In this section, we analyze the social impact of financial crises and IMF adjustment programs in East Asia. Over the past 20 years, East Asian countries were remarkably successful in reducing poverty and achieving high employment growth. The present crisis, however, has jeopardized this hard-won reputation for economic performance that has provided better living standards for Asians and offered millions of people in other regions the hope of rescuing themselves from poverty. The financial meltdown in Asia, currently translating into rising social and political unrest, has resulted in more people being thrown out of employment and joining the ranks of the poor.

As the Asian Crisis is still unraveling, only limited information is currently available for assessing its social impact. Given this difficulty, this paper will focus only on its impact on employment, real wage, income distribution, and poverty. To examine how the financial crisis affects these social variables, we consult the past records of countries that experienced a currency crisis and received conditional financial assistance from the IMF. In order to accurately measure the impact of both the financial crisis and the IMF adjustment program, we have to evaluate the performance of program countries in comparison with the performance that would have prevailed in the absence of the crisis and adjustment program. In other words, we have to evaluate whether the IMF programs were associated with better or worse social outcomes than would otherwise have occurred. It is very difficult both conceptually and practically to identify the hypothetical reference point and to disentangle the effects of IMF programs from those of other factors. In this section, we first briefly discuss several methodologies for evaluating the effects of IMF programs. Then we will analyze the social impact of IMF programs in past program countries from 1973 to 1994. Based on these empirical results and specific East Asian characteristics, we will try to assess the social impact of the recent financial crisis and IMF programs in East Asia.

III. 1. Methodology to Evaluate the Programs in a Cross-country Framework

A number of previous studies have tried to assess the effects of Fund programs based on cross-sectional country data. The methodology in evaluating IMF programs can be classified into three categories: the "before-after" approach; the "control-group" approach; and the "modified control-group" approach.²² The first and most popular method is the "before-after" approach, which compares performance during a program with that prior to the program. It uses non-parametric statistical methods to evaluate whether there is a significant change in some essential variables over time. Therefore, while easy to employ and seemingly objective, this approach often gives biased results due to the assumption that had it not been for the program, the performance indicators would have taken their pre-crisis period values.

The "control-group" methodology attempts to overcome some of the limitations of the "before-after" approach. Here, the behavior in key variables in the program countries was compared to their behavior in non-program countries (a control group). Thus it implicitly assumes that only the imposition of the IMF program itself distinguishes the group of program countries from the control group. The external environment is assumed to affect program and no-program countries equally.

The third methodology is the so-called "modified control-group" approach, which consists of regressions that control for differences in initial conditions and policies undertaken in program and non-program countries. That is, this approach identifies the differences between program and non-program countries in the pre-program period, and then controls these differences statistically in order to find out the isolated impacts of the programs in the post-reform performance.

A substantial body of research has adopted one of these approaches to assess the impact of IMF programs. In particular, since the primary purpose of the IMF program is to assist the member country in restoring a sustainable balance of payments, reducing inflation and creating the conditions for sustainable income growth, most of the studies focused on evaluating how successfully these primary macroeconomic goals have been achieved (see Goldstein and Montiel (1986), Khan(1990), and Conway(1994)). However, little investigation has been conducted on the analysis of the social impact of IMF programs. A notable exception is a study by Garuda (1998) that conducts an extensive cross-country investigation into the distributional effects of IMF programs in 39 countries from 1975 to 1991.

III.2. Evaluation of the Social Impact of the IMF Programs, 1973-1994

We examine the social impact of IMF programs, using data of all developing countries that received stand-by and extended arrangements from mid-1973 to mid-1994. During this period, 88 non-OECD countries received financial support from the IMF at least once and the total number of programs amounted to 455.²³ In order to avoid "double counting" of economic crises or IMF programs, we pay special attention to the programs, which were continued from the previous year. That is, in our sample, a consecutive approval of programs or a program of more than one year in length is counted as only one program and is identified by the first year of the program. This procedure yields a total of 313 programs.

For each program in our sample, we estimate the social outcomes following the "before-after" approach, and then compare them to the average outcomes of non-program countries following the "control group" approach. We focus on two key social outcomes: employment and real wage on the one hand; poverty and income distribution on the other hand. The changes in these social variables are measured over the period of three years preceding, and one to five years following the approval of the IMF program. We also construct a control group of "tranquil" observations. If a country had not been subject to any IMF program within a window of plus/minus five years surrounding a specific year, it is counted as a non-program country in that specific year. We use all these observations as our control group of non-program "tranquil" observations. We have not tried to control statistically the differences between program and non-program countries as in the "modified control group" approach.

