Chapter 6: Closing the Jobs Gap in the Southern African Customs Union

Joannes Mongardini, Tamon Asonuma, Olivier Basdevant, Alfredo Cuevas, Xavier Debrun, Lars Engstrom, Imelda Flores Vazquez, Vitaliy Kramarenko, Lamin Leigh, Paul Masson, and Genevieve Verdier
Published Date:
April 2013
  • ShareShare
Show Summary Details
Lamin Y.M. Leigh and Imelda M. Flores Vazquez 

The Southern African Customs Union (SACU) region is facing an unemployment crisis of enormous proportions. Available statistics indicate that the official unemployment rate in SACU is between 20 and 50 percent and is largely a youth phenomenon (Figure 6.1). To provide jobs for those now jobless and for new entrants to the labor force, SACU members would have to increase employment by an estimated 10 million full-time positions over the period 2012–21. Even this increase would leave the ratio of employment to the working-age population at below 50 percent—lower than that currently observed in many other countries.

Figure 6.1Unemployment Crisis in the Southern African Customs Union Compared with Other Regions

Sources: International Labor Organization database; and country labor force surveys.

Note: Data for 2010 for countries in the Southern African Customs Union; all other countries 2008 estimates.

How did SACU’s unemployment situation become so dire, given that, by and large, its member countries have registered reasonably strong GDP growth in recent decades? What are the characteristics of labor markets in SACU and how do these characteristics affect job creation in the region? Has the education system in SACU delivered the skills that are in demand in the labor market? What has been the impact of unions and the centralized wage bargaining system in SACU on labor market outcomes? Have demographic pressures played a role? What does the ongoing structural economic transformation in SACU bode for job creation? Using a combination of empirical techniques and country case studies, this chapter pulls together various strands to distill key messages for policymakers in the region on how best to close the huge jobs gap.

The diversity of the economies in SACU precludes a simple solution to the unemployment problem. SACU’s labor markets are fairly segmented, and this duality arises from a combination of formal and informal sectors, urban and rural labor markets, and a “good jobs sector” and a “bad jobs sector.” Employment in the good jobs sector is usually rationed—wages are institutionally set above the competitive market-clearing level because firms set “efficiency wages,” whereas the unemployed or underemployed remain in the bad jobs sector. Factors such as minimum wages, strong unions in SACU, and the insider-outsider phenomenon1 have contributed to further duality in the labor market. This chapter’s preliminary findings suggest that no single available measure can address the unemployment problem in SACU. Only a combination of carefully designed initiatives, including prudent wage policies, measures to address the skills mismatch in the labor market, and faster growth, is likely to make significant inroads in unemployment.

The rest of the chapter is organized as follows. The next section summarizes the main features of the unemployment data, and is followed by a section that presents the empirical analysis of the factors that could explain the high level of structural unemployment in SACU countries. This section also analyzes the empirical relationship between unemployment and income inequality in SACU countries and discusses the implications for inclusive growth policies. The final section summarizes the policy messages and offers concluding remarks.

The Unemployment Data

This section examines the unemployment data in SACU and other selected regions in sub-Saharan Africa (SSA) and summarizes the key features. The cross-country analysis includes SACU, other regions in SSA, and a selected group of countries outside SSA that provide insight into structural unemployment issues. Given the weaknesses prevalent in labor market data in SSA, the discussion first summarizes a few observations from the cross-country unemployment data. The data are based on three different sources: 2

  • the World Bank’s World Development Report (WDR) database;

  • labor force and household surveys of individual countries; and

  • the out-of-employment/population ratios computed from the International Labor Organization’s (ILO’s) database (calculated as the proportion of the population between 14 and 65 years old that is currently without a job).3 This measure counts students, stay-home spouses, discouraged job seekers, and all individuals not willing or able to work. Although imprecise, this measure captures both the willingness to work on the part of individuals and willingness to hire by firms.

The unemployment data (Figure 6.2) yield seven key observations:

  • The unemployment rates in WDR vary significantly from the estimated out-of-employment rates for both SACU and other regions in SSA, thus highlighting the weaknesses of the unemployment statistics across the region.

  • Almost all data sources show that the unemployment rates in SACU are the highest in SSA (Figure 6.3) followed by those of natural-resource-rich economies. The non–natural-resource-rich economies in sub-Saharan Africa have, on average, lower unemployment rates. The analysis conducted for this chapter of labor force surveys and their quality suggests that unemployment is better measured in SACU, which could partly explain why the unemployment rates are higher in SACU than in other regions of SSA.

