008), Christiano, Eichenbaum and Rebelo (2009) and Auerbach and Gorodnichenko (2010, 2011) …nd that economic downturns tend to raise the gov- ernment spending multiplier on private consumption and output. Such …ndings have particular relevance for emerging market countries when seen in the con- text of studies documenting …scal policy in developing countries to be procyclical, particularly the investment component of public spending.7 Among many others, Kaminsky, Reinhart and Végh (2004) and Alesina, Campante and Tabellini (2008) …nd that …scal policy in developing countries has tended to be overwhelmingly pro- cyclical, partly because of political incentives for governments to be more generous and run large de…cits in good times. Taken together, these recent …ndings imply that emerging market countries may not be maximising the ‘ bang for the buck’of …scal policy by loosening …scal discipline during boom times. This paper explores how the size of …scal multipliers may change depending upon the stage of the growth cycle. In contrast to Auerbach and Gorodnichenko (2010, 2011) who investigate how the multiplier changes depending upon whether the economy is in a technical recession or not, this paper explores how the size of multiplier may change depending upon the stage of the growth cycle. This ap- proach is more relevant for emerging market countries whose economic cycles are characterised by less recessions in the classical sense and where absolute declines in economic activity are rarer episodes. Emerging market countries tend to expe- rience more growth, as opposed to business, cycles.8 The paper also contributes to the literature by investigating how output multipliers for government purchases may di¤er for di¤erent components of government spending. A number of …ndings 4 See Perotti (2005) and Galí, Lopéz and Vallés (2007) and Ramey (2009). 5 In the standard neoclassical model expansionary government spending tends to crowd out private consumption due to the negative wealth e¤ect on consumers induced by higher future tax payments. 6 See Tagkalakis (2008) and Bachmann and Sims (2011). 7 See Gavin and Perotti (1997) and Alesina, Campante and Tabellini (2008). 8 See Stock and Watson (2002) and Aguiar and Gopinath (2007). 4 emerge: The paper …nds that …scal policy in Malaysia has been largely procyclical over the past 25 years. The time-varying correlation coe¢ cients show that procyclicality accelerated following the Asian …nancial crisis in 1997/98. The e¤ectiveness of …scal policy varies over the growth cycle. Fiscal multipli- ers tend to increase (decrease) during periods of depressed (high) economic growth. The relative change in the size of state dependent …scal multipli- ers also depends on the component of government spending. Additionally, changes in tax policy are less e¤ective in stimulating economic activity than direct government spending. Government investment generates …scal multipliers twice as large as gov- ernment consumption spending in growth expansions and growth recession states. During a growth recession government investment spending gener- ates a real output multiplier of around 2:7, compared with a multiplier value of around 1:8 for government consumption spending. There is also evidence that government consumption crowds out private consumption spending in ‘normal’times. Multiplier values for tax cuts are below 0:3 for real output and private consumption. These results can be interpreted as evidence in support of the argument that higher …scal outlays are particularly e¤ective as a stabilisation tool in a sharp economic downturn. These results con…rm conjectures in Alesina, Campante and Tabellini (2008), providing empirical backing for the idea that procyclicality in expenditures reduces the e¤ectiveness of …scal actions in emerging market countries. The remainder of this paper is organised as follows. Section 2 presents the econometric speci…cation, Section 3 sets out the data, model and structural iden- ti…cation scheme, Section 4 assesses …scal policy in Malaysia and tests for pro- cyclicality. Section 5 discusses the results from the model, including the size of the …scal multipliers and policy implications. Section 6 concludes. 5 2 Econometric speci…cation Most recent studies have focused on the size of …scal multipliers in a recession, classically de…ned, see Auerbach and Gorodnichenko (2010, 2011). However, for fast-growing emerging markets, absolute declines in economic activity tend to be the result of very rare sudden crisis episodes. Emerging markets economic cycles are typically characterised less by recessions in the technical sense and more by growth expansions and growth contractions. For Malaysia, over our sample period (1981:1 to 2010:4), there have been only three recessions in the classical de…nition sense.9 This limits the number of observation in the alternate regime. This paper therefore explores how the size of multiplier may change depending upon the stage of the growth cycle. Recessions in this paper are de…ned as growth contractions. It is important to note that all classical recessions involve a growth contraction, but a growth contraction need not necessarily imply a classical recession. To allow for di¤erentiated responses in boom and recession episodes this paper employs a regime switching vector autoregressive model where transitions across states are smooth. The approach in this paper is similar to the smooth transi- tion autoregressive (STAR) models developed in Granger and Terasvirta (1993). Moreover, following recent developments in the STVAR literature the model not only allows for di¤erential dynamic responses but also di¤erential contemporane- ous responses to structural shocks.10 Xt = (1 F (zt 1 ) E (L)Xt 1 + F (zt 1 ) R (L)Xt 1 + "t (1) "t N (0; t) (2) t = E (1 F (zt 1 ) + R F (zt 1 ) (3) 9 The classical de…nition of the recession is de…ned as two quarters of absolute decline in real output. 10 As noted in Auerbach and Gorodnichenko (2010), the advantage of STVAR models over their structural VAR contemporaries is that such models better exploit variation in the degree of a particular regime so that estimation and inference for each regime is based on a larger set of observation. Estimating a SVAR model for each regime separately may seriously limit the amount of observations in a regime, biasing the coe¢ cient values. 