## _cr12235

## Source details

**Canonical URL:** [_cr12235](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2012/_cr12235.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2012/_cr12235.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2012/_cr12235.pdf.json)

---

### Executive summary — key findings
- Despite sustained high growth over five decades, poverty and inequality remain high in Botswana and other SACU countries; unemployment rate of about 18 percent.
- Reducing income inequalities can significantly increase the duration of growth spells; SACU countries could almost double the duration of growth periods if they achieved inequality levels similar to selected countries at comparable development levels.
- Expansion of welfare programs in Botswana reduced poverty but has been relatively less effective at targeting the very poor compared with other high middle-income countries.
- Policies targeting inequalities at the source (for example, early investments in human capital of the poor) are most effective; carefully crafted fiscal redistributive policies can also be effective while limiting negative effects on growth.
- High unemployment in Botswana and SACU reflects interaction of government employment policies (and their effect on reservation wages) and skill mismatches; addressing low effective cost of capital that favors capital-intensive sectors is also needed to encourage labor-intensive activity.
- Initial results for the two chapters were presented to the Botswana authorities in Gaborone in January and May 2012.

### I. Inequality and growth in the SACU region — context and empirical framing
- Real GDP per capita (PPP, constant 2005 international dollars) in Botswana: from around $3,500 in 1980 to close to $12,500 in 2010; average annual growth rate of 4.3 percent.
- Income inequality measures and poverty (selected figures):
  - Gini coefficient: 0.54 (1985/86) and 0.61 (1993/94).
  - National poverty headcount: 30.6 percent (2002/03) to 20.7 percent (2009/10).
  - Absolute number below poverty line: around 500 thousand (2002/03) to about 373 thousand (2009/10).
  - Rural headcount: 44.8 percent (2002/03) to 25.5 percent (2009/10).
  - Urban headcount: 10.6 percent (2002/03) to 14.0 percent (2009/10).
- Cash earnings account for approximately 73.5 percent of total income (BCWIS 2009/10).
- Note on measurement: part of the decline in provisional poverty estimates may reflect differences between the 43 percent increase in the cost of the consumption basket and the accumulated CPI inflation of around 58 percent over 2002/03–2009/10.
- Methodology:
  - Applies Berg and Ostry (2011) proportional hazard approach to assess how income inequalities affect growth spell duration.
  - Definition of growth spells: periods of real GDP per capita growth of at least 5 years, beginning with an upbreak >2 percent and ending with a downbreak or sample end.
  - Empirical implication: lower income inequalities → longer growth spells on average.

### I.B. Comparative analysis and targeting performance
- Growth incidence (preliminary BCWIS 2009/10 vs 2002/03 HIES): middle percentiles (percentiles 15–75) experienced faster real consumption per capita growth than the lowest 15 percent or highest 25 percent.
- Botswana social safety nets:
  - Spending about 3.2 percent of GDP on social safety nets.
  - Coverage about 20 percent of all poor households (World Bank 2010 Public Expenditure Review for Botswana).
- International lessons summarized (selected outcomes):
  - Brazil: Bolsa Familia contributed to Gini decline from around 0.594 (2001) to 0.539 (2009).
  - Chile: fiscal policy rule plus progressive redistribution and targeted information systems supported sustained growth with declining inequality.
  - Indonesia: macro stability, fiscal discipline, trade liberalization, predictable agricultural policies, and labor-intensive infrastructure investments supported diversification and poverty reduction.

### II. Closing the jobs gap — principal findings
- Main drivers of high unemployment in Botswana and SACU:
  - Government employment policies that affect reservation wages.
  - Skill mismatches between labor supply and labor demand.
  - Low effective cost of capital encouraging capital-intensive activity.
- Key policy implications:
  - Adopt prudent public employment policies to avoid upward pressure on reservation wages and crowding out private employment.
  - Align university, tertiary, and vocational curricula with demand for skilled labor.
  - Address low effective cost of capital to shift incentives toward labor-intensive sectors and job creation.

### Employment, unemployment statistics, and data quality
- Official unemployment rates in SACU range between 20 to 50 percent (Leigh and Flores, 2011).
- Youth unemployment rates in SACU are about twice the average official unemployment rates.
- Labor force surveys: on average about 40 percent of the unemployed in SACU are long-term unemployed (6 months or longer).
- In South Africa about 15 percent of the labor force are discouraged job seekers; in Botswana the proportion of discouraged job seekers is about 25 percent; in Botswana the number of discouraged job seekers is even higher than the number actively seeking work.
- Recommendation: governments in SACU need to invest in improving unemployment data quality to better design and monitor job creation policies.