  • Youth unemployment rates in SACU are about twice the average official unemployment rates. Labor force surveys also show that, on average, about 40 percent of the unemployed in SACU are long-term unemployed (unemployed for six months or longer).

  • The structure of labor markets in SACU tends to be somewhat different from labor markets in other parts of SSA. With the exception of Botswana and Lesotho, the average portion of the labor force in the agricultural sector in SACU is about 20 percent, reflecting the prevalence of large-scale commercial farming, whereas the rest of SSA is largely dominated by small-scale and subsistence farming (Figure 6.4). By way of comparison, the labor force surveys of Ethiopia and Ghana report that 80 percent and 50 percent of employed workers, respectively, are in the agricultural sector. This could suggest that SACU has higher-quality jobs than the rest of SSA, despite its higher unemployment rates, and also suggests that higher labor productivity can coexist with higher unemployment, in line with the efficiency wage theory.4

  • The share of people outside the labor force is a more reliable indicator of unemployment than the official unemployment rate. Thus, this analysis computed the out-of-employment rates as another indicator of unemployment levels. These rates are significantly higher than the official unemployment rates for all other countries covered in this chapter, in SSA including SACU.5 Like the overall unemployment rate, SACU countries have the highest out-of-employment rates, reflecting a larger proportion of discouraged job seekers in the region or people who have not embarked upon a job search because they do not see promising prospects. In South Africa, about 15 percent of the labor force is made up of discouraged job seekers. In Botswana, the proportion of discouraged job seekers is about 25 percent, and the number of discouraged job seekers is even higher than the number of individuals actively seeking work, which is the official unemployment rate. By contrast, only about 2 percent of people out of the labor force in Chile are discouraged workers. These results suggest that the out-of-employment rates in SACU countries are driven largely by the number of discouraged job seekers compared with other regions.

  • A comparison of the labor force composition in SACU with that of other regions shows that female labor force participation rates tend to be lower in SACU countries. This level probably reflects higher reservation wages for females in SACU countries than other regions in SSA. It could also reflect better employment conditions in SACU countries. High unemployment in SACU, together with low labor force participation rates for females, has resulted in very low ratios of employment to working-age population.

  • Overall, one key message from the data analysis is that governments in SACU, and more generally in SSA, need to invest more in improving the quality of unemployment data. Implementing policies to enhance job creation and monitoring the effectiveness of those policies require higher-quality statistics.

Figure 6.2Cross-Country Unemployment Rates

Source: International Labor Organization database; and World Bank, World Development Report database.

Note: Data for the Southern African Customs Union are for 2010; all other sub-Saharan African countries are 2008 estimates. Data for countries outside sub-Saharan Africa are 2009 estimates.

Figure 6.3Southern African Customs Union: Unemployment Rate Estimates, 2010


Sources: International Labor Organization database; World Bank, World Development Report database; and IMF country desks.

Figure 6.4Employment in the Agricultural Sector, 2010


Source: International Labor Organization database.

Empirical Analysis of the Data

This section reports on the results of our empirical analysis on the factors that could explain the persistently high unemployment rates in the SACU region.

Unemployment and Growth—Employment-Output Elasticity

Any meaningful discussion of unemployment needs to look at the role of economic growth in reducing unemployment. A commonly held view is that significant growth acceleration is required for unemployment to be reduced substantially. Table 6.1 provides estimates, using equation (6.1), of the employment-output elasticities for the 33 countries in our cross-country sample including the SACU region. The results show that the employment-output elasticity β averaged about 0.4 and the constant term, α, in the panel regression was consistently negative. The latter is significant because it signals the role of factors other than GDP growth in employment creation.6

Table 6.1Estimated Employment-Output Elasticity Using Panel Data
Total sample−3.221 (t = −3.04)0.473 (t = 2.57)
Excluding CEMAC−1.923 (t = −2.07)0.419 (t = 2.33)
Excluding WAEMU−2.078 (t = −2.83)0.454 (t = 2.41)
Excluding CEMAC and WAEMU−1.518 (t = −2.01)0.401 (t = 2.04)
Source: IMF staff calculations.Note: CEMAC = Central African Economic and Monetary Community; WAEMU = West African Economic and Monetary Union. The values in parentheses are the t-statistics.
Source: IMF staff calculations.Note: CEMAC = Central African Economic and Monetary Community; WAEMU = West African Economic and Monetary Union. The values in parentheses are the t-statistics.