6 exp( zt ) F (zt ) = ; >0 (4) 1 + exp( zt ) var(zt ) = 1; E(zt ) = 0 (5) The matrix Xt is vector containing the endogenous variables with "t a normal error term. The model allows for two di¤erences in the propagation mechanism via (a) contemporaneous via di¤erences in covariance matrices for disturbances E and R and (b) dynamic via di¤erences in lag polynomials E and R : Variable z is an index, which is normalised to have unit variance so that is scale invariant, with positive z indicating an expansion. Assuming that > 0, R (L) and R describe the behaviour of the variables in a growth recession (F (z) 1); and E and E (L) as describing the behaviour of the system in a strong growth expansion (1 F (z) 1): Economic theory or empirics unfortunately o¤ers no guide on the choice of index z. Thus following Auerbach and Gorodnichenko (2010, z is set equal to four-quarter moving average of output growth. The key advantages of using this measure of z are: (a) it is possible to use the full sample for estimation, which makes the estimates as precise and robust as possible; (b) it is possible to consider dynamic feedback e¤ects from policy changes to the state of the regime. Said di¤erently, the model accounts for the fact that policy shocks could alter the regime, shifting the economy from a recession to an expansionary phase.. It is possible to estimate the parameter using Newton Raphson grid search. Estimating R (L); E (L); R and E and simultaneously suggests a point esti- mate for between 8 and 13:5: These parameter values suggest that the model is best described as a model switching regimes at sharp thresholds. This is consistent with Aguiar and Gopinath (2007), who showed that business cycles in emerging market countries tend to be more volatile and more prone to sudden shifts, partic- ularly when compared to economic cycles in the advanced economies. Auerbach and Gorodnichenko (2010), for instance, calibrate = 0:8 for the U.S. and = 1:5 for a panel of OECD countries, which is consistent with smooth transitions be- tween expansions and recessions. However, for the data used in this paper we select = 5: This is because we prefer relatively smooth transitions between regimes than suggested by the grid search values, which amount to moderate val- 7 ues of : To better understand the type of nonlinearity involved Figure 1 plots the time pro…le of the transition function, F (zt ), over the range of zt . 3 Structural identi…cation and data This paper uses a unique quarterly dataset from Malaysia to identify government consumption and government spending shocks. The data runs from 1981:1 till 2010:4. The identi…cation scheme is based on a generalised version of reduced form models in, amongst many others, Fatás and Mihov (2001), Blanchard and Perotti (2002), Perotti (2005), Cimadomo, Hauptmeier and Kirchner (2010), Corsetti, Meier and Müller (2010), Auerbach and Gorodnichenko (2010, 2011) and Bach- mann and Sims (2011). The identi…cation approach is based on a recursive iden- ti…cation scheme that aims to identify …scal policy shocks on their timings within the system. As in Blanchard and Perotti (2002) the basic model is computed using quarterly data and is composed of Xt = [Gt ; Tt ; Yt ]: The variable Gt represents government spending shocks. Just as output multipliers for government purchases may dif- fer according to the regime in which they occur, they can also di¤er for di¤erent components of government spending.11 Leeper, Walker and Yang (2009), for ex- ample, have shown how the short-run multiplier for government investment shocks are smaller than those for government consumption innovations, but are more e¤ective in the longer-term.12 Two types of government spending shocks are iden- ti…ed: consumption and investment.13 The government consumption de…nitions follow Perotti (2005a). Thus, government consumption spending is de…ned as gov- 11 See Perotti (2005a, 2007). 12 See Romp and de Haan (2007) for an in-depth survey on public capital and economic growth. 13 Much of the recent literature, to capture the stance of …scal policy, have utilised government consumption spending as de…ned as current real terms spending on goods and services. This is because such a measure contains little automatic cyclical component, thus helping to identify government spending shifts over and beyond those changes caused as a natural consequence of the business cycle. Furthermore, government purchases have no direct link to productivity, thus, limiting the number of channels through which government spending in‡ uences the real economy. This measure of …scal policy is also less likely to be in‡ uenced by the automatic stabilisers channels, which operate primarily through tax revenue and transfer payments. Moreover, in economic downturns, it is politically easier to cut capital expenditure (public investment) than government consumption of goods and services, such as health care. For these reasons, the baseline model in this paper includes government consumption spending. 8 ernment consumption minus social security contribution and capital consumption allowances. Government investment spending is de…ned as development expendi- ture from the Ministry of Finance government spending accounts. The data on tax revenues, Tt , are also taken from the same source, and is de…ned as the sum of total indirect taxes , direct taxes on households, social security contributions and other capital transfers received. Finally Yt is the cyclical component of real output extracted by a Hodrick-Prescott …lter. Following the de…nition laid out in Stock and Watson (2002), growth below trend is classi…ed as a growth recession.14 The real GDP data is taken from Abeysinghe and Gulasekaran (2004). The ordering of the variables in Xt assumes that shocks in tax revenues and real output have no contemporaneous e¤ect on government spending. As argued in Blanchard and Perotti (2002), this identifying minimum-delay assumption may be a sensible description of how government spending operates given that in the short-run government spending may be unable to adjust to spending in response to changes in the …scal and macroeconomic conditions. On the other hand, Barro and Redlick (2011) argue that the government spending shock in a structural VAR is likely to be endogenous, as a higher GDP leads to higher taxes and therefore to more government spending. However, Barro and Redlick (2011) use yearly data and their argument, thus, is unlikely to hold at the quarterly frequency. Due to decision lags, contemporaneous discretionary government spending is unlikely to respond within a quarter to any news about the economy. The identi…cation of tax shocks is more problematic in the framework outlined in this paper. As noted in Blanchard and Perotti (2002) identi…cation of tax shocks depend on purging innovations in revenues of automatic responses to output. This could be achieved by imposing a contemporaneous coe¢ cient on the elasticity of revenue with respect to real output. However, as noted in Auerbach and Gorodnichenko (2010, 2011) this elasticity is likely to vary over the cycle, thereby, introducing a bias of unknown magnitude and direction in our time-varying speci…c estimates. Auerbach and Gorodnichenko (2011) have shown that output responses to tax shocks in di¤erent regimes are sensitive to the assumed elasticity. Unlike monetary policy decisions, changes in government spending and taxes are typically decided and communicated well in advance of their implementation. 