Determinants of Private Investment in Thailand

Transcription

1 UNIVERSITY OF WARWICK Determinants of Private Investment in Thailand ID: /2013 Department of Economics EC902: Econometrics A

2 Abstract This paper studies the determinants of private investment in Thailand to understand the recovery of private investment in Thailand after the financial crisis. The analysis begins with the time series properties of data are tested which following by two diagnosis tests. The estimation of the regression reveals that private investment index is affected negatively be real minimum lending rate, real exchange rate volatility, and the existence of financial crisis; while real GDP, capacity utilisation rate, and real effective exchange rate encourage private investment. 1. Introduction Thailand is one of the countries that were damaged largely by the East Asian financial crisis. Before the financial crisis on , private investment in Thailand was supported by financial liberalization on early 1990s. It can be seen clearly from fiure1 that during the Asian financial crisis in 1997 and 1998, private investment in Thailand decrease almost half from early However, there has be a recent revival in private investment since early 2002, and it takes Thailand over a decade to be able to achieve the same level of private investment since before the crisis. Moreover, it is important to study the determinants of private investment, since it plays a crucial role in generating growth of developing countries. Jongwanich and Kohpaiboon (2008) claimed that investment determines the capital accumulated rate which is a key condition for the expansion of production capacity and economic growth in long period, although investment expresses a smaller portion of aggregate demand than consumption usually does Private Investment Index PII Q1 1996Q1 1997Q1 1998Q1 1999Q1 2000Q1 2001Q1 2002Q1 2003Q1 2004Q1 2005Q1 2006Q1 2007Q1 2008Q1 2009Q1 2010Q1 2011Q1 2012Q1 Figure1: Private investment index 2. The Analytical Framework and the Model There are many studies on the determinants of private investment which can be described as private investments are determined by economic and institutional issues. In the case of a developing country, Thailand, these issues are described as follow

3 Domestic Demand stated by Cardoso (1993) and Oshikoya (1994), one of the key factors to determine private investment in developing countries is domestic demand or market potential. It is expected to have positive effect on private investment. As actual output rises, it implies the increase in demand which would attract firms to increase their production capacity to be able to capture the higher demand. Cardoso (1993), after applying a panel data analysis, suggests that the GDP growth would encourage private investment in Argentina, Brazil, Chile, Columbia, Mexico, and Venezuela during the period of the study which is Furthermore, the results from the study by Oshikoya (1994) which using the OLS estimation in Africa from 1970 to 1988 presents that the rise in real GDP growth rate provides a positive effect on private investment. Rate of Return on Investment Mallikamas, Thaicharoen, and Rodpengsangkaha (2003) comment that return rate on investment could be the main essential element on preferred level of investment. It is undeniable that a higher rate of return, which is corresponding to the higher marginal productivity of capital, is likely to attract firms to retrieve new capital at the margin because it would provide more profit if invest more. Their study also suggest that marginal return could be considered more applicable to decision on investment in theory, such as firms could reduce the cost and still receive practical average return, even firms are in the duration of excess capacity and inactive output growth. According to the previous study, the marginal return rate on investment cannot be observed directly. However, it should be correlated positively with the capacity utilisation in all likelihood, as the payoff to increase extra capacity should be small when capacity utilisation is small, and the results show that the increased growth rate of capacity utilization results in a rise in the growth rate of private investment. Cost of Capital One way to measure the cost of capital is the real Minimum Loan Rate (MLR). Mallikamas, Thaicharoen, and Rodpengsangkaha (2003) approximate real MLR by deducting headline inflation from nominal MLR. With the historically belief, as the supply of external financing, the corporate division has been greatly reliant on bank loans. Therefore, they believe that a properly precise measure of the cost of external financing for most Thai firms could be provided by the real MLR. Exchange Rate private investment also can be influenced by the real exchange rate to both positive and negative directions. Depreciation could lessen the aggregate demand because it reduces wealth and the real income of private sector. Therefore, firms might want to reconsider the potentials of future demand which might lead to the delay of their plan to invest. Moreover, the real cost of imported capital goods could increase because of the depreciation which, consequently, affect the private investment. Nevertheless, depreciation increases the price of tradable goods, which under an assumption that it will be equal across the world (Kipici and Kesriyeli, 1997), relative to the price of imported inputs. Thus, the investment in tradable goods should be able to make more profit, and if the positive effects has more influence over the negative impacts, then the total investment could rise (Agenor, 2001; Bleaney and Greenaway 2001; Jongwanich and Kohpaiboon 2008). A research by Jongwanich and Kohpaiboon (2008) claim that real exchange rates in the long run are statistically