### Causes and structural diagnostics of unemployment
- Empirical correlations and regression findings (selected exact figures preserved):
  - Employment-output elasticity β averaged around 0.4.
  - Estimated regression line (job creation vs employment-output elasticity): y = 0.0198x + 1.538; R² = 0.2066.
  - Estimated regression line (effective cost of capital vs unemployment rate): y = －4.1461x + 30.153; R² = 0.0882.
  - Hazard ratio: Botswana = 0.7; other SACU countries have hazard ratios significantly above 1.
  - Reported average GDP per capita growth during spells: Namibia = 2¼ percent; South Africa = 3 percent.
- Panel regression highlights (Appendix II.1 and summary):
  - ∆Wages in excess of productivity(-1): 0.84*** (4.45) ; 0.76*** (3.93) ; 0.72*** (3.77).
  - ∆Union density(-1): 0.22** (2.23) ; 0.17** (2.37) ; 0.15** (2.19).
  - ∆Skills mismatch index(-1): 0.65*** (4.22) ; 0.59*** (4.36) ; 0.55*** (4.11).
  - ∆Effective cost of capital(-3): −0.34** (−2.056) ; −0.28** (−2.340) ; −0.33** (−2.182).
  - ∆Labor law restrictiveness(-2): coefficient sign as expected but insignificant.
  - Adj. R-squared: 0.61 ; 0.54 ; 0.59 (number of observations: 220 ; 220 ; 220).
  - Statistical notation: * significant at 10%; ** significant at 5%; *** significant at 1 %.
- Synthesis: key drivers of high unemployment in SACU are wage policy distortions (public sector wage growth exceeding productivity), union density, skill mismatch, and a low effective cost of capital biasing toward capital-intensive activity. Hiring and firing costs and welfare benefits are not significant determinants in panel regressions.

### Skills, education, and rural dimensions
- Education access:
  - SACU countries rank among the best in primary school completion rates in sub-Saharan Africa.
  - Expansion of access reduced inequalities in years of education, but qualitative differences in education contribute to income disparities.
- Skills mismatch:
  - Constructed Skills Mismatch Index (SMI) shows substantial mismatch correlated with unemployment.
  - Despite large spending on education, SACU has produced graduates whose skills are not in demand by the private sector.
  - For Botswana unemployment rates are highest among college graduates; for South Africa highest among unskilled workers.
- Rural employment and land:
  - Large fraction dependent on subsistence agriculture: about ¾ of population in Namibia and Swaziland; ½ in Botswana.
  - Low productivity and limited rural investment constrain human capital returns; unequal land access sustains low-skilled agricultural labor force.

### HIV/AIDS and human capital constraints
- HIV/AIDS prevalence and effects:
  - SACU countries share among the highest prevalence rates in the world (Table 3 reported prevalence rates among 15–49 age group; selected entries):
    - Swaziland: 125.9
    - Botswana: 224.8
    - Lesotho: 323.6
    - South Africa: 417.8
    - Namibia: 713.1
    - World average: 1.9
    - World median: 0.4
    - World st. dev.: 4.4
  - Higher HIV prevalence associated with fewer years of education for affected groups and intergenerational transmission of disadvantage.
- Policy implication: public provision of education and health can correct market failures and externalities; donor assistance could mitigate fiscal risks and help attain quality goals.

### Inequality, growth spell duration, and quantified potential gains
- Empirical finding: higher income inequalities are associated with shorter growth durations; reduction in the Gini coefficient from the 50th to the 60th percentile is associated with 50 percent longer growth period.
- Experiments applying Berg and Ostry (2011) methodology to SACU:
  - Set each country’s inequality to its lowest historical level.
  - Set inequality to the average level encountered in countries of similar development.
- Outcome summary: for most spells, average duration could have increased from about 5–8 years up to 15 years and above (Namibia, Botswana); even marginal improvements in income distribution could substantially prolong growth spells.

### Redistribution and policy design to strengthen growth
- Core priorities:
  - Reduce inequalities through early investments in human capital (education, health).
  - Improve targeting efficiency of existing welfare programs to reach the very poor.
  - Carefully design fiscal redistributive measures to limit adverse incentives on work and investment.
- Complementary policies:
  - Promote private sector development so new skills match vacancies.
  - Reform public sector wage policy to enhance fiscal sustainability and reduce labor-market distortions.
  - Streamline tax incentives that lower the effective cost of capital and bias toward capital-intensive activities.

