An analysis of the nature of unemployment in Sri Lanka

THE WORLD BANK Internal Discussion Paper AsIA REGIONAL SERIES Report No. IDP 42 An Analysis of the Nature of Unemployment in Sri Lanka William T. Dickens and Kevin Lang July 1989 The views presented here are those of the author, and they should not be interpreted as reflecting those of the World Bank. An Analysis of the Nature of Unemployment in Sri Lanka by William T. Dickens U. C. Berkeley and NBER and Kevin Lang Boston University and NBER prepared for the World Bank AN ANALYSIS OF THE NATURE OF UNEMPLOYMENT IN SRI LANKA William T. Dickens and Kevin Lang Abstract Since the 1950's, Sri Lanka has had a significant chronic unemployment problem. Depending on how unemployment is defined and what time period is being examined, it varies from the low teens to over twenty percent of the labor force. This rate of unemployment is two to three times higher than what is normally found in other developing countries. Most of Sri Lanka's unemployment is concentrated among young people who are looking for their first job. Unemployment duration is very long with typical spells lasting four years or more. The purpose of this paper is to investigate the nature of the unemployment problem in Sri Lanka, which policies could best address it and which are the areas for further research. One conclusion of this paper is that, contrary to a generally held view, unemployment in Sri Lanka is not a problem of the educated youth, but a problem of the youth. While unemployment rates increase with the level of education, those who are "highly educated' are a relatively small proportion of the unemployed. Only 9Z of those unemployed in 1985-86 had A level education or more, and about one-third had 0 level education or more. A second conclusion of this paper is that the nature of and the reason for unemployment vary accordin2 to the level of education. The more educated (i.e. those with college education A or 0 levels) are queuing for Government jobs which pay above market wages (except in the case of the highest grades which represent a relatively small proportion of total employment in Government), provide job security and pension after retirement. The reason why that the unemployed do not take other employment while waiting is that, in hiring, the Government gives preference to the unemployed. Among the less educated there is queuing for high wage private sector jobs, but the reason why they do not take other employment while queuing needs to be investigated further. One reason may be that many remain unemployed while studying for 0- level examinations--even though they are finished with formal schooling--to increase their chances of being employed in the Government since few in Government have less than O-level qualifications. On the other hand, the Government employs a very large friction of the educated population including over half of those with 0 or A level education and nearly three-fourths of those with college degrees or more. Another reason may be that the private sector hiring policies mirror government policies in favoring the unemployed. Having identified the existence of attractive government jobs as one of the major reasons for unemployment in Sri Lanka, the paper first considers how changes to government hiring and personnel policies might affect the level of unemployment. Next, it reviews how changes in industriaL and trade policies, as well as changes in development priorities (such as emphasis in the development of agriculture -- the lowest wage sector in the economy) could also contribute, in the longer term, to reduce unemployment by creating new high-wage earning opportunities. Finally, it suggests that current government regulations of employment in the private sector are counter-productive as far as employment is concerned, and a hindrance to the sector's development. AN ANALYSIS OF THE NATURE OF UNEMPLOYMENT IN SRI LANKA Table of Contents Page No. I. Introductdn.................................................. 1 II. The StrLcture of Employment in the Rural. Urban and Estate Sectors................................................ 2 Self-Employment............................................... 5 Government Sector Employment.................................. 7 Earnings...................................................... 9 III. The Structure of Unemployment..................... ........... 15 Unemployment Rates by Sex and Sector.......................... 16 Unemployment Rates by Age..................................... 16 Unemployment and Self-Employment.............................. 17 Unemployment Rates by Education: A First Look................ 19 Unemployment Rates by Education: A Second Look............... 20 Unemployment Duration......................................... 23 Summary and Conclusions....................................... 25 IV. The Unemployment Proble...................................... 25 Other Explanations............................................ 25 Unemployment Among the Educated............................... 29 Unemployment Among the Less Educated.......................... 32 Further Research.............................................. 34 V. Policy Proposals.............................................. 36 Government Personnel Practices and Employment Programs........ 36 Industrial azid Trade Policies................................. 41 Further Results............................................... 45 VI. Conclusions................................................... 47 LIST OF TABLES 1 The Distribution of Employment by Sector and Industry......... 52 2 Distribution cf Manufacturing Employment in Sri Lanka......... 53 3 Self-Employment and Unpaid Family Workers as a Percent of Labor Force, Population and Employment by Age.............. 54 4 Public and Corporation Sector Employment by Age and Education..................................................... 55 5 Median and Mean Earnings by Class of Worker and Sector........ 56 6 Unemployment Rates in Sri Lanka by Sex and Sector............. 57 7 Unemployment Rates in Sri Lanka by Age and Sector............. 58 8 Unemployment and Self-Employment Rates, Selected Cohorts, 1971-1986..................................................... 59 Page No. LIST OF TABLES (cont...) 9 Unemployment Rates in Sri Lanka by Age and Education 1969/70, 1973, 1985/86........................................ 60 10a Unemployment Rates by Age and Education (Urban Males)......... 61 10b Unemployment Rates by Age and Education (Rural Males)......... 61 10c Unemployment Rates by Age and Education (Urban Women)......... 62 10d Unemployment Rates by Age and Education (Rural Women)......... 62 11 Predicted Unemployment Duratio by Education (Rural youth, 15-24), 1973.................................... 63 I. Introduction If there is unemployment, wages must exceed their market- clearing level in at least some sectors. Either because it is not profitable for employers to lower wages or because social or governmental pressures prevent them from doing so, wages do not fall to elimiaate the oversupply of labor. However, for this to be the sole explanation for unemployment, it must be that wages do not fall in any sector. If at least some sectors appear to have flexible wages, unemployment occurs because some workers prefer to queue for jobs in which wages are relatively high. In this case, there are three broad policies which may be pursued to lower unemployment rates depending on the specific cause. These are aggregate demand policies, wage reductions in high wage sectors, or policies aimed at increasing the cost of queuing for high wage jobs. In this report, we analyze the labor market in Sri Lanka and argue tiat the clearest benefits are to reducing the benefits associated with queuing. There may also be some gaih from reducing excess compensation in certain government jobs. We argue that there is widespread queue unemployment. Moreover, in contrast to the prevailing view in the literature, this queue unemployment is not particularly concentrated among educated workers. There is disagreement in the literature as to whether government employment continues to be "overcompensated." We find that in the lower paying grades which make up the vast majority of.public sector employment, compensation continues to -2- exceed the "market-clearing" level. In addition, while queuing workers aspire to high paying government jobs, they also aspire to high paying jobs in the private sector. Regular jobs in the private sector pay significantly more than casual jobs although they comprise only a small fraction of employment. The next section presents an overview of the structure of employment. Since open unemployment among prime-age workers is rare in developing countries, we examine levels of self- employment, a frequent area for disguised unemployment. We also look at the importance of the public sector in providing jobs, especially for educated youth. We then examine wage rates in different sectors of the economy and argue that there are large wage differentials which encourage queuing. We then examine the structure of unemployment in Sri Lanka. Finally, we present an analysis of various policies designed to reduce unemployment. While we conclude that unemployment arises in Sri Lanka for reasons which are similar to those in the classical Harris-Todaro model -- some sectors pay higher wages than others --, our suggested policies are nt aimed primarily at removing those wage differentials which may exist for sound economic reasons. Instead we concentrate on proposals designed to reduce the desirability of being unemployed while waiting for employment in the high wage sector. II. The StructUre of Employment in the Rural, Urban and Estate Sectors -3- In Sri Lanka an important distinction is made between the rural and estate sectors.1 The estate set-tor accounts for 8.5% of the employed population and is comprised of relatively large landholdings, formerly under foreign control and now mostly government owned. Not surprisingly, almost all employment in the estate sector is in agriculture (see Table 1). In addition, education levels are low relative to the rest of the island; the median worker has not completed grade four2 and over a quarter have no schooling at all. Unemployment rates, particularly for women, are lower than elsewhere on the island while female labor force participation (67%) is nearly equal to that of men (79%) and is much higher than for the island as a whole (32%). The rural sector accounts for nearly three-quarters (72.4%) of employment. While the majority of employment is in agriculture, there is significant employment in both rural industry and services (table 1). The median worker appears to have passed roughly grade 6, but about 10% obtained no schooling. About one-fifth (19%) of workers are employed in the urban sector. Not surprisingly, the bulk of employment is in services and to a lesser extent industry (see Table 1), but a significant minority of workers in the urban sector are engaged primarily in agriculture (11%). Levels of education are relatively high; the IThe results reported in this subsection are drawn from the Labor Force and Socio-Economic Survey, 1985/86, unless otherwise noted. 