4 significant, which also show that when real exchange rate depreciate for 1%, there is a 5% rise in private investment in the long run. Investor s Confidence Serven (2002) found that real exchange rate volatility obstructs the growth of private investment in small and open economies, as exchange rate volatility are related with inconsistent fluctuates in the relative investment profitability. Furthermore, Mallikamas, Thaicharoen, and Rodpengsangkaha (2003) suggest that in developing country like Thailand, it needs to import the content of investment. Thus, the real exchange rate volatility make the cost of new capital becomes uncertain. One reason is that investors are mostly risk-averse, so investors would require higher return premiums due to high exchange rate volatility which would hinder spending on new capital. Political Instability In the past decade, Thailand has faced many situations show that Thailand has a serious political instability; for example, the coup d'état in 2006, the blockade of Thailand s main airports in 2008, and the political unrest in 2010 that tens of thousands protesters paralyse parts of central Bangkok and burnt down parts of the city. Therefore, it is important to consider the influence of political instability. Svensson (1998) applies samples of 101 developing countries during the period of to study the effect of unstable government and polarized political structure which leads to a less of infrastructure investment, and it results in the lower private investment level. There is another research by Gyimah-Brempong and Traynor (1999) using not only a simultaneous equations model, but also an approach of dynamic panel data estimation to study the connection between political instability and economic growth in countries in Sub-Saharan Africa. This research suggests that the political instability has a negative effect on demand in domestic by indirectly reducing economic growth via declining the capital accumulation in the long-run. From the discussion and the findings of previous studies above, the empirical model is developed to estimate the determinant of private investment in Thailand is as followed: PII = f(rgdp, CAPU, RMLR, REER, CRISIS, POL) + random disturbance(ԑ) Where PII is the private investment index, RDGP is the real gross domestic product, REER is the real exchange rate, CAPU is the capacity utilisation rate, CRISIS is a time dummy variable capturing the period that there is a financial crisis in East Asia that affect Thai economy, and POL is the dummy variable presenting political instability in Thailand. In addition, to include the effect of real exchange rate volatility into the estimation, not only the real effective exchange rate variable, but also the dummy variables are excluded from the model to avoid multicollinearity problem. Therefore, another model is created. PII = f(rgdp, CAPU, RMLR, REER_VOL+POL) + random disturbance(ԑ) A quarterly data set, which covered the period of the first quarter of 1995 to the fourth quarter of 2012, used in this project is collected from the Bank of Thailand website. In order to achieve a detailed estimation, the quarterly data is used instead of annual data. However, there are some modifications on these variables, and the detailed of each variable is followed:

