This Selected Issues paper for Thailand highlights the effect of higher global interest rates on Thailand and the relationship between financial crises and long-term potential growth. Since the Asian crisis, Thailand has adopted an inflation targeting regime, and has intervened in the foreign exchange market to prevent excessive baht volatility. The monetary tightening in the United States in 1994 has been followed by heightened bond market volatility and a widening of emerging countries’ credit spreads.

Abstract

This Selected Issues paper for Thailand highlights the effect of higher global interest rates on Thailand and the relationship between financial crises and long-term potential growth. Since the Asian crisis, Thailand has adopted an inflation targeting regime, and has intervened in the foreign exchange market to prevent excessive baht volatility. The monetary tightening in the United States in 1994 has been followed by heightened bond market volatility and a widening of emerging countries’ credit spreads.

III. Financial Crises and Growth1

A. Introduction

The effects of currency crises on output are highly diverse. While the average country suffers output and growth losses, a significant fraction of crises leads to growth accelerations. This paper documents this diversity and identifies some of the macroeconomic determinants that are correlated with stronger post-crisis growth recoveries.

1. The Asian financial crisis of the late 1990s resulted in significant output contractions in the five crisis countries: Indonesia, Malaysia, Philippines, South Korea, and Thailand, and it took all of these countries several years to recover to their pre-crisis levels. However, the depth of the crisis and the speed of recovery differed widely across these countries. For example, while real per-capita GDP in Indonesia dropped by over 14 percent in 1998 and barely reached its pre-crisis level in 2004, the Philippines lost less than 3 percent in 1998 and surpassed its pre-crisis GDP by 2000. Thailand’s experience is somewhere in the middle, with a loss of just over 11 percent and a recovery to pre-crisis GDP before the end of 2002. In fact, looking beyond the specific case of the Asian crisis, a broader look at cross-country experiences indicates that the effects of financial crisis are often much less severe, sometimes even leading into growth accelerations. What are possible sources of the different outcomes across countries? This is one of the questions that this chapter attempts to address.

2. Perhaps more important than the short- and medium-term aftermath of the crisis, however, is the second question this chapter is concerned with, namely, do financial crises affect countries’ long-term growth potential? That is, do countries embark on growth profiles in the long term that are flatter or steeper than their pre-crisis growth paths? Both are conceivable outcomes. A financial crisis may be the unavoidable outcome of unsustainably, high pre-crisis growth rates, as may occur in episodes of speculative bubbles; a return to the lower, fundamentals-driven growth rates would then be the expected outcome. But crises may also trigger structural reforms to remedy the structural weaknesses that contributed to the crisis, thereby allowing the country to move to higher long-term growth.

3. Although a key focus of this paper is on the Asian crisis, and specifically its implications for Thailand’s growth, the questions posed above are difficult to answer within the narrow context of the Asian crisis, given the small sample and the short-time period that has elapsed since then. This chapter therefore takes a cross-country perspective to provide some answers to these questions. Lessons for Thailand are then examined in a case study. The outline is as follows. In Section B, some of the conceptual issues surrounding crisis effects are discussed and descriptive statistics of crises effects presented. Using a dynamic panel dataset, a first set of empirical analyses then examines the long-term implications of financial crisis. Focusing on crisis events only, a second set of regressions analyzes the correlates of growth recovery with key macroeconomic variables, including fiscal and monetary policy indicators. Section C considers the specific case of Thailand and its growth outlook, and Section D concludes.

B. Financial Crises and Growth

Conceptual issues and descriptive statistics

4. Financial crises, currency and/or banking, are typically associated with large losses in economic activity and severe and prolonged recessions. This view has been reinforced by the recent Asian financial crisis in 1997/98 in which all of the affected countries suffered large output losses. A first glance at the data indicates, however, that this dire view of financial crises is not fully reflected by the international experience. Table 1 provides data on the medium-term growth effects of 183 currency crises during the 1980s and 1990s. As an indicator of the severity of crisis effects, the difference in average real, per-capita GDP growth rates in the five years leading up the to crisis and those in the five years following the crisis, including the crisis year, is calculated. Both the median and mean are shown for a number of subsamples, including by decade, level of development and region.