2 and probably not grade three but the data are insufficient to ascertain the precise median. -4- median worker has passed grade 8 or 9 and only 7% have no education. There are two striking findings to which we will return in the course of our analysis. First, as many previous researchers have remarked, the country has remained surprisingly rural. This reflects a deliberate government policy to foster rural development by opening new lands for settlement to relieve rural population pressures and by providing high levels of public services to the rural sector. Second, the fraction of employment in industry seems low. Manufacturing employs only 12% of workers, another 4% are employed in construction with the rest of employment in "industry" being made up of workers in mining and quarrying utilities. Within manufacturing, 36% of employment is in food, beverages and tobacco, 25% in textiles, wearing apparel, and leather products, 12% in chemicals, petroleum, rubber and plastics, and 10% in non-metallic mineral products (Table 2). These figures include both privately and publicly owned manufacturing concerns. Government and semi-government employment accounted for 22% of manufacturing in 1981. Private sector "employees" accounted for less than two-thirds of manufacturing employment with the rest made up of self-employed workers, employers and unpaid family workers (Labor Statistics, 1986, p. 17). -5- Within manufacturing, a large fraction of workers (30%) are employed in what are commonly referred to as informal sector establishments, defined as those employing fewer than five workers. In addition, an undetermined fraction of manufacturing employees are casual rather than regular employees. The Labor Force and Socio-Economic Survey, 1985/86 estimates that there were about 1.3 million regular employees in Sri Lanka, accounting for a little over one-fourth of all employment. We estimate from the Census of Public Sector and Corporation Sector Employment, 1985 that just over half of these were in the government sector so that regular private sector employment accounted for less than 13% of employment and only about 11% of the labor force. Deriniyagala et al (1976) estimate that in 1971 only about 1% of the labor force had regular employment in the private modern sector. We are unable to obtain a precise estimate; however, it is unlikely that more than about five percent of workers had regular employment in private modern manufacturing establishments. Self-employment In order to cast light on the nature and magnitude of unemployment in Sri Lanka, we must first examine the level and nature of self-employment. In developing countries high levels of unemployment are likely to induce disguised unemployment in the form of informal sector employment or self-employment rather -6- than open unemployment. We have already noted that a substantial fraction of manufacturing employment is in the informal sector. Since unemployment rates decline rapidly between the twenty to twenty-four and the twenty-five to twenty-nine age groups, an obvious explanation is that as individuals reach the age when they feel obliged to support their own households, they lower their expectations and accept whatever job is available to them. Table 3 reports the fraction of self-employment and employment as an unpaid family worker for different age groups. There is a steady increase with age in the fraction of the population which is self-employed and a steady decline in the fraction engaged in unpaid family labor. Moreover, there is a big jump in self-employment and a somewhat smaller but notable decline in unpaid family work as we move from the nineteen to twenty-five category to the twenty-six to thirty-five category. This is consistent with the view that as they enter the stage of family formation, workers seek out acceptable employment rather than waiting for the most desirable jobs. However, we show in our discussion of unemployment, below (section III, Unemployment and Self-Employment), that the timing of increases in self- employment and decreases in unemployment within a cohort appear to be quite different. This finding casts doubt on the view that self-employment arises because unemployed workers reduce their aspirations. -7- Government sector employment Employment in the public (roughly central government) and corporation sectors (roughly independent public corporations including but not limited to public controlled enterprises) accounts for about one-sixth of employment in the rural and urban sectors. As shown in Table 4, government sector3 employment is negligible for workers under age twenty, since government policy prohibits hiring individuals below that age.4 For those over twenty, government sector employment rises rapidly to account for one-fifth of exployment for those in their late 20s and one- quarter of employment for those age 30 to 44. Government sector employment becomes small again for those past the retirement age of 55. Table 4 also shows the importance of government sector employment for the highly educated. While only 2% of workers who did not complete grade five and only 12% of those with some secondary education are employed in the government sector, over half of O-level employment and nearly three-quarters of employment for those with degrees is in the government sector. Moreover, employment of the most qualified segment of the work force is highly concentrated in the central government rather 3The term government sector employment is often used to refer to central government employment. We use it here to include both public sector and publicly-owned corporations (the corporation sector). It excludes employment on the estates. 4Report on Consumer Finances and Socio Economic Survey 1981/82, part I, p. 126. -8- than public corporations. Over half of the labor force with degrees are employed in the central government. Nearly three- fourths hold some form of government employment. Because of the importance of the government sector as an employer of highly-educated labor, the question naturally arises as to whether the education standards used in allocating government sector jobs result in government sector workers being over-qualified for the positions they fill. We can bring only indirect evidence to bear on this question. In the public sector, roughly 65,000 workers are employed in either agriculture or unskilled labor, a further 82,000 are employed as skilled workers and 70,000 as clerical workers (Census of Public Sector and Corporation Sector Employment 1985, pp viii-xiii). In contrast, 10,000 workers have not passed grade 5, 85,000 have passed grade 5 but not O-levels and 211,000 have passed O-levels. It appears to us that there may be room for reducing the uwa of educational qualifications in allocating certain public sectors jobs, but a more certain statement would require more detailed information. In public corporations, 97,000 workers are employed either in agriculture or as unskilled workers. A further 114,000 are employed as skilled workers and 84,000 as clerical workers. This contrasts with 22,000 who have not passed grade 5, 129,000 with passes at grades 5 - 9 and 125,000 who have passed O-levels. Although only circumstantial, we see somewhat stronger evidence -9- of the excessive use of educational qualifications here. Earnings We address three issues regarding compensation in Sri Lanka. The first is whether self-employment represents disguised unemployment. In this case we would expect self-employed workers to have relatively low earnings. The second is whether unemployment arises because of wage differences among different sectors of the economy. The third is whether there is a large private return to education. We address these issues in turn. Self-employed workers seem to have somewhat higher earnings, on average, than do employees. Glewwe (1988) reports that in 1981/82 per capita expenditure on food was slightly higher among the self-employed than among employees. The mean is somewhat misleading. Table 5 reports both the median and mean incomes of different classes of worker by sector of the economy. Except in the estate set,tor, where the self-employed constitute less than 2% of employment, the median self-employed worker earns less than the median regular employee despite the fact that the mean self-employed worker earns more than the mean regular worker. Thus there is evidence that the high mean earnings of self-employed workers reflects high earnings among a relatively small fraction of them. We suspect that a substantial fraction of self-employed workers earn significant returns on their land and/or capital but that those with little productive wealth have -10 - low earnings. Typically, self-reported income among the self- employed includes the return to their capital as well as to their labor. An individual who farms his own land will therefore tend to have "high" labor earnings which reflect the rental value of the land. Perhaps the most significant aspect of table 5 is the high earnings of the self-employed relative to casual employees. Since, as we have already noted, most regular employees are in the government sector, the choice for most workers appears to be between casual employment and self-employment. Moreover, since government employs a large fraction of the educated labor force, it appears to us that the self-employed are probably drawn disproportionately from the better-off workers among the relatively less-educated. Under these circumstances we suspect that straightforward comparisons of the earnings of different classes of workers are not informative about "true" wage differentials. We do not know what the same person would earn in the different sectors. However, without careful statistical analysis, with multiple controls, we cannot be sure. We note in passing that table 5 also underscores the large disparities in earnings among sectors. It appears that social ard legal restrictions have prevented workers (largely Indian Hindus) from leaving the estates to seek more remunerative employment elsewhere. Government policy aimed at providing a high level of services in rural areas also appears to have -11- limited migration from rural to urban areas. In a sample of 7927 households, only four people said they had moved to an urban area for employment reasons (Report of Consumer Finances ..., 1981/82, part 1, p. 55). While once it might have been possible to discuss a government/nongovernment wage differential, the relative salaries earnings of low skill and senior administrative government employees have moved in such different ways in recent years that it is necessary to consider them separately. It appears that lower-ranked government employees are well paid relative to their counterparts in the private sector. In the only occupation for which we could get a direct comparison, in 1984 starting salaries for unskilled engineers were 13% above the private sector minimum. Semi-skilled engineers had salaries 14% above private sector minimums and skilled engineers 13% (Labor