Changes in Employment and Real Wage

To analyze the effects of IMF programs on employment and real wage, we use data on manufacturing employment and wage growth rates available from The World Bank’s World Tables. The data were compiled for the period of 1968 to 1994 to examine the lagging effects of IMF programs. The data covers 1,306 observations for employment and 1,157 for real wage, of which a total of 138 and 126 observations respectively, correspond to IMF program years.²⁴
 
 

Figure 3.1 shows the changes in employment and real wage growth rates before and after the initiation of IMF programs. In panel (a) of Figure3.1, we plot the behavior of average employment growth rate at the onset of the IMF programs; during the three preceding years; and in each of the following five years. For comparison, we include the straight line in the panel, which indicates the average employment growth rate during the tranquil period that did not have an IMF program within a window of plus/minus five years.

We can clearly see that employment growth rates in program periods were significantly lower than those in non-program periods throughout the periods surrounding the initiation of programs. It is not hard to understand why the employment growth rate prior to the initiation of programs was lower than those in non-program periods: it indicates aggravated economic conditions prior to the crisis. But what is surprising is the fact that the employment growth rate did not recover its pre-crisis level even five years after the crisis. The average employment growth rate was 3.2 percent in the initial program year, which was essentially identical to the average of the three years before the program. As programs proceeded, the employment growth rate fell to 2.5 percent in the year following the program and then remained, after fluctuating for the next three years, at 2.4 percent in the fifth year. What caused the slow recovery of employment growth rates will be discussed after examining the behavior of real wage growth rates.

Panel (b) of Figure 3.1 portrays the change in real wage growth rates. The growth rates dropped a little in the program year and then further declined in the year following the crisis. But it increased substantially following the second year after the program. Since most IMF programs contained measures to restrain wage bills by such measures as wage freezes, reduced work hours, and cuts in fringe benefits, the initial drop in wage growth seemed inevitable (Sisson, 1986). The real wage growth in subsequent years is also consistent with the changes in output and inflation over the period of IMF involvement. Schadeler et al, (1995) shows that output growth rates declined in the program year and then subsequently recovered over the years following. The initiation of the programs was in general accompanied by lower inflation rates.²⁵ Higher output growth and lower inflation certainly contributed to higher real wage growth in subsequent years after IMF involvement.

Considering the strong surge of real wage and output growth, the relatively weak performance of employment growth after IMF programs is surprising. It implies that even after output growth rates, exchange rates, interest rates, etc. recover their pre-crisis level, one cannot expect the same recovery for employment. This fact, ironically, may be the result of labor productivity increases due to the adjustment program. After the crisis, program countries implement various structural reforms to enhance economic efficiency. Among them, increasing labor productivity by cutting over-employment is usually a primary objective. In other words, the reform has the same short-run effect as a laborsaving technology progress. Therefore, even after output demand returns to its pre-crisis level, labor demand is not fully recovered in the short-run. Only after positive externality from enhanced labor productivity is materialized can employment growth rates be significantly increased. In any case, the weak performance of employment growth indicates that unemployment rates can remain at a higher level for a long period after economic crises and IMF programs. This has a very important policy implication that we will discuss in section IV.2.

Changes in Income Distribution and Poverty

When we analyze distributional effects of IMF programs, the quality of cross-country data on income distribution raises a serious concern regarding the reliability of estimated results. A database recently constructed by Deininger and Squire (1996) considerably mitigates the data constraint faced in previous works. Deininger and Squire reviewed major studies on income distribution that had been conducted during the last 40 years and then constructed a fairly accurate and comparable data set across countries and time.

From their "high quality" database, we focus on two indicators of income distribution: Gini coefficients and the income share of the lowest quintiles. Our data set covers the period from 1968 to 1994 and consists of 322 observations of Gini coefficients and 274 observations of the lowest quintile income share for the sample of developing countries. Among total observations, 29 and 25 observations for Gini and the quintile income shares correspond to the IMF program years, respectively.
 