Table 1.

Medium-Term Effects of Currency Crises: Difference in Average Per-Capita GDP Growth 5 Years Post- and Pre-Crisis

(median/mean, number of crises in parentheses)

article image

5. Two aspects stands out. First, in general, the effects of crises on growth are relatively small. The average loss in growth after crisis is only about 0.05 percentage points, with a median loss of 0.11 percentage points. This is particularly surprising in light of a commonly held view that speculative bubbles are the source of crises, with “artificially” elevated pre-crisis growth rates. Second, there is a large heterogeneity in outcomes. The growth losses that were observed in East Asia are larger (in absolute terms) by several orders of magnitude compared to the full sample mean, both during the 1990s, reflecting the Asian crisis, but also during the earlier decade. The table also indicates that currency crises have become less harmful over time overall, mostly reflecting the trend in developing countries, while emerging markets in East Asia and Latin America went the opposite way. In summary, large differences in outcomes can be observed, both over time, across regions, and across income levels.

6. To gain a conceptual understanding of the potential growth outcomes, it is helpful to consider the diagram in Figure 1 which depicts three stylized growth outcomes. Clearly, many other outcomes are possible, but as will be seen below, these are some of the empirically relevant ones. Case 1 in this figure takes a benign view of financial crises: while output dips at the time of the crisis, it recovers after a transition period to the pre-crisis trend level, thus involving no permanent loss, although growth rates will be above trend/pre-crisis growth during the transition period. Case 3 depicts an outcome at the lower end of the spectrum: output drops to, and stays at, a lower trend level without ever recovering, thus involving large permanent losses. Case 2 is somewhere in between: GDP recovers somewhat after the crisis, but not fully, leading to higher than pre-crisis growth rates in the short and medium term and to lower than pre-crisis trend growth in the long run, thus incurring permanent losses, albeit smaller than in case 3.

Figure 1.
Figure 1.

Possible Post-Crisis Recoveries

Citation: IMF Staff Country Reports 2006, 019; 10.5089/9781451836837.002.A003

7. Figure 2 presents the empirical counterpart to Figure 1 for the five countries most affected by the Asian financial crisis in the late 1990s. A casual glance at the graph suggests that all three cases depicted in Figure 1 are represented in the data. Corresponding to the more benign case 1 in Figure 1, Korea appears to have quickly recovered back to its pre-crisis growth path within only a few years after the crisis, helped by the fact that the drop at the crisis was relatively small. On the other end of the spectrum, corresponding to case 3, is Indonesia. Malaysia, the Philippines, and Thailand are somewhere in between, catching up to some extent, but not all the way to the pre-crisis trend, thus corresponding to case 2 in Figure 1.

Figure 2.
Figure 2.

Real Per-capita GDP, 1990-2003

(1998 = 100)

Citation: IMF Staff Country Reports 2006, 019; 10.5089/9781451836837.002.A003

8. The conceptual and empirical considerations in Figures 1 and 2 suggest important lessons for the estimation of the effects of financial crisis on output and growth. First, comparing pre- and post-crisis growth rates may not yield much information on the changes in long-term trends, given that even without changes in the trend, as in the Korean case, growth rates may be higher than pre-crisis for several years due to catch-up. Second, the fairly large differences in growth outcomes across countries, even conditional on the “same” crisis as shown in Figure 2, opens up the possibility that country-specific factors, such as institutions and policies, have a possibly important influence on how the country is affected by a financial crisis.

9. The empirical analyses in the following two subsections attempt to highlight some possible explanations for differences across outcomes. The main objective of this paper is to provide an analytical framework which allows for a meaningful discussion of Thailand’s recent experience and its outlook. As such, much of the work in this paper draws from the existing literature, although some of the results obtained here extend and modify some of the existing work. Some of the main contributions on which this chapter builds are Barro (2002) and Barro and Lee (2003) who focus on the medium-term growth implications of currency and banking crises. Their main finding is that crises have little medium-term impact, but as will be seen below, considering a longer-term horizon challenges this view. Park and Lee (2002) take an event-based view and focus on the recovery profile after a crisis; their methodology is used in the section on short- and medium-term recovery below. Gupta, Mishra, and Sahay (2003) take a similar approach, although they focus on medium-term output losses only, rather than tracing out the recovery path.

