The Government Spending Multiplier in Turkey

ABSTRACT

This study aims to measure the size of the government spending multiplier in Turkey for post-2001 financial crisis period within a structural VAR framework. The analysis demonstrates that a positive shock to government spending tends to increase output, tax, and real interest rate on impact and the size of the fiscal multiplier is relatively large at first few quarters. The fiscal multiplier reaches a peak value of 1.5 at second quarter and then starts to diminish. Furthermore, investigating the effects of the components of government spending reveals the fact that government investment expenditures, rather than consumption expenditures, have a profound impact on output at first few quarters.

KEY WORDS:

• Government spending multiplier
• SVAR

JEL CLASSIFICATION:

• E62
• H30

Acknowledgments

I would like to thank Zafer Yükseler, Ercan Türkan, Hakan Kara, Mustafa Kılınç, Hande Küçük Yeşil, and two anonymous referees for useful comments and suggestions. The article represents the views and analysis of the author only and does not represent those of the Central Bank of the Republic of Turkey. Thus, any error is mine.

Notes

^1. Auerbach and Gorodnichenko (2012, 2013)) assess the sensitivity of fiscal multipliers to state of the business cycles. Christiano, Eichenbaum, and Rebelo (2011) analyze the effectiveness of fiscal policy when monetary policy is at zero lower bound.

^2. Woodford (2011) uses simple analytical frameworks in order to investigate the size of the fiscal multiplier under a neoclassical benchmark and a New Keynesian benchmark. He also shows how the size of fiscal multiplier changes under alternative assumptions about the degrees of monetary accommodation.

^3. See Hemming, Kell, and Mahfouz (2002) for information about the theoretical and empirical literature on the effects of fiscal policy on the economy.

^4. Comparison of these techniques can be found in Perotti (2005) and Caldara and Kamps (2008).

^5. The determinants of fiscal multiplier are clearly explained in Batini, Eyraud, and Weber (2014), who classified these factors into two categories, namely structural characteristics (trade openness, labor market rigidity, the size of fiscal multipliers, the exchange rate regime, the debt level, public expenditure management, and revenue administration) and temporary factors such as the state of the economy and the existence of monetary accommodation.

^6. The rationale behind this short sample period can be attributed to the more stable feature of the Turkish economy compared to the pre-crisis period.

^7. General government covers central government, local governments, social security institutions, extrabudgetary funds, revolving funds, the unemployment insurance fund, and the general healthcare insurance.

^8. The fiscal stimulus package was of a temporary nature and comprised many measures regarding revenues and expenditures. Revenue measures covered private taxes, taxes on companies, consumption taxes on goods and services, and other revenue measures. On the expenditure side, Turkey took several measures including public consumption and investment, transfers to households and companies, contributions to employment and social security premiums, transfers to the rest of the public sector, and other expenditures.

^9. A deterioration in the budget balance and debt stock in 2009 mainly stems from two channels: First, automatic stabilizer effects due to the economic contraction and, second, a countercyclical discretionary fiscal policy via a fiscal stimulus package.

^10. We also observe a similar upward trend for aggregated budget data, namely primary spending to GDP ratio and total revenue to GDP ratio (Table 1).

^11. One may expect that a deficit-financed increase in government purchases has a more profound impact on output than a tax-financed increase in government spending.

^12. Suppose that government implements an expansionary fiscal policy by increasing public expenditures and this leads to higher inflation. How should the Central Bank react to an increase in inflation? If the Central Bank chose to keep inflation under control via an increase in interest rates, this may offset some of government spending’s expansionary effects.

^13. Ex ante real interest rates are defined as$r_t=(1+i_t)/(1+{\pi }_{t+4}^e)-1$, where $r_t$ and $i_t$ represent real and nominal interest rates, respectively, and ${\pi }_{t+4}^e$ denotes 12-month ahead CPI expectation. Nominal interest rates are average annual compounded interest rates at the Treasury’s TL-denominated zero-coupon security auctions.

^14. HEGY Test results are available from the author upon request.

^15. Cointegration equations (in the form of tax relation) show that the coefficients for government spending and output have positive sign and they are statistically significant. As it is expected, when output increases tax also increases due to automatic stabilizer effect. Moreover, an increase government spending leads to higher tax revenues. This might reflect the fact that an increase in government spending motivates the output and so tax revenues, which are positively associated with output.