 

Figure 3.2 plots the behavior of Gini coefficients and the lowest quintile’s income share. Panel (a) of Figure 3.2 shows that countries initiating IMF programs experienced gradual deterioration of income distribution over the period of the initial program year and the following two years after the programs. The average Gini coefficients increased to 42.8 in the program year from 42.4 in the preceding three years and then further increased to 43.7 in the two years after the programs. However, over the longer-run period, income distribution showed substantial improvement. Gini coefficients dropped to 41.5 in the three years and to 39.9 in the five years following the programs. Hence, on average, income distribution improved well over the level in the pre-program years and approached close to the level of the non-program tranquil period.

The lowest quintile’s income share shows a similar pattern to those of Gini coefficients. Panel (b) of Figure 3.2 shows short-term deterioration and long-term improvement of income distribution in terms of the lowest quintile’s income share. However, there exists an important difference. The immediate adverse effects of IMF programs on income distribution are more visible in the quintile indicator. The lowest quintile’s share dropped on average to 5.72 percent at the initiation of the program from 6.16 in the preceding three years. Then it fluctuated for the next two years, eventually increasing to 6.12 in the five years following programs. Note that while in the five years after the crisis, long-run income distribution measured by Gini coefficients improved far more than the level in the pre-program years, the income share of the poor increased only close to the level in the pre-program years.

Although the estimated magnitude of the distributional impact of the IMF program may vary depending on sample countries, we think the pattern of short-term deterioration and long-term improvement of income distribution is quite a robust phenomenon.²⁶ The initial deterioration of income distribution can be attributed to government policy changes. The stabilization programs in general consist of contractionary monetary and fiscal policies and real exchange devaluation. Since these policy changes immediately lead to increases in bankruptcies, unemployment and the slow growth of real wage, there is likely to be a severe deterioration of income distribution. Poverty is aggravated as prices of items such as food, public transportation, and energy, which account for a large share of the consumption of low-income households, rise. In the long run, however, distribution started to improve when successful programs led to increases in foreign capital inflows, investment, and output growth.

There are other channels through which the IMF stabilization programs can affect income distribution. First of all, fiscal constraints have significant effects on income distribution and poverty through changes in both revenue and expenditure. IMF programs typically require the government to increase its revenues and/or decrease its outlays so as to reduce its overall deficit. Increase in taxes on income or imported luxury goods would influence income equality more favorably. The distributional effects of reduction in government expenditure depend on where the specific reductions are made. Workers in the public sectors as a whole tend to experience a decline in real wage or salary earnings with the downsizing of the public sector. Reduction of social expenditures - particularly subsidies to the poor, such as food subsidies - results in more perverse distributional effects (Sisson, 1986).

Monetary and credit policy also affects income distribution in various ways. A credit crunch and tight monetary policy hurts small and medium-sized firms more severely than large firms, which negatively impacts income equality (Johnson and Salop, 1980). Increasing real interest rates has an additional effect by redistributing income from borrowers to lenders, which is likely to render relative gain for households in the richest quintile, considering their interest-bearing asset holdings. Real depreciation of exchange rates causes a relative increase in the price of traded goods, leading to increases in the incomes of producers in the export and import-competing sectors.

III.3. Impact of the Asian Crisis on Unemployment and Real Wage

For the last two decades, Asian countries enjoyed virtually full employment prior to the crisis. As shown in Table 2.1, the unemployment rates in Indonesia, Thailand, and Korea were remarkably low, at less than 3-4% during the 1990s.²⁷ But their performance has drastically deteriorated since the crisis began. Bankruptcies due to credit crunches, contractionary fiscal and monetary policy, and the lift of legal restrictions on lay-offs, have contributed to a rapid increase in unemployment. Unemployment rates have been rising faster in these countries than in Mexico in 1994. According to the estimates reported in ILO (1998), the unemployment rate in Indonesia would reach 8 to 10% compared with about 5% in 1996. In Thailand, it is expected to increase from 1.54% in 1996 to 5.6% in 1998. In Korea, the unemployment rate had already risen drastically from 2.0% in October 1997 to 6.7% in April 1998. Given that Korea had not fully gone through economic restructuring yet, the unemployment rate is expected to go up even higher, to 7.5% by the end of the year. Since the extent of the crisis was so unexpected and drastic, there exists a pessimistic view that the recovery may not be as rapid as that of Mexico and Argentina following the tequila crisis in 1994. Moreover, the stylized pattern of employment changes discussed in section III.2 showed that unemployment rates are likely to remain high, if not higher, for a long time.