Long-run growth

10. Data on financial crises, both currency and banking, are a key element of this chapter. The source of currency crisis data is Gupta, Mishra, and Sahay (2003) who use a majority rule based on five currency crisis definitions from the four papers by Milesi-Ferretti and Razin (1998; 2 definitions), Frankel and Rose (1996), Berg and Pattillo (1999), and Goldstein, Kaminsky, and Reinhart (2000). Data on banking crises are from Demirgüç-Kunt and Detragiache (2005). All remaining data series are from the World Bank’s World Development Indicators except for data on education attainment which is from the Barro-Lee dataset.

11. Table 2 presents the regression results for dynamic panel regressions using the General Method of Moments (GMM) methodology developed by Arellano and Bond (1991). The regression setup is very similar to that used by Barro and Lee (2003) using lagged GDP, schooling, trade openness, investment, government size, and democracy as control variables. The regressions here also include dummies for currency and banking crises. Although Barro and Lee use 2SLS and a different measure of trade openness, their results are very similar to those obtained here. In particular, columns (1) to (3) replicate their regressions, including in various combinations, contemporaneous, and lagged values of dummies for currency and banking crises. The negative coefficient on lagged per-capita GDP is evidence of conditional convergence; also as expected, trade openness, investment rates, and democracy are all significantly and positively correlated with growth.

Table 2.

Dependent Variable: Per-Capita GDP Growth, 1980-2000 1/

article image
Note: Numbers in parentheses are robust standard errors. Significance levels are *10%; **5%; ***1%. All regressions are estimated by GMM and include time dummies and a constant.

Non-overlapping five-year averages.

Numbers are p-values for the Arellano and Bond (1991) test of no second-order autocorrelation in residuals.

12. As in Barro and Lee, currency and banking crises are found to be negatively correlated with growth around the time of impact, but positively when lagged by one five-year period. In light of the previous section’s discussion, this is not surprising: on average, countries suffer output losses at the time of the crisis, but growth is likely to be elevated during the recovery period. Barro and Lee conclude that there is no evidence of a long-run impact of financial crisis on growth. This conclusion may be misleading in several ways, however. First, growth recoveries in the medium term provide little information on the longer-term effects of a crisis—the simple picture in Figure 1 and the related discussion suggest that elevated growth rates in the short to medium term are consistent with any long-term outcome. The results from columns (1) to (3) therefore still leave open the question of whether a country permanently suffers from a crisis. Second, the time horizon of one lag, that is, about five years, is arguably too short to measure long-term effects of a crisis.

13. A fuller picture emerges from considering longer horizons, as done in columns (4) and (5) which also include twice-lagged crisis dummies and so measure, on average, the growth effects 10 years after a crisis. The coefficients for both currency and banking crises are statistically significant (column 5) and negative. Financial crises appear to induce a growth pattern similar to case 2 in Figure 1—namely, a negative growth shock at the time of the crisis, elevated growth rates during the following recovery period, and a convergence to growth rates in the long term that are below those the country experienced around the time of the crisis.

14. Barro and Lee’s (2003) conclusion that the “analysis found no evidence that financial crises had effects on growth that persisted beyond a five-year period” (p. 83) is thus to be interpreted with caution; indeed, the reverse is likely to be true. Nonetheless, a variant of Barro and Lee’s conclusion may still hold true: the results do not preclude the possibility that average pre-crisis growth rates were “artificially” high (that is, above trend growth)—using raw growth data without detrending, as is done here, therefore still allows for the fact that the observed negative effect on future long-term growth simply reflects a return to a country’s “true” potential growth rate. As such, a crisis may well represent a correction of excessively high pre-crisis growth rates, rather than a permanent loss in a country’s growth potential, and in this sense, crises may indeed have little impact on a country’s long-term growth. As is discussed in Section C, this appears to be the case in Thailand.