14 Robustness tests indicate that the fundamental conclusions remain unaltered should one use the detrended GDP growth rate. 9 The e¤ects of …scal policy, thus, show up immediately in interest rates and other …nancial variables. Thus, it is argued that such models su¤er from the inability to recover the ‘true’structural shock given the absence of news on future govern- ment spending. This has been termed the ‘ anticipation’or ‘non fundamentalness 15 . problem’ As a consequence it can be the case that the estimated innovations of a VAR are such only with respect to the information set of the practitioner, but not of the private sector. The implication being that government spending shocks may be misestimated. One way in which this problem could be overcome is by including observations on the present value of government spending forecasts di- rectly in the speci…cation. Such data is currently lacking for Malaysia. However, Perotti (2011) has shown how even without the exclusion of an expectations term the estimates from structural models need not be unduly biased. Perotti (2005), based on data for a group of OECD countries, has shown that the identi…cation scheme and choice of variables in the speci…cation outlined should provide unpredictable government spending shocks. It is worth pointing out that whether …scal shocks are truly anticipated or not matter only if anticipated and unanticipated …scal policy actions have di¤erent e¤ects. There is currently little agreement on this issue. Anticipation e¤ects are unlikely to undermine the central point of this paper, which is to examine the size of the changes in …scal multipliers given shifts in economic conditions. That being whilst anticipated …scal policy might bias the estimated impulse response functions, it is not clear whether and why such e¤ects change over the cycle. In addition to the three baseline variables - government spending, taxes and output - in the …scal policy literature real private consumption and an interest rate measure are also included in the model. The inclusion of consumption is based on the ongoing discussion on the crowding out e¤ects of government spending on private consumption.16 The impact of …scal policy on interest rates is among the most debated issues in macroeconomics, and a key issue in times of high de…cits (Perotti, 2005a). The treasury bill rate is used as the interest rate measure in the model. This also has the added advantage of being an information variable. Numerous studies have documented the superior predictive power of the term structure for real activity and in‡ ation relative to a single measure of short-term 15 See Perotti (2011). 16 See Galí, Lopéz and Vallés (2007), Ramey (2009) and Corsetti, Meier and Müller (2010). 10 interest rates. Bianchi et al. (2009) use the term structure as a proxy for in‡ation expectations in the U.K. Additionally, Fama (1990), Plosser and Rowenhorst (1994) and Adrian, Estrella and Shin (2010) all …nd a link - to varying degrees - between the term structure and in‡ ation/real activity. 4 Pro-cyclical government spending Alesina, Campante and Tabellini (2008) show government spending in emerging market countries to be much more pro-cyclical than in the advanced economies. This section explores the extent to which changes in government spending have been historically correlated with domestic economic conditions. The correlation coe¢ cients are based on the cyclical components which are extracted by using a bandpass …lter. This has the advantage of allowing one to extract cyclical ‡uctuations at business cycle frequencies (6 - 32 quarters). Eco- nomic theory also implies that changes in the persistence in …scal policy actions may lead to altering real economic outcomes.17 For this reason, correlation co- e¢ cients are also estimated using frequencies greater and less than the standard business cycle frequencies. The rolling correlation is calculated from a linear version of the reduced form VAR(p) from equation (1), Xt = (L)Xt 1 + "t ; where E("t "0t ) = (6) where Xt is a n 1 vector containing government spending and real output ‡ uctu- ations. This model forms the basis for the factor model estimated in subsequent sections. Following Stock and Watson (2005), the spectral density matrix of quarterly government spending is given as Shp (!) = C(ei! ) C(e i! )0 =2 , where C(L) 1 = [I (L)] is the moving average of the reduced form model. The implied spectral density matrix is j1 + ei! + e2i! + e3i! j2 SY Y (!) = [sij (!)]; so that sij (!) is the cross variable spectrum between variable i and variable j at frequency !. Using bandpass-…ltered series changes the spectral density matrix. In this case the spectral matrix of real house prices is jb(ei! )=(1 ei! )j2 SY Y (!) = [sij (!)]; where b is the bandpass …lter, so that jb(ei! )j2 = 1 for ! 0 ! ! 1 ; where the fre- quencies ! 0 and ! 1 correspond to periodicities of between six and 32 quarters, with 17 See Monacelli and Perotti (2006). 11 jb(ei! )j2 = 0 otherwise. The contemporaneous correlation, denoted ij ; between variables i and j can be estimated by R sij (!)d! ij = R 1=2 R 1=2 (7) sii (!)d! sjj (!)d! It is widely acknowledged in the economics profession that government spending as a fraction of GDP ought to remain constant over the cycle (Alesina, Campante and Tabellini, 2008). Following such a rule would lead to an implicit adoption of a counter-cyclical …scal policy rule. Figure 3 reports the rolling correlation between government spending as a fraction of GDP and real output growth, along with the one-standard deviation credible sets. The estimates show that …scal policy in Malaysia has become increasingly pro-cyclical over the last 25 years. This fact appears consistent at all three business cycle frequencies. The estimates show that during the 1980s and early 1990s …scal policy became progressively more procycli- cal, although not to any signi…cant degree. Following the Asian …nancial crisis, however, the median estimate shows the procyclicality of government spending rose. By 2007 the correlation coe¢ cient had reached 0:5. The estimates imply that government spending has grown in excess of real output growth since the Asian …nancial crisis in 1997/98. 5 State-dependent …scal multipliers The response functions are calculated using Koop, Pesaran and Potter (1996) non- linear impulse response functions are de…ned as: IRFt+h = Et [yt+h j"t ; t] Et [yt+h j"t ] (8) where h is the forecast horizon, yt+h contains the forecasts of the endogenous vari- ables at horizon h, t represents the current information set and "t is the current disturbance term. In the linear case, the impulse response functions are time in- variant, so that t = 0. In the non-linear case the impulse response functions are conditioned on a particularly history. In the linear case, the expectation of the path of output following the government spending shock, conditional on future shocks, is equal to the path of output when future shocks are set to their expected values. Therefore, future shocks can be set equal to zero for convenience. This is 12 not the case for linear models. Future shocks are drawn from some distribution and their e¤ects averaged out over a large number of draws. The non-linear re- sponse functions are calculated using a six step Monte Carlo procedure set out in Weise (1999). 