One important thing to note is that the crisis had diverse impact on unemployment across different groups. In Thailand and Indonesia, the wave of lay-offs affected urban white-collar workers the most (Tambunlertchai, 1998 and Azis, 1998). In general, however, the crisis hit marginal workers such as women; young workers; the less educated; recent school dropouts; and first-time job seekers, the hardest (Kim, 1998). Table 3.1 and Table 3.2 clearly exhibits this pattern in Korea. Table 3.1 shows the changes in employment by gender, age, and schooling. Between April 1997 and April 1998, employment declined by 3.8% among men, but by 7.1% among women. Young workers aged between 15 to 29 accounted for the lion’s share of job destruction, especially young female workers. Jobs for those with no high school diploma were destroyed by 11.1%, whereas the employment of college graduates increased by 7.2%. The increase is not surprising, because it reflects the deterioration of jobs. Displaced college graduates are settling for jobs that used to be taken by high school graduates. We also see that employment of older workers, particularly older women who were more likely to be forced to accept early retirement, declined more compared with primary workers. This pattern is consistent with the internal labor market hypothesis that marginal workers - young, female, less experienced, less-educated workers - rather than primary workers are more likely to bear the burden of adjustment to external shocks. Its policy implications will be discussed in section IV.
 
 

(units: thousand, %)
 

Age	April 1996	April 1997	April 1998	D 96/97 (%)	D 97/98 (%)
Total
All	20,743	21,219	20,127	476 (2.3)	-1,092 (-5.1)
15/19	394	398	335	4 (1.0)	-63 (-15.8)
20/29	4,775	4,811	4,162	36 (0.8)	-649 (-13.5)
30/39	6,100	6,007	5,915	-93 (-1.5)	-92 (-1.5)
40/49	4,621	4,825	4,802	204 (4.4)	-23 (-0.5)
50/59	3,000	3,161	2,973	161 (5.4)	-188 (-5.9)
60+	1,852	2,017	1,939	165 (8.9)	-78 (-3.9)
Men
All	12,349	12,446	11,976	97 (0.8)	-470 (-3.8)
15/19	151	150	137	-1 (-0.7)	-13 (-8.7)
20/29	2,528	2,513	2,178	-15 (-0.6)	-335 (-13.3)
30/39	3,969	3,867	3,841	-102 (-2.6)	-26 (-0.7)
40/49	2,836	2,893	2,912	57 (2.0)	19 (0.7)
50/59	1,819	1,910	1,805	91 (5.0)	-105 (-5.5)
60+	1,045	1,112	1,103	67 (6.4)	-9 (-0.8)
Women
All	8,395	8,773	8,151	378 (4.5)	-622 (-7.1)
15/19	243	248	198	5 (2.1)	-50 (-20.2)
20/29	2,248	2,299	1,985	51 (2.3)	-314 (-13.7)
30/39	2,131	2,139	2,074	8 (0.4)	-65 (-3.0)
40/49	1,784	1,933	1,890	149 (8.4)	-43 (-2.2)
50/59	1,181	1,251	1,169	70 (5.9)	-82 (-6.6)
60+	807	904	836	97 (12.0)	-68 (-7.5)
Schooling
No HS Diploma	7,637	7,715	6,870*	78 (1.2)	-845 (-11.1)
HS Diploma	9,009	9,163	8,582*	154 (1.6)	-581 (-6.2)
College Diploma	4,098	4,341	4,675*	243 (6.4)	334 (7.2)

Note: *; projected number
Source: National Statistical Office, Korea, The Economically Active Population Survey, Cited from Kim (1998).
 
 

(units: thousand)

Source: National Statistical Office, Korea, The Economically Active Population Survey.
 

Table 3.2 examines the changes in employment by industry, occupation and work hours. It shows there have been substantial retrenchments, especially in manufacturing and construction industries. To a lesser degree, employment in retail and service sectors decreased, while the agricultural and fishery industries gained in employment. This implies that displaced workers and unsuccessful job seekers in the primary sector are involuntarily settling for inferior employment in the rural or the urban informal sector. No doubt this trend will increase underemployment. Underemployment will also rise due to the fact that unpaid family workers and part-time workers gained employment whereas regular workers lost it. The influx of displaced workers into the rural or the urban informal sectors and the decline of regular jobs will reduce the already low average income in those sectors even more, and are likely to increase the number of people below the poverty level.
 
