Appendix Private Investment Behavior in Thailand
- Erik Offerdal, Kalpana Kochhar, Louis Dicks-Mireaux, Jianping Zhou, Mauro Mecagni, and Balázs Horváth
- Published Date:
- December 1996
This appendix presents results of an empirical investigation of the determinants of private investment in Thailand. During the 1970s and the first half of the 1980s, the ratio of private investment to GDP (in constant prices) remained more or less unchanged at about 20 percent. Since the second half of the 1980s, however, there has been a surge in private investment, which reached the equivalent of about 33 percent of GDP by 1993 (Figure 14). The discussion attempts to examine empirically the factors underlying the behavior of private investment in Thailand during the period 1970–93.
Figure 14.Private and Public Investment Rates
Sources: Thai authorities; and IMF staff estimates.
Private investment in developing countries is typically specified as being a function of some or all of the following variables:32 (1) changes in demand or economic activity, consistent with the accelerator model of capital accumulation; (2) relative prices of capital or labor (or both) to measure profitability; (3) the availability of financing to capture the effects of credit rationing and financial repression; (4) public sector investment to examine whether there is complementarity between public and private investment or whether public sector investment crowds out the private sector by preempting scarce physical and financial resources; (5) macroeconomic instability, proxied by the variability in inflation or real effective exchange rates; and (6) general measures of uncertainty, sometimes proxied by the debt-service ratio to measure the impact of the “debt overhang” on private investment behavior. In addition, in the case of Thailand, where a large part of the surge in private investment since the mid-1980s was due to the large inflows of foreign direct investment and where foreign borrowing has been an important source of financing for domestic investment, unit labor costs in Thailand relative to its trading partners and foreign interest rates are included as explanatory variables.
The estimated equation thus takes the following general form:
LIPt = α0 + α1ACTYt + α2DINTt + α3CREDt + α4FINTt + α5LRULCt + α6LIPUBt + α7UNCERTt + α8LTOTt + α9FXt + α10LIPt-l + €t, (1)
where LIP is the logarithm of the ratio of private investment to GDP in constant prices;33ACTY refers to the activity variable; DINT stands for the domestic real interest rate to proxy for the cost of capital; and CRED measures the availability of domestic financing; FINT refers to the foreign real interest rates, adjusted for expected exchange rate changes; LRULC stands for the logarithm of the index of unit labor costs in Thailand relative to its trading partners; and LIPUB is the logarithm of the ratio of public sector investment to GDP in constant prices; FX stands for international reserves and serves as a proxy for the availability of foreign exchange; UNCERT refers to the various possible proxies of macroeconomic instability or uncertainty, such as the variance in inflation rates and debt-service ratios; and LTOT stands for the logarithm of the external terms of trade.
Testing for Stationarity
Before turning to the specific form of the investment function for Thailand, the key time-series properties of the variables to be included need to be examined. In particular, it is necessary to test whether the data are generated by stationary time-series processes.34 Using nonstationary time series, except under special circumstances, could lead to spurious results in the sense that regressing strongly trended variables on each other could overstate the “true” explanatory power of the regressors. Statistical tests for stationarity essentially involve testing whether |α| < 1 in the equation yt = αyt-1 + Vt, or testing for a unit root in the process for yt.
Indeed, the rationale for using the ratio of private investment to GDP as the dependent variable is to ensure stationarity. In the case of Thailand, however, as can be seen from Figure 14, there appears to have been a onetime “structural break” in the behavior of private investment around the mid-1980s. At first glance, there appears to be an indication that even the ratio is strongly trended, particularly in the later part of the sample period, suggesting the presence of a unit root.35 The index of relative unit labor costs displays similar nonstationary behavior (see Figure 5 of the text) in the sample. All other variables considered as part of the general set of regressors were found to be stationary.
For the two variables for which casual observation suggests the presence of unit roots, “correlograms” were plotted (Figure 15) showing correlations between the series and its own lagged values. If a time series is stationary, the autocorrelations would be expected to decline quickly to values that are not significantly different from zero.Figure 15 shows that both the investment ratio and relative unit labor costs are strongly autocorrelated, that the autocorrelations die out within five to seven lags, and that the pattern of autocorrelations is consistent with the data being generated by the autoregressive process of order 1 (or an AR(1) process).
Figure 15.Private Investment Ratio and Relative Unit Labor Costs: Sample Autocorrelations
Source: IMF staff estimates.
Next, tests for unit roots of the type suggested by Perron (1989) were performed. Perron (1989) showed that standard tests may not reject the hypothesis of a unit root when the “true” data-generating mechanisms are that of stationary fluctuations around a trend that contains a onetime break. He derived a test that can distinguish between a process with a unit root and one that is a trend-stationary process but with a “permanent” break in the structure of the time series. These tests suggest that the hypothesis of unit roots for the two series in question can be rejected at the 1 percent significance level.36
Specification of Thailand’s Investment Function
The choice of the activity or variable is influenced by the need to reduce the possibility of simultaneity biases that can arise because investment affects growth in the current period. In the case of Thailand, three economic activity variables were experimented with—lagged real GDP growth (GRTH), export volume growth (EXGRTH), and real GDP growth in partner countries (FGRTH).
INT was the domestic lending rate adjusted by actual inflation as measured by the consumer price index.
CRED was measured as the real growth of private sector credit. A priori, it is not clear whether a variable measuring the availability of financing should be included in the investment equation. On the one hand, although Thailand has had a financial system that is closer to the free market end of the developing country spectrum, government intervention remained an important feature of financial markets in Thailand until the early 1990s, when interest rates began to be deregulated. Although the lending ceilings may not have been binding for prime borrowers, there is evidence that smaller enterprises may have, at various times, had to queue for rationed credit. On the other hand, the scope for intervention has effectively been constrained by the relatively open capital account, especially for inflows. Thus, interest rate ceilings could not be set so low as to persistently push domestic interest rates below the corresponding international rates. The inclusion of this variable in the investment function is, therefore, essentially an empirical issue and will be discussed further below.