5.1 Regime-switching impulse responses functions The impulse responses are expressed in terms of their deviation from steady-state. Since the IRFs are nonlinear, they will depend on the initial value of the index z and the size of the government policy shock. For example, the more deep the initial recession, and the less positive the spending shock, the less important future regime shifts out of the recession will be. Therefore, we must specify the initial conditions and the size of the policy experiment in order to estimate the response functions. Figures 4 - 6 illustrate the response of the macroeconomy and the size of the …scal multipliers to an unanticipated 1 percent increase in government spending (investment and consumption) and a 1 percent cut in personal income taxes. The real output and private consumption multiplier values are calculated as in Woodford (2011), which allows the size of the multiplier value to depend on the persistence of …scal shocks.18 Finally, the responses from the nonlinear model are compared to a linear model. Three key …ndings are uncovered: 1. In general, Figures 4(a) and 5(a) illustrate the nonlinear framework to have estimated public expenditure to be more e¤ective in growth recessions than normal’times. This is most clearly elucidated by the response of real output in ‘ to changes in government consumption spending in expansionary and recession- ary episodes. Figure 5(a) shows that during expansionary periods discretionary government consumption spending has a crowding out e¤ect on real output. In times of depressed economic growth, however, the 6-to-12 month response of real 18 Changes in the hysteresis of government spending may, theoretically, have important impli- cations for the size of the …scal multipliers. The more persistent a budget de…cit is expected to be, the more likely it is that current individuals will not be around before taxes are raised, and therefore the larger is the wealth e¤ect and the multiplier. On the other hand, Monacelli and Perotti (2006) showed in a New Keynesian dynamic general equilibrium model with perfect foresight that more persistent positive government spending shocks lead to a fall in real output. Although, this has the disadvantage of tying the model to historical experience concerning the persistence of shocks, and therefore may not apply to policies either less or more permanent. 13 output to a government consumption shock is positive. Figure 4(a) shows that the size of the response of real output and private consumption to discretionary government investment spending is greater during periods of depressed economic growth. Finally, the response of economic activity in Figure 6(a) shows that there is little di¤erence in the to a 1 percent tax cut in either state of nature. Fixed coe¢ cient structural VARs tend to predict a rise in private consumption in response to a government spending disturbance. In contrast, event based studies, such as Ramey and Shapiro (1998) and Ramey (2009), report a decline in private consumption to a discretionary spending shock. This, it has been pointed out, is due to …scal anticipation e¤ects which lead to mismeasurement of the timing of spending shocks. The …ndings in this paper imply the response of private consumption to discretionary government spending in …xed coe¢ cient models is dependent upon the component of government expenditure and the nature of the regime. For instance, the estimates show that under an expansionary growth state, discretionary government consumption crowds out private consumption. 2. It is instructive to look at the multiplier values. Comparing the response of real output and private consumption to government investment and government consumption shocks in Figures 4(b) and 5(b), the results show that government stimulus spending is more e¤ective - as measured by the size of the multiplier val- ues - when the focus is on discretionary government investment spending. Con- forming to (old) Keynesian theory a 1 dollar rise in government investment spend- ing leads to a more than 1 dollar rise in real output. Figure 4(b) illustrates a peak median real output multiplier of around 2:7 during the recessionary state and a multiplier of around 2 during normal time to a government investment shock. The corresponding multiplier values for private consumption are 1:2 and 1:4, re- spectively. That the response of private consumption is more damped relative to real output is consistent with the permanent/life-cycle hypothesis. In contrast to Leeper, Walker and Yang (2009), the estimates appear to display little evidence of a negative short-run multiplier in response to a government investment shock. Figure 5(b) reports the dynamic multiplier values to discretionary government consumption spending. Relative to the response for government investment spend- ing, the long-run con…dence intervals for real output and private consumption are 14 much wider, implying greater uncertainty. The dynamics multiplier values for real output during to government consumption innovations are negative in normal times. However, in the recessionary state there is a short-run (six month) positive rise in real output, with a peak multiplier value of around 2 after six months. The e¤ect on real output is completely dissipated after one year. In the short-run discretionary government consumption innovations have a pos- itive impact on private consumption, with multiplier values of around 1: Bouakez and Rebel (2003) show that, even in a neoclassical model, private consumption may rise following a public consumption shock if the two are complements. Our esti- mates show, however, that the positive e¤ect of government consumption spending on private consumption dissipates within six months in either regime. In the ex- pansionary state private consumption becomes crowded in the medium-term to discretionary government consumption spending. That government consumption spending has a negligible long-run e¤ect on private consumption for Malaysia is consistent with Kwan (2007), which showed private consumption to be completely crowded out in the long-run. Cochrane (1994) and Bachmann and Sims (2011) note that changes in private consumption spending proxy for news that consumers receive about future productivity. More ect persistence/permanent movements in consumption and income re‡ correspond- ing shifts in the long-run productivity potential of the economy. Such shifts are potentially una¤ected by animal spirits. On the other hand, if …scal policy con- tained no news about future fundamentals and the relationship between …scal pol- icy and subsequent activity only re‡ected animal spirits, this would be consistent with a transitory response of private consumption to a discretionary …scal policy action. The persistent almost random walk like long-run response of the real output and private consumption multiplier to government investment spending in Figure 4(b) is consistent with investment spending having a permanent impact on real economic activity. As a contrast, the multiplier values illustrated in Figure 5(b) show that government consumption spending has a transitory impact on real output and private consumption, implying no permanent e¤ect on income. Finally, the rise in interest rates to a government investment shock, as opposed to the fall recorded for government consumption spending, is consistent with im- 15 proved optimism over future economic prospects in response to new discretionary government investment spending. 