15. Considering the effects of crisis on investment rates can provide added insight into the effects of crises on growth. As the regressions in Table 2 indicate, the investment-to-GDP ratio is one of the key explanatory variables in the growth regressions. Table 3 presents the results from dynamic panel regressions very similar to those in Table 2, except that now investment-to-GDP is the dependent variable. Crises trigger a sharp reduction in investment rates at impact, but have no statistically significant effect in the medium or long term. That is, investment rates under medium and long horizons are not significantly different than in the absence of crises. In the context of East Asia, Sarel (1995) has argued that both physical capital accumulation and productivity growth account for East Asia’s strong growth performance prior to the crisis. The results in Table 3 do not preclude the possibility of a long-term effect on growth, but they suggest that changes in average growth rates are likely to come from changes in productivity growth rather than from reduced capital accumulation.

Table 3.

Dependent Variable: Investment/GDP, 1980-2000 1/

article image
Note: Numbers in parentheses are robust standard errors. Significance levels are *10%; **5%; ***1%. All regressions are estimated by GMM and include time dummies and a constant.

Nonoverlapping five-year averages.

Numbers are p-values for the Arellano and Bond (1991) test of no second-order autocorrelation in residuals.

Short- and medium-term recovery

16. By considering five-year averages, the analyses in the previous section abstract from shorter-term fluctuations and thus take a longer-term view. However, independent of whether or not financial crises have long-term growth (or level) implications, it is clear that they typically inflict severe economic short-term pain by inducing sharp losses in economic output during the year(s) of the crisis. Thus, even if economies eventually adjust and escape such crises with little long-term damage, an important question to the policymaker remains as to how the adjustment can be made as quickly and painlessly as possible.

17. To shed light on this question, rather than including all observations, as was done in the previous section, this section focuses on crisis events only and considers in more detail the economic adjustment during the five years following a crisis. To maximize the sample size, the focus is on currency crises, for which more observations are available; twin crises, that is, contemporaneous banking crises, are controlled for via a banking crisis dummy.

18. The methodology closely follows Park and Lee (2002). For each of the five years following the onset of a crisis, average growth since the crisis year T is calculated and regressed on a number of variables, thus providing an understanding of countries’ adjustment profile, and the evolution of its determinants, over time. The setup is parsimonious in order to focus on the main policy indicators, namely, fiscal and monetary. Financial (capital flows) and investment variables are also included in order to measure the extent to which domestic and international investors regain confidence in the economy. World economic growth, time dummies, trade-weighted real exchange rate developments, trade (the sum of exports and imports as a percentage of GDP), the country’s per-capita GDP at the time of the crisis, and its average growth rate during the five years prior to the crisis are also included to control for world economic conditions as well as the country’s economic development and business cycle.2

19. Table 4 presents the results for the growth regressions. Perhaps surprisingly, the coefficient on the pre-crisis growth variable is almost always positive. Although it is never statistically insignificant, it does suggest that countries with high growth rates do not suffer more strongly than countries that enter a crisis with more modest growth. The factors determining depth of the crisis (column 1) and its recovery (columns 2–6) change during the adjustment process. Countries that are able to avoid drops in investment moderate the negative growth impact at time T, with other factors, including fiscal and monetary policy, playing no role. As the country emerges from the crisis, the degree to which investment is maintained plays a smaller role, becoming insignificant from T+2 onwards. Fiscal and monetary expansions, in turn, start to play a more important role throughout the remainder of the recovery, although their quantitative importance peaks at around T+3. Capital flows play no statistically significant role until T+5. Twin crisis, that is, those that involve both a currency and a banking crisis, are particularly severe, although the added effect of banking crises appears to dissipate relatively quickly within two years after the crisis.

Table 4.