To what extent does the low cost of capital influence labor market outcomes in SACU? Specifically, have the roles of capital and labor in SACU been distorted over the years as large sections of the population were excluded from economic activity and production became more capital intensive because of the low cost of capital?7Figure 6.5 shows that the lower the cost of capital (measured using the benchmark bank lending rate), the higher the unemployment rate.8

Figure 6.5Employment-Output Elasticity, Job Creation, and the Effective Cost of Capital

Sources: International Labor Organization database; and IMF staff calculations.

Note: Member countries of the Southern African Customs Union (SACU) have lower estimated employment-output elasticity and employment growth compared with other middle-income countries. SACU countries also have generally lower effective costs of capital, which seems to be associated with high unemployment rates across the region.

Overall, both the highly significant constant term (αs) in the estimated panel regressions and the low effective cost of capital in SACU suggest that structural distortions in the SACU labor market may be contributing to persistently high unemployment, explaining why the labor market is not clearing. The next section reports the results of a series of estimations exploring the likely role of such structural distortions in SACU’s labor market through empirical techniques using both correlation and panel regression analyses.

Wage Policy and Labor Market Outcomes

The analysis shows that public sector real wage growth in excess of productivity is closely correlated with the unemployment rate in SACU (Figure 6.6).9 In fact, SACU’s real wages in excess of productivity gains are significantly higher than for other countries in the sample. The size of the public sector and higher public sector wages do influence labor market outcomes in an economy, including the private sector’s ability to create jobs. The bloated public sector in SACU, in which governments typically account for 40–60 percent of total national employment, lures job seekers with greater job security and higher wages, thus distorting labor market outcomes. Historically, government hiring practices in SACU have typically inflated wage expectations. These practices have also placed a premium on liberal arts and social sciences degrees over skills in demand in the private sector, thus influencing education choices and contributing to the skills mismatch in the labor market (see the discussion below on the role of skills mismatch in SACU’s labor market.) Other benefits of a public sector job include stability, reputation, and long-term security, among others. However, inferring causality from real wages to unemployment outcomes in SACU is difficult based on bivariate correlations. Thus, the panel regression analysis analyzes the causal relationships between unemployment and its macroeconomic determinants. The estimated panel regressions, which use various panel estimation techniques (Table 6A.1), show that high real wage growth above productivity in SACU tends to result in persistently high unemployment, including by encouraging informality in the SACU economies.

The high real wage growth, which outpaced labor productivity growth in SACU, partly reflects the outcomes of its centralized collective bargaining framework.10 This wage bargaining system not only contributes to the weak link between pay and productivity, but also reduces the response of the real wage to fluctuations in the business cycle. Additionally, the higher real wage puts upward pressure on labor costs and causes firms to substitute capital for labor, thereby increasing the marginal productivity of labor.

Union Density and Unemployment

Over the years, unions have played a pivotal role in SACU. Their emphasis on workers’ rights is well placed and bodes well with enhancing more inclusive growth. SACU’s union density is also relatively high compared with other countries in our sample. This said, if the job market is mainly dominated by a highly unionized government sector, sometimes this tends to give rise to voluntary unemployment in the nonunionized sectors. The high degree of correlation between unionization and unemployment (Figure 6.6) suggests that high union density in SACU may be contributing to unemployment outcomes. The panel regression also supports this negative impact of union density on SACU’s overall unemployment rate.

Figure 6.6Unemployment Rates and Labor Market Indicators: Correlation Analysis, 2000–09

Sources: International Labor Organization database; World Bank, World Development Report database; and IMF staff calculations.

Note: CEMAC = Central African Economic and Monetary Community; SACU = Southern African Customs Union; WAEMU = West African Economic and Monetary Union. The unemployment rate seems to be positively correlated with the wage-productivity gap, union density, and skills mismatch in the labor market. However, the unemployment rate seems to have little association with the welfare benefits or with restrictiveness of labor laws. Moreover, SACU countries generally have low minimum wages compared with peer countries.

Skills Mismatch in the Labor Market and Unemployment

The analysis shows that the skills mismatch in SACU is highly correlated with the region’s unemployment rate. The skills mismatch index is calculated by taking the difference between the skill demand and supply for each country in the sample. Following Estevao and Tsounta (2011), the skills mismatch index for each country i at time t is constructed using equation (6.2):

in which j is the skill level; Sijt is the percentage of the population with skill level j at time t in country i (skill level supply), and Mijt is the percentage of employees with skill level j at time t in country i (skill level demand).