5 -Monthly Private Investment Index and Components (Seasonally Adjusted, 3-month moving average with a base year of 2000) are calculated to get quarterly private investment index as a proxy of private investment. -Real gross domestic product is calculated from the expenditure on gross domestic product at 1988 prices with a measure as millions of Baht by the office of the National Economic and Social Development Board. - Real effective exchange rate indices are collected and recalculated the base year to 2000 as a proxy of real exchange rate. Weighting currencies by trade weights is one common method to calculate the real effective exchange rate (Chinn, 2005). -Monthly percentage of industrial capacity utilisation rate exclude liquor is used, as there was an observation of an irregularity in liquor production in 2000 due to the concession of a major liquor producer is expected to reach the expiration Mallikamas, Thaicharoen, and Rodpengsangkaha (2003). Then the monthly data is recalculated to get the quarterly data. -Real minimum lending rate (MLR) is calculated by subtracting off the headline inflation from nominal minimise MLR of the commercial banks, and used as the cost of capital. -The most widely method used to measure the exchange rate volatility is the annual standard deviation of the nominal monthly exchange rate, which also be applied by the IMF Akhtar and Hilton (1984) suggest one of the main measures of exchange rate volatility is the standard deviation which can be calculated over a one-year period to acquire exchange rate volatility. Ex_Vol = [NER NER ] In this study, the standard deviation is calculated over a quarter or 3 months, so m is 3. REER is used instead of NER (nominal exchange rate), and REER quarterly average is used instead of NER (average of nominal exchange rate in year t) -The time dummy variable capturing the period that Thailand face with the financial crisis from 1997Q2 to 1999Q1 which is one and zero otherwise. -The dummy variable representing the period that Thailand has a political instability which is one from 2006Q1 to 2010Q2 and zero otherwise. 3. Econometric Analysis and Results Unit root test The dataset in this study is a time series data; therefore, before proceeding to the regression, the unit root test is used to test if the dataset is stationary or nonstationary (Sakr, 1993). Hill, Griffiths, and Lim (2011) suggest that it is important to do so because if nonstationary series are applied in regression analysis, results of the regression from unrelated data might still be significant.

6 This situation is called spurious regression. One of many methods to test the unit roots is Augmented Dickey-Fuller test which can ensure that the errors are not correlated. Δy = α + γy + a Δy + v γ expresses the hypotheses for stationary and nonstationary. When the series is nonstationary, γ would equal to zero which can also say that this series has a unit root. Corradi, (2013) states that M is a number of lagged differences to place in the ADF regressions which is chosen by the lowest number of selection criteria: AIC. In this dataset, the variables considered are all nonstationary except REER_VOL. Therefore, the ones that are nonstationary are converted to the first difference form. Consequently, the ADF tests indicate that variables are stationary when they are in difference form; therefore, these series are called integrated of order one (I(1)) (Hill, Griffiths, and Lim, 2011). Table1: Unit Root Tests Variable Test Specification ADF statistics PP statistics PII C,T First different of PII (D(PII)) C,T -5.66*** -5.55*** RGDP C,T * First different of RMLR (D(RGDP)) C,T -4.48*** *** CAPU C,T First different of RDGP (D(CAPU)) C,T *** *** RMLR C,T ** First different of REER (D(RMLR)) C,T -7.25*** -9.94*** REER C,T First different of CAPU (D(REER)) C,T -5.12*** -7.76*** REER_VOL C,T -5.77*** -5.85*** *,**,***denotes statistical significance at 10, 5 and 1 percent respectively Multicollinearity Woodridge (2006) stated that when there are high but not perfect correlation between at least two explanatory variables, it is called Multicollinearity. Greene (2008) suggested that in the model with multiple regressions, the variance of the kth least squares coefficient estimator is as shown below. Then the highly a variable correlated with other explanatory variables in the model, the higher the variance of the model will be, which might lead to the lower of t-statistic from the higher standard error (square root of variance). In the end, this could result in the rejection of null hypothesis, and that variables are left out of the model.

7 Var b X = σ (1 R ) (x x ) Greene (2008) also mentioned a diagnosis statistic for each coefficient in the regression that is called a variance inflation factor (VIF) which is equal to. The VIF for one variable presents the increase in variance of estimated coefficient which is able to be attributed that this particular variable is not orthogonal to others of this model. In addition, Simon (2004) does not only note that for every k predictors in the multiple regression model, there is a VIF, but also claims that there is a rule in which if the VIFs are more than 4, there need to be a further investigation. If the VIFs are more than 10, this is the signs of severe multicollinearity which need to be corrected. Table2: Variance Inflation Factor VIF Model 1 Model 2 Included observations Variables C NA NA DLOG(RGDP)(-1) D(CAPU) D(RMLR) D(REER)(-1) CRISIS LOG(REER_VOL)(-2) POL It can be seen that in this study, the VIF of each variables are less than 4, so it can be concluded that there is no multicollinearity problem. Autocorrelation According to Wooldridge(2006), when errors are correlated across time, it is called Autocorrelation. For example, when u >0, then the error in the next period of time or u could be positive too, on average. Therefore, corr(u, u )>0 which mean these errors are serial correlated. Corradi (2013) claimed that when the errors are correlated, this leads to a misspecified dynamic model. Hill, Griffiths, and Lim (2011) provide the consequences; if the autocorrelation problem is ignored, the standard error which normally computed from the least squares estimator will be no longer correct. This could lead to the misleading confidence intervals and hypothesis tests. One way to test for the autocorrelation is to use the Breusch-Godfrey test which can be used with AR(q) serial correlation. The null hypothesis is no autocorrelation, and it is tested by F-test.. If the data is quarterly and is not seasonally adjusted, then it is sensible to test for correlation between u and u. Then if the null hypothesis cannot be rejected, it can be concluded that it has no autocorrelation (Wooldridge, 2006)