Statistics, 1986). While it is possible that the private sector minimum was not binding, casual evidence suggests that it probably was for at least the unskilled group. The private sector average was only 8% above thca government sector starting salary for unskilled engineers. The comparable figures are 22% for semi-skilled engineers and over 50% for the skilled group. This pattern is found more generally. In central government, workers in manual and clerical and allied grades appear to earn more than their private sector counterparts while those in higher grades appear to earn less. There is some possibility that the -12- earnings/seniority profile is less steep in government than in the modern private sector. We are unfortunately not in a position to verify this. It is probably the case that most central government workers earn more that their private sector counterparts. Almost all government workers (97%) are in the lowest 25% of the salary grades. In addition, central government salaries are not taxed and central government workers receive pensions. Finally, the prestige associated with working for the government is generally quite high. .Ln sum, high salaries for most government workers appear to be sufficient to create a queue for government employment. The major exception is for professionals and high level administrators who appear to be significantly underpaid. This suggests a need to increase the salaries of those workers in the highest salary classes. However, as pointed out above, there may be severe constraints on government's ability to use salary to attract more qualified workers since government already employs a large majority of the highly educated work force and may have a near monopoly in certain specialized areas. An important issue is whether the short supply of highly educated worker results from education being unattractive or from restrictions on the number of students accepted. There appears to be no shortage of applicants for A-level and university educations. At first blush, the value of a university degree is confirmed by the 1981/82 Socio Economic Survey (p201). Among -13- those who had passed their degrees the median workers earned about 60% more than those who had only passed A-levels. However, those who had passed A-levels actually earned less than those who had passed O-levels and the return to a degree relative to 0- levels was less than 50%. Although this return seems low for a developing country, it is greater than or equal to returns typically observed in developed countries. Further, since nearly three-fourths of all workers with degrees are employed by the government, and since there are many benefits associated with government employment beyond the wage, this 50% figure probably substantially understates the total return -- pecuniary and non- pecuniary -- of obtaining a college education. Again, there is no shortage of people willing to take the time to study for a degree. It is worth noting that the return to a university degree has dropped sharply in recent years. In 1973, the median worker with a degree earned almost two and a half times as much as the median worker who had passed O-levels (Socioeco.,iomic Survey, 1981/2, p. 201). By 1978/79 this figure had fallen to less than 50%. The primary problem with assessing the structure of compensation in Sri Lanka is the lack of statistical analysis with multiple control variables. Government workers have different levels of education than nongovernment workers. The more educated.are more likely to be in the government sector, -14- etc. These overlapping differences make it difficult to assess the "true" effects of critical variables -- what an individual would earn in the different sectors. With this caveat in mind, we suggest the following conclusions to the questions posed at the beginning of this section. We see little evidence that self-employment represents disguised unemployment. While the median self-employed worker earns less than the median regular employee, we suspect that this difference is accounted for by differences in education between the two groups. On the other hand, the self-employed appear to earn more than casual employees. Unemployment could arise either because individuals queue for high wage (primarily government) jobs or because individuals are unemployed while trying to raise their educational credentials. In fact, it appears that there is a substantial premium for most government employment which could account for much unemployment. On the other hand, it appears the return to high levels of education may have dropped significantly in recent years. However, this has apparently not been sufficient to eliminate queues for higher levels of education. These conclusions -- that self-employment is not disguised unemployment, that most government sector workers earn more than their private sector counterparts, and that the returns to education have fallen -- must be considered tentative until -15- substantiated by more thorough analysis. III. The Structure of UnemployMet There is no single ongoing survey which gives consistent longitudinal data on unemployment rates in Sri Lanka. Instead, we must rely on a variety of sources using different definitions of unemployment. In this section we rely on unemployment data from five sources -- the 1969/70 Socio-Economic Survey (reported in a 1971 ILO study), the 1971 and 1981 censuses, a 1973 study by the Marga Institute and the 1985/86 Labor Force and Socio- Economic Survey. The Socio-Economic Survey, 1969-70 was a survey of 9694 households. We could not determine the precise definition of unemployment that was used, but the unemployed consisted of those between 15 and 55 who were not primarily engaged in keeping house or studying and who were looking for work. The Census of Population, 1971 was, in principle, a total enumeration of the population. The definition included all those who desired and were available for work whether or not they were actively looking. Thus the definition includes discouraged workers. The 1981 Census of Population is thought to have undercounted the unemployed (Korale, 1987, p. 21) by leaving unemployment as a residual. All those who were not employed, students, retired, unable to work or keeping house and had sought -16- work within the last 30 days were treated as unemployed. The 1973 Marga Institute study covered 13831 households including 15945 youths age 15-24. The sample design was chosen to overrepresent districts with high levels of unemployment. The definition of unemployment was not clearly presented. The Labor Force and Socio Economic Survey, 1985-86 was a study of 11897 households. It defined persons as unemployed if they had been economically active (employed, actively looking or a discouraged worker) for a major part of the last 12 months and were mostly unemployed during that period. UeMlovment rates by sex and sector Open unemployment (including self-employed and unpaid family workers employed fewer than twelve hours a week and available for work) in Sri Lanka in 1985/86 was calculated at 14.1%. As reported in table 6, unemployment rates were substantially higher for women than for men and lower in the estate sector than elsewhere in the ecoromy. Unemployment rates by age Table 7 gives unemployment rates for different age groups and sectors of the economy. It is evident that unemployment is concentrated among the young. Open unemployment for workers over age thirty or thirty-four falls to levels comparable with the United States and western Europe. On the other hand, almost half -17- of fifteen to nineteen year olds in the urban sector are unemployed. Moreover, in the urban sector unemployment declines only gradually from that peak, remaining at 16.5% for thirty to thirty-four year olds and only reaching single digits above that age. Unemployment rates for children under age fifteen are relatively low in the rural and urban sectors where most are still in school but is nearly 20% for children ten to fourteen in the estate sector. Unemployment and Self-Employment In our discussion of self-employment in section II, we noted that there is a sharp increase in self-employment in the late 20s and early 30s. While this evidence is suggestive of the view that unemployment declines as individuals become pressed by family needs and accept whatever employment is available, it is possible that what we observe is not an age effect but a cohort effect. Some transformation of the economy may have caused a steady reduction in self-employment over time. Table 8 allows us to assess this possibility by showing unemployment rates and self-employment rates (including unpaid family work) for different age cohorts over a period of fifteen years. Unemployment rates drop sharply as the cohorts age at least for the younger cohorts. There is also a smaller but sharp increase in the rate of self-employment. For example, the group which was age fifteen to nineteen in 1971 had a 40% unemployment -18- rate. By 1981, this group (now twenty-five to twenty-nine) had an unemployment rate of only 21%. At the same time the percentage self-employed rose from 17% to 22%. Thus it appears that about one-quarter of the decline in unemployment can be accounted for by the growth in self-employment. Unfortunately, the places where unemployment declines and self-employment grows do not coincide. Between 1971 and 1981 the growth in self-employment was five to seven percentage points-for every cohort. Between 1981 and 1985/86 growth rates of self- employment were, if anything, somewhat higher for the older cohorts. While some of the growth may be accounted for by different definitions used in the censuses and labor force survey, the lack of correlation between declining unemployment and rising self-employment is evident. Differences among the surveys make it difficult to determine with confidence whether the level of self-employment has changed over time. Between 1971 and 1981, the percentage of 25 to 29 year olds who were self-employed increased by four percentage points from 18% to 22%. However, the percentage of self-employed at other age levels remained essentially unchanged. On the other hand self-employment increased sharply between the 1981 Census and the 1986 Labor Force Survey. For example, in 1981, 23% of 30 to 34 year olds were self-employed compared with 29% in 1986. It is, of course, possible that this reflects differences between the surveys rather than a real change. -19- Unemployment rates by education: A first look A "Well-known fact" about Sri Lankan unemployment is that it is heavily concentrated not only among the young but among the educated young. In part, this maj reflect the fact that the young are more educated, but as table 9 shows, even among the young, education and unemployment rates are positively correlated. A well known ILO report from 1971 (ILO, 1971) reported that among both fifteen to nineteen year olds and twenty to twenty-four year olds, unemployment rates rose with education levels. As shown in table 9, 63% of twenty to twenty-four year olds who had passed O-levels were still unemployed. A Marga Institute (1977) study of rural youths age fifteen to twenty-four reported a 74% unemployment rate for those with O-levels although the unemployment rate declined somewhat for more educated youths. The Aost recent results reveal a similar pattern. The last three columns of table 9 report unemployment rates in 1985/86 for three different age groups of younger workers. The estimated unemployment rates are lower for more educated workers than those reported in the ILO and Marga Institute studies, perhaps because of somewhat different definitions of unemployment in the three studies. However, the essential result that unemployment rates are generally higher for those with more education up to and including A-levels than for less educated workers remains true in the latest data. Perhaps the most striking result is that even -20- among twenty-five to twenty-nine year olds, where the overall rate of unemployment is 17%, the rate for those who have passed A-levels is 36%. The statement that unemployment is concentrated among educated youth must be clarified. Even among twenty-five to twenty-nine year olds, most of whom will have completed their education, the fraction having passed A-levels is only 6% and only slightly more than 1% have a degree or higher. The median level of education for twenty-five to twenty-nine year olds is grades 8 or 9 so that the roughly 18% who have passed 0-levels form the great majority of the highly educated. Of unemployed youths age fifteen to twenty-nine, only about one-tenth have passed A-levels. Almost one-quarter have passed O-levels. There is some reason for interpreting these results with caution. As shown in table 6, unemployment rates are much higher for women than for men. Since Sri Lankan women get more education than men, this could cause a spurious relation between education and unemployment. Similarly, table 6 shows that unemployment rates are much lower in the estate sector than in the rest of the economy. Since education levels also tend to be much lower in the estate sector, failure to control for sector of the economy may again result in a spurious relation between education and unemployment. Unemployment rates by education: A second look -21- Tables 10A through 10D give unemployment rates by age and education for young workers divided into four groups -- urban males, rural males, urban women and rural women. The number of youths in the sample from the estate sector was too small to permit a separate analysis of that group. Breaking the sample up this way significantly diminishes the relation between education and unemployment. The results for urban twenty-five to twenty-nine year old males are most striking. Unemployment peaks for those with education through grades five to seven. Since the median level of education is grades eight to nine, among urban males in this age group, unemployment can be said to be concentrated among those with relatively little education. The picture for rural males is somewhat different. By age twenty-five to twenty-nine unemployment rates are relatively low for all education groups except for a surprising "blip" for those with A-levels. In assessing the importance of the higher unemployment rate among those who have passed A-levels, it must be remembered that even among twenty-five to twenty-nine year olds, only about 6% of the population has passed A-levels. Consequently, those with A-levels make-up only a small fraction of the unemployed. Unemployment is actually somewhat more prevalent among those with the median level of education than among those who have passed O-levels. Among urban women age twenty-five to twenty-nine, -22- unemployment is high and fairly uniform for those with schooling through grades one to nine. Those with O-levels have a somewhat higher rate. Among twenty-five to twent.y-nine year olds, it is only the result for rural women which accord with the general finding that unemployment is educated unemployment. As shown in table 10D, for this group unemployment rates rise uniformly with education through A-levels. The essential point to take away from tables 10A through 10D is that, at least for twenty-five to twenty-nine year olds, the general finding that unemployment is concentrated among those with relatively high levels of education is due to a large extent to the failure to control for sector and sex. To the extent that the correlation is a reality for this age group it seems to be due mostly to the experience of rural women. These women make up over one-third of unemployment in the twenty-five to twenty-nine age group. The results of controlling for sex and sector are less dramatic for younger workers but still evident. In particular, for fifteen to nineteen year olds, the sharp difference in unemployment rates between those with the median level of education, grades eight and nine, and those with O-levels (who form the large majority of youths with above average levels of education), more or less disappears for urban workers. While overall those with O-levels have a 55% unemployment rate and -23- those with grade eight or nine education have a 39% unemployment rate, the gap between the two rates is 1 percentage point for urban males, and 6 percentage points for urban women. Similarly for twenty to twenty-four year olds, in the urban sector those with A-levels have lower not higher unemployment rates than those with O-levels. Overall Sri Lankan unemployment is much better understood as youth unemployment than as educated youth unemployment. Unemoloyment duration One of the striking aspects of Sri Lankan unemployment is the length of unemployment spells. In 1981/82, the median unemployed person had been unemployed for about fifteen months and more than one in ten had been unemployed for more than four years (Report on Consumer Finances...., p. 161). Unfortunately, it is impossible to project the distribution of eventual time spent before employment based on these data without making very strong assumptions. Fortunately the Marga Institute (1977) study, although somewhat dated, provides information which allows us to make reasonable estimates. The study obtained both time unemployed for those unemployed and time unemployed prior to getting a job for those employed. Since virtually all the unemployed were first-time job seekers, we can use these data to calculate the probability of obtaining employment in a period conditional on ~7 -24- not having yet obtained a job.5 We use these estimates to calculate the eventual distribution of completed unemployment durations. Table 11 gives the estimated distribution of unemployment duration based on the Marga Institute data. The results are reported separately by education group. The findings are striking for a number of reasons. First, eventual unemployment duration is extremely long. We calculate that nearly half of the unemployed will take more than five years to get a job. Second, although there is some slight tendency for those who have passed O-levels to have longer expected unemployment durations, the effect is not large. The principal fazt that comes through from table 11 is that unemployment duration is very long for all education groups except those with no school. Moreover, even among those with no education, unemployment duration is comparable to that of those with education for individuals who do not obtain a job within the first year. The very long unemployment durations oblige us to sound a further note of caution about the comparison of unemployment rates across education groups. Holding age constant, more educated individuals will have started their job queuing more 5From tables 14 and 25, we calculate the fraction of workers who were unemployed for at least t periods who found a job in period t+1, that is the number of individuals who reported finding a job in period t+1 divided by the sum of the number who report finding a job in period t+l and the number who reported being unemployed for at least t+1 periods. In technical jargon, we calculate the Kaplan-Meier estimates of the hazard rates. -25- recently. Consequently they will tend to have higher unemployment rates. Given the time it takes to find employment, this bias is not trivial. Summary and conclusions Unemployment in Sri Lanka is concentrated among the young, broadly defined. Within this group, unemployment is higher for women than for men and higher in the urban sector and lower in the rural sector. Because education is higher in the grous that have the highest unemployment rates, unemployment appears to be "educated" unemployment. However, it appears to be more appropriately viewed as youth unemployment. Among youths and young adults, unemployment levels are high and durations are long for virtually all education levels. IV. The Unemployment Problem The problem of unemployment in Sri Lanka is one that has received a great deal of attention. Consequently many explanations have been proposed. In this section we first discuss the proposed explanations that we find unconvincing and then go on to describe our favored explanation. Other E,,mlanations Harris and Todaro describe unemployment in developing countries as resulting from urban-rural wage differences. -26- Workers choose between working for low wages in the rural sector or migrating to the city for a chance at a higher wage job. Those who migrate but do not find jobs are unemployed. By assumption there is no way for these aspiring workers to search for urban jobs while working in rural areas.6 This is clearly not the problem in Sri Lanka where the bulk of unemployment is located in rural areas beyond the urban fringe. Thus it does not seem possible to argue that queuing is necessitated by differences in the location of rural and high wage jobs. Harris-Todaro type unemployment may be a partial explanation for unemployment since unemployment rates are higher in urban areas. However, it can not explain the majority of the observed unemployment which is located in rural areas. Sri Lanka is unusual among developing countries in having a highly developed social welfare system. Its high literacy rates and excellent health statistics are the result. Some have argued that its chronic problems with unemployment are another result. We do not find this convincing. Certainly any program which reduces the cost of being unemployed is likely to have some effect on the unemployment rate. However, we doubt that unemployment would be reduced much even if there were a very large reduction in the extent of social welfare benefits. First, many of the unemployed are children of relatively well-off 6 Or at least their probability of being chosen must be lower than that of someone who migrates. -27- families. Given that families with unemployed young adults will necessarily be poorer than similar families where young adults are working it is surprising how evenly distributed across different wealth groups unemployment is (Marga Institute, 1977, table 11; Visaria, 1979, p. 5). We believe that the children of these better off families would be unemployed even if there were no free food, no free medical care and no free education. Second, there has been a significant