3. Cuts in taxation also appear less e¤ective in stimulating real output and private consumption than government spending. The size of the short-run multipliers for real output and private consumption are considerably smaller than their government spending equivalents. Figure 6(b) shows that in an economic downturn a 1 ringgit cut in direct taxation leads to around a 0:25 sen rise in private consumption after six months. The analogous short-run estimate for normal times is 0:12. That the impact of a tax cut for private consumption doubles in e¤ectiveness during times of slow economic growth is consistent with the idea that the proportion of liquidity constrained individuals, with reduced access to …nancial markets, rises during an economic downturn. In contrast to private consumption, the short-run real output multiplier to a discretionary tax shock is negative in both states of nature. According to neoclassical predictions in Baxter and King (1993) a tax cut may lead private consumption and real output to move in opposite directions. In this instance, a tax cut leads to a rise in private wealth. This has the e¤ect of pushing in the labour supply curve, leading output (and employment) to fall. However, the same positive wealth e¤ect implies that private consumption must rise. Thus, consumption and real output move in opposite directions. After 6-to-9 months the real output multiplier rises, reaching a peak multiplier value of around 0:15: Within the year, however, the positive stimulus has completely dissipated. The long-run e¤ect of a tax cut on real output and private consumption is negligible, as re‡ected in the posterior error bands. These …ndings are perhaps unsurprising given the narrow tax base in Malaysia, coupled with the fact that a signi…cant proportion of government revenues are attained from commodities. There is scant evidence of the supply-side e¤ects arising from a tax cut predicted by standard neoclassical growth models. Standard real-business-cycle (RBC) models predict that a tax cut lead agents to substitute leisure for work, pushing out the labour supply curve, raising real output and investment. Finally, the decline in public spending to a tax cut shock illustrated in Figure 6(a) is consistent with the starve the beast’ hypothesis in Romer and Romer (2009). ‘ This suggests that 16 s further reducing corporate income tax in Malaysia, unless it’ matched by savings from a rationalisation of the …scal incentives system, may lead to a reduction in public spending. It must be noted that the di¤erence between the regime-based multipliers may be exaggerated due to the assumption that the parameters in the two states - growth and expansion - are constant. Regimes are disentangled based on strong growth expansions and recessions. Hence, as noted in Auerbach and Gorod- nichenko (2010), one should interpret reported magnitudes of the multipliers for the two regimes as bounds from polar settings rather than routinely encountered values. More realistic situations will fall between these values. If there is a positive probability of the economy shifting from a recessionary to expansionary phase in future periods, then the multipliers starting in recession (or expansion) should be a mix of those estimated for the separate regimes . Economic downturns are usually characterised by a rise in the number of liq- uidity constrained households and looser monetary policy conditions. Galí, López rule of thumb’(ROT) con- and Vallés (2007) have shown that as the fraction of ‘ sumers, which are characterised by their inability to save or borrow, increases in government spending will have more of an impact on aggregate demand. Under such circumstances it would be optimal for government spending to counteract the decline in economic activity and private consumption by increasing govern- ment consumption. This would have a positive e¤ect on the disposable income of individuals by pushing out of the demand curve for goods and services. These e¤ects are likely to be larger in emerging market economies, which in general have less sophisticated …nancial markets whilst containing greater frictions. As reporetd for the OECD countries in Auerbach and Gorodnichenko (2011) the results support predictions in the IS-LM-AS model. During an economic downturn (upturn), the AS curve ‡attens (steepens) implying greater slack (tightness) in the economy. In such a scenario a …scal stimulus that pushes out the IS curve would have a larger real economic impact compared with an environment where the economy is operating near capacity (steeper AS curve). Recent empirical evidence, mainly based on the advanced economies, shows …scal multiplier to vary over the traditional business cycle. Christiano, Eichenbaum and Rebelo (2009) 17 derive real output …scal multipliers that exceed a value of two and sometimes three in a deeply depressed economic environment and rising ROT consumers. Inconjunction with this …nding Hall (2009), Corsetti, Meier and Müller (2010), Erceg and Lindé (2010) and Woodford (2011) have all reported the multiplier value to increase during times of high …nancial stress. Eggertsson and Woodford (2003) have also previously shown that the size of …scal multipliers increase when monetary policy is in a liquidity trap. The …ndings also corroborate a consensus which has begun to emerge in the …scal policy literature regarding the growth-enhancing e¤ect of government in- vestment spending. General equilibrium models (see Baxter and King (1993), Pappa (2005) and Straub and Tchakarov (2007)) and in the recent empirical lit- erature (see Auerbach and Gorodnichenko (2010) and Cimadomo, Hauptmeier and Kirchner (2010)) …nd that government investment spending tends to generate larger …scal multipliers than discretionary government consumption innovations. Conventional explanations imply that government investment spending not only generates the usual aggregate demand e¤ect but also a positive additional aggre- gate supply e¤ect through enhancing production and the marginal productivity of labour (Auesher, 1989). The positive and persistent response of private consump- tion to government investment is consistent with standard neoclassical predictions, which imply that shocks about future productivity induce positive comovement in macroeconomic variables and persistent responses in private consumption.19 5.2 Policy Discussion The …ndings in this paper clearly underscore the case for countercyclical …scal policy, which implies preserving and strengthening …scal bu¤ers in good times. However, …scal risks in Malaysia have grown over recent years. Malaysian …s- cal policy has become increasingly pro-cyclical, setting the stage for large de…cits during the recent crisis. Additionally, the negative multiplier values for the re- sponses drawn for some of the estimates imply that not all types of government expenditures have credible stimulative e¤ects when the economy is in an economic slump. The results show that discretionary government investment spending to 19 See Cochrane (1994) and Barksy and Sims (2011). 