  • Skill level supply. World Bank educational attainment data are used to construct skill level supply using primary education (as low-skilled), secondary education (as semiskilled), and college and tertiary education (as highly skilled).

  • Skill level demand. Skill level demand is approximated by the percentage of employees in three key sectors: mining and construction (to proxy low-skilled workers), manufacturing (for semiskilled workers), and government and financial services (for highly skilled workers).11

The results support the basic conclusion from the analysis of labor force surveys for SACU countries, which shows that skills mismatch is an important issue in its labor markets (Figure 6.6). SACU countries generally have a high rate of schooling for primary and secondary education, reflecting their high spending levels on education compared with other regions in SSA. However, this high rate of schooling has not yet translated into greater private sector–type skills because it has produced graduates whose skills are not in demand in the private sector.12 Increases in tertiary education would eventually help to meet the demands of the private sector and, over time, reduce the skills mismatch in the labor market. The required type of tertiary education closely mimics specialized advanced education, which supplies firms with highly skilled workers to create more employment. The estimated panel regressions (Table 6A.1) suggest that the mismatch between the skills the labor force possesses and the skills firms seek explains part of the high unemployment in SACU. Although many governments in the region have spent generously educating their youth, firms regularly cite the lack of suitable skills among job applicants as a constraint to hiring. In Botswana, unemployment rates are highest among college graduates, although for South Africa it is highest among unskilled workers. The former suggests that the education systems have not been successful in producing graduates with marketable job skills. Improving the quality of education spending to support public-private partnerships for skills development, vocational and technical training, and information and communications technologies skills, including the recent graduates’ internship program in Botswana, will over time reduce the skills mismatch and thus the overall unemployment rate.

Welfare Benefits and Unemployment

The analysis suggests that welfare benefits are not closely correlated with unemployment in SACU (Figure 6.6).13 Although other regions in SSA have, on average, higher welfare spending than that in SACU countries, they have significantly lower unemployment rates. In fact, the data show that welfare benefits in SACU are not only, on average, lower than in other regions in SSA, they also are not associated with increased levels of voluntary unemployment through their impact on the reservation wage of workers—the minimum wage that workers usually demand to formally reenter the labor force. The panel regressions also support this broad finding. This is an important result with implications for public policy in SACU: if unemployment is not affected by welfare spending, then welfare programs can be used to help the unemployed and discouraged workers without fear of the policy leading to a higher unemployment rate. These findings are consistent with Kingdon and Knight (1999), which also rejects the voluntary unemployment hypothesis through the impact of the replacement ratio (benefit-wage ratio) on the unemployment rate.

Labor Market Regulations and Unemployment

Labor market regulations typically hamper job creation. Specifically, hiring and firing costs can negatively influence employers’ decisions to hire new employees. Figure 6.6 suggests a low degree of correlation between hiring and firing costs14 and unemployment rates in SACU. In fact, overall, SACU countries have lower hiring and firing costs despite having the highest unemployment rates in SSA. The countries in the non–natural-resource-rich group have relatively high hiring and firing costs, yet their unemployment rates are, on average, lower than SACU’s. Thus, despite relatively low hiring and firing costs, the unemployment rates for SACU remain high; in other countries in SSA, high hiring and firing costs coexist with low unemployment rates, although higher-quality unemployment data in SACU could be the cause of this disparity.

The estimated panel regressions (Table 6A.1) also show that hiring and firing costs in SACU are not a significant determinant of overall unemployment in SACU. The result suggests that the high unemployment rate in SACU is not closely associated with labor hiring and firing rules. In fact, minimum wages (as a share of average wages in the economy) in SACU are on the low side relative to some other countries, although the small sample size, caused by data limitations, precludes the drawing of a general conclusion (Figure 6.6). Beyond this, the estimated panel regressions suggest that the high unemployment rate has little to do with the restrictiveness of the labor laws.15

The Role of Demographic Factors

The analysis does not demonstrate that demographic pressures have affected unemployment in SACU. Figure 6.7 shows that the rate of population growth is trending downward (below 1 percent) across SACU countries (this compares with an estimated 2½ percent annual population growth for the whole of SSA). Although the size of the working population as a ratio of total population is projected to increase as the impact of HIV/AIDS dissipates in the region, these ratios would generally remain low by standards in other regions in SSA. Moreover, the demographic variable is not significant in the estimated panel regressions.