8 Table3: Breusch-Godfrey Serial correlation LM test Model 1 Model 2 Lag F-statistics p-value Durbin- Watson F-statistics p-value Durbin- Watson *** *** ** *** *,**,***denotes statistical significance at 10, 5 and 1 percent respectively Table3 shows that in model 2, the null hypothesis is rejected, so it can be implied that the least squares standard error are not correct. Therefore, the standard error has to be corrected by computing the HAC (heteroskedasticity and autocorrelation consistent) standard errors or Newey- West standard errors (Hill, Griffiths, and Lim, 2011). Table4: HAC standard errors and covariance (Bartlett kernel, Newey-West fixed bandwidth=4) Variable Coefficient Least square standard errors HAC standard errors Std. Error t-value p-value Std. Error t-value p-value C DLOG(RGDP)(-1) D(CAPU) D(RMLR) CRISIS D(REER)(-1) POL The HAC standard errors provided by HAC (Newey-West) coefficient covariance matrix are different from the first set of standard errors obtained form least squares as shown in table4. Next, the new HAC standard errors will be used to proceed to the regression analysis. After considering the unit root tests and two diagnosis tests, the two models are: Δlog(PII) = c + β Δlog(RGDP) + β Δ(CAPU) + β Δ(RMLR) + β Δ(REER) + β CRISIS + β POL + ε Δ(PII) = c + α Δlog(RGDP) + α Δ(CAPU) + α Δ(RMLR) + α log(reer_vol) + α POL+ε To achieve the research objectives, Ordinary Lease Square (OLS) is used to estimate the relationship between dependent variable and explanatory variables on time series data.

9 Dependent variable Explanatory variables Table 5: Summary of variables used in the model and the expected signs of coefficients Variable Symbol Model1 Model2 Unit of Coefficient Coefficient measurement Expected signs of coefficients Private Investment Index PII Real GDP RGDP β α Million Baht positive Capacity Utilisation Rate CAPU β α percent positive Real Minimum Lending Rate RMLR β α percent negative Real Effective Exchange Rate REER β - - positive Dummy variable of financial crisis CRISIS β - - negative Real Effective Exchange Rate REER_VOL - α - negative Volatility Dummy variable of political POL β α - negative instability To test the hypotheses, t-statistic is used to analyze the significance of each explanatory variable. The hypothesis assumption can be derived below: Model1 H : β, β, β,, β = 0 H : β, β, β,, β 0 Model2 H : α, α, α,, α = 0 H : α, α, α,, α 0 If these coefficients are statistically significant the explanatory variables do affect private investment. To demonstrate, if the p-value is less than 0.01, 0.05 and 0.10, the explanatory variables significantly affect the dependent variable at 99%, 95% and 90% confidence level respectively. On the other hand, if p-value is more than 0.01, 0.05 and 0.10, the explanatory variables have no significant effect on the dependent variable at 99%, 95% and 90% confidence level respectively.