 

(units: %)
 

Source: National Statistical Office, Korea, The Economically Active Population Survey.
 
 

(units: thousand, %)
 

Source: National Statistical Office, Korea, The Economically Active Population Survey.
 

Table 3.3 shows the changes in participation rates by gender and age. Between 1997 and 1998, participation rates declined by 0.5 percent among men but by 2.8% among women. Age differences do not seem to exist even though the decline is slightly more visible among older workers. Considering the extent of gender discrimination in the Korean labor market, it is no wonder that participation rates among female workers, who were more likely to be second-income earners in a family, dropped significantly more than among male workers. Table 3.4 reports the distribution of unemployment and unemployment rates. We can see that the unemployment rates of young workers (15-29 years old) are the highest and that they account for 42% of total unemployment in April 1998. But it is important to note that primary workers, not just marginally attached workers, are also losing jobs on a large scale, indicating the severity of the crisis. In terms of growth rates, unemployment rates increased faster for primary workers. For example, unemployment rates of workers aged between 40 to 49 tripled from 1.7% to 5.4% within a year.
 
 

Note: percentage change compared with the same period in the previous year.
Source: Korea Development Institute, Monthly Economic Outlook.
 

Table 3.5 reports the changes in real wage in Korea. It is noteworthy that the growth rate of nominal wage, which used to be about 10% per year, dropped sharply after the crisis. In the first quarter of 1998, nominal wage did not increase at all. On the other hand, the inflation rate increased significantly following the substantial currency devaluation. As a result, real wage decreased by 8.9% in the first quarter of 1998. In section III.2 we see that the growth rates of real wage should recover soon after the sharp initial fall. At this moment, it is premature to tell whether real wage in Korea will follow this general trend. The freeze in nominal wage that Korea achieved in the first quarter of this year was not only due to the decline in labor demand after the crisis. It was mainly a temporal outcome negotiated in the Tripartite Committee, which consists of representatives from the government, workers’ and employers’ organizations. Whether the Tripartite Committee can fully accomplish its mission is very uncertain. As the restructuring goes on and mass lay-offs begin, it is likely that labor unions will protest against their unfair suffering. Then labor strikes and nominal wage hikes will follow.

III.4. Impact of the Asian Crisis on Income Distribution and Poverty

The rapid economic growth in East Asia significantly reduced the number of people living under the absolute poverty line. However, even before the crisis began, there had been widespread concern that the accelerating trend towards globalization in the 1990s could exacerbate the prevailing income distribution. This concern is now being reinforced. The current crisis may reverse the trend of equitable distribution in the region. In this section, we provide a summary of trends in income distribution in the three worst-affected Asian countries and then discuss the impact of the present crisis on their income distribution.

Trends in Inequality and Poverty

To see the changes in income inequality from before the crisis, we look at the data on Gini coefficients and the quintile shares of total national income. Though methods of collection, degree of coverage, and specific definitions of personal income may vary among counties, Tables 3.6 and 3.7 depict a general trend of income inequality in Indonesia, Korea and Thailand.
 
 

Note: 1) Inc = Whether the Gini coefficient is calculated based on income(I) or expenditure (E)
2) Pers = Whether the recipient unit is the person (P) or the household(H).
3) Gross = Whether the income reported is gross(G) or net of taxes(N)
Source: Deininger and Squire (1996); and for the Korean data of 1993 and 1996, National Statistical Office, Social Indicators of Korea 1997.
 
 

Note: See notes to Table 4.1.
Source: Deininger and Squire (1996).

Table 3.6 shows that Indonesia made steady progress in reducing income inequality during the past two decades. Gini coefficients increased a little in the 1970s, reaching a peak in 1978. From then on, they declined consistently until 1993, the year up to which data is available. In the Republic of Korea, Gini coefficients showed an increasing trend from 29.8 in 1969 to 39.1 in 1976, and then continued to drop to 29.5 in 1996. Hence, according to the data, Indonesia and Korea have succeeded at least in preventing serious deterioration in income distribution over the last three decades. In contrast to the good performance of Indonesia and Korea, Thailand experienced a persistent deterioration in income distribution despite high-income growth. Gini coefficients steadily increased from 41.7 in 1975 to 51.5 in 1992. The share of income of the lowest quintile decreased from 0.049 to 0.037 during the same period. According to the UN (1998), the deterioration of income distribution can be attributed to a widening income differential between the urban and rural poor.