The foreign interest rate (FINT) was measured as the U.S. prime lending rate adjusted by consumer price inflation in the United States and for the actual depreciation of the baht vis-à-vis the U.S. dollar.37
The relationship between private and public investment in Thailand was examined by including the ratio of public investment to GDP (LIPUB) and the ratio of public infrastructure expenditure to GDP (LIPUBI) in the estimated equations.38
FX was measured as total international reserves (minus gold), expressed in months of imports.
The terms of trade (LTOT) were measured as the logarithm of the ratio of the price of exports to that of imports.
The two measures of instability or uncertainty that were used are VINFL, the variance of month-to-month consumer price inflation rates, and LDSER, the logarithm of the debt-service ratio.
A dummy variable was included to capture the impact of the shift in emphasis in the Board of Investment in the post-1986 period toward a proactive export- and private-sector-oriented strategy.
The equation was estimated with annual data from 1970 to 1993 using ordinary least squares; the standard errors shown are heteroscedastic-consistent estimates.39 The approach was to start by using the complete set of explanatory variables in the general equation (1) and to eliminate variables that were statistically insignificant. During this process, attention was also paid to the regression diagnostics, especially to the evidence on serial correlation and homoscedasticity of the residuals, since nonspherical errors in an equation that includes a lagged dependent variable give rise to inconsistent estimators.
The results of the different specifications of the equation are presented in Table 13. Key results of the specification search are the following. The coefficient on lagged real GDP growth, GRTH, was insignificant (Specification 1), as was that on the volume of export growth (not reported). The activity variable that was most accurately estimated and performed the best was FGRTH, the rate of real GDP growth in partner countries, and so it was used in subsequent specifications.40 Thailand has always been a relatively open economy. This fact, coupled with the sharp acceleration in the growth in manufactured exports since 1986, is probably the reason for this finding.
|Durbin’s. H for serial correlation||0.54||–0.71||–0.36||–0.37||0.03||0.01||0.13||0.02|
|Augmented Dickey-fuller test statistic||–5.39||–5.17||–4.77||–4.73||–4.29||–4.29||–4.16||–4.31|
|F-statistic (zero slopes)||16.22||23.09||17.88||21.65||24.49||30.05||35.04||50.41|
|Jarque-Bera normality test statistic||0.97||1.64||0.27||0.17||0.35||0.34||1.00||0.26|
|Breusch-Pagan heteroscedasticity test statistic||10.29||13.51||11.08||8.40||5.34||–5.57||3.97||2.09|
As mentioned above, the inclusion of the CRED variable is primarily an empirical issue, since from a theoretical standpoint no clear-cut justification can be made in the case of Thailand for either including or excluding this variable. A key problem with including the volume of credit as an explanatory variable in an investment equation is that, if credit is in fact not constrained, then credit growth is determined by investment demand. In a statistical sense, there is the possibility that “causality” runs from the dependent variable to the independent variable, rendering the coefficient on the explanatory variable difficult to interpret.
The direction of causality was tested using “Granger-causality” tests. Granger (1969) introduced a concept of causality in which, broadly speaking, a variable y is said to be “Granger caused” by another variable x if current values of y can be predicted with better accuracy by using past values of x.41 Testing for Granger causality essentially involves setting up a vector autoregression, in which all the variables of the system are expressed as linear functions of their own and each other’s lagged values.42F-tests are then computed to test whether lagged values of any of the other variables enter a given equation significantly. Here, the vector autoregression was set up using the variables that enter the “preferred” specification.
Such tests reveal that, in the case of Thailand, the direction of causality seems to be from the growth of credit to investment rather than the other way around, implying that the coefficient on CRED in the estimated equation can be interpreted in the conventional way. However, the serially correlated and heteroscedastic residuals in Specifications 1 and 2 shown in Table 13 are evidence of equation misspecification. After some experimentation, it was found that residual serial correlation was no longer significant when the CRED variable was omitted from the regression. The exclusion of this variable had little significant impact on the coefficients on real growth in partner countries, relative unit labor costs, and real interest rates. However, the coefficient on the lagged dependent variable increased slightly while that on the public investment rate increased markedly, as did its statistical significance.
This latter result is indicative of collinearity between the CRED variable and LIPUB and is consistent with the finding (discussed below) of a negative coefficient on LIPUB.
Specification 3 shows that, besides the foreign growth variable, the two variables that are significant are LRULC and the ratio of public investment to GDP (LIPUB).43 The coefficient on the relative unit labor cost variable implies that when Thailand’s unit labor costs rise relative to its trading partners, private investment declines. The growing share of foreign-controlled investment in private investment and the increasing integration of the Thai economy into the world economy account for this relationship.
The negative coefficient on LIPUB suggests substitutability between public and private investment in Thailand, at least in the short term. This result may appear surprising, given that the provision of infrastructure has been a central theme of Thailand’s development plans before the 1980s, and suggests that the coefficient needs to be interpreted carefully. In particular, it is useful to distinguish between investment spending and the stock of capital. The relatively well-developed stock of infrastructure was no doubt instrumental in supporting the boom in private investment in the late 1980s. It is also true, however, that during the second half of the 1970s and the early 1980s there was a rapid and somewhat uncoordinated expansion in public investment. Although much of this investment was concentrated in infrastructure-related sectors, any positive impact on private investment is likely to have materialized only with relatively long lags. In any event, such a relationship would not necessarily be captured by the contemporaneous correlation between public and private investment.
Figure 14 plots the public and private investment rates over the sample period. It shows that the negative relationship between public and private investment is, for the most part, explained by periods in which declines in public investment were accompanied by increases in private investment. This is particularly evident in the period since 1985, when the government began to significantly reduce and rationalize public investment spending; the resulting fiscal consolidation served to accommodate the acceleration in private investment in an environment of overall macroeconomic stability. The coefficients on these two variables were relatively robust to the addition and deletion of other variables in subsequent specifications.