reduction in the level of food support since the late 1970s, and there has been no evidence of a concomitant reduction in unemployment. Another explanation commonly heard for unemployment in Sri Lanka is that the systeu of free education encourages over- education and that this results in too many youth with unrealistic career expectations who refuse to take the available jobs. Given the simple relation betweet education and unemployment shown in Table 9 this seems like a plausible argument. However, as we demonstrated =hove, the actual relation between education and unemployment is much more complicated and less dramatic than Table 9 would lead one to believe. Further, there is little evidence that youths' job expectations are unrealistic. The majority of the more highly educated youth do eventually find good government jobs. While this pattern of employment may reflect a deliberate government policy to head off unrest among over educated youth, we still do not knov what the equilibrium level of educational attainment or unemployment would -28- be if the government did not hire these workers. This last point suggests another one of the common explanations for Sri Lanka's unemployment problems -- queuing for government jobs. We accept this as a partial explanation for the phenomena but the explanation is incomplete for two reasons. First, as noted above, a large fraction of the unemployed are people with education levels below those required of government workers. Once we control for sex, age and location, the unemployment rate of those with O-levels is nearly the same as those with eight or ninth grade or those with sixth or seventh grade educations. The argument that unemployment represents queues for government jobs is incomplete in another sense as well. As noted above, high inflexible wages in one sector are not sufficient to cause unemployment. Either wages must be rigid in all sectors or there must be some advantage to being unemployed as opposed to working in a market clearing sector.7 An important contribution of this paper is the identification of institutions which make queuing preferable to work in *ither sectors for some workers. Our policy recommendations flow largely from these observations. Before turning to this point, we must, however, address one final explanation for unemployment. All of the above explanations suggest that unemployment is 7 It may also be the case that the "unemployed" simply have reservation wages above the market clearing rate but below the wages in the high wage sector. -29- a supply phenomena -- that if workers chose to find work they could. The traditional alternative explanation is that there is deficient demand for labor either because of inadequate investment or inadequate aggregate demand. If this is the case then the solution .o the unemployment problem is to increase demand through fiscal or monetary policy and/or to stimulate capital formation. We are sympathetic to this view and believe that demand policies can not be ignored. Nor can the demand side effects of fiscal policies aimed at curing supply side problems and structural imbalances be assumed away without serious consequences. However, Sri Lanka's current budget deficit and balance of payment problems preclude the exclusive use of these tools in addressing the problem. Further, we strongly believe that there are labor market policies which could increase employment and create a climate which will encourage investment and entrepreneurial effort even in the absence of additional fiscal stimulus. Unemployment Among the Educated Our analysis of unemployment considers workers with different levels of education separately. For those with degrees, and to a lesser extent those with A-levels, eventual employment with the government is virtually guaranteed. For these workers unemployment is arguably entirely due to queues for government jobs. But why don't these aspiring government workers -30- take jobs in the private sector while they wait for government jobs? As we will argue below, we believe that jobs in the private formal sector are also rationed. It is probably also the case that, for many of these unemployed workers, their reservation wage is below the government wage but above that in the best alternative job they could easily obtain. However, we believe there are many workers who would be willing to take low wage jobs while waiting for government jobs. The reason they do not is the hiring preference given to the unemployed. Under the "job bank" programs of the Late , 70s this was an explicit policy. Government hiring was used to reduce unemployment; so naturally those targeted were the unemployed. Government jobs are generally good paying, are always stable, offer opportunities for advancement, have pensions, and are considered to be high status. They are very desirable. Showing a preference for the unemployed in hiring creates a strong inducement to be unemployed. To the extent that current government policies still favor the hiring of the unemployed, this incentive continues. We do not have direct evidence on the extent to which the hiring practices utilized by government favor the unemployed. However, past policy appears to reflect a deeply felt belief that government should hire the unemployed rather than those who have jobs already. In a speech entitled "The Government's Approach to Solving the Unemployment Problem in Sri Lanka," -31- the then Secretary to the Minister of Plan ImpLementation, expressed his disgust at employed individuaLs trying to get government jobs: "I will give you an example which I have been repeating and is worthy of repetition. The People's Bank and the Bank of Ceylon advertised a thousand vacancies for clerks. Each Bank received 45,000 to 50,000 applications and it took their staff over six months to process these applications. They ultimately found that out of the 50,000 about 15,000 to 17,000 applicants were already gainfully employed. It may be that in a better society we can improve the jobs people have but the stunning situation today is unemployment - the large number of people who have NEVER been employed." This statement, if typical, suggests a set if values which will lead to implicit or explicit priority for unemployed workers. It also indicates that it is not feasible for the unemployed to accept employment and apply for government jobs under the pretense of being unemployed. For those with O-level education the probability of getting a government job is still very high and the reasons for queuing rather than taking other jobs while waiting for a government job are mostly the same. Also, any lower probability of being hired is made up for by the larger difference between wages paid for those with O-levels in government service versus those earned in the private formal sector and particularly the agricultural sector. For the younger unemployed there is another consideration -- the government will not hire workers who are younger than nineteen. Thus many youth who finish their O-levels before their nineteenth birthday can not begin to be employed -32- before they are nineteen. Unemoloyment Among the Less Educated For those with only a secondary education or less, probability of government employment is very low. Despite this we see unemployment rates for this group which are comparable to those for the more highly educated. We do not think it likely that these young people are queuing for government jobs. There are two other explanations which are complementary. The first is that many may be unemployed while studying for O-level examinations even though finished with formal schooling. Thus they may not have the education required for government employment now but they may still aspire to government employment at a later date. For those who are less than nineteen there is no reason to rush. Second, it is likely that there is queuing for formal private sector jobs as well as those in the government sector. The development literature has long accepted the notion that formal sector jobs in developing countries pay more than jobs in agriculture where it is believed the labor market clears. Many explanations have been given for the wage differences paid in the formal sector and many of these explanations are the same as those that have recently been proposed to explain unemployment and inter-firm and inter-industry wage differences in developed countries (Dickens and Lang, 1988). Theoretical studies of -33- "efficiency wages" explain the wage differences as being due to different needs to motivate workers or to prevent turnover or to attract a higher quality workforce when direct observation of individual 'worker quality is difficult. "Rent- extraction" or "insider-outsider" models explain unemployment and wage differences in terms of workers' bargaining power, particularly when acting collectively. Finally, "rent-sharing" or "expense preference" models explain wage differences as being due to employer preferences for spending money on good quality happy workers beyond the needs for strict profit maximization.8 A large body of empirical work on U.S. labor markets,9 and a growing literature on foreign countries10 suggests that such explanations are necessary for a complete picture of how even developed economies operate. We thus expect that outside of the agricultural labor force, and perhaps a few other sectors using day labor, all jobs are rationed. It is less clear in these cases what the reasons for queuing is. It could be the search advantage an unemployed worker has though this seems unlikely given the durations of unemployment we observe. While the rural/urban distinction does 8 See Katz 1986 or Akerlof and Yellen 1987 for reviews of this literature. 9 See Dickens and Lang 1988 for a recent review. 