18 be a more e¤ective way of stimulating economic activity than changes in tax policy or discretionary changes in government consumption spending. The larger returns to government investment spending, particularly when com- pared with estimates for the advanced economies, are perhaps unsurprising given that returns to capital are often higher in emerging market countries relative to the advanced economies. However, Figure 8 shows that investment spending has declined in Malaysia over the last decade. Correspondingly Figure 7 illustrates government spending plus transfers and subsidies to have become increasingly more procyclical and expansionary. Reducing the number of investment projects or slowing down their implementation is often politically and socially easier to im- plement than reducing current outlays such as subsidies. The patterns of spending voracity e¤ect’ and in- in Malaysia over the past decade are consistent with the ‘ , dicative of a political agency problem. In Malaysia, a multi-ethnic society, rent seeking is high and access to infor- mation is limited including on contingent liabilities.20 Due to the political agency problem, voters have tended to demand higher utility for themselves in good times, which was very evident between 2002 and 2008 (see Figure 7). These demands came in the form of pressures to maintain, and in some cases increase, subsidies for food and petrol in the pre-2008 crisis period, pressures to postpone implementation of GST (government sales tax) despite the need to expand the tax base and reduce oil dependence and volatility of government revenues, and an increased unwilling- ness to pay and pressures to freeze, lower or eliminate toll on major highways. Past governments have been generous moving away from the path of …scal consolidation in good times. As a consequence, subsidies and debt increased between 2002 and 20 Contingent liabilities in Malaysia emanate from a number of factors, including government exposure to large scale PPPs infrastructure projects. The expansive toll roads program was the main source in the past. More recently, the new wave of projects under the Economic Transformation Program (ETP) may have increased government exposure in a considerable way. The government can expect to face large pay out if a number of unfavorable factors coincide such as low ridership, higher than anticipated costs or change in tra¢ c ‡ ows. An additional source s of coningent liabilities is subsidised lending and government’ commitment to various subsidies which will still have to be paid even in the event of a downturn. The World Bank (2000) has estimated that if ridership was about half of the projections, then the expected costs to the Government for one of the projects could be as high as about two-thirds of project costs. To date, to our knowledge, there is no comprehensive system to account for the Malaysian government’ s contingent liabilities. 19 2008. Subsidies increased by 642 percent, whilst the corresponding …gure for the stock of debt is 109 percent. As noted in the introduction, it is the role of …scal policy during periods of slower growth or crises times that should be of relevance to policymakers in emerg- ing market economies. The …ndings in this paper suggest that a large …scal stim- ulus during an economic downturn like the 199/98 Asian …nancial crisis - precisely when the number of balance-sheet/liquidity constrained agents were on the rise - could have loosened economic conditions and precipitated a faster economic recov- ery. These …ndings provide evidence in support of the argument that higher …scal outlays are particularly e¤ective as a stabilisation tool in a sharp economic down- turn. Thus, a recession could be further aggravated if it forces the government to retrench spending in the midst of the downturn. s A key criticism of the International Monetary Fund’ advice, during the 1997/98 …nancial crisis that ravaged countries in South-East Asia, including Malaysia, was the stress put on …scal tightening. These countries were seen as sacri…cing long-run growth in order to show signs of …scal discipline. Tight …scal policy implemented s during the crisis reduced the country’ net worth, by cutting productive govern- ment spending and causing a downward shift in their long-run growth trajectory. The implication of the …ndings is that expansionary …scal policy during crisis periods may also help prevent long-term growth damage to the economy. The estimates in this paper show that government investment spending has long-run e¤ects on real economic growth in Malaysia. These results appear more consistent with the predictions of (old) Keynesian theory, and hard to reconcile with those of the neoclassical paradigm. Eggertsson and Krugman (2010) recently argued that a …scal expansion - even through an increase in (public) debt - could help alleviate a problem caused by high (private) debt and a resulting deleveraging shock due to a sharp economic downturn or economic (…nancial) crisis. 6 Conclusion This paper investigates changes in the size of …scal multipliers over stages of the growth cycle, which is more relevant for emerging market countries, for Malaysia. 20 We also contribute to the literature by investigating how output multipliers for government purchases may di¤er for di¤erent components of government spending. This paper shows that …scal policy in Malaysia has become increasingly pro- cyclical over the last 25 years, especially after the 1998 East Asian …nancial crisis. Using a non-linear model to estimate the dynamic relationship between discre- tionary …scal policy and economic activity, it is established that the size of …scal multipliers tend to increase (decrease) during periods of slow (faster) economic growth. The returns to discretionary government spending also di¤er depending upon the components of government spending. Multiplier values are larger when the focus is on public investment, as opposed to consumption. A 1 Malaysian Ring- git rise in government investment spending leads to a maximum median output multiplier of around 2:7 during the recessionary state and a around 2 in normal times. The corresponding multiplier values for private consumption are 1:2 and 1:4, respectively. Our estimates suggest that government investment spending has a permanent impact on the real economic activity, while government consumption spending in Malaysia has a transitory long-run impact. Finally, changes in tax policy are less e¤ective in stimulating economic activity than direct government spending. These results con…rm conjectures in the literature by providing empiri- cal backing for the idea that procyclicality in …scal policy reduces the e¤ectiveness of …scal actions in emerging countries. References Quarterly real GDP estimates [1] Abeysinghe, T. and R. Gulasekaran (2004), ‘ , for China and ASEAN4 with a forecast evaluation’ Journal of Forecasting, 23: 431 - 447. [2] Aghion, P., G-M. Angeletos, A. Banerjee and K. Manova (2010), ‘ Volatility , and growth: Credit constraints and the composition of investment’ Journal of Monetary Economics, 57(3): 246 - 265. [3] Aguiar, M. and G. Gopinath (2007), ‘ Emerging Market Business Cycles: The , Cycle is the Trend’ Journal of Political Economy, 115: 69 - 102. Why is …scal policy often [4] Alesina, A., F. Campante and G. Tabellini (2008), ‘ , procyclical’ Journal of the European Economic Association, 6(5): 1006 - 1036. 