Dependent Variable: Average Per-Capita GDP Growth During k - 1 Post-Crisis Years

article image
Note: Numbers in parentheses are robust standard errors. Significance levels are *10%; **5%; ***1%. All regressions are estimated by OLS and include a constant and time dummies.

20. While macroeconomic demand management via fiscal and monetary policy is unlikely to be effective for boosting long-term growth, the lesson that emerges from these regressions is that they may play an important role during the recovery from a financial crisis. Of course, countries that are sufficiently robust to avoid confidence losses in the first place and can maintain high investment will be in a better position. Similarly, in the medium to long term, restoring both domestic and external confidence is crucial for growth. In this context, a word of caution is appropriate: while both fiscal and monetary expansionary policies are positively correlated with growth recoveries, an important possible link is not reflected in these regressions: to the extent that excessive fiscal and monetary policies may lower confidence, these may negatively influence capital inflows, for example. Hence, the positive coefficient may be upward biased. Nevertheless, modest expansionary policies are likely to help the recovery process.

21. The same regressions are repeated with investment-to-GDP as the dependent variable, the results are presented in Table 5. Fiscal and monetary policy appear much less relevant for investment. The key correlates of investment are trade and capital flows. That is, the external performance appears to matter strongly for investment flows. While the connection between investment and growth recovery does not appear very strong (see Table 4), it is likely to be important for long-run growth prospects as suggested by the earlier long-run growth regressions. Adjustments in the real exchange rate (specifically, devaluations, represented by decreases in the real exchange rate indicator) also help.

Table 5.

Dependent Variable: Investment/GDP During k - 1 Post-Crisis Years

article image
Note: Numbers in parentheses are robust standard errors. Significance levels are *10%; **5%; ***1%. All regressions are estimated by OLS and include a constant and time dummies.

Discussion

22. Overall, from a long-term perspective, the above analysis suggests that financial crises are costly, and more so in the case of a twin crises, that is, the contemporaneous occurrence of currency and banking crisis. Whether or not crises affect a country’s long-term growth potential cannot be unambiguously determined from the regressions—the long-term growth reductions that were found may be a lowered growth potential, but they may also represent the unavoidable correction of pre-crisis growth rates that were unsustainable. In the latter scenario, therefore, growth reductions are to be interpreted as returns to the country’s growth trend, rather than a reduction in the trend.

23. Although the long-term implications leave room for interpretation, it is clear that crises are costly, resulting in large average output losses when they occur and requiring several years of recovery. The second set of regressions provides some guidance to the policymaker as to how the recovery process can be helped. Expansionary fiscal and monetary policies are likely to help, as are mechanisms that allow for an adjustment of real exchange rates and a free flow of goods and capital. It is crucial, of course, to limit the negative effects of a crisis in the first place—here, countries that maintain high investment rates do better. Maintaining high investment rates is likely to be a function of the degree of confidence that market participants have in the economy, and as such, prudent and stable macro policies are generally likely to be helpful in mitigating the negative effects of crises.

C. The Case of Thailand

Recent growth experience

24. On the background of the cross-country analyses, this section examines in more detail Thailand’s recent macroeconomic experience, with particular focus on the implications of the financial crisis of 1997/98 for Thailand’s growth outlook. Figure 2 depicts per-capita GDP outcomes between 1990 and 2003 for the five East Asian countries hit by the crisis. The divergence of possible outcomes in the full sample of crisis, as documented in Table 1, is also reflected in the smaller context of the Asian crisis. While Korea could be interpreted as following the path of case 1 in Figure 1, quickly resuming its pre-crisis growth path, Indonesia can be seen to occupy the opposite end of the spectrum, experiencing the largest drop and suffering both a level and a growth decline, with Malaysia and the Philippines taking a similar path. While Thailand is recovering well particularly in recent years, it is not doing quite as well as Korea. Indeed, no country other than Korea appears to be resuming its pre-crisis path. This outcome is, of course, consistent with the negative crisis dummy coefficients shown in Table 2. As mentioned above, however, these comparisons pertain only to the growth path that was observed during the years prior to the crisis—as will be seen in the next section, there is little evidence that Thailand in particular has suffered a growth loss in comparison to its long-term growth average.