Table 3.7 presents cross-country comparisons of income distribution. In general, countries in Asia appear to be more egalitarian than those in Latin America. The relationship between growth and equity is not clear. Countries such as Taiwan and Korea have successfully combined the reduction of inequality with high-income growth. The superior performance of these countries is in contrast to that of countries such as Hong Kong, Mexico and Malaysia, all of which had high economic growth rates but failed to reduce income inequality.
 
 

 
 
 
 
 

	Headcount Index(percentage of the poor) (number of poor, millions)
Indonesia	1976	1981	1987	1993	1996
Total:   Urban   Rural	40.1 (54.3) 38.8 (10.0) 40.3 (44.2)	26.9 (40.6) 28.1 (9.3) 26.4 (31.3)	17.4 (30.3) 20.1 (9.7) 16.1 (20.8)	13.7 (25.9) 13.4 (8.7) 13.8 (17.2)	11.3 (22.5) 9.7 (7.2) 12.3 (15.3)
Thailand	1975	1981	1988	1992	1996
Total	30 (12.4)	23 (11.0)	22 (11.9)	13 (7.5)	-
Korea	1975	1980	1985	1990	1995
Urban	20.0	14.4	14.2	10.5	7.4

Note: Poverty estimates are based on country-specific poverty lines.
Source: Data for Indonesia are from Statistical Yearbook of Indonesia; for Korea, Chanyong Park and Meesook Kim, Current Poverty Issues and Counter Policies, Korean Institute for Health and Social Affairs, recited from ILO(1998); for Thailand, UN(1998).

In addition to income distribution and relative poverty, another important issue focuses on the extent and magnitude of absolute poverty. Until the recent crisis, all three countries enjoyed improving living standards as the population living in poverty fell substantially. Table 3.8 demonstrates that all three Asian countries reduced poverty as a result of remarkable growth rates. In Korea for instance, the level of absolute poverty decreased from 21.5% in 1975 to 8.5% in 1995.²⁸ However, despite the impressive success of these countries in reducing income distribution and poverty during the past two decades, a substantial body of the population still lives below the poverty line, particularly in the rural areas of Indonesia and Thailand. Some 22 million Indonesian people were still living below the officially defined poverty line in 1996. The poor in Indonesia are predominantly located in the rural and agricultural sectors. Similarly in Thailand, though the absolute population living below the poverty line continued to decline, poverty is much higher in the rural areas, particularly among less-educated households, agricultural workers, and large families.
 

Distributional Impact of the Asian Crisis

Although we do not have precise statistics or information on the evolution of poverty and income distribution at this stage, the current economic crisis is considered to have already had significant adverse effects on equitable growth in this region.

The immediate impacts of economic crises and IMF programs on income distribution were increases in unemployment and inflation. The increases in unemployment and underemployment directly aggravated poverty. The total number of unemployed has increased in unprecedented numbers in this region and will continue to pile up. The newly unemployed are obviously suffering a drastic drop in income and living standards. Loss of jobs or reallocation to low-wage occupations has led to a sizeable increase in the number of people living below the poverty line. Hence, the decline in job opportunities has definitely contributed to increased poverty.

Price increases lowered the real wages of those still employed, exacerbating poverty even more. Average annual inflation reached 44 percent in Indonesia in May 1998, and around 11 percent in Korea and Thailand. Because nominal wages did not adjust to offset the effect of price increases and social income compensation from social safety nets was minimal, real income of a typical household declined almost by the full extent of the price increases. The price increases of specific commodities also had a great impact on household consumption. They had a differentiated impact on households, depending on the shares of food and necessary items in a household’s consumption basket. Because the price increases concentrated on the items that account for a large share of the consumption of low-income households, they further adversely affected income distribution. For example, food constitutes 70 percent of the total expenditure of the households in the lowest income in Indonesia and 55 percent in Thailand. The corresponding expenditure shares for the top decile households are 35 and 21 percent, respectively (Gupta, et al, 1998). Thus, price increases in these countries would have more significant adverse impacts on the consumption of the poor.