21 Fiscal Multipliers in Recessions [5] Auerbach, A. and Y. Gorodnichenko (2010), ‘ , and Expansions’ NBER Working Paper No. 17447. Measuring the Output Respone- [6] Auerbach, A. and Y. Gorodnichenko (2011), ‘ , ses to Fiscal Policy’ NBER Working Paper No. 16311. [7] Azrai, E. M. and A. G. Zeufack (2011), ‘ Malaysia: Postcrisis Growth Prospects Depend on Restoring Fiscal Discipline and Private Sector Con…- , dence’ ed. M. K. Nabli in The Great Recession and Developing Countries: Economic Impact and Growth Prospects, The World Bank. Con…dence and the Transmission of Gov- [8] Bachmann, R. and E. Sims (2011), ‘ , ernment Spending Shocks’ NBER Working Paper No. 17063. Macroeconomic e¤ects from government [9] Barro, R. and C. J. Redlick (2011) ‘ , purchases and taxes’ forthcoming Quarterly Journal of Economics. Fiscal Policy in General Equilibrium’ Amer- [10] Baxter, M. and R. King (1993), ‘ , ican Economic Review, 83(3): 315 - 334. An [11] Blanchard, O. J. and R. Perotti (2002), ‘ empirical characterisation of the dynamic e¤ects of changes in government spending and taxes on output’ , Quarterly Journal of Economics, 117(4): 1439 - 1469. Why does private consumption rise after [12] Bouakez, H. and R. Norman (2003), ‘ , a government spending shock’ Bank of Canada Working Paper No 2003-48. When is the government [13] Christiano, L., M. Eichenbaum and S. Rebelo (2009), ‘ , spending multiplier large?’ NBER Working Paper No. 15394. [14] Cimadomo, J., S. Hauptmeier and M. Kirchner (2010), ‘ Transmission of gov- , ernment spending shocks in the euro area: Time variation and driving forces’ European Central Bank Working Paper Series No. 1219. [15] Cochrane, J. (1994), ‘ , Shocks’ Carnegie-Rochester Conference Series on Pub- lic Policy, 41: 295 - 364. What determines government [16] Corsetti, G., A. Meier and G. J. Müller (2010), ‘ , spending multipliers’ mimeo Debt, Deleveraging, and the Liquid- [17] Eggertsson, G. and P. Krugman (2010), ‘ , ity Trap’ mimeo Is [18] Erceg, C. J. and J. Lindé (2010), ‘ there a Fiscal Free Lunch in a Liquidity , Trap’ CEPR Discussion Paper No. 7624. International Economic Update’ [19] Federal Reserve Bank of Dallas (2009), ‘ Understanding the E¤ects of [20] Galí, J., J. D. Lopéz and J. Vallés (2007), ‘ , Government Spending on Consumption’ Journal of the European Economic Association, 5: 227 - 270. 22 [21] Granger, C. W. and T. Terasvirta (1993), Modelling Nonlinear Economic Relationships, New York: Oxford Economic Press. By [22] Hall, R. E. (2009), ‘ how much does GDP rise if the government buys more , output’ Brookings Papers on Economic Activity, 2: 183 - 231. When it Rains it Pours: [23] Kaminsky, G., C. Reinhart and and C. Végh (2004), ‘ , Procyclical Capital Flows and Macroeconomic Policies’ NBER Macroeco- nomic Annual, MIT Press. The Direct Substitution between Government and Pri- [24] Kwan, Y. K. (2007), ‘ , vate Consumption in East Asia’ in ed. by T. Ito and A. Rose Fiscal Policy and Management in East Asia, NBER (EASE), Volume 16. Government Investment [25] Leeper. E., T. B. Walker and S-C. S. Yang (2009), ‘ , and Fiscal Stimulus in the Short and Long Runs’ NBER Working Paper No. 15153. Fiscal Policy, the Trade Balance, and [26] Monacelli, T. and R. Perotti (2006), ‘ , the Real Exchange Rate: Implications for International Risk Sharing’ mimeo [27] Pappa, E. (2005), ‘ New-Keynesian or RBC Transmission? The E¤ects of , Fiscal Shocks in Labour Markets’ CEPR Discussion Paper No. 5313. Estimating the e¤ects of …scal policy in OECD countries’ [28] Perotti, R. (2005a), ‘ , mimeo. [29] Perotti, R. (2005b), ‘Fiscal Policy in Developing Countries: A Framework , and Some Questions’ The World Bank, Development Research Group, Policy Research Working Paper 4365. In , [30] Perotti, R. (2007), ‘ Search of the Transmission Mechanism of Fiscal Policy’ NBER Macroeconomics Annual, 22: 169 - 249. Expectations and Fiscal Policy: An Empirical Investiga- [31] Perotti, R. (2011), ‘ , tion’ mimeo. [32] Ramey, V. A and M. Shapiro (1998), ‘ Costly capital reallocation and the ef- , fects of government spending’ Carnegie-Rochester Conference Series on Pub- lic Policy, 48: 145 –149. Identifying Government Spending Shocks: It’ All in [33] Ramey, V . A. (2009), ‘ s , the Timing’ NBER Working Paper No. 15464. Do [34] Romer, D. and C. Romer (2009), ‘ Tax Cuts Starve the Beast? The E¤ect , of Tax Changes on Government Spending’ Brookings Papers on Economic Activity, 40(1): 139 - 214. [35] Romp, W. and J. de Haan (2007), ‘ Public Capital and Economic Growth: A , Critical Survey’ Perspektiven der Wirtschaftspolitik, 8(s1): 6 - 52. Has the Business Cycle Changed and Why?’ [36] Stock, J. and M. Watson (2002), ‘ , NBER Macroeconomics Annual, M. Gertler and K. Rogo¤ (eds), MIT Press. 23 Assessing the Impact of a Change in the [37] Straub, R. and I. Tchakarov (2007), ‘ Composition of Public Spending’ECB Working Paper No. 795. The e¤ects of …scal policy on consumption in recessions [38] Tagkalakis, A. (2008), ‘ , and expansions’ Journal of Public Economics, 92: 1486 - 1508. The Asymmetric E¤ects of Monetary Policy: A Nonlin- [39] Weise, C. L. (1999), ‘ , ear Vector Autoregression Approach’ Journal of Money, Credit and Banking, 31(1),85 - 108. Simple Analytics of the Government Expenditure Mul- [40] Woodford, M. (2011), ‘ , tiplier’ American Economic Journal: Macroeconomics, 3(1): 1 - 35. Malaysia: Public Expenditure Review’ Washington D.C. [41] World Bank (1999), ‘ , 24 S m ooth Tra nsition Function 1 .0 0 0 .7 5 0 .5 0 0 .2 5 0 .0 0 -0 .2 -0 .1 6 -0 .1 2 -0 .0 8 -0 .0 4 0 0 .0 4 0 .0 8 0 .1 2 0 .1 6 0 .2 -0 .1 8 -0 .1 4 -0 .1 -0 .0 6 -0 .0 2 0 .0 2 0 .0 6 0 .1 0 .1 4 0 .1 8 Figure 1 Real Output Growth Weight on Recession Regime - F(z) 7.5 0.00 5.0 -0.25 2.5 0.0 -0.50 -2.5 -5.0 -0.75 -7.5 -10.0 -1.00 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Figure 2: Weight on recession regime and growth recessions Cyclical fre que ncy: 6-36 Cyclical fre que ncy: 12-48 Cyclical fre que ncy: 6-24 1.00 1.00 1.00 0.75 0.75 0.75 0.50 0.50 0.50 0.25 0.25 0.25 0.00 0.00 0.00 -0.25 -0.25 -0.25 -0.50 -0.50 -0.50 -0.75 -0.75 -0.75 -1.00 -1.00 -1.00 1984 1987 1990 1993 1996 1999 2002 2005 2008 1984 1987 1990 1993 1996 1999 2002 2005 2008 1984 1987 1990 1993 1996 1999 2002 