Chapter 13
The Impact of Social Policy on the Thai Economy

Bangorn Tubtimtong, Polpat Kotrajaras, and Bundit Chaivichayachat

INTRODUCTION
DATA AND METHODOLOGY
EMPIRICAL RESULTS
CONCLUSION
References

INTRODUCTION

In fiscal year 2004, the government of Thailand allocated its budget to “five strategic uses” to meet the expected economic growth target of 5.5–6.0 percent. The five strategies were for economic recovery and development; restructuring the economy to improve competitiveness; social development, poverty reduction, and upgrading the quality of life; foreign affairs and security in national development; and national administration. Approximately 39.3 percent of the total budget of 1,028 billion baht was allocated to a number of social development projects, such as educational reform, universal health care, protection of natural resources and the environment, the prevention and suppression of drug abuse, and poverty reduction. Details of the budget allocation for fiscal year 2004 are shown in Table 13.1.

It can be seen from the table that the first priority of the allocation was given to social spending in order to improve the people's quality of life. This spending was to stimulate the growth of the economy and to increase domestic consumption, investment, and exports. The National Economic and Social Development Board (2003), using survey methods, reported the positive effects of the budget on village and urban community fund management and development from the One Tumbon One Project (OTOP) and the universal health care project. This grassroot policy contributed approximately 0.72 percent to the growth of gross domestic product (GDP) (to 5.3 percent) in 2002, while private consumption increased by about 0.85 percent of total consumption. The Thailand Development Research Institute (TDRI), using simulation methods under different scenarios, forecasted the same outcomes from the effects of fiscal policy in stimulating the economy.

The objective of this chapter is to analyze the impact of fiscal policy, focusing on the impact of social spending on GDP. The hypothetical test here is that social spending shocks should have a positive effect on output and its components, and a positive tax shock should have a negative effect. To carry out these tests a VAR model and generalized impulse response functions are used, in the tradition of Blanchard and Perotti (1999), to characterize the dynamic effects of social spending shocks and tax shocks.¹ Tax shocks are included in the model because the VAR approach allows for interaction among variables, such as GDP, taxation, and spending. Secondly, a comparison is made of the effectiveness of social and economic spending in stimulating output. The second objective aims to increase the efficiency of the budget allocation by changing the type of spending. Economic spending should have a stronger impact than social spending. Thirdly, a comparison is made of the dynamic response of each constituent of social policy in stimulating the economy, such as spending on educational reform, universal health care, and poverty reduction. Finally, an analysis is made of the impact of social policy on the composition of GDP, such as consumption, investment, exports, and imports.

¹ For a review of the literature on the effectiveness of fiscal stimuli on economic activity, see Hemming et al. (2002), and for Japan, Matsuoka (1996).

DATA AND METHODOLOGY

Quarterly data for GDP, taxation, and government spending from 1995 to 2004 are taken from the Bureau of Budget (2004). The average share of taxation is approximately 14.97 percent of GDP, while government spending averages 17.53 percent of GDP during that period. This pattern remains the same for the whole period. The allocated budget for social spending has the highest share in the total budget, at 42.50 percent (Figure 13.1). It is used to provide for community and social services and is divided into capital spending and current spending.² Current expenditure tends to stimulate

² Capital spending includes the acquisition of fixed assets, such as land and buildings, which contribute to capital formation. Current spending consists of general administrative expenses, such as salaries, wages, personal expenses, and other procurements, which are not related to capital formation.

consumption more than investment. During the sample period, 1995–2004, the average share of economic spending is 24.76 percent, on health it is 7.82 percent, on social security and welfare it is 6.24 percent, and on housing it is 4.38 percent of the total budget.

The methodology used here is applied from Blanchard and Perotti (1999). They use a structured VAR to characterize the dynamic effects of changes in government spending and taxes on output. A dummy variable is used to identify the period of the shock and a structured VAR to investigate the relationships among the residuals of the variables in the model during different time periods. Institutional information about tax and transfer systems and the timing of tax collections are used to identify the restrictions of the structural VAR, which is then used to look at the dynamic effects of government defense spending on output and its components.

In the empirical work in this chapter, the Blanchard and Perotti model is applied to study the dynamic effects of the Thai government's social spending on output and its components. However, the generalized impulse response function is used instead of the traditional impulse response function, in which the shocks from the innovation are based only on information of past shocks in the initial period. Because unexpected shocks can occur during the response process, the generalized impulse response has been constructed to take care of these unexpected shocks during the response process. They are added in the form of a random vector, which includes both historical shocks and the new arbitrary shocks to the model. In this way, the problem of reordering responses from the model can be avoided.

The specification for the VAR model, with the generalized impulse response function, is as follows:

where v_t is an arbitrary shock and w_t–;₁ is a historical shock. The dynamic effect of the shocks is estimated from the VAR model with the generalized impulse response function. This new pattern includes both historical shocks and arbitrary shocks.

Model 1

The first model looks at the dynamic effects of total government spending on output, and the effect of tax shocks on both output and spending. The specification is as follows:

Y_t = A(L)Y_t−11 U_t

where , X_t is output, T_t is taxation, and GEX_t is government spending.

Model 2

The second model looks at the dynamic effect of changes in social spending on output. Total government spending is divided into social spending and non-social spending. The specification for the second model is as follows:

Y_t = [X_t, T_t, GEX_SOC_t, GEX_NSOC_t]

where GEX_SOC_t is social spending and GEX_NSOCt is non-social spending.

Model 3

The third model analyses the comparative effects of social spending shocks and economic spending shocks. In this model, total spending is broken down into economic spending and non-economic spending. Therefore, the specification is as follows:

Y_t = [X_t, T_t, GEX_ECO_t, GEX_NECO_t]

where GEX_ECO_t is economic spending and GEX_NECO_t is non-economic spending.