The real interest rate (RDINT) entered the equation with a negative sign but was less robust than LRULC and LIPUB across specifications. Indeed, the decline in private investment in the mid-1980s is attributable to the sharp increase in interest rates when a conservative monetary policy stance was adopted (see Section V of the text).
The measure of instability (VINFL) was statistically insignificant (Specification 3). This is attributable to the fact that Thailand has had low and stable inflation for virtually all of the sample, with only very short-lived accelerations following the two oil shocks. The availability of foreign exchange (FX) (Specifications 3 and 4) was also statistically insignificant. Despite the secular decline that Thailand experienced in its external terms of trade during the period under study, the coefficient on LTOT was found to be statistically insignificant (Specifications 3, 4, and 5). The debt-service ratio was similarly found to be statistically insignificant (Specifications 3, 4, 5, and 6). Finally, the foreign interest rate variable, which was consistently insignificant, was omitted.44
Specification 8 in Table 13 was chosen as the preferred equation. All the variables in that specification have the expected sign and are statistically significant. The diagnostics on the regression suggest that the hypotheses of serially uncorrelated and homoscedastic errors cannot be rejected. The coefficient on the activity variable suggests that a 1 percentage point increase in trading partners’ real GDP growth would result in an increase in the ratio of private investment to GDP of 0.7 percent. The coefficient on real domestic interest rates suggests that a 1 percentage point increase in real interest rates would lower the ratio of private investment to GDP by about 0.1 percent, while an increase in public investment by the equivalent of 1 percentage point of GDP would decrease the private investment rate by 0.5 percent. The coefficient on relative unit labor costs suggests that a 1 percentage point decrease in Thailand’s relative unit labor costs would result in an almost equivalent increase in the ratio of the private investment rate to GDP.Figure 16 shows that the ratio of private investment implied by the estimated equation tracks the actual ratio quite closely and accurately predicts the important turning point in 1986.
Figure 16.Actual and Fitted Private Investment
Source: IMF staff estimates.
Once the “preferred” specification was arrived at, a recursive estimation procedure was implemented to test the robustness of the parameter estimates over time. In view of the possibility that the structure of the economy has undergone significant changes over the past two and a half decades, the assumption of time-invariant parameters may not be supported by the data.Figure 17 shows the behavior of the estimated coefficients when the sample begins in 1972 and ends in the year shown on the horizontal axis (forward recursive estimates). The charts show that the estimated coefficients are relatively stable over time. The exception is the coefficient on the lagged dependent variable, which goes from being close to zero in the first part of the sample to being significantly different from zero in the later part of the sample, implying that the degree of autocorrelation in the private investment ratio has increased over time. This finding is not inconsistent with the apparent “regime” change characterized by the increase in the private investment rate to a permanently higher level, and with the fact that the later part of the sample covers the period of transition from one “regime” to the other.
Figure 17.Recursive Estimates (Forward Recursions)
Source: IMF staff estimates.
Relative Contributions of Explanatory Variables
Table 14 shows the contributions of the explanatory variables to the estimated change in private investment during the postadjustment period (1987–93).45 The decline in real GDP growth in partner countries and the increase in public investment in the early 1990s were negative influences on private investment in the postadjustment phase, while the fall in real lending rates throughout this period had a positive impact on private investment. The decline in relative unit labor costs made a significant positive contribution, accounting for about a fourth of the acceleration in private investment. The large share of the lagged dependent variable suggests an increase in “inertia” in the evolution of private investment—a development that appears, prima facie, to be inconsistent with an increasingly dynamic economy. A possible explanation could be that there was a change in the composition of investment—toward high-technology industries—as a result of the production relocation that took place from other East Asian countries, and that the technology gap resulted in higher costs, at least initially, of adjusting the capital stock.
|Explanatory Variables||Estimated Coefficient||Estimated Change in I/GDP due to:1||1987–93 Share of Estimated Change due to:2||Long-run Elasticities with respect to:|
|Real GDP growth in trading partners||0.026||–1.3||–12.0||0.08|
|Real domestic interest rates||–0.005||0.7||6.5||–0.02|
|Relative unit labor costs||–0.220||3.0||26.8||–0.68|
|Ratio of public investment to GDP||–0.182||–1.0||–8.5||–0.56|
|(in constant prices)|
|Lagged dependent variable||0.676||9.8||87.1|
|Actual change in dependent variable||11.4|
|Share of actual change explained by regression||98.5|
|Share of actual change explained by residual||1.5|
The movement in relative unit labor costs in Thailand appears to be an important variable in explaining the evolution of private investment, especially in the post-1986 period. The decline in relative unit labor costs in Thailand (Figure 18) during this period was driven by the currency realignments in the area and by rising labor costs in other countries. With the yen appreciation between 1985 and 1988, production was increasingly shifted to Thailand (and to Malaysia and the other NIEs). Subsequent appreciations of the currencies of some of the NIEs, rising wage costs in Japan, Korea, and Taiwan Province of China, and the depreciation of the baht along with the U.S. dollar were also instrumental in shifting production to Thailand. To examine the importance of this variable in explaining the behavior of private investment since the second half of the 1980s, a simulation exercise was performed in which relative unit labor costs were assumed to be unchanged at the average level in 1984–85.46Figure 19 shows the actual and simulated behavior of private investment. Clearly, the evolution of relative labor costs in Thailand has been a very important factor in private investment behavior, particularly since 1989. Had labor costs in Thailand relative to its trading partners remained at their level in 1984–85, private investment would, other things being equal, have been lower by the equivalent of over 4 percentage points of GDP.
Figure 18.Unit Labor Costs in Thailand and Selected Countries
Sources: International Monetary Fund, World Economic Outlook (various issues); and IMF staff estimates.