Figure 6.7Countries in the Southern African Customs Union: Demographic Projections, 2015–50

Source: World Bank database.

Summary of Results

Pulling the threads together, the balance of evidence suggests that SACU’s high unemployment rate is largely driven by public sector wage policies, which distort labor market outcomes, including through their impact on education choices that cause skills mismatches. As noted, government hiring practices have typically inflated wage expectations and placed a premium on graduates with liberal arts or social science degrees over actual skills in demand in the private sector. Thus, the wage structure of the civil service in SACU is distorting the overall labor market and creating voluntary unemployment as graduates line up to get public sector jobs commensurate with their high reservation wages. This situation has also exacerbated the misalignment between labor productivity and real wages established in this chapter, thus discouraging employment creation. All of these factors have important policy implications because it is the confluence of government hiring practices and the public sector wage policy, reinforced by a less flexible wage bargaining process, that gives rise to distortions in the labor market, suggesting the need for a fundamental change in SACU. SACU countries should change their public sector wage policies not only to enhance fiscal sustainability, but also to reduce the associated distortions in their labor markets.

The prevailing wage rates in SACU cause excess demand for skilled labor and, for some countries, an excess supply of unskilled labor. Targeted interventions in key sectors, combined with a comprehensive reform of the education system, are needed to create conditions for rapid economic growth with job creation. Private sector–led approaches to improving technical and vocational training, in which governments standardize the curriculum and accrediting program, are the most promising route for education system reform. Korea’s experience in reducing the heavy emphasis on university education and promotion of tertiary education and vocational training is relevant in this regard. Similarly, the Japanese experience of improving labor market outcomes of training through close links with industry, continuous curriculum development, and the introduction of new programs focused on skills requirements of the job market, can help meet the demand for skilled labor in both the tradable and nontradable sectors (Treichel, 2010). It is encouraging to note that governments in the SACU region have already began to take some of these initiatives to address the skill mismatch.

The significance of the effective cost of capital variable in the estimated unemployment panel regressions supports the view that policies in SACU could be biased toward capital-intensive sectors at the expense of labor-intensive sectors. In particular, SACU’s wide-ranging tax incentives have resulted in a low effective tax rate on capital compared with labor (Figure 6.8).16 Since economic liberalization policies began in the early 1980s, countries in SACU have put in place a series of tax incentives aimed at supporting capital-intensive sectors. Over the years, tax incentives have proliferated and led to what is now a very complex incentive regime. The analysis presented here shows that the plethora of tax incentives in SACU has produced low effective tax rates on capital, which favors capital-intensive activities. Streamlining tax incentives for capital will raise the effective cost of capital and reduce the associated distortions on job creation.

Figure 6.8Average Effective Tax Rates on Labor and Capital, 1990–2005


Sources: Organization for Economic Cooperation and Development database; national accounts; and IMF staff calculations.

In addition to the above empirical analysis, the authors explored Botswana and Chile—two natural-resource-rich economies—both of which are deemed to be success stories for prudent economic management but have had different unemployment outcomes. Both economies are characterized by relatively low inflation, fiscal discipline, institutional strength, good infrastructure, and high standards of governance. Despite these similarities in fundamentals, Botswana continues to have a double-digit unemployment rate, well above the world average, whereas Chile has generally kept its unemployment rate below 10 percent (Figure 6.8). What explains this divergence in unemployment outcomes between the two economies?

  • Unlike Botswana, Chile has made much progress in diversifying its economy away from mineral resources, thereby making the economy more resilient to shocks and limiting the Balassa-Samuelson effect from the tradable to the nontradable sector through wages. In particular, over the years, Chile’s service sector has expanded in both value added and in share of total employment.

  • Through sound fiscal policy, Chile has also reduced the size of the government (Figure 6.9) and has maintained a composition of government spending that favors economic growth and job creation.

  • Chile has also delivered better education outcomes, that is, the quality of its skilled employees is higher compared with Botswana or SACU more generally.

Figure 6.9Unemployment Rate, Government Spending, and Wage Bill

Sources: World Bank, World Development Indicators; and IMF staff calculations.

Note: SACU = Southern African Customs Union; SD = standard deviation. SACU’s unemployment is well above the global average in contrast to Chile’s, because Chile has generally maintained a small government and a prudent public sector wage policy.