Model 4

The fourth model focuses on the dynamic effects of some of the more important components of social spending on output, such as education, health, housing, and social security expenditures. The specification of the fourth model is as follows:

Model 4a: Y_t = [X_t, T_t, GEX_SOS_t, GEX_NSOS_t]

Model 4b: Y_t = [X_t, T_t, GEX_EDU_t, GEX_NEDU_t]

Model 4c: Y_t = [X_t, T_t, GEX_HEA_t, GEX_NHEA_t]

Model 4d: Y_t = [X_t, T_t, GEX_HOU_t, GEX_NHOU_t]

where GEX_SOS_t is government spending on social security, GEX_NSOSt is government spending on non-social security, GEX_EDU_t is government spending on education, GEX_NEDU_t is government spending on non-education, GEX_HEAt is government spending on health, GEX_NHEA_t is government spending on non-health, GEX_HOUt is government spending on housing, GEX_NHOU_t is government spending on non-housing.

Model 5

The final model assesses the dynamic effects of social spending on the components of GDP, such as consumption, investment, exports, and imports. The VAR specification for

this model is as follows:

Model 5a: Y_t = [X_t, T_t, GEX_SOC_t, GEX_NSOC_t, CONP_t]

Model 5b: Y_t =[X_t, T_t, GEX_SOC_t, GEX_NSOC_t, INV_t]

Model 5c: Y_t =[X_t, T_t, GEX_SOC_t, GEX_NSOC_t, EXPORT_t]

Model 5d: Y_t =[X_t, T_t, GEX_SOC_t, GEX_NSOC_t, IMP_t]

where CONP is consumption, INV is investment, EXPORT is exports, and IMP is imports.

EMPIRICAL RESULTS

The results for the generalized impulse response analysis are presented in the appendix, organized under each model specified above.

For Model 1, a unit government spending shock increases GDP by about 46 billion baht in the second period, but the positive effect lasts only for one period, and then turns negative. The dynamic effect on total tax revenue has a pattern consistent with the path for GDP. The small positive effect lasts only two periods and then declines.

Model 2 looks at the dynamic effects of social government spending shocks on GDP. In this model, total government spending is broken down into social and non-social spending. The social spending shock on GDP is effective only in the second period and the impact is very small. A similar result was found for the non-social spending shock. Overall, the results are insignificant.

Model 3 captures the dynamic effects of economic spending shocks on GDP, the objective being to stimulate the economy. The positive effect of economic spending should be higher than the government's social spending effect. However, the results indicate a low response for GDP of only 53 billion baht in the second period, and 4.4 billion baht in the third. This is a smaller effect than for social spending shocks. Therefore, if the government wants to substitute social spending with economic spending, the result will not be so effective.

Model 4 investigates the dynamic effects of shocks to different subcomponents of total social spending, such as social security, education, health, and housing. For Model 4a, the dynamic effect of a shock to social security and welfare spending on GDP is quite strong, and it can stimulate GDP for the whole forecast period. Therefore, the government should target this component of spending to help stimulate GDP. The effect of spending on education on GDP (Model 4b) is insignificant, and stimulates GDP only at the beginning of the shock period although the budget allocation to education has the highest share in total social spending. The effects of a shock to health spending (Model 4c) on GDP are all negative for the whole period, so that current expenditure, such as the universal health care project, cannot be used to stimulate GDP. Rather, it has to be capital expenditure.

Finally, as far as Model 4d is concerned, the dynamic effects of a shock to housing expenditure on GDP is negative and insignificant, despite the fact that the government would like to use such expenditure to stimulate the economy.

Model 5 focuses on the dynamic effects of a social spending shock on the components of GDP. For consumption, there is a positive effect except in the second and third periods when the effects are negative. The greatest effect occurs after one year. This result is not strong enough to support the Ricardian equivalence outcome, which predicts negative consumption in the future. For investment, according to the classical theory, there should be a “crowding out” effect from increasing government spending. However, the results indicate only one negative effect in the third period while the other periods are positive. Therefore, social spending policy can be used to stimulate investment. For exports and imports, it can be concluded that after the first and second periods, social spending shocks stimulate both very strongly. This could lead to balance of trade problems.

CONCLUSION

Blanchard and Perotti (1999) concluded from their study that when government spending increases, output increases, and when taxes increase, output falls. The multipliers are small. As far as the composition of GDP is concerned, they concluded that the impact of spending shocks on private consumption is positive but that private investment is crowded out. Exports and imports also fall.

From the present empirical analysis it can be concluded that when government spending increases, output also increases, but with very insignificant effects. The multipliers are very small in every case, including social spending shocks, economic spending shocks, and shocks to spending on education, health, and housing. The exception is when social security and welfare spending is shocked.

As far as the components of GDP are concerned, the impact of social spending shocks is positive on consumption, investment, exports, and imports. Therefore, the policy implication from this chapter is that social spending policy is effective in stimulating consumption, investment, exports, and imports but the overall effect is diminished by the high import leakage. If the government wants to use social spending policy, the distinction between productive and unproductive spending must be clear and it should try to reduce the impact on imports in order to stimulate the economy.

References

Blanchard, O., and R. Perotti. 1999. An empirical characterization of the dynamic effects of changes in government spending and taxes on output. NBER Working Paper 7269.

Bureau of the Budget 2004. Thailand Budget in Brief.

Hemming, R., M. Kell, and S. Mahforz. 2002. The effectiveness of fiscal policy in stimulating economic activity: A review of the literature. IMF Working Paper WP/02/208.

Matsuoka, M. 1996. Measuring the effect of fiscal policy in Japan. Mimeo, Daiwa Institute. The National Economic and Social Development Board. 2003. The evaluation of the grass root and social security policy.