10 Dependent variable: D(PII) Sample: 1995:3 2012:4 Independent 1 variables Table6: Regression Results for Private Investment Determinant Equations Standard error t-statistic p-value 2 Standard error t-statistic p-value Constant * DLOG(RGDP)(-1) ** * D(CAPU) *** *** D(RMLR) ** *** D(REER)(-1) *** CRISIS *** LOG(REER_VOL)(-2) *** POL *,**,***denotes statistical significance at 10, 5 and 1 percent respectively 1 2 Adjusted R-squared F-statistic Prob(F-statistic) Durbin-Watson statistics Akaike info criterion The results which can be seen from table6 suggest that the estimated coefficients for all explanatory variables take their expected signs and are statistically significant except one dummy variable, POL, which is found to be not statistically significant. First, in model1 which is the main model, let other variables equal, higher real GDP and REER seem to lead to higher private investment index in the following quarter; a one percent increase in real GDP and REER encourage private investment index by 0.2 and respectively. In addition, a higher capacity utilisation rate would increase the private investment index by 0.005, all else equal. On the other hand, not only a percent higher in real MLR, but also the presences of financial crisis deter the rise in private investment index by Moving to the result from model2, the REER volatility is added to capture the additional effect through investor s confidence. The results are similar to the main model; the estimated coefficient takes its expected sign which is negative, and it is statistically significant. An increase in REER volatility would discourage the private investment index by Conclusion, Limitation and Recommendations for Further Research The project examines determinants of private investment in Thailand in order to understand issues that obstruct the recovery from the financial crisis. The result finds that real GDP, capacity utilisation, and real effective exchange rate have positive effect on private investment index, while real minimum lending rate and the presence of financial crisis effect the private investment

11 negatively. Therefore, policy makers should encourage beneficial investment environment such as, real effective exchange rate or real minimum lending rate. There are some constraints occurred on the ground of the restriction of the time used to conduct the research which can be described as; the hypotheses applied to investigate the determinant of private investment in Thailand are covered only some factors, not all relevant theories and hypothesis. Therefore, the future research could include other possible variables from previous studies such as leverage ratio or Tobin s q. Secondly, this examination considers only the determinants of private investment in Thailand. Moreover, the results from the analysis could not provide comparison among countries. From this reason, it would be better for the future research to investigate the determinant of private investment in Thailand in comparison with other countries in order to see the outcomes among different economic environments.

12 Appendix GDP at 1988 Price Q1 1996Q2 1997Q3 1998Q4 2000Q1 2001Q2 2002Q3 2003Q4 2005Q1 2006Q2 2007Q3 2008Q4 2010Q1 2011Q2 2012Q3 GDP at 88 P Q1 1996Q1 1997Q1 1998Q1 1999Q1 2000Q1 2001Q1 2002Q1 2003Q1 2004Q1 2005Q1 2006Q1 2007Q1 2008Q1 2009Q1 2010Q1 2011Q1 2012Q1 SET REER PII Capacity Utilisation 1995Q1 1996Q1 1997Q1 1998Q1 1999Q1 2000Q1 2001Q1 2002Q1 2003Q1 2004Q1 2005Q1 2006Q1 2007Q1 2008Q1 2009Q1 2010Q1 2011Q1 2012Q1 CAPU Real Minimum Lending Rate Q1 1996Q1 1997Q1 1998Q1 1999Q1 2000Q1 2001Q1 2002Q1 2003Q1 2004Q1 2005Q1 2006Q1 2007Q1 2008Q1 2009Q1 2010Q1 2011Q1 2012Q1 RMLR

13 EVIEWS results Model1 Dependent Variable: DLOG(PII) Method: Least Squares Date: 03/17/13 Time: 16:28 Sample (adjusted): 1995Q3 2012Q4 Included observations: 70 after adjustments HAC standard errors & covariance (Bartlett kernel, Newey-West fixed bandwidth = ) Variable Coefficient Std. Error t-statistic Prob. C DLOG(RGDPL) D(CAPU) D(RMLR) D(REERL) CRISIS POL R-squared Mean dependent var Adjusted R-squared S.D. dependent var S.E. of regression Akaike info criterion Sum squared resid Schwarz criterion Log likelihood Hannan-Quinn criter F-statistic Durbin-Watson stat Prob(F-statistic) Model2 Dependent Variable: DLOG(PII) Method: Least Squares Date: 03/17/13 Time: 15:26 Sample (adjusted): 1995Q3 2012Q4 Included observations: 70 after adjustments Variable Coefficient Std. Error t-statistic Prob. C DLOG(RGDPL) D(CAPU) D(RMLR) LOG(REER_VOLL2) POL R-squared Mean dependent var Adjusted R-squared S.D. dependent var S.E. of regression Akaike info criterion Sum squared resid Schwarz criterion Log likelihood Hannan-Quinn criter F-statistic Durbin-Watson stat Prob(F-statistic)