Figure 19.Actual and Simulated Private Investment
Source: IMF staff estimates.
The key messages that emerge from this analysis are the following. First, while some of the surge in private investment in the second half of the 1980s could be ascribed to fortuitous exogenous developments, policy actions taken by the Thai authorities—including the devaluation of the baht and the elimination of export taxes and granting of export incentives—contributed significantly. Second, the decline in interest rates following the period of relatively tight monetary conditions in the immediate aftermath of the devaluation was also important. Prudent financial management and a largely internally consistent policy mix were also instrumental in translating the nominal devaluation into a sustained real depreciation and in enhancing policy credibility. Third, flexible labor markets (in terms of real wage flexibility and labor mobility) were important factors underlying the developments in relative unit labor costs and, in turn, the increase in investment. Fourth, the surge in private investment could be accommodated without a significant acceleration in inflation through most of the period because of the simultaneous fiscal consolidation, in which public expenditures (especially on investment) were rationalized.
Finally, although not all of these effects are explicitly accounted for in the estimated investment equation because of the lack of measurable proxies, a hallmark of macroeconomic developments in Thailand has been the high priority accorded to overall macroeconomic stability. This, combined with the pro-business orientation of economic policymakers (despite several changes in government), resulted in relatively low policy uncertainty and has proven to be conducive to private investment, both domestic and foreign.47
- Search Google Scholar
- Export Citation
- Search Google Scholar
- Export Citation
- Search Google Scholar
- Export Citation
Anand, Ritu, and Swedervan Wijnbergen, 1989, “Inflation and the Financing of Government Expenditure: An Introductory Analysis with an Application to Turkey,” World Bank Economic Review, Vol. 3 (January), pp. 17–38.
Barro, Robert J., 1989, “Economic Growth in a Cross Section of Countries,” NBER Working Paper 3120 (Cambridge, Mass.: National Bureau of Economic Research, September).
Barro, Robert J., and Jong-WhaLee, 1993, “International Comparisons of Educational Attainment,” Journal of Monetary Economics, Vol. 32 (December), pp. 363–94.
Bercuson, Kenneth, and Linda M.Koenig, 1993, “The Recent Surge in Capital Inflows to Three Asian Countries: Causes and Macroeconomic Impact,” Occasional Paper 15 (Kuala Lumpur: South East Asian Central Banks, Research and Training Centre).
Bertrand, Trent, and LynSquire, 1980, “The Relevance of the Dual Economy Model: A Case Study of Thailand,” Oxford Economic Papers, Vol. 32 (November), pp. 480–511.
Bhattacharya, Amarendra, and Johannes F.Linn, 1988 “Trade and Industrial Policies in the Developing Countries of East Asia,” World Bank Discussion Paper 27 (Washington: World Bank).
Bruno, Michael, 1985, “Aggregate Supply and Demand Factors in OECD Unemployment: An Update,” NBER Working Paper 1696 (Cambridge, Mass.: National Bureau of Economic Research).
Bruno, Michael, and WilliamEasterly, 1995, “Inflation Crises and Long-Run Growth,” NBER Working Paper 5209 (Cambridge, Mass.: National Bureau of Economic Research).
Bruno, Michael, and JeffreySachs, 1985, The Economics of Worldwide Stagflation (Cambridge, Mass.: Harvard University Press).
Dean, Judith M., SeemaDesai, and JamesRiedel, 1994, “Trade Policy Reform in Developing Countries since 1985,” World Bank Discussion Paper 267 (Washington: World Bank).
Elias, Victor, 1991, “The Role of Total Factor Productivity on Economic Growth,” background paper for the World Development Report (unpublished; Washington: World Bank).
Fischer, Stanley, 1993, “The Role of Macroeconomic Factors in Growth,” Journal of Monetary Economics, Vol. 32 (December), pp. 485–512.
Goldsbrough, David, SharminiCoorey, LouisDicks-Mireaux, BalazsHorvath, KalpanaKochhar, MauroMecagni, ErikOfferdal, and JianpingZhou, 1996, Reinvigorating Growth in Developing Countries: Lessons from Adjustment Policies in Eight Economies, Occasional Paper 139 (Washington: International Monetary Fund, July).
Granger, C.W.J., 1969 “Investigating Causal Relations by Econometric Models and Cross-Spectral Methods,” Econometrica, Vol. 37 (July), pp. 424–38.
Heller, Peter S., Richard D.Haas, and Ahsan S.Mansur, 1986, A Review of the Fiscal Impulse Measure, Occasional Paper 44 (Washington: International Monetary Fund, May).
International Monetary Fund, various years, Government Finance Statistics Yearbook (Washington).
International Monetary Fund, various years, International Financial Statistics (Washington).
International Monetary Fund, various years, World Economic Outlook (Washington).
King, Robert G., and RossLevine, 1994, “Capital Fundamentalism, Economic Development and Economic Growth,” Carnegie-Rochester Series on Public Policy, Vol. 40 (June), pp. 259–300.
Knight, Malcolm, NormanLoayza, and DelanoVillanueva, 1992, “Testing the Neoclassical Theory of Economic Growth: A Panel Data Approach,” IMF Working Paper 92/106 (Washington: International Monetary Fund).
Mankiw, N. Gregory, DavidRomer, and David N.Weil, 1992, “A Contribution to the Empirics of Economic Growth,” Quarterly Journal of Economics, Vol. 107 (May), pp. 407–38.
McCarthy, F. Desmond, J. PeterNeary, and GiovanniZanalda, 1994, “Measuring the Effect of External Shocks and the Policy Response to Them: Empirical Methodology Applied to the Philippines,” Policy Research Working Paper 1271 (Washington: World Bank, March).
Nehru, Vikram, and AshokDhareshwar, 1993, “A New Database on Physical Capital Stock: Sources, Methodology, and Results,” Revista de Análisis Económico, Vol. 8 (No. 1), pp. 37–59.