Unemployment and Income Inequality in the SACU Region

High structural unemployment in many countries has hindered the ability of governments to achieve more-inclusive growth. The literature shows that countries with better Gini coefficients (greater income equality) tend to be those with more diversified economies and lower rates of unemployment, which provide a robust foundation for more sustainable growth in the long term (Mocan, 1995). Those countries are generally able to improve the living standards of their populations for longer periods, even if their output growth is not as impressive as that in many natural-resource-rich economies. This leads to the question of the relationship between the unemployment rate and income inequality in SACU.

To address this question, the analysis takes a panel that includes three SACU countries and other selected middle-income countries, decomposes unemployment into its structural and cyclical components, and investigates their impact on income distribution, controlling for the effect of inflation. If marginal workers with relatively low skills are laid off first during an economic downturn, and if these workers are at the bottom part of the income distribution, temporary increases in unemployment are expected to worsen income inequality. However, the loss of income owing to transitory unemployment of a family member may be offset by unemployment insurance and welfare benefits, especially given the growing incidence of dual earners in families. Thus, it may take longer spells of unemployment to have a marked impact on annual family incomes.

Following Mocan (1995) to investigate whether long-term and short-term unemployment have differential impacts on income inequality in SACU, actual unemployment is decomposed into its trend and cyclical components. Because the hypothesis of a unit root is rejected for unemployment across SACU countries, the conventional way to determine structural (long-term) unemployment is to regress the unemployment rate on a constant, and linear and quadratic trend terms. The fitted values represent the long-term (structural) unemployment, whereas the trend deviations illustrate cyclical unemployment. As a robustness check, structural unemployment is obtained in two ways. First, the Hodrick-Prescott (HP) filter is applied to obtain the structural component of unemployment. Second, the Kalman filter technique is applied, which allows an estimate to be made of the trend at all points in the sample using all the observations. Structural unemployment obtained from the HP filter and from fitting linear and quadratic trends are broadly similar. The benchmark is taken to be the structural unemployment data series obtained from the fitted values of linear and quadratic trends. However, as Table 6A.2 illustrates, the results obtained from models with other measures of structural unemployment are similar to the ones obtained from the model with standard decomposition.

Table 6A.2 shows the estimation results of the models in which changes in income shares are regressed on inflation, and on structural and cyclical unemployment.17 The first panel presents the results when structural unemployment is obtained by fitting linear and quadratic trend terms to actual unemployment. The second panel is the model in which structural unemployment is obtained through the HP filter, and the third panel is the case in which structural unemployment is obtained through the Kalman filter. In all specifications, an increase in structural unemployment is associated with an increase in the fourth-highest and highest income quintiles with a negative impact on the first three income quintiles, but a change in cyclical unemployment has no impact on the income share of this group. Thus, the results provide some evidence that an increase in structural unemployment is associated with an increase in the income share of the richest 40 percent of the population, and with a decrease in the share of the bottom 60 percent of the population.

These results suggest that, although policies that aim to prevent a worsening in income inequality by combating cyclical downturns have validity, sustained GDP growth in SACU cannot by itself improve income inequality if it is not associated with a reduction in long-term structural unemployment. The results show that reductions in structural unemployment substantially improve income distribution. To the extent that better education outcomes in SACU contribute to a reduction in structural unemployment (as partly inferred from the estimated unemployment–skills mismatch function), they reduce income inequality, thus having the potential to make growth more inclusive.18 The policy implication is that for SACU countries and countries with similar structural unemployment–income equality dynamics, policies that lead to more sustained reduction in structural unemployment would enhance more inclusive growth. Policies may include incentives to employers to hire less-skilled workers in addition to training programs for workers who face stagnant wages and longer spells of unemployment or hysteresis effects.

Conclusions and Policy Lessons

Job creation is a key challenge for policymakers in the SACU region. This chapter analyzes the factors that have contributed to SACU’s high unemployment rate. Although there is some diversity in labor market conditions among the SACU countries, the broad conclusion from the empirical analysis and case studies suggests that no single measure can address the unemployment problem. Only a combination of carefully designed initiatives and faster growth are likely to make significant inroads into unemployment.

Closing the huge jobs gap in SACU would require faster economic growth and fundamental changes in public sector wage policy, not only to enhance fiscal sustainability but also to reduce the associated distortions in the labor markets. The analysis suggests that significant growth acceleration is required to put a substantial dent in the SACU unemployment rate. The confluence of government hiring practices and the public sector wage policy, reinforced by less flexible wage bargaining processes, has also given rise to distortions in the labor market. Education policies in SACU urgently need to be aligned with the skills needs of the private sector. The quality of education spending could be improved to support public-private partnerships for skills development, focus on vocational and technical training, and build skills in information and communications technologies. It is encouraging to note that governments in the SACU region have already began to take some of these initiatives to address the skill mismatch in their labor market.