25. Given the empirical exercises in the previous section, it is worthwhile examining for each of the crisis countries some of the macro variables that were included in the regressions. Figure 3 presents the time-series for investment, the real exchange rate, trade, government consumption, M2/GDP, and capital flows for the five crisis countries during 1990–2003.3

Figure 3.
Figure 3.

Selected Macroeconomic Variables, 1990-2003

Citation: IMF Staff Country Reports 2006, 019; 10.5089/9781451836837.002.A003

26. All countries suffered drops in investment, and their ranking on this dimension broadly corresponds to the countries’ relative growth recoveries: Korea has maintained a relatively high investment ratio, and also Thailand’s recent growth spurt is consistent with its recovery in investment. By contrast, the other countries, after a brief bounce-back, suffered continuing declines in investment. The evidence is less clear on the trade dimension, with almost all countries (except for Indonesia) maintaining stable trade paths. Similarly, the empirical evidence for a positive effect of expansionary fiscal policy on growth recovery is not clearly borne out in the Asian crisis context; if anything, the reverse appears to be the case, lending support to the previously discussed potential role of fiscal policy for shaping investor perceptions and building confidence. Finally, M2/GDP appears to have contributed to Korea’s recovery, where M2/GDP has been fairly expansionary since the crisis, in contrast to the other countries.

Growth outlook

27. This section attempts to look forward and to answer the question as to what Thailand’s long-term potential is. This section employs several different methods, including nonstructural trend estimation techniques and growth accounting, all of which produce broadly similar results. To extract Thailand’s long-term trend, two alternative methods were used: log-linear trend estimation, where the log of per-capita GDP is regressed on time, and Hodrick-Prescott filtering4 (with λ=100 for annual values). Both methods included a shift dummy for 1997 to allow for trend shifts resulting from the crisis, and the trend estimation additionally included a slope dummy from 1997 onwards to allow for structural changes in growth rates. The actual series as well as the two trend series are shown in Figures 4 (levels) and 5 (growth rates).

Figure 4.
Figure 4.

Actual and Potential Per-Capita GDP

(In thousands of 1998 prices)

Citation: IMF Staff Country Reports 2006, 019; 10.5089/9781451836837.002.A003

Note: The HP-trend includes a crisis shift dummy; the log-linear trend includes a shift and a slope dummy. See SI chapter on financial crises and growth for details.
Figure 5.
Figure 5.

Actual and Potential Annual Per-Capita GDP Growth

(In annual percent)

Citation: IMF Staff Country Reports 2006, 019; 10.5089/9781451836837.002.A003

28. Both methods agree that Thailand grew at an average rate of about 5 percent during the time period. Although potential growth may have increased somewhat during the early 1990s, most of the growth was above potential and thus unsustainable, as evidenced by the financial crisis. Neither method lends support to the hypothesis that the crisis has induced a significant deviation in growth from Thailand’s long-term historical average.5 The collapse in 1997/98 is thus best interpreted as a correction of the above-potential growth rates that prevailed during the run-up to the crisis, rather than as a change in Thailand’s long-term trend. Although such estimations are backward-looking and are therefore ill-suited for forecasting purposes, they are suggestive that, in the absence of unexpected changes, and to the extent that historical structural relationships extend into the future, Thailand may well continue to grow at a rate of about 5 percent.6

29. Given their atheoretical nature, these trend decomposition methods lack theoretical content; however, and it is therefore difficult to formally include economic assumptions in the forecast. An alternative approach with more economic context is growth accounting. This approach starts from the assumption that the aggregate production process follows a Cobb-Douglas production function of the form GDP = TFP · Kα · H1–α, where K is the physical capital stock, H is the human capital stock, and α denotes the share of physical capital. Growth can then be decomposed as

ΔGDPGDP=ΔTFPTFP+αΔKK+(1α)ΔHH.