Perron, Philippe, 1989, “The Great Crash, the Oil Price Shock, and the Unit Root Hypothesis,” Econometrica, Vol. 57 (November), pp. 1361–1401.
Robinson, David, YanghoByeon, and RanjitTeja, WandaTseng, 1991, Thailand: Adjusting to Success—Current Policy Issues, Occasional Paper 85 (Washington: International Monetary Fund, August).
Sarel, Michael, 1995, “Demographic Dynamics and the Empirics of Economic Growth,” Staff Papers,International Monetary Fund, Vol. 42 (June) pp. 398–410.
Schadler, Susan, MariaCarkovic, AdamBennett, and RobertKahn, 1993, Recent Experiences with Surges in Capital Inflows, Occasional Paper 108 (Washington: International Monetary Fund, December).
Serven, Luis, and AndrésSolimano, 1994, Striving for Growth After Adjustment: The Role of Capital Formation (Washington: World Bank).
Summers, Robert, and AlanHeston, 1991, “The Penn World Table (Mark 5): An Expanded Set of International Comparisons, 1950–88,” Quarterly Journal of Economics, Vol. 106 (May), No. 2, pp. 327–68.An updated version of the Penn World Tables, Mark 5.6 (available from NBER Publications, 1050 Massachusetts Avenue, Cambridge, Mass. 02138) was used for the regressions in this Occasional Paper.
Tinakorn, Pranee, and C.Sussangkarn, 1994, “Productivity Growth in Thailand” (unpublished; Bangkok: Thailand Development Research Institute).
World Bank, 1991, World Development Report (New York: Oxford University Press).
World Bank, various years, World Tables (Baltimore, Md.: Johns Hopkins University Press).
Throughout this paper, the adjustment period in Thailand is treated as the period between 1981 and 1986, which also coincides with the period during which Thailand’s adjustment efforts were supported by the IMF and the World Bank. Thailand had three IMF stand-by arrangements—in June 1981, November 1982, and June 1985—together with drawings under the IMF’s compensatory and contingency financing facility (CCFF) and buffer stock financing facility, and two structural adjustment loans from the World Bank (in May 1982 and March 1983). Owing to a marked improvement in Thailand’s economic circumstances, particularly the balance of payments position, the Thai authorities decided to cancel the last IMF stand-by arrangement as of January 1. 1987.
This specification is generally associated with theories of endogenous growth, but it is also consistent with the transitional dynamics of a neoclassical growth model that incorporates human capital.
Assuming a constant returns-to-scale production function, TFP growdi is calculated by subtracting from output growth the contributions made by the accumulation of capital and growth in the labor force, with the further assumption that the factor shares in the production function are roughly equal to the factor shares in income (that is, 0.4 and 0.6 for capital and labor, respectively). A set of alternative estimates was also prepared using a production function that includes human capital, following the approach of Mankiw, Romer, and Weil (1992). Several other studies have estimated the coefficients of the production function using crosscountry data (for example. Knight, Loayza, and Villanueva, 1992; World Bank. 1991; and Elias,1991). Estimates of TFP using these various approaches are typically highly correlated.
In practice, it is difficult to construct measures of variables that are directly controlled by policies. Researchers therefore proxy policies or distortions by using variables that can be the outcome of policies, but that can also be affected by other (non-policy-related) factors. Thus, the policy implications of the observed linkages between these proxy variables and growth should be drawn with caution. Also, the estimated equations are in no sense “structural equations”; the estimated coefficients can be interpreted only as measures of partial correlations, after controlling for the effects of some other measures of macroeconomic policies and structural conditions.
By contrast, secondary school enrollment rates in Thailand are low and have not increased much over time. As the economy moves toward industries requiring more skilled labor, this may prove to be an impediment to sustained growth. Recognizing this, the authorities have begun to emphasize secondary and higher education in the allocation of public education expenditures.
The factors accounting for the terms of trade deterioration, most of which occurred between 1974 and 1982, are discussed in Section IV.
The first measure of TFP growth—the Solow residual—is Calculated by assuming a constant returns-to-scale production function and a factor share of 0.4 for capital. The second-—the Mankiw-Romer-Weil residual—includes human capital accumulation and an alternative measure of labor force growth—age-adjusted population growth (Sarel. 1995)—and assumes factor shares of 0.33 for each of the three factors of production. Because the labor input measure does not take account of cyclical variations in hours worked, the estimates of TFP incorporate a substantial cyclical component. To corree: for this, the contributions of TFP growth shown in the chart are calculated as five-year moving averages. Alternative estimates of Thailand’s TFP growth, which make adjustments for improvements in the quality of labor and capital inputs, suggest that TFP growth during this period was somewhat lower than shown here (Tinakorn and Sussangkam, 1994).
The terms of trade effect is derived as the net impact on the balance of payments from changes in expont or import prices (or both), assuming that trade volumes are unchanged.
The estimated impact of global demand shocks is measured by the deviation in world export volume growth from its trend value, and the interest rate effect is measured as [lie impact of changes in world interest rates on interest payments on the previous year’s stock of debt at variable interest rates. The measures of the size of exogenous shocks are based on relatively sirong assumptions of other things being equal and should therefore be seen as yielding broad orders of magnitude rather than precise calculations of the impact of shocks.
The estimated change in the real output gap (measured by the difference between actual growth and an estimated trend growth rate) is multiplied by the initial share of imports in GDP in order to calculate a measure of the effect of economic “compression” on the balance of payments. Changes in export performance and import intensity are measured in terms of deviations of actual exports (imports) from simple trade functions, assuming constant income elasticities with respect to partner-country demand (domestic demand). Thus, the exercise involves comparing actual outcomes with a simple counierfaclual derived by an extrapolation of past economic relationships (estimated over the preceding five-year period).