Based on this study, other policy changes that would generate faster job creation include the following:

  • Economic diversification would create labor-intensive sectors and enhance the economy’s potential to create more jobs, including through limiting the traditional Balassa-Samuelson effects on the nontradable sectors. Thus, policies also need to focus on measures to improve the investment climate and reduce the costs of doing business in SACU to boost such new job-creating sectors, especially in the context of the ongoing structural economic transformation in the region.

  • Targeted government intervention in key nontradable sectors with high employment multipliers will also help. The segmentation of SACU’s labor markets argues for well-targeted intervention in certain sectors.

Finally, the governments in SACU urgently need to invest in strengthening statistics for their labor markets and unemployment. This chapter has established that there are significant variations between unemployment indicators in SACU from various sources and estimated out-of-unemployment rates, highlighting severe weaknesses in unemployment data. Monitoring the effectiveness of job-creation policies in SACU, and in sub-Saharan Africa more generally, would require higher-quality unemployment statistics.

Appendix 6A.

The baseline sample consists of 33 countries, but most do not have continuous time series data on the key variables used to estimate the panel. Thus, the final panel has only 11 countries that have continuous time series data on the key variables in the estimated regression countries, including the five SACU members. The sample period is 1990–2009. As noted in Table 6A.2, panel regression techniques were employed: pooled-regression, fixed effects estimator, and Arellano and Bond’s generalized method of moments regression estimation technique. The latter method implicitly addresses endogeneity issues inherent in the estimated unemployment equation. As expected, the wage-productivity gap variable has the correct estimated positive coefficient, with a rising wage-productivity gap leading to an increase in the unemployment rate after a one-year lag. Union density also raises the unemployment rate after a one-year lag and increase in skills mismatch. In contrast, a lowering of the effective cost of capital raises the overall unemployment rate with a three-year lag. As noted in the main text, the labor law variable has the correct estimated coefficient but is insignificant in all the estimated panel regressions.

Table 6A.1Estimated Panel Regressions
Explanatory variablesPooled regressionFixed effects regressionDynamic estimation (Arellano and Bond)
ΔWages in excess of productivity(−1)0.84***0.76***0.72***
ΔUnion density(−1)0.22**0.17**0.15**
ΔLabor laws restrictiveness(−2)0.0230.0150.008
ΔSkills mismatch index(−1)0.65***0.59***0.55***
ΔEffective cost of capital(−3)−0.34**−0.28**−0.33**
Error-correction mechanism(−1)−0.056***−0.049***−0.037***
Adjusted R20.610.540.59
Number of observations220220220
Source: Authors’ calculations.Note: Entries in parentheses are the calculated t-statistics.* significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent.
Source: Authors’ calculations.Note: Entries in parentheses are the calculated t-statistics.* significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent.
Table 6A.2Structural Unemployment and Income Inequality for Selected Middle-Income CountriesEstimation Method: Dynamic Panel Data Modeling using the Arellano-Bond Estimator
Structural Unemployment from Fitted Trend
Explanatory variablesLowest quintileSecond quintileMiddle quintileFourth quintileHighest quintile
Structural unemployment−0.038**−0.058**−0.073**0.049**0.216**
Cyclical unemployment−0.0260.041**−0.0140.0020.068
Durbin-Watson statistic2.342.262.472.252.55
Source: IMF staff calculations.Note: Entries in parentheses are the calculated t-statistics.* significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent.
Structural Unemployment from Hodrick-Prescott Filter
Explanatory variablesLowest quintileSecond quintileMiddle quintileFourth quintileHighest quintile
Structural unemployment−0.040*−0.049**−0.061*0.0390.186**
Cyclical unemployment−0.0240.047**−0.017−0.00040.077
Durbin-Watson statistic2.352.
Source: IMF staff calculations.Note: Entries in parentheses are the calculated t-statistics.* significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent.
Structural Unemployment from Kalman Filter
Explanatory variablesLowest quintileSecond quintileMiddle quintileFourth quintileHighest quintile
Structural unemployment−0.029**−0.037**−0.039**0.0220.120*
Cyclical unemployment−0.0360.082−0.027−0.0030.135
Durbin-Watson statistic2.312.22.332.172.40
Source: IMF staff calculations.Note: Entries in parentheses are the calculated t-statistics.* significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent.
Source: IMF staff calculations.Note: Entries in parentheses are the calculated t-statistics.* significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent.