10 See for example Lang et. al (1988), Gerlach (1989), Blanchflower and Oswald (1987a&b), Blanchflower, Oswald and Gattett (1988), and Carruth and Oswald (1986). -34- not seem to be important for the bulk of Sri Lanka's unemployment, it may be for the less well educated workers who aspire to better paying jobs. As we noted above, in urban areas there is no evidence of a consistent relation between unemployment and education. However, in the rural sector, where there are fewer job opportunities outside of the agricultural sector, unemployment does seem to be significantly lower among the less educated. Thus it may be that unemployment among this group is of the Harris-Todaro type. A final possibility is that private sector hiring policies may mirror government policies in favoring the unemployed. To the extent they do, they would create the same incentives for unemployment. Since the private sector in Sri Lanka is so heavily regulated, and as a result dependent on the good favor of the government, it would not surprise us if the private sector followed government employment policy. Further Research Our study of Sri Lanka's unemployment problem has turned up what we believe to be some major culprits producing joblessness. However it also raises a number of questions. First, it is difficult for us to identify from our reading exactly how current government hiring takes place. There is much in the literature on the "job bank" policies of the past but little on how workers are currently chosen. A complete analysis of this process is necessary to confirm our suspicion that current policies still -35- give preference to unemplcyed workers. Such an analysis might also allow one to identify other similarly dysfunctional practices. Our speculations on what direction this might go are included in the section on policy recommendations below. Another area which warrants further study are the hiring and compensation practices of private sector firms and the extent to which they may contribute to creating a "demand for unemployment." It is possible that private employers avoid raiding each other's employees to the point where they too are hiring only from the unemployed. We do not know to what extent the government has pressured private employers to help the fight against unemployment by hiring the unemployed. It would also be interesting to extend to Sri Lanka some of the studies of wage determination mechanisms which have been done for the U.S. and a few other countries. Studies of inter- industry and inter-firm wage differences and their relation to industry and firm characteristics, studies of the relation between wages and application/vacancy ratios and if possible detailed studies of relatio;ns between wages and worker characteristics in different employments could shed light on the question of whether the modern private sector in Sri Lanka pays above market clearing wages. To complement the study of hiring practices, there should be a study of the behavior of the unemployed. We find it difficult to believe that individuals are truly unemployed for a -36- median period of five years. Consequently, it would be useful to know in what activities the unemployed are actually engaged. Finally, it would be most useful to consider whether the explanation we propose could contribute more to the understanding of inter-group differences in unemployment and medium term fluctuations. If there have been any important changes over time in the probabilities of different groups obtaining govarnment or other high wage employment or in the productivity of queuing this should be rfflected in the time series of those groups' unemployment. To consider such questions would require a larger study including direct analysis of primary data. V. Policy Proosals Having identified the existence of attractive government jobs as one of the major reasons for unemployment in Sri Lanka, and government hiring policies as a concomitant, we will first consider how changes to government personnel policies might affect the level of unemployment. Government Personnel Practices and Employment Proarams The first attribute of government jobs that make them so attractive is the high wages they pay. Obviously, lowering wages in the government will tend to reduce the extent of queues for such jobs. However, we do not believe that it is feasible or desirable in a country where employment is dominated by peasant -37- agriculture to reduce wages in the government to eliminate all queues. The government has the same motivations as the private sector for paying above market- clearing rates in at least some jobs. Since there are already complaints among the most highly qualified government workers about their wages, it may be that Sri Lanka has gone as far in reducing wages of government employees as it can without seriously damaging morale and/or retention. This being said, we are not impressed with the case that has been made for any general increase in pay for government workers. As noted above, it appears that the vast majority of government workers are still paid substantially more than those in alternative employment in the private and particularly the agricultural sector. To the extent that motivation and retention are a problem for the most highly educated workers, judiciously administered plans for merit pay increases and promotions would best solve the personnel problems and help keep the government budget deficit from getting worse. Besides reducing wages, a second way to make queuing less attractive is to reduce the probability of being drawn from the queue. In this case that would mean making fewer government positions available by reducing the amount of government employment. In the long run we would expect this to reduce the fraction of workers queuing for government jobs. However, any reduction in government employment would tend to increase unemployment In the short run. Ultimately, the size of the -38- government is a social and political issue which must be resolved only in small part with respect to the role of government employment in the generation of unemployment, particularly since other policies besides reducing the size of the government are likely to have greater efficacy in addressing the unemployment problem. The question of the size of government along with the issue of pay for highly trained workers raises an additional issue -- whether Sri Lanka's workforce is overeducated. We are convinced that it is not. The small size of the private sector and the difficulty the government is having with pay for the best educated workers may have a common source -- a shortage of college educated workers, managers and entrepreneurs. Reducing the size of the government would be one way to address this problem. An alternative would be to require less education of applicants for many government positions. Yet another alternative would be to increase the number of students permitted to attend university. The high salaries earned by such workers in the private sector suggests that the social return to college education may be very high. Increasing the supply of such workers might serve the dual purpose of allowing an expansion of the private sector and reducing the pressure on government salaries for the highly trained. Reducing government pay and/or reducing the size of that sector are both drastic measures which may reduce unemployment -39- only at a substantial cost in some other area. However, even with a large high wage sector there is no need for unemployment if there is no return to queuing relative to working in another job. We believe that the best way to address the problem of unemployment in Sri Lanka is to change the way the government hires workers. In particular we believe that the preference for unemployed workers should be replaced with a normal preference for experienced workers with demonstrated skills and ability. This should have two salutatory affects in addition to removing the incentive for workers to be unemployed while waiting for government jobs. First, the government should be able to hire better qualified and more experienced workers. Second, the high government salaries would help not hinder the rest of the economy. If people must get work experience to compete for good government jobs they will be willing to work for less and will try harder to find private sector jobs which will allow them to accumulate the necessary experience. This will create a large pool of eager workers for private sector expansion. We do not know to what extent the increased supply of workers to the private sector will be absorbed. There are legitimate concerns with respect to aggregate demand and capital formation. Particularly in a time of fiscal retrenchment. However, we are confident that there are productive uses to which these aspiring workers could be put. In case the private sector does not do it we suggest some government programs that might -40- below. The hiring preference for unemployed workers is only the most important of several government personnel policies which might be amended. We have already commented on the deleterious effect of the governments refusal to hire workers who are less than nineteen. There should be no age requirements beyond reasonable experience and education requirements. Although the major current reason for queues is probably the hiring preference for the unemployed, once this is eliminated there are other institutional changes which might help reduce queues further. Reducing any advantage to unemployed job search will reduce queues, and the less uncertainty there can be about who is likely to get those jobs that are available, again the smaller will be the queue. To accomplish the end of making unemployed job search no more productive than employed search, the process of applying for and obtaining government jobs should be made as simple and streamlined as possible. To the degree possible the process should be centralized, thereby eliminating any incentive for shopping around or moving to areas where one's prospects of finding a government job may be higher. To accomplish the second end of reducing uncertainty about government hiring, every attempt should be made to standardize the review process for jobs. If possible applicants should be assigned an application score early in the process and then given as good an idea as -41- possible how likely they are to be hired. To the extent that private formal sector employers follow practices similar to those of the government, they should be encouraged to modify them. The less advantage the unemployed have over the employed in finding good jobs, the less incentive there will be for individuals to remain unemployed as opposed to seeking the best job available at the time. Since the absorptive capacity of the economy is an open question it may be desirable for the government to provide short- term low wage employment for any youth who are interested. Experience in the program could count towards the experience requirements for permanent government jobs. From our reading of a number of sources it seems clear that such youth could be profitably put to work improving and repairing rural infra- structure and in helping to improve cultivation of export crops. Such a rural youth job corps program could put the surplus in the pay of regular government jobs to work to accomplish some important work to improve export earnings and the rural standard of living. The added cost to the government could be very low. Industrial and Trade Poligies Most of the modern theories of unemployment which would explain the existence of queues for jobs in the private sector would also imply that there would be a potential to increase national income through industrial and trade policies targeted at -42- increasing employment in the high wage part of the private sector.11 Put simply, if there are factors which set the wages in the formal sector above those in other sectorr of the economy it is likely that the marginal product of labor in the formal sector is also higher. If this is the case, then policies which lead to workers moving from low marginal product/wage sectors to high marginal product/wage sectors will increase average productivity and national income. Such reasoning suggests a major reason for concern about the small size of the private formal sector in Sri Lanka. Empirical analysis of wage patterns in many countries suggests that one important correlate of high wages is capital intensity. Thus policies which promote