Trend output was calculated using the Hodrick-Prescolt filter, a univariate trend-extraction algorithm that is based on actual real GDP. The filter calculates trend GDP by minimizing the variation of actual GDP around a trend, subject to certain assumptions about the variance of the cyclical component relative to that of the trend component.
Unit labor costs in the manufacturing sector in Thailand were compared with a trade-weighled average of unit labor costs in the United States. United Kingdom, Germany. France, the Netherlands. Hong Kong, Singapore, Korea. Taiwan Province of China, and Malaysia.
Total long-and short-term inflows averaged about 10 percent of GDP annually between [9H7-88and 1993.
Another factor was rising protection is I pressures between the United States and Japan. By the mid-1980s. such pressures, in conjunction with the appreciation of the yen, led to production shifting from Japan to Thailand, a country with which Japan has had long-standing close cultural and business lies.
The fiscal impulse measure used here is similar to that used in the IMF’s World Economic Outlook (see Heller, Haas, and Mansur, 1986). It is calculated by separating the actual budget balance into two components: a cyclically neutral component and a fiscal stance component. The cyclically neutral component is defined by assuming that government expenditures increase proportionately with trend nominal output and that government revenues increase proportionately with actual nominal output. The fiscal stance component—the difference between the cyclically neutral and the actual budget balance—then captures the full effect of automatic stabilizers and discretionary changes in fiscal policy. The fiscal impulse is the annual change in the fiscal stance measure, expressed as a share of GDP. A negative number would indicate a contractionary demand impulse emanating from fiscal policy, and a positive number would indicate an expansionary demand impulse.
Using the approach set out in Anand and van Wijnbergen (1989), the budget constraint can be described as δb=(r-n)b+d-s. where b represents the public debt-to-GDP ratio, d is the primary deficit as a share of GDP. s represents seigniorage revenues (including the inflation tax) as a share of GDP. r stands for the average real interest rate on debt, and n is the real GDP growth rate. Since the aim of the exercise is to assess whether fiscal polies was judged to be sustainable at the time of adjustment, rather than with the benefit of hindsight, the debt calculations are based on a three-year moving average of actual interest rates and growth rates. The calculations also assume, for each year’s estimate, that the real exchange rate is expected to be constant.
For the purposes of illustration, the sustainable fiscal balance was recalculated assuming that all external financing is on commercial terms. Given the relatively small share of external debt on concessional terms in Thailand, ihe impact of this assumption is marginal.
See Robinson. Byeon, and Teja (1991). The Thai capital account has been quite open, especially for inflows, with a few important limitations. First, for prudential reasons, limits are imposed on the open foreign exchange position of commercial banks. Second, banks are not allowed to make foreign-currency-denominated loans, and. until the early 1990s, residents were not permitted to hold foreign currency deposits at domestic banks. Third, until the elimination in 1990 and 1992 of ceilings on interest rates, bold lending and deposil rates were subject to regulatory ceilings. Various controls on capital outflows limited the extent of arbitrage in response to interest rate differentials. Fourth, foreign borrowing is subject to a withholding at rates that are varied from lime to time to regulate capital inflows.
Although the evidence is fragmentary and somewhat outdated. Bertrand and Squire’s analysis suggests that there is little open unemployment in rural Thailand: that wage differentials between agriculture and manufacturing do not appear to be out of line with productivity differentials: that minimum wages in the formal sector and wages of unskilled labor in certain manufacturing activities are not significantly different from each other and that they tend to move together over time; and that participation rates are high, and unemployment rates tow. among migrant labor moving into the urban labor market.
Despite the limited effective coverage of minimum wages, revisions to die minimum wage do tend to raise the entire salary structure of the formal sector. However, as will be seen below, wage increases have thus far not generally been out of line with productivity developments, particularly in the manufacturing vector. The potential distortions that can arise from such regulations do not appear to have been relevant in Thailand during the period under study.
Open unemployment is defined to include those who were looking for work and those who were available and willing to work.
Bruno (1985) and Bruno and Sachs (1985) calculated wage gap measures by assuming that the underlying production technology is Cobb-Douglas in nature, with unitary elasticity of substitution between capital and labor. Under these assumptions, the issue of estimating the warranted wage is one of calculating average productivity at full employment. In contrast, the measure used here does not necessarily imply that the absence of a wage gap is appropriate from the point of view of full employment, only unchanged employment.
The change in warranted wages is given by δ (WIPc) = δ (Productivity) + δ (Pm/Pc Where W is the nominal wage. Pc is the consumer price index, and Pm is the output deflator.
It is given by
δ (WIPc) ≤ δ (ULC($)* + δ (Productivity) + δ (Nominal exchange rate) -δ(Pc)
Wage gap measures can be strongly influenced by the choice of the base year; however, experiments with alternative base years were chosen in the posladjuslmenl period and yielded similar measures of the wage gap.
Lessons from various studies of adjustment programs are summarized in Appendix 1 of Goldsbrough and others (1996).
The spread between deposit and lending rates in Thailand, before 1990, averaged about 6.5 percentage points, compared with 1-2 percentage points in Malaysia and Korea, 3-4 percentage points in Indonesia and Singapore, and 5–6 percentage points in Sri Lanka and the Philippines.
Although quantitative restrictions—including bans on automobile imports, and the application of domestic content rules—were used (especially during the 1970s), the coverage of quantitative restrictions has generally been relatively low—less than 5 percent of tariff lines—and has remained stable during the period under consideration.
The adjustment efforts of 1985-86 included several actions to raise public enterprises’ prices (especially for oil and water supply) so as to improve profitability.
Although efforts to adjust to the external shocks and to correct the growing fiscal and external current account imbalances started in the early 1980s, at the beginning of what has been called the “adjustment” period throughout this paper, the major policy actions that can be most closely linked with the growth performance of the 1987-93 period did not take place until the mid-1980s.