The insider-outsider theory is related to the conflict of interest between insiders and outsiders in the labor market. “Insiders” are incumbent employees whose positions are protected by labor turnover costs. “Outsiders” enjoy no such protection; they could be unemployed or working in the informal, competitive sectors of the labor market.

These data are based on the latest available unemployment series from the World Bank’s World Development Report (WDR) database) and in a few selected cases from IMF country desks. In some cases, the data point may not be the most up to date, and the latest estimate of unemployment from the labor force and household surveys may be more representative of the actual unemployment rate. The authors used the three indicators for the level of unemployment precisely because of the weaknesses in unemployment data series in SSA, including in the SACU countries. The three series could be considered to be a lower bound, a possible mean, and an upper bound of unemployment rates in these economies. The high unemployment rate in SACU could reflect higher-quality unemployment data in SACU.

The idea of the efficiency wage theory is that it may benefit firms to pay workers a higher wage than their marginal revenue product because doing so might lead to increased productivity from the worker.

The out-of-employment rate is the percentage of people of working age without a job whether they are out of the labor force, discouraged unemployed individuals, or unemployed and actively looking for a job.

The explanatory powers of the estimated panel regressions are reasonably high, with adjusted R2 for the four regressions in Table 6.1 being 0.61, 0.58, 0.57, and 0.54, respectively.

The persistently high unemployment rate in some SACU countries, specifically Botswana and Namibia, also reflects the fact that although the mining sector generates growth it does not create much employment because of the sector’s capital intensity.

A growth accounting exercise for Botswana suggest that much of the real GDP growth since 2000 was based on capital deepening.

This chapter uses the consumer price index–based real wage measure instead of the GDP-deflator– based real wage, and productivity growth is proxied by an adjusted output per capita for the manufacturing and construction sectors.

See supporting econometric evidence in Klein (2012) using micro/industry-level data.

Although the Estevao and Tsounta (2011) method of estimating skill supply is reasonably robust based on educational attainment, the measures of skill demand and skill intensity do have some weaknesses, including treating the mining sector as low-skilled in the skill-intensity spectrum when most of the mining sector employees in SACU are at least medium- to highly skilled.

The current structure of the education system in many countries in SSA is in part a legacy of the colonial period through which senior civil servants encouraged students in the 1960s to study liberal arts and social sciences so they could help them run the public sector. Public servants’ salaries and benefits were made very generous to attract the best. This was exacerbated during the post-independence era, giving rise to unsustainable public wage bills as they made public sector employment very attractive in order to get the needed skilled civil servants. The distorted public sector wages resulted in an education system that produced graduates for the civil service but who did not have modern skills that firms demand in the current labor market.

Most countries in the sample did not have continuous time series data on unemployment benefits; therefore, the analysis used welfare benefits data series that come directly from the fiscal accounts under spending.

Hiring and firing costs are based on World Bank data.

Neither the welfare benefits variable nor the HIV/AIDs dummy was significant in the estimated panel regressions. Although HIV/AIDs did put a dent in labor force growth in the early 1990s, efforts by governments to address the pandemic attracted donor support, including from the William J. Clinton Foundation and the Bill and Melinda Gates Foundation, and seem to have made significant progress on the rate of new infections, especially among pregnant women.

The average effective tax rate is calculated following Mendoza, Razin, and Tesar (1994), which is a method of producing effective tax rates using data on actual tax payments and national accounts. The main advantage of this method is that it is less stringent on data requirements than other methods and easily applicable to cross-country work.

The sample consists of three SACU members (Botswana, Namibia, and South Africa) and eight other middle-income countries from three regions: Eastern Europe, Latin America, and Asia. The sample period is 1990–2009, and Arellano and Bond’s generalized method of moments panel regression estimation technique was used. Data for Lesotho and Swaziland were not available.

These results are broadly consistent with the results of the growth incidence curves in the October 2011 Regional Economic Outlook for sub-Saharan Africa (IMF, 2011b), which found changes in the coefficients on the level of education are broadly consistent with changes in per capita consumption of the poorest quartile of the distribution for a selected group of economies in sub-Saharan Africa.


    Other Resources Citing This Publication