employment in high wage industries will run contrary to at least one development tradition -- promotion of labor intensive technologies. Policies favoring high-wage capital-intensive industries would be a major change of direction for Sri Lanka which has stressed investment in agriculture, the lowest wage industry of all. We would argue that the best hope for prosperity in Sri Lanka can be had by giving future plans to expand the availability of agricultural land second priority to the development of the infrastructure required for a modern formal sector. We hold to this view despite the obviously greater 11 See Dickens and Lang (1988b) or Katz and Summers (forthcoming) for a discussion of this point in the context of developed economies. -43- initial employment effects to be had by following the traditional path. It is our opinion that by developing higher wage industries, particularly for export, the wealth that will be made available for future development and the creation of domestic demand for consumer goods will promote more rapid growth. With respect to trade, these theories suggest that countries which export high wage goods and import low wage goods should prosper. Trade policy should promote this end. By emphasizing the development of domestic agriculture to replace imports Sri Lanka has pursued exactly the opposite policy. We believe that future efforts should be aimed at promoting manufactured exports. We wish to make it clear that we are not advocating a rejection of the idea of basing trade on comparative advantage. Instead, we wish mainly to stress that relative prices are an imperfect indicator of the opportunity cost of switching from one product to another. Trade policy should be based on opportunity costs not relative prices. In general the opportunity cost of producing a good in the high wage sector is less than the price of the good. This analysis points to the use of policies which will further the natural development of the private formal sector. The question of which industries Sri Lanka has.a comparative advantage in is beyond the scope of this paper. However, we suggest that a natural starting point is to consider the relatively high wage industries which are already in place. -44- Finally, with respect to industrial and trade policy, we believe that some of the biggest gains to be made can be accomplished by promoting a strong, dynamic, independent private sector. We have already identified what we believe to be one current hindrance to the development of such a sector -- the lack of highly trained technicians, managers and entrepreneurs. To increase their availability, either government hiring of these workers should be cut back or more students should be admitted to an enlarged university system. We also believe it likely that current government regulation of employment in the private sector is counter-productive and a hindrance to the sector's development. In particular, the stringent regulation of layoffs discourages bold entrepreneurial action of the sort that is likely to have the biggest payoff. Government regulation of private sector employment should be reviewed with an eye towards making it easier for entrepreneurs to innovate and develop new businesses. To this same end, the government might consider making capital more easily available to start-up firms. It might also consider developing programs to help individuals acquire entrepreneurial skills and to connect them with networks of technical workers and venture capitalists. Such an approach could aid greatly in increasing the absorption of labor freed for private sector use by the elimination of government hiring preferences for the unemployed. -43- Further Results The major remaining question which we have not been able to address in this work is the extent to which the economy of Sri Lanka will be able to absorb the new job seekers who would be unleashed by the institutional changes we propose. We suspect that the rural labor market is flexible enough that a large number of rural youth will be able to find employment.12 We are less sure about the opportunities for expanding employment in urban areas without additional policies aimed at expanding private sector manufacturing and services. It would require a major research effort to answer these questions. Such an effort would require estimation of elasticities of labor demand for as many sectors of Sri Lanka's economy as possible. It is not clear that the data for a first best analysis of this sort are available. However, it is possible that the necessary information could be elicited through an employer survey and/or by examining the behavior of equivalent sectors in other countries where more extensive data are available. To ascertain the effect of the changes we propose on reservation wages it would be best to construct a simple model of 12 From our reading of the literature we suspect that at most times the wages in the rural sector of Sri Lanka reflect a surplus above subsistence and are relatively flexible. Whether or not this is the case is a question which deserves further study. It is possible that rural unemployment represents surplus labor in the presence of a subsistence or customary wage floor. -46- the queuing process. This along with information on the wage gap between available work and work in the government or high wage sector would allow us to estimate the extent of surplus in those high wages and thus the maximum expense a worker would be-willing to undertake to secure such a job. We would also need to know labor supply elasticities for youths from different backgrounds. As with the demand elasticities it might be possible to establish these numbers with surveys and/or through the use of data from other countries. Information on demand and supply could be combined in the context of the queuing model to provide a framework for analyzing questions of absorptive potential. Such work would undoubtedly be aided by a more careful examination of where unemployed youth ultimately find employment. The ideas presented above also suggest some other questions. It is remarkable that there is really no detailed analysis of wages in the government and private sector for truly comparable workers. The overall statistics on average wages for broad groups in these different sectors is of little real value when we consider the variety of occupations and worker attributes. Such a survey should be an essential element in any revision of government salary setting. We suspect on the basis of circumstantial evidence that shortages of highly trained technicians, managers and entrepreneurs are an important factor limiting the size of the -47- private sector. Whether or not this is the case is a serious question which deserves to be treated in detail. Analysis of the markets for each type of skill should be carried out and attempts to estimate the returns to each type of education made. Finally, to the extent that the private sector will be unable to employ the influx of aspiring young workers, it may be desirable to have the government provide short term employment of the type described above. A study aimed at identifying projects on which people could be put to work and cost/benefit analysis of the projects would be a necessary prerequisite to the planning of such a program. VI. conclusions Sri Lanka is a country with a predominantly agricultural economy with 81% of its employment in rural areas. The country has only a small private formal sector which comprises about 13% of employment. On the other hand, government employs a very large fraction of the population including over half of those with 0 or A level educations and nearly three-fourths of those with college degrees or more. Although existing published data are not all we might want, it is clear from them that there are large differences in the wages paid in the different sectors. Sri Lanka has a significant chronic unemployment problem. Depending on how unemployment is defined and what time period is examined it varies from the low teens to over twenty percent. -48- The overwhelming majority of this unemployment is concentrated among young people who are looking for their first job. Unemployment duration is very long with typical spells lasting four years or more. Although it appears from a simple break-out of unemployment by education that it is mostly a problem of the more educated, a closer examination shows that once sex, sector and age are controlled for the relation between education and unemployment disappears for urban youth and is significantly weakened for rural youth. Of the many explanations that have been proposed for Sri Lanka's unemployment problem we fiiil arguments that it is primarily Harris-Todaro urban unemployment, that it is due to overeducation or overly generous social programs, or that it is due only to inadequate aggregate demand unconvincing. Although we are sympathetic to the view that the unemployment may be due to queues for government jobs, we find the explanation as it has been made in the literature incomplete for two reasons. First, much of the unemployment is concentrated among people who have little or no chance of obtaining government jobs. There must be another explanation for unemployment among these workers. Also, high wages in the government sector by themselves can not explain unemployment. There must also be an explanation for why workers prefer to queue for government jobs while unemployed rather than by taking whatever job they can while continuing to seek higher wage employment elsewhere. -49- In our view the unemployment experience of the more highly educated and the less well educated require somewhat different explanations. We believe that the more educated are queuing for government jobs and that the reason why they do not take other employment while doing so is the governments preference for hiring the unemployed. Among the less educated we believe that there is queuing for private sector high wage jobs. There are now many good explanations as to why profit maximizing firms may pay above market clearing wages and this may generate some unemployment. We are not sure why those queuing for private sector jobs do not take less well paying jobs. Some possible explanations are the advantages of unemployed job search, the unavailability of any unrationed jobs in urban areas, and the possibility that private sector hiring practices may mirror those of the government. We discuss several ways of dealing with these problems. Government salaries could be lowered, or the size of the government could be reduced. Both of these policies could be expected to reduce equilibrium unemployment in the long run. Whether or not these policies are adopted, we expect that unemployment could be reduced by eliminating the government's preference for hiring unemployed workers and by streamlining the hiring system. It may also be advisable to encourage private employers to adopt similar policy changes if they currently follow government practices. -50- We expect that if the government were to drop its hiring preference for the unemployed and replace it with a normal preference for