Serven and Solimano (1994) have presented a detailed discussion of the theory and empirics of private investment behavior in developing countries.
Using the level of private investment as the dependent variable could lead to spurious results with aggregate demand or the economic activity variable acting as a trend.
A time series is said to be strictly stationary if all of its moments (mean, variance, and so forth) are independent of time. Typically, the weaker concept of stationarity is used, which only requires that the first two moments are invariant with respect to time and that the autocorrelations depend only on the length of the lag between observations and not on the points in time at which these observations are made.
It should be noted, however, that by definition the ratio of private investment to GDP is inherently stationary in that it cannot grow without limit.
The test statistics for the private investment ratio and the index of relative unit labor costs are estimated to be -5.77 and -8.32, respectively. The 1 percent critical value of the lest statistic for the appropriate sample size, as calculated by Perron (1989), is -4.45, suggesting that the null hypothesis of the presence of a unit root can be rejected.
The expected rate of dépréciation was proxied by the actual depreciation. In practice, since the baht was effectively pegged to the U.S. dollar throughout this period, this adjustment should not be expected to significantly affect the sign of the estimated coefficient.
Public infrastructure investment was defined as capital expenditure by the central government and state enterprises on power, transport, and communication, and that by the central government on education, health, and agriculture.
As is well known, estimation by ordinary least squares is less sensitive to specification errors in small samples than is estimation using instrumental-variables techniques. Further, instrumental-variables estimators are highly sensitive to the (often arbitrary) choice of instruments. This being said, care was taken to avoid potential simultaneity biases, particularly in the choice of the variable measuring economic activity, by using lagged real GDP growth or, in the case of Thailand, real GDP growth in trading-partner countries
These results do nut imply that domestic activity levels, which are accounted for by expressing the dependent variable as an investment rate, are unimportant for private sector investment.
It is clear from this definition that this concept of causality does not necessarily imply an “event-outcome” relationship between the two variables.
The variables that were included in the system are those that were statistically significant in the specification search outlined below.
The results were very similar when public infrastructure investment was used. However, since the data on total public investment were judged to be more reliable, total public investment was used in the specification search.
The dummy variable for the post-1986 period was always insignificant, possibly because any influence of the shift toward active export promotion, through the devaluation and the elimination of export taxes, is being captured in the relative unit labor cost variable. The relative price of capital goods (proxied by the ratio of the fixed investment deflator to the GDP deflator) was also used instead of the domestic real interest rate but proved to be insignificant in all specifications.
The last IMF stand-by arrangement, scheduled to expire in March 1987, was canceled at the end of 1986. The decomposition was also calculated for the period 1972-87 and for the “preadjustment” (pre-1980) and “adjustment” (1980-86) periods separately. However, in view of the very small degree of variation in the private investment ratio and in most of the explanatory variables (with the exception of relative unit labor costs) before 1986-87, this exercise did not result in a meaningful decomposition and is not reported here.
This is a “ceteris paribus” simulation, in which all other coefficients and variables are assumed to remain invariant in this counterfactual scenario. This is clearly a strong assumption, but the results are nevertheless useful in gauging the importance of the evolution of particular variables in the estimated equation.
The finding that the various measures of uneertainly that were included were always insignificant is consistent with this conclusion.
Recent Occasional Papers of the International Monetary Fund
146. Thailand: The Road to Sustained Growth, by Kalpana Kochhar, Louis Dicks-Mireaux, Balazs Horvath, Mauro Mecagni, Erik Offerdal, and Jianping Zhou. 1996.
145. Exchange Rate Movements and Their Impact on Trade and Investment in the APEC Region, by Takatoshi Ito, Peter Isard, Steven Symansky, and Tamim Bayoumi. 1996.
144. National Bank of Poland: The Road to Indirect Instruments, by Piero Ugolini. 1996.
143. Adjustment for Growth: The African Experience, by Michael T. Hadjimichael, Michael Nowak, Robert Sharer, and Amor Tahari. 1996.
142. Quasi-Fiscal Operations of Public Financial Institutions, by G.A. Mackenzie and Peter Stella. 1996.
141. Monetary and Exchange System Reforms in China: An Experiment in Gradualism, by Hassanali Mehran, Marc Quintyn, Tom Nordman, and Bernard Laurens. 1996.
140. Government Reform in New Zealand, by Graham C. Scott. 1996.
139. Reinvigorating Growth in Developing Countries: Lessons from Adjustment Policies in Eight Economies, by David Goldsbrough, Sharmini Coorey, Louis Dicks-Mireaux, Balazs Horvath, Kalpana Kochhar, Mauro Mecagni, Erik Offerdal, and Jianping Zhou. 1996.
138. Aftermath of the CFA Franc Devaluation, by Jean A.P. Clément, with Johannes Mueller, Stéphane Cossé, and Jean Le Dem. 1996.
137. The Lao People’s Democratic Republic: Systemic Transformation and Adjustment, edited by Ichiro Otani and Chi Do Pham. 1996.
136. Jordan: Strategy for Adjustment and Growth, edited by Edouard Maciejewski and Ahsan Mansur. 1996.
135. Vietnam: Transition to a Market Economy, by John R. Dodsworth, Erich Spitäller, Michael Braulke, Keon Hyok Lee, Kenneth Miranda, Christian Mulder, Hisanobu Shishido, and Krishna Srinivasan. 1996.
134. India: Economic Reform and Growth, by Ajai Chopra, Charles Collyns, Richard Hemming, and Karen Parker with Woosik Chu and Oliver Fratzscher. 1995.
133. Policy Experiences and Issues in the Baltics, Russia, and Other Countries of the Former Soviet Union, edited by Daniel A. Citrin and Ashok K. Lahiri. 1995.
132. Financial Fragilities in Latin America: The 1980s and 1990s, by Liliana Rojas-Suárez and Steven R. Weisbrod. 1995.
131. Capital Account Convertibility: Review of Experience and Implications for IMF Policies, by staff teams headed by Peter J. Quirk and Owen Evans. 1995.