Given data on GDP, K, H and α, TFP can be calculated as a residual.7

30. This equation can be used in two ways. Under a long-term horizon, growth theory predicts a constant K/GDP ratio once a steady state has been reached.8 Long-run per-capita steady-state growth can thus be expressed as

ΔGDPGDP=ΔTFPTFP/(1α)+ΔHH.

Predicting long-term growth then requires making assumptions on growth in TFP, K and H, and on the value of α.

31. Assuming constant employment and labor force participation rates, growth in H is the sum of population and per-worker human capital growth. Population growth has converged to about 0.7 percent in recent years, while human capital per worker has grown at about 0.8 percent during the last five years. Human capital growth is thus estimated at 1.5 percent. As the growth accounting equation shows, historical TFP calculations, and consequently TFP predictions to the extent that they are based on historical experience, are largely a function of α, given time series for K and H. It turns out, however, that average TFP growth rates are very similar, around 3.1 percent, during 2000-04 for all three values of α that were used.9 Nevertheless, even small differences in TFP growth are magnified. Table 6 summarizes the resulting long-term growth predictions. According to these calculations, Thailand may approach per-capita growth rates between 5 and 6.1 percent in the long term, close to Thailand’s historical long-run average.10

32. These projections look very far into the future. Over a medium-term horizon, the assumption of a constant K/GDP-ratio is likely not met. Staff estimates that the investment-to-GDP ratio may reach 30-35 percent by 2010, resulting in average capital accumulation rates of around 5.5 percent during that time period. A continuation of recent TFP growth and human capital growth of 1.5 percent would result in medium-term growth between 5 and 5.7 percent, between Thailand’s historical growth and the long-term growth rates estimated above.

33. All of these projections omit many potentially relevant features. In particular, all of the above methods extrapolate from the past and implicitly assume that past structural relationships extend into the future, which may not be appropriate. By the same token, however, projecting higher growth rates than those above would require significant structural breaks with Thailand’s past experience over the last four decades. It is noted that the projections above are based on the continuation of recent high TFP growth rates, which may be considered optimistic given that historical TFP growth was between 1 and 2 percentage points lower.

D. Conclusion

34. The goal of this chapter has been to examine countries’ experience with and recovery from financial crises, with a special application to the case of Thailand. There are several findings. Growth experiences after financial crises vary significantly across countries, with some countries experiencing growth accelerations. However, the average country suffers, with growth rates below those prior to the crisis even 5 to 10 years after the event. However, the general cross-country analyses do not reveal whether crisis permanently lower countries’ growth potential or whether they simply correct unsustainable paths. In the case of Thailand, the answer appears to be the latter—while growth rates have come down from the exuberant rates in the 1990s that led into the crisis, the rates since then seem perfectly in line with Thailand’s longer growth experience. Looking ahead, various methods suggest that such growth patterns are also likely to continue in the future.

References

  • Arellano, Manuel, and Stephen Bond, 1991, “Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Data,Review of Economic Studies 58, 277297.

    • Search Google Scholar
    • Export Citation
  • Barro, Robert, 2002, “Economic Growth in East Asia Before and After the Financial Crisis,in Korean Crisis and Recovery, David Coe and Se-Jik Kim, eds.

    • Search Google Scholar
    • Export Citation
  • Barro, Robert, and Jong-Wha Lee, 2003, “Growth and Investment in East Asia Before and After the Financial Crisis,Seoul Journal of Economics 16, 83118.

    • Search Google Scholar
    • Export Citation
  • Berg, Andrew, and Catherine Pattillo, 1999, “Are Currency Crises Predictable? A Test,IMF Staff Papers 46, 107138.

  • Bils, Mark, and Peter J. Klenow, 2000, “Does Schooling Cause Growth?,American Economic Review 90, 11601183.

  • Demirgüç-Kunt, Asli and Enrica Detragiache, 2005, “Cross-Country Empirical Studies of Systemic Bank Distress: A Survey,IMF Working Paper 05/96.

    • Search Google Scholar
    • Export Citation
  • Frankel, Jeffrey, and Andrew Rose, 1996, “Currency Crashes in Emerging Markets: An Empirical Treatment,Journal of International Economics 41, 351366.