In addition to the severe adverse effects of rising unemployment and inflation, poverty could be further aggravated by "social income poverty" (Ranis and Stewart, 1998). During a crisis, higher prices and fewer employment opportunities deprive people of primary (or private) income. Moreover, a crisis reduces secondary (or social) income from the state via public works or income transfers (e.g. unemployment benefits). Although it is not clear whether total government expenditure itself was reduced after the crisis, social expenditure for education, public health, and social services was negatively affected. In Thailand for example, the government budget in 1997 was reduced by 32%, 15%, and 11% for social services, public health, and education sectors, respectively, after the crisis (Siamwalla and Sobchokchai, 1998). The cuts in social expenditure have had an immediate, adverse effect on social incomes of the poor, and will also have long-run consequences on the private incomes of all economic agents. Since it is expenditure for education and health care that has a significant effect on human capital formation, the cuts in these social sectors can hurt long-run growth potential and prolong the adverse poverty situation over a long period.

The combined effects of higher price increases, job losses and reduced social expenditures indicate that the crisis will have a deep adverse effect on (absolute and relative) poverty in these Asian economies. However, its impact on overall income inequality is rather ambiguous. The impact of job losses on income inequality is hard to predict: it depends on the composition of job losses. If the crisis hurts urban middle class workers more severely than those in the upper and bottom quintiles, how the Gini coefficients will change is not clear.

Moreover, not all population groups lose from the crisis. Some households will gain from exchange rate depreciation. Incomes of those engaging in export and tourism sectors can improve. The sharp increase in interest rates can benefit those holding a larger stock of interest-bearing assets. Diverse impacts of the crisis on income distribution imply that the increase of Gini coefficients during the crisis will be marginal. The cross-country evidence of Section III.2 confirms this prediction. It shows that income inequality tended to increase immediately after the IMF programs but that the degree of deterioration was not substantial. Individual country experience also supports the prediction that the crisis aggravates poverty significantly, but that the change in overall income distribution is relatively small. For instance, according to Hernandez and Mayer(1998), the Gini coefficient in Chile worsened only marginally during the 1982-83 economic crisis (from about 52 in 1980-81 to about 55 in 1982-83) even though poverty indices deteriorated significantly. The share of population below the poverty line increased from 33 percent in 1981 to about 58 percent in 1983. This finding has an important policy implication for building a social safety net during a crisis. In view of the significant deterioration of poverty and the minimal rise in overall income inequality, welfare policy should be targeted to the core group of the poorest and hardest-hit victims instead of trying to maximize the number of beneficiaries.

In sum, although the short-term deterioration of poverty and income distribution is inevitable, the longer-run impact of the crisis on income distribution is less clear. It surely depends on the nature and implementation of government policy in handling the crisis. The cross-country evidence in section III.2 shows the possibility of income distribution improving with the recovery of economic growth in the long run. However, without adequate government policies, we cannot expect the level of income equality to soon recover to what it was before the crisis in Asia.

22. For discussions on the methodology of evaluating the effects of the IMF programs, see Khan(1990), Killick(1995), Killick and Malik(1993), Killick, Malik, and Manuel(1993), and Corbo and Fisher(1995).

23.The 455 programs approved during the sample period consist of 345 stand-by arrangements,  42 extended fund facility (EFF) arrangements, and 44 arrangements under structural adjustment facility (SAF) or enhanced structural adjustment facility (ESAF). The remaining 21 cases were combined programs of more than two arrangements.

24.We have excluded some extreme observations such that annual growth rate of employment or real wage is  higher than  50 percent  or lower than –50 percent. The results are basically identical when these observations are included.

25. The stabilization effects of IMF programs appear depending on exchange rate regime: in countries with nominal exchange anchors, which could add a strong credibility to the stabilization packages, inflation  rates fell dramatically from the first program year.  See Shadeler et al, (1995, part I) for more details.

26. Garuda (1998) claims that distributional effects of IMF programs may depend on a country’s pre-program economic situation.  He finds evidence of a significant relative improvement in income distribution in the program countries in which external imbalance prior to the program initiation is not severe,  while countries that experienced the most severe pre-program external problems showed deterioration in income distribution relative to non-program countries with equally severe conditions.  We have not tried the same experiment because the sample size becomes too small with the further classification of data.

27. On the labor market front during the pre-crisis period, the performance of Indonesia had been less impressive among these countries. But it is widely pointed out that the underemployment problem is most severe in Thailand and therefore, her unemployment rate is significantly underestimated.

28. See Whang and Lee (1997) for detailed analysis of changes in income distribution and poverty in Korea.