130. Challenges to the Swedish Welfare State, by Desmond Lachman, Adam Bennett, John H. Green, Robert Hagemann, and Ramana Ramaswamy. 1995.
129. IMF Conditionality: Experience Under Stand-By and Extended Arrangements. Part II: Background Papers. Susan Schadler, Editor, with Adam Bennett, Maria Carkovic, Louis Dicks-Mireaux, Mauro Mecagni, James H.J. Morsink, and Miguel A. Savastano. 1995.
128. IMF Conditionality: Experience Under Stand-By and Extended Arrangements. Part I: Key Issues and Findings, by Susan Schadler, Adam Bennett, Maria Carkovic, Louis Dicks-Mireaux, Mauro Mecagni, James H.J. Morsink, and Miguel A. Savastano. 1995.
127. Road Maps of the Transition: The Baltics, the Czech Republic, Hungary, and Russia, by Biswajit Banerjee, Vincent Koen, Thomas Krueger, Mark S. Lutz, Michael Marrese, and Tapio O. Saavalainen. 1995.
126. The Adoption of Indirect Instruments of Monetary Policy, by a staff team headed by William E. Alexander, Tomás J.T. Baliño, and Charles Enoch. 1995.
125. United Germany: The First Five Years—Performance and Policy Issues, by Robert Corker, Robert A. Feldman, Karl Habermeier, Hari Vittas, and Tessa van der Willigen. 1995.
124. Saving Behavior and the Asset Price “Bubble” in Japan: Analytical Studies, edited by Ulrich Baumgartner and Guy Meredith. 1995.
123. Comprehensive Tax Reform: The Colombian Experience, edited by Parthasarathi Shome. 1995.
122. Capital Flows in the APEC Region, edited by Mohsin S. Khan and Carmen M. Reinhart. 1995.
121. Uganda: Adjustment with Growth, 1987–94, by Robert L. Sharer, Hema R. De Zoysa, and Calvin A. McDonald. 1995.
120. Economic Dislocation and Recovery in Lebanon, by Sena Eken, Paul Cashin, S. Nuri Erbas, Jose Martelino, and Adnan Mazarei. 1995.
119. Singapore: A Case Study in Rapid Development, edited by Kenneth Bercuson with a staff team comprising Robert G. Carling, Aasim M. Husain, Thomas Rumbaugh, and Rachel van Elkan. 1995.
118. Sub-Saharan Africa: Growth, Savings, and Investment, by Michael T. Hadjimichael, Dhaneshwar Ghura, Martin Mühleisen, Roger Nord, and E. Murat Ucer. 1995.
117. Resilience and Growth Through Sustained Adjustment: The Moroccan Experience, by Saleh M. Nsouli, Sena Eken, Klaus Enders, Van-Can Thai, Jörg Decressin, and Filippo Cartiglia, with Janet Bungay. 1995.
116. Improving the International Monetary System: Constraints and Possibilities, by Michael Mussa, Morris Goldstein, Peter B. Clark, Donald J. Mathieson, and Tamim Bayoumi. 1994.
115. Exchange Rates and Economic Fundamentals: A Framework for Analysis, by Peter B. Clark, Leonardo Bartolini, Tamim Bayoumi, and Steven Symansky. 1994.
114. Economic Reform in China: A New Phase, by Wanda Tseng, Hoe Ee Khor, Kalpana Kochhar, Dubravko Mihaljek, and David Burton. 1994.
113. Poland: The Path to a Market Economy, by Liam P. Ebrill, Ajai Chopra, Charalambos Christofides, Paul Mylonas, Inci Otker, and Gerd Schwartz. 1994.
112. The Behavior of Non-Oil Commodity Prices, by Eduardo Borensztein, Mohsin S. Khan, Carmen M. Reinhart, and Peter Wickham. 1994.
111. The Russian Federation in Transition: External Developments, by Benedicte Vibe Christensen. 1994.
110. Limiting Central Bank Credit to the Government: Theory and Practice, by Carlo Cottarelli. 1993.
109. The Path to Convertibility and Growth: The Tunisian Experience, by Saleh M. Nsouli, Sena Eken, Paul Duran, Gerwin Bell, and Zühtü Yucelik. 1993.
108. Recent Experiences with Surges in Capital Inflows, by Susan Schadler, Maria Carkovic, Adam Bennett, and Robert Kahn. 1993.
107. China at the Threshold of a Market Economy, by Michael W. Bell, Hoe Ee Khor, and Kalpana Kochhar with Jun Ma, Simon N’guiamba, and Rajiv Lall. 1993.
106. Economic Adjustment in Low-Income Countries: Experience Under the Enhanced Structural Adjustment Facility, by Susan Schadler, Franek Rozwadowski, Siddharth Tiwari, and David O. Robinson. 1993.
105. The Structure and Operation of the World Gold Market, by Gary O’Callaghan. 1993.
104. Price Liberalization in Russia: Behavior of Prices, Household Incomes, and Consumption During the First Year, by Vincent Koen and Steven Phillips. 1993.
103. Liberalization of the Capital Account: Experiences and Issues, by Donald J. Mathieson and Liliana Rojas-Suárez. 1993.
102. Financial Sector Reforms and Exchange Arrangements in Eastern Europe. Part I: Financial Markets Fund Intermediation, by Guillermo A. Calvo and Manmohan S. Kumar. Part II: Exchange Arrangements of Previously Centrally Planned Economies, by Eduardo Borensztein and Paul R. Masson. 1993.
101. Spain: Converging with the European Community, by Michel Galy, Gonzalo Pastor, and Thierry Pujol. 1993.
Note: For information on the title and availability of Occasional Papers not listed, please consult the IMF Publications Catalog or contact IMF Publication Services.