    • Search Google Scholar
    • Export Citation
  • Goldstein, Morris, Graciela L. Kaminsky and Carmen M. Reinhart, 2000, “Assessing Financial Vulnerability: An Early Warning System for Emerging Markets,” (Washington: Institute for International Economics).

    • Search Google Scholar
    • Export Citation
  • Gupta, Poonam, Deepak Mishra and Ratna Sahay, 2003Output Response to Currency Crises,IMF Working Paper 03/230 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Milesi-Ferretti, Gian Maria, and Assaf Razin, 1998, “Current Account Reversals and Currency Crises: Empirical Regularities,IMF Working Paper 98/89 (Washington, DC: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Park, Yung-Chul, and Jong-Wha Lee, 2002, “Financial Crisis and Recovery: Patterns of Adjustment in East Asia, 1996-99,ADB Institute Research Paper 45-2002.

    • Search Google Scholar
    • Export Citation
  • Sarel, Michael, 1995, “Growth in East Asia: What We Can and What We Cannot Infer From It,IMF Working Paper 95/98 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
1

Prepared by Martin Schindler.

2

Other variables were also experimented with, including a capital control variable and measures of fiscal volatility, none of which turned out significant in any of the regressions and therefore, in the interest of parsimony, were omitted from the final regressions.

3

No capital flows data were available for Korea. A money supply measure that may be more representative of actual liquidity conditions is M2a, which includes promissory notes in addition to M2. However, the BoT only provides this measure from 1994 onwards and cross-country availability of this measure is limited. Given that the analyses in this chapter utilize a sample with long time-series data for a large number of countries, the more standard M2-measure is used throughout. However, it is noted that for the 1994–2000 period, at a correlation coefficient of .97, the two time-series are highly correlated in Thailand, indicating that the results presented here are unlikely to be substantially affected by the choice of the money supply measure.

4

The Hodrick-Prescott filtered time series was created by first regressing the log of per-capita GDP on time and a crisis shift dummy, adjusting the actual time series by removing the trend shift, filtering the adjusted data by applying the HP-filter, and finally adding the trend shift back into the filtered data.

5

The apparent contradiction to the discussion in Section B is resolved by noting that Figure 2 compares post-crisis growth with growth in the 1990s, not with long-term growth as is done here.

6

An important qualification applies, however, due to the limited numbers of observation since the crisis, making it difficult to reliably estimate post-crisis trend growth.

7

The capital stock is constructed as Kt = (1–δ)·Kt-1 + It, where It is investment in year t and where a depreciation rate of δ = 0.05 is assumed. The initial capital stock K1960 is taken from Bosworth and Collins (2003). For α, a variety of parameters were used. A constrained regression of GDP on K and H (with the coefficients summing to one) indicates α = 0.4. Sarel (1997) uses an average value of 0.33, consistent with the standard value chosen in the macro literature. For robustness purposes, results for a value of 0.25 are reported as well. The human capital data are constructed as in Bosworth and Collins (2003), except that a 10 percent return on schooling is assumed (see Bils and Klenow, 2000).

8

This is easy to see from the capital accumulation function from the previous footnote, which can be rewritten as Kt/Kt-1 = 1 – δ + It-1/Kt-1. In a steady state, both the growth rate of K and the ratio I/GDP must be constant. This means one can express I = sGDP, where s > 0 is a constant. Consequently, GDPt/Kt-1 = (Kt/Kt-1 – (1 – δ))/s = constant, and so Kt/Kt-1 = GDPt/GDPt-1.

9

TFP growth in earlier periods differs more drastically across values α; for the whole period 1960–2004, average TFP growth ranges from 1.1 percent (α = 0.4) to 2 percent (α = 0.25).

10

However, it is worth emphasizing that these rates are based on the recent years’ TFP growth performance. A return to Thailand’s long-run TFP growth would lower the growth estimates to between 2.7 and 3.5 percent.

Thailand: Selected Issues
Author: International Monetary Fund