workers who have demonstrated ability in previous work experience, a large fraction of the currently unemployed youth would begin actively seeking employment of any sort which would supply them with the requisite job experience to obtain government employment. This would be a boon to the private sector -- a large supply of eager young workers willing to work for relatively low wages in exchange for valuable work experience. There is an open question as to how many of these aspiring workers the economy could accommodate. To the extent that it is less than complete the government may wish to establish a program of short-term low-wage employment for young workers. Beyond personnel policies, promoting the private formal sector through trade and industrial policies can potentially increase national income. Such policies would require a complete reversal of Sri Lanka's past public investment and trade strategy from a concentration on low wage employment in agriculture to the development of high wage industrial employment. Finally, we suggest several areas for further research. Our analysis of the unemployment problem suggests the importance of further study of private and government sector hiring policies. To further validate the theory it would be interesting to see to what extent it can explain historical changes and more -51- detailed sectoral differences in unemployment rates. The policy recommendations all raise the question of the absorptive capacity of the economy and further study of this issue is in order. If the government wishes to develop a program to employ young workers it will be necessary to identify and evaluate particular projects. Finally, a side issue in much of our analysis has been conditions in the market for the most highly educated. The extent of demand for these types of workers and the effect of expanding supply on growth in the private sector are another set of important questions worthy of further consideration. -52- lBL 1 the Distribition of Emploment by Sector an Irdustry Agriculture 49.3% 11.2% 54.3% 92.2% Irdustry 18.8% 27.9% 18.3% 2.9% Services 27.9% 58.1% 22.7% 4.0% Uhkriun 4.0% 2.8% 4.7% 0.9% Total 513.2 97.7 371.7 43.8 eployment in (10, 000s) Percent of 100% 19.0% 72.5% 8.5% all euployment Source: labor Fbroe ard Socio-Economic Survey, 1985-86, p. 49. -53- TAEB 2 Distributio of tu ploynt in Sri Ianka ISIC Cede and Industry 31 - Fod, Beve:ages a~d Tcbaco 35.93 32 - Tetiles, Wearing Apparel ard Lathe.r 11.56 33 - Wood, Wood Pr«dts, rnitUre 7.31 34 - Paper, Paper Proucts, Printing 2.45 and Publishin 35 - c ails, Petroleum, R~bber and 11.60 Plastic Products 36 - Nn-Mqe-all c Mineral Products ept 10.27 Petrolem and oal 37 - Basic Metal 0.64 38 - Pabricted Metal Products Machinery 4.74 and Egpnt 39 - Other manufacturing 1.90 Source: Census of Industry, 1983, p. A4. -54- TM2 3 Self-~oymr ard U~padd Fa~ily Wod~rs as a Pei:ent of Tahn For:, Populati and Epl~yent by Age %~~ ~ ofppltoQ f labor foc a f eoo n 14 - 18 1 5 4 22 5 31 19-25 5 8 8 13 n 18 26 -35 12 4' 19 6 21 6 36-45 15 2 23 3 24 3 46 -55 17 2 29 3 29 3 O~er 55 18 1 48 3 48 3 All ages 7 3 20 .8 23 8 Source: Report on o er Finaes and Socio Econemic Survey 1981/82, part I, p. 124. -55- 'M= 4 Public and Coxporation Sector Eployment by Age and M1axtion % of Ovoyn % of labor for All 9 7 8 6 10 - 19 <.5 <.5 <.5 <.5 20-24 4 4 3 3 25 -29 10 10 8 7 30 - 34 13 12 12 11 35 - 39 14 12 13 11 40 -44 14 11 13 10 45 -49 14 9 14 8 50 -54 13 7 13 7 >54 2 2 2 2 < Grada 5 1 1 1 1 Grades 5 - 9 5 7 4 6 Paser 0-level 32 19 26 15 Passed A-level 36 19 25 13 Dgree or more 55 18 52 17 * ~oughly central government iloyment # Public corporations Scurces: Census of Public Sector and arporation Sector Employment, 1985, pp vi-ix. Iabor Force Survey, pp. 127-34, 192-9. -56- MEBLE 5 Median ard Man EMrnings by Class of Worker ard Sector (fs.) B@Mau nlye Casual EloesSelfapoe Urban 1144 694 1056 pzal 976 561 808 Estate 216 151 600 eD Urban 1672 794 1854 Aural 1218 688 1540 Estate 466 392 942 Souzce: Report On COnsuMer Finances and Socio Eonmic Suzvey 1981/82, part I, p. 212. -57- Mk= 6 Un~mployment Rt:. in Sri Ianka by Sex ard Setor Au &D All island 14.1 10.8 20.8 Utan 19.5 15.8 27.9 rLral 13.2 9.5 21.3 Estate 7.8 9.0 6.4 Soure: Labr Force and Socio-Enic Survey - 1985/86, p. 218 -58- MEE 7 UmWployvmnt Rates in Sri Lanka by Age and Setor aa All ages 14.1 19.5 13.2 7.8 10 - 14 9.5 12.9 8.2 19.9 15 - 19 31.7 42.6 28.9 27.9 20 - 24 30.4 38.9 30.2 11.7 25 - 29 16.7 22.8 16.4 4.5 30 - 34 10.2 16.5 9.3 3.0 35 - 39 5.7 8.5 5.4 1.7 40 - 44 5.5 6.2 5.3 5.4 45 - 49 3.6 4.8 3.5 0.7 50 - 54 2.2 4.4 1.8 - 55 - 59 4.6 6.0 4.5 1.5 60 - 64 2.6 9.3 1.3 - 65 & over 3.0 1.4 3.1 9.0 - Figures not reported, sample too small. Source: Labor Force and Socio-Econaic Survey - 1985/86, p. 218 -59- ThE 8 tU~mploya~mt and Self-Eployent Rates, Selected OCIorts, 1971-1986 221 221 I9M 2MG AM2 126ff2A 15- 19 40 21 10 17 22 29 20 -24 35 12 6 16 23 34 25 -29 20 8 6 18 24 30 30 - 34 11 5 3 22 27 39 35 - 39 7 3 2 26 30 40 40 - 44 6 3 5 28 34 47 Sources: Dpiarenit of labor, Sri Lanka, Iabr Statistis. 1986, pp. 11-12, Labor Far and Socio-Econoic Survey, 1985-86, pp. 115-6, 139-42. -60- Dm 9 ea.1poyn td:s in Sri Ianka by Age and carticn 1969/70, 1973, 1985/86 1969/70 1973 (xural nl1985/8 j15-19 20- 2"4 2":2 No Ysoo1 23 8 30 21 10 4 ade 1-4 22 13 7 Pr~mary 34 15 43 G:ad 5-7 31 23 11 Gade 8-9/ 46 39 62 39 40 20 Seceamy Pas 0- 92 63 74 55 43 25 level Pas~ A- 69 68 54 52 36 level Dmgr ar 45 53 16 higer Soures: ILD (1971), Vol 1., p.28; arga Institute (1977), table 4; Labor Force and Socio-Ecnic Survey - 1985/86, pp. 123-6,188-91. -61- TBLE 10A Uhmployment Mtes by Age ard Mcation (Uzban Males) 15-1220-.!24 2-.22 No school 32 20 8 Grades 0-4 23 20 10 Grades 5-7 34 28 24 Grades 8-9 51 38 19 Passed O-level 50 40 17 Passed A-level 33 10 Degree or nore 21 4 Souce: labor prce atd Socio-Eonaic Survey - 1985/86, pp. 127-30, 192-5. TABE 10B unemployment Rates by Age assd Education (Rural Males) 15-12 20-2 2522 No school 12 8 4 Grades 0-4 20 10 6 Grades 5-7 28 18 7 Grades 8-9 26 28 12 Passed O-level 39 29 10 Passed A-level 66 36 26 Degree or more 4 Source: labor Force and Socio-Econmic Survey - 1985/86, pp. 131-4, 196-9. -62- TME 100 hmploymt Pates by Age and aduaticn (Urban Wanen) M=l 20.24 22 No adool 45 9 3 Grades 0-4 24 34 34 Grades 5-7 56 52 36 Grades 8-9 60 49 35 Passed 0-level 66 51 42 Passed A-level 48 26 Degree or re 14 Source: labr FMe and Socio-ECoMic Survey - 1985/86, pp. 127-30, 192-5. TBBE 10D Unemployment Rates by Age and Education (Mral Women) 1512 20!!2d 25-22 No sciool 9 10 7 Grades 0-4 23 18 11 Grades 5-7 34 31 14 Grades 8-9 49 54 34 Passed O-level 65 57 45 Passed A-level 48 75 57 Degree or more 36 Soure: labor Force and Socio-Eonmic Survey - 1985/86, pp. 131-4, 196-9. -63- %mE 11 Prodict UYeuploymeDt Duration by Eu tim (Mural youth, 15-24), 1973 < i3r. 1-2 yrM. 2-3 yr=. 3-4 yr=. 4-5 yr. > . M No scol 61% 2% 3% 9% 2% 22% 163 Primry 16% 8% 10% 10% 10% 46% 1893 Middle 10% 8% 12% 13% 14% 43% 4080 0-level 7% 10* 11% 11% 6% 55% 1227 A-level 12% 10% 13% 5% 11% 49% 112 Based n data in Marga Institute (1977), tables 14 and 25. -64- MEFENCE Akerlof, George A. afd Yellen, Jant, eds., Efficiency Waae Model of the Labor Market, Camridg : »ridge University Press, 1986. Blandifmær, D.G., Osald A.J. "Insider Powr in Wege Datenatiæ: Part I," mimo Yarch, 1987a. , "Internal Influee Upn Pay Settl s: Nw Survey Evidn," C»ntre for #abour Ecnmiæ, ISE nisæsiæ paper #275, March 1987b. Blanchfloss., D.G., Oswald A.J. and Gattett, .D., "Inser Pwer in Wage Determinatio," ina June 1988 Crruth, A. A. and Oswald A.J., "Vage Infleibility in Britain," <E LSE DP #258, Oct. 1986. Draniyajala, Dore, R.P. and Little, A. W., "C rtificatin and qp~oyment in Sri Ianka," National Institute of un~pbl he, 1986. Dicken, William T. and Iang, K~in, "e of Segmente Iabar Mar~t ery A_i_ Riew and 78 (May 1985), 129-34. ,"y It Matters at Wa Trade: A Case fr Active Policy," in Dickens, W. T lysn, L., and Zysman, J. eds, The Elyn Eects of Internatimal Trade. B~st: Be1iLiger, 1988b. Gerlach, 1~bt, "Sectoral Vage Patterns, Irdivirual Earnings and ¶e Efficiency Wage Hypothsis," Uhiversity of Hanver, 'r~blinhe, 1988. Gløe, P., "Self-Employent In "næ in Sri Lanka, 1969-70 to 1981-82," in van Ginneken, W., ed., Trends in Emlvment and Labur Inm., Geneva, IIO, 1988. International Labor Office, Matchir Emlayment Oppruities and Expetatons Genava: ILD, 1971. Katz, Lawrence F. "Efficiency Vage Ueories: A Partial Evaluation," NBE M Annual. 1986, mbridge, M: MIT Press, 1986. and Swn=s, Lawrenæ H. "Industry Rents and Industrial Policy, Brookins Papers on Economic Aivity 1989, fortber nirig. Krale, R. B. M., Unmlovment and Weaes: A Case Stud~ of Sri Lanka. Clmo: Deprt d of cns ard Statistics, 1987. -65- Iang, Kevin, arquez, Qstavo an Pfmua, Pilar, "n=ies ofW Determinatiæ: Evidence fr= Chile and Venezuela," Rt University, 1988, u~published. ~aga Institute, "Pattern of Job Expectatins and Eployent Jing »ral Yoth," presented to Asian aegnal Seminar e Rural Employnit Prmctin, C01ombo, June 1977. Sri Iana, C*ral Bnk oft Cyl, Ræcrt Cnsumer FinanEs and Socio EcnMic Survy 1981/82, Part I, 1984. Sri Iana, t at of Cnsus and Statistics, Cnsus of Istry, UME, Prell~inary Report, 1986. , onsus of Public Sector an! C grporatir Sector Emplomnt 1985, 1987. -, abor Force and Socio-Emnomic Survey - 1985/86. Preliminary Reort, 1987. Sri Lanka, Departmnt of iabor, Labor Statistics. 1986. 1987. Visaria, Pravin, "Incidence of Unenployment and Characteristics of the Unployed in 69-70," Working Paper No. 7, Joint ESaP-IBD project n the evaluatio of Asian data en ire distributii, gust 1979. Wewrisooria, Dr. Wicrema, Secretary to the Ministry of Plan IpleMn*ati , in Pro dins of the Seminar m A Practicl Apoc Tø~ards Solvi the Uneployment Prmblem in Sri ,a. Sri Lanka Endatin SUtite, JUly 15, 1978, Ministry of Youth Affars and Employmenxt. ASIA REGION DISCUSSION PAPER SERIES 11 Title Author Originatoi IDP2 The Labor Force Participation of Women in the Republic of Korea: Evolution and Policy Issues C. Grootsert May 1987 P. Iqbal XDP15 The Role of Exchange Rate Policy in Sang-Woo Nam May 1988 D. Leipail Four East Asian Countries 78841 IDP28 The Small-Scale Enterprise Credit Program (S.S.E.P.) Under the Second and Third Calcutta Urban Development Projects F. Kahnert March 1988 F. Kahnerl (CUDP II and CUDP III) - An Assessment 76376 IDP35 1l Improving Tax Policy Advice: Lessons and B. Fleisig June 1989 H. Fleisil Unresolved Issues from Asia Experience 76375 IDP36 21 Direct Taxes and Fiscal Policy Issues: A. Virmani June 1989 H. Fleisi An Illustration for East Asia 76375 IDP37 Z1 Comnodity Taxation in Selected Countries Z. Shalizi June 1989 H. Fleis in South East and East Asia 76375 IDP38 ;/ Tax Analysis in Developing Country R. Musgrave June 1989 H. Fleisi Settings 76375 IDP39 Indonesia: External Shocks, Policy Sadiq Ahmed June 1989 Kyle Pete Response and Adjustment Performance 74270 IDP42 An Analysis of the Nature of Unemployment W. T. Dickens July 1989 R. Zagha in Sri Lanka and Kevin Lang 78049 1/ Extra copies may be obtained from the Asia Information Service Center. 2l Not released yet.