14 Correlation Matrices Model1 DLOG(RGDPL) D(CAPU) D(RMLR) D(REERL) CRISIS POL DLOG(RGDPL) D(CAPU) D(RMLR) D(REERL) CRISIS POL Model2 DLOG(RGDPL) D(CAPU) D(RMLR) LOG(REER_VOLL2) POL DLOG(RGDPL) D(CAPU) D(RMLR) LOG(REER_VOLL2) POL VIF Model1 Variance Inflation Factors Date: 03/16/13 Time: 18:45 Sample: 1995Q1 2012Q4 Included observations: 70 Coefficient Uncentered Centered Variable Variance VIF VIF C 2.75E NA DLOG(RGDPL) D(CAPU) 1.25E D(RMLR) 9.94E D(REERL) 9.62E CRISIS POL 9.31E Model2 Variance Inflation Factors Date: 03/16/13 Time: 18:46 Sample: 1995Q1 2012Q4 Included observations: 69 Coefficient Uncentered Centered Variable Variance VIF VIF C 9.03E NA DLOG(RGDPL) D(CAPU) 1.18E D(RMLR) 9.63E LOG(REER_VOLL2) 2.74E POL AR(1)

15 Bibliography Akhtar, M. & R.Spence Hilton (1984), Effects of Exchange Rate Uncertainty on German and U.S. Trade, Federal Reserve Bank of New York. Quarterly Review. Vol 9, Bleaney, M. & Greenaway, D., The Impact of Terms of Trade and Real Exchange Rate Volatility on Investment and Growth in Sub-Saharan Africa. Journal of Development Economics, 65 (2), pp Cardoso, E., Private Investment in Latin America. Economic Development and Cultural Change, [Online] 14 (4), pp Available at: [Accessed 12 March 2013]. Chinn, Menzie D., A Primer On Real Effective Exchange Rates: Determinants, Overvaluation, Trade Flows And Competitive Devaluation. Open Economies Review, 17 (1), pp Corradi, V., Econometrics A-EC902, EC902 Econometrics A. University of Warwick, unpublished. Greene, William H., Econometric Analysis. 6th ed. New Jersey: Pearson Education, Inc. Gyimah-Brempong, K. & Traynor, TL., Political Instability, Investment and Economic Growth in Sub-Saharan Africa. Journal of African Economies, 8 (1), pp Hill, R., Griffiths, E. and Lim, C., Principles of Econometrics. 4th ed. Hoboken: John Wiley & Sons, Inc. Jongwanich, J. & Kohpaiboon, K., Private Investment: Trends and Determinants in Thailand. World Development, 36 (10), pp Kipici, A.N. & Kesriyeli, M., The Real Exchange Rate Definitions and Calculations. Ankara: Central Bank of the Republic of Turkey. Mallikamas, R.P. Thaicharoen, Y. & Rodpengsangkaha, D., Investment Cycles, Economic Recovery and Monetary Policy. Bangkok: BOT Symposium. Oshikoya, W.T., Macroeconomic Determinants of Domestic Private Investment in Africa: An Empirical Analysis. Economic Development and Cultural Change, [Online] 42 (3), pp Available at: [Accessed 11 March 2013].

16 Sakr, K., Determinants of Private Investment in Pakistan. IMF Working paper, International Monetary Fund. Serven, L., Real Exchange Rate Uncertainty and Private Investment in Developing Countries. Review of Economics and Statistics, The World Bank. Simon, L.J., Detecting Multicollinearity Using Variance Inflation Factors. [Online] Available at: ml [Accessed 7 March 2013]. Svensson, J., Investment, Property Rights and Political Instability: Theory and Evidence. European Economic Review, [Online] 42 (7), pp Available at: [Accessed 10 March 2013] Wooldridge, Jeffrey M., Introductory Econometrics A Modern Approach. 3rd ed. Mason: Thomson Higher Education.