2005 2008 Figure 3: Rolling correlation between government spending and business cycle fluctuations Confidence Interval Recession Regime Confidence Interval Expansion Regime Confidence Interval for Linear Model G response to G shock G response to G shock G response to G shock 1 .0 1 .0 1 .4 Recessi on Confi dence 0 .8 Expansi on 0 .8 1 .2 Li near 1 .0 0 .6 0 .6 0 .8 0 .4 0 .4 0 .6 0 .2 0 .2 0 .4 0 .2 0 .0 0 .0 0 .0 - 0.2 - 0.2 - 0 .2 - 0.4 - 0.4 - 0 .4 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 T response to G shock T response to G shock T response to G shock 3 .0 2 .0 2 .0 2 .5 1 .5 1 .5 2 .0 1 .0 1 .0 1 .5 1 .0 0 .5 0 .5 0 .5 0 .0 0 .0 0 .0 - 0.5 - 0 .5 - 0.5 - 1.0 - 1.0 - 1 .0 - 1.5 - 1.5 - 1 .5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Y response to G shock Y response to G shock Y response to G shock 4 3 3 3 2 2 2 1 1 1 0 0 0 -1 -1 -1 -2 -2 -2 -3 -3 -3 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 C response to G shock C response to G shock C response to G shock 2 .0 1 .2 5 2 .0 1 .0 0 1 .5 1 .5 0 .7 5 1 .0 0 .5 0 1 .0 0 .2 5 0 .5 0 .5 0 .0 0 0 .0 - 0 .2 5 0 .0 - 0 .5 0 - 0.5 - 0 .5 - 0 .7 5 - 1.0 - 1 .0 0 - 1 .0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 i response to G shock i response to G shock i response to G shock 0 .75 1 .0 0 2 .0 0 .50 0 .7 5 1 .5 0 .25 0 .5 0 1 .0 0 .2 5 0 .00 0 .0 0 0 .5 - 0 .2 5 - 0 .2 5 0 .0 - 0 .5 0 - 0 .5 0 - 0 .7 5 - 0 .5 - 0 .7 5 - 1 .0 0 - 1 .0 0 - 1 .0 - 1 .2 5 - 1 .2 5 - 1 .5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Figure 4(a): Responses to government investment spending Confidence Interval Recession Regime Confidence Interval Expansion Regime Y multiplier to Gov. Inv. shock Y multiplier to Gov. Inv. shock 3.0 3.0 Recession Expansion Confidence Interval 2.5 2.5 2.0 2.0 Multiplier Value Multiplier Value 1.5 1.5 1.0 1.0 0.5 0.5 0.0 0.0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 C multiplier to Gov. Inv. shock C multiplier to Gov. Inv. shock 1.50 1.50 1.25 1.25 Multiplier Value Multiplier Value 1.00 1.00 0.75 0.75 0.50 0.50 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Figure 4(b): Fiscal multipliers to a government investment spending shock Confidence Interval Recession Regime Confidence Interval Expansion Regime Confidence Interval for Linear Model G response to G shock G response to G shock G response to G shock 1 .2 5 1 .2 5 2 .0 Rec es s ion Confi denc e 1 .0 0 Ex pans ion 1 .0 0 1 .5 Linear 0 .7 5 0 .7 5 1 .0 0 .5 0 0 .5 0 0 .5 0 .2 5 0 .2 5 0 .0 0 .0 0 0 .0 0 - 0 .2 5 - 0 .5 - 0 .2 5 - 0 .5 0 - 1 .0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 T response to G shock T response to G shock T response to G shock 3 2 2 2 1 1 1 0 0 -1 0 -2 -1 -1 -3 -2 -2 -4 -3 -3 -5 -4 -6 -4 -5 -7 -5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Y response to G shock Y response to G shock Y response to G shock 2 .0 1 .5 1 .5 1 .5 1 .0 1 .0 1 .0 0 .5 0 .5 0 .5 0 .0 0 .0 0 .0 - 0 .5 - 0 .5 - 1 .0 - 0 .5 - 1 .0 - 1 .5 - 1 .0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 C response to G shock C response to G shock C response to G shock 2 .0 1 .5 1 .5 0 1 .2 5 1 .5 1 .0 1 .0 0 1 .0 0 .7 5 0 .5 0 .5 0 0 .5 0 .2 5 0 .0 0 .0 0 .0 0 - 0 .5 - 0 .2 5 - 0 .5 - 0 .5 0 - 1 .0 - 1 .0 - 0 .7 5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 i response to G shock i response to G shock i response to G shock - 0 .0 0 .0 0 0 .2 - 0 .2 - 0 .2 5 - 0 .0 - 0 .4 - 0 .5 0 - 0 .2 - 0 .6 - 0 .4 - 0 .7 5 - 0 .8 - 0 .6 - 1 .0 0 - 1 .0 - 0 .8 - 1 .2 5 - 1 .2 - 1 .0 - 1 .4 - 1 .5 0 - 1 .2 - 1 .6 - 1 .7 5 - 1 .4 - 1 .8 - 2 .0 0 - 1 .6 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Figure 5(a): Responses to government consumption spending Confidence Interval Recession Regime Confidence Interval Expansion Regime Y multiplier to Gov. Cons. shock Y multiplier to Gov. Cons. shock 3 2 2 1 1 Multiplier Value Multiplier Value 0 0 -1 -1 -2 -3 -2 -4 Recession Expansion Confidence Interval -5 -3 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 C multiplier to Gov. Cons. shock C multiplier to Gov. Cons. shock 2.0 1.5 1.5 1.0 1.0 0.5 0.5 0.0 Multiplier Value Multiplier Value 0.0 -0.5 -0.5 -1.0 -1.0 -1.5 -1.5 -2.0 -2.0 -2.5 -2.5 -3.0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Figure 5(b): Fiscal multipliers to a government consumption spending shock Confidence Interval Recession Regime Confidence Interval Expansion Regime Confidence Interval for Linear Model G response to Tax shock i response to Tax shock G response to Tax shock 0 .0 0 2 0 .0 0 2 0 .0 0 4 Rec es s ion Confidenc e Ex pans ion 0 .0 0 1 0 .0 0 0 0 .0 0 2 Li near 0 .0 0 0 - 0 .0 0 1 0 .0 0 0 - 0 .0 0 2 - 0 .0 0 2 - 0 .0 0 2 - 0 .0 0 3 - 0 .0 0 4 - 0 .0 0 4 - 0 .0 0 4 - 0 .0 0 6 - 0 .0 0 5 - 0 .0 0 6 - 0 .0 0 6 - 0 .0 0 8 - 0 .0 0 7 - 0 .0 0 8 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 T response to Tax shock T response to Tax shock T response to Tax shock 0 .5 0 0 .2 5 0 .2 5 0 .2 5 0 .0 0 0 .0 0 0 .0 0 - 0 .2 5 - 0 .2 5 - 0 .2 5 - 0 .5 0 - 0 .5 0 - 0 .5 0 - 0 .7 5 - 0 .7 5 - 1 .0 0 - 0 .7 5 - 1 .0 0 - 1 .2 5 - 1 .2 5 - 1 .0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Y response to Tax shock Y response to Tax shock Y response to Tax shock 0 .1 5 0 .1 5 0 .2 0 0 .1 5 0 .1 0 0 .1 0 0 .1 0 0 .0 5 0 .0 5 0 .0 5 0 .0 0 0 .0 0 - 0 .0 0 - 0 .0 5 - 0 .0 5 - 0 .0 5 - 0 .1 0 - 0 .1 0 - 0 .1 0 - 0 .1 5 - 0 .1 5 - 0 .1 5 - 0 .2 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 C response to Tax shock C response to Tax shock C response to Tax shock 0 .0 6 0 .0 4 0 .0 4 0 .0 4 0 .0 2 0 .0 2 0 .0 2 0 .0 0 0 .0 0 0 .0 0 - 0 .0 2 - 0 .0 2 - 0 .0 2 - 0 .0 4 - 0 .0 4 - 0 .0 4 - 0 .0 6 - 0 .0 6 - 0 .0 6 - 0 .0 8 - 0 .0 8 - 0 .1 0 - 0 .0 8 - 0 .1 0 - 0 .1 2 - 0 .1 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 i response to Tax shock i response to Tax shock i response to Tax shock 0 .0 2 0 .0 2 0 .0 2 0 .0 0 0 .0 0 0 .0 0 - 0 .0 2 - 0 .0 2 - 0 .0 2 - 0 .0 4 - 0 .0 4 - 0 .0 4 - 0 .0 6 - 0 .0 6 - 0 .0 6 - 0 .0 8 - 0 .0 8 - 0 .1 0 - 0 .1 0 - 0 .0 8 - 0 .1 2 - 0 .1 2 - 0 .1 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Figure 6(a): Responses to 1 percent tax cut shock Confidence Interval Recession Regime Confidence Interval Expansion Regime Y multiplier to Tax shock Y multiplier to Tax shock 0.3 0.25 Recession Expansion Confidence Interval 0.20 0.2 0.15 Multiplier Value Multiplier Value 0.10 0.1 0.05 0.0 0.00 -0.05 -0.1 -0.10 -0.2 -0.15 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 C multiplier to Tax shock C multiplier to Tax shock 0.30 0.25 0.25 0.20 0.20 0.15 Multiplier Value Multiplier Value 0.15 0.10 0.10 0.05 0.05 0.00 0.00 -0.05 -0.05 -0.10 -0.10 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Figure 6(b): Fiscal multipliers to a tax cut shock 0.250 0.225 0.200 0.175 %GDP 0.150 0.125 0.100 0.075 0.050 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 Figure 7: Govenment consumption spending plus transfers as %GDP 0.50 0.45 0.40 % of total spending 0.35 0.30 0.25 0.20 0.15 0.10 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Figure 8: Investment expenditures as % of total spending