^16. This kind of VAR system is commonly used and called AB model in the literature (See Amisano and Giannini (1997)). It is also quite common to use in the fiscal VAR studies following the seminal paper of Blanchard and Perotti (2002). The advantage of using SVAR approach is that it deals with the simultaneity (endogeneity) problem in the system. As shown in model equations fiscal policy variables affect output contemporaneously and vice versa. Different from the recursive approach (Cholesky ordering), Blanchard and Perotti’s SVAR approach enables to impose restrictions rather than zero on the model parameters. Since, the size of fiscal multiplier is sensitive to the elasticity assumptions imposing on model parameters one need to take into account this fact.

^17. Discretionary fiscal policy contains both decision lags and implementation lags, which implies that policymakers do not make a decision in order to minimize fluctuations in the business cycle in a very short time period. It takes some time for policymakers to understand what happens to economy, take fiscal measures that are needed, and implement these decisions. Therefore, the frequency of the data is very important for identification of the fiscal policy shocks. Using quarterly data instead of annual data can help to eliminate the second channel (discretionary fiscal policy). This is because a period of three months is a very short time period for policymakers to take fiscal measures due to decision and implementation lags in fiscal policy. This is the identification strategy that takes into account decision lags in fiscal policy.

^18. While Perotti (2005) and Perotti (2004) include nominal interest rate in their models, Batini et al. (2012) use real interest rate.

^19. Note that the definition of government spending used in this study does not cover interest payments and other transfer expenditures.

^20. Consumption-based tax revenues are the primary components of total tax revenues in Turkey, and, on average, indirect taxes constitute two-thirds of total central government tax revenues (67%) for the period of 2006–2013. Some studies estimate tax elasticity in Turkey, such as Çebi and Özlale (2012) and Çulha (2012), who find the weighted average of tax elasticity as 1.07 and 1.2, respectively.

^21. A detailed survey on fiscal multipliers can be found in Spilimbergo, Symansky, and Schindler (2009) and Batini, Eyraud, and Weber (2014).

^22. RATS 6.0 is used for SVAR estimation and impulse response functions. Impulse response functions are presented with one-standard deviation confidence intervals which are calculated by Monte Carlo simulations. Additionally, since we work with growth rates of the variables, impulse responses are accumulated to find the responses of level of each variable to a shock in level of government spending.

^23. To be more precise, we find that real interest rate tends to fall on impact (statistically insignificant), which in turn leads to an increase in output response at first few quarters. After that real interest rate starts to rise and limits the initial expansionary effects of government spending. An increase in interest rate may result from two effects: First, following an increase in government spending Central Bank may increase interest rate in order to respond to an inflationary effect. Second, an increase in government spending may increase risk premium, which results in an increase in interest rate. Since concerns on debt sustainability have been reduced during the estimation period, we think that the first channel plays an important role in increasing interest rates.

^24. Higher tax revenues may stem from two channels: First, cyclical changes may affect tax revenues automatically. An increase in government spending may increase the output, which in turn raises private spending and so does the tax revenues. Second, the government may raise taxes in a discretionary way in order to finance the budget deficit resulting from increasing government spending.

^25. To understand the reason why real interest rate falls after a government investment shock, we estimate a five-variable VAR model by including both nominal interest rate and expected inflation instead of real interest rate. The impulse response analysis shows that the response of real interest rate is shaped mostly by nominal interest rate, which declines following a government investment shock.

^26. The size of the tax multiplier is beyond the scope of this article.

^27. To convert elasticities to TL changes we divide elasticities to sample average of government spending to GDP ratio.

^28. Consistent with the most of the literature, such as Blanchard and Perotti (2002) for US and de Castro and Hernández De Cos (2008) for Spain, government spending reacts persistently to its own shock in Turkey. However, in contrast to these studies, we do not observe a persistent response for output. One possible explanation for this result can be attributed to a positive and persistent behavior of real interest rate following a government spending shock.

^29. We also explore the effects of a positive shock to government purchases of goods and services on the output. Fiscal multiplier is found to be 1.3.

^30. The findings of this study are in line with other studies such as Bouakez, Guillard, and Roulleau-Pasdeloup (2014) who find that the stimulative effects of government investment are higher than the ones of government consumption in the short run. Moreover, Hernández De Cos and Moral-Benito (2013) find that the size of government spending multiplier appears to be larger and above 1 during recessions in Spain. For example, they find that the peak values of government investment and that of government consumption multiplier are 4.21 and 1.56, respectively, in recession times. Herbert (2014), who focuses on US, Germany, and France, can also be given as another example for high values of government investment and consumption multipliers.