### Quantitative summaries and regression coefficients (selected exact entries)
- Duration model (summary regression coefficients):
  - Inequality (GINI): −0.05
  - Income per capita at the beginning of the growth spell: −0.11
  - Debt liabilities: 0.00
  - FDI liabilities: 0.02
  - Change of inflation within spell: −0.01
  - First lag of US interest rate change: −0.24
  - Overvaluation of exchange rate: 0.00
  - Polity IV autocracy measure: −0.13
  - Trade liberalization: 0.66
  - Terms of trade growth: 0.01
  - Constant: 5.20
- Regional unemployment comparative figures (selected outside-Africa comparators):
  - Chile: 9.99%
  - India: 3.26%
  - Thailand: 2.61%
  - Malaysia: 4.27%
  - Jamaica: 17.38%
  - Trinidad and Tobago: 13.51%

### Policy recommendations — actionable priorities
- Reduce inequalities:
  - Early investments in human capital of the poor (education, health).
  - Improve targeting efficiency of welfare programs to reach the very poor.
  - Carefully designed fiscal redistributive measures that limit adverse growth effects.
- Reduce structural unemployment:
  - Prudent, restrained public employment policies to avoid upward pressure on reservation wages.
  - Align tertiary and vocational education curricula with private-sector demand; public-private partnerships for skills development.
  - Tackle distortions from a low effective cost of capital (streamline tax incentives) to incentivize labor-intensive sectors.
- Other measures:
  - Economic diversification to create labor-intensive sectors and improve investment climate.
  - Targeted government intervention in key non-tradable sectors with high employment multipliers.
  - Urgent investment in strengthening labor market/unemployment statistics for monitoring and policy design.

*Source: Executive Summary and selected chapter material from IMF staff report _cr12235.*

### Executive Summary ......................................................................................................

### _cr12235 - Executive Summary

### Executive summary — key findings
- Despite sustained high growth over five decades, poverty and inequality remain high in Botswana and other SACU countries, and labor market policies have shortcomings evidenced by a high unemployment rate of about 18 percent.
- Reducing income inequalities can significantly increase the duration of growth spells; SACU countries could almost double the duration of growth periods if they achieved inequality levels similar to selected countries at comparable development levels.
- Expansion of welfare programs in Botswana reduced poverty but has been relatively less effective at targeting the very poor compared with other high middle-income countries.
- Policies that target inequalities at the source (for example, early investments in human capital of the poor) are the most effective to reduce inequalities and promote growth; carefully crafted fiscal redistributive policies can also be effective while limiting negative effects on growth.
- High unemployment in Botswana and SACU reflects the interaction of government employment policies (and their effect on reservation wages) and skill mismatches; addressing low effective cost of capital that favors capital-intensive sectors is also needed to encourage labor-intensive activity.

*Initial results for the two chapters were presented to the Botswana authorities in Gaborone in January and May 2012.*

### I. Inequality and growth in the SACU region — background and context
- Real GDP per capita (PPP, constant 2005 international dollars) in Botswana: from around $3,500 in 1980 to close to $12,500 in 2010, implying an average annual growth rate of 4.3 percent.
- Unemployment rate: approximately 18 percent (remained nearly constant across surveys).
- Income inequality: Gini coefficient reported as 0.54 in 1985/86 and 0.61 in 1993/94.
- Poverty headcount (preliminary BCWIS 2009/10 vs 2002/03):
  - National headcount decline from 30.6 percent in 2002/03 to 20.7 percent in 2009/10.
  - Absolute number below poverty line: around 500 thousand in 2002/03 to about 373 thousand in 2009/10.
  - Rural headcount decline from 44.8 percent in 2002/03 to 25.5 percent in 2009/10.
  - Urban headcount increased from 10.6 percent in 2002/03 to 14.0 percent in 2009/10.
- Cash earnings (excluding business profits, unearned cash income, own produce, wages in kind, aid and school meals) account for approximately 73.5 percent of total income (BCWIS 2009/10).
- Expansion of social programs likely contributed to poverty reduction: School Feeding Program, Vulnerable Group Feeding Program, Old Age Pension program, Orphan Care program, Program for Destitute Persons.
- Note on measurement: part of the decline in provisional poverty estimates may reflect differences between the 43 percent increase in the cost of the consumption basket and the accumulated CPI inflation of around 58 percent over 2002/03–2009/10.

### I.A. Inequality-growth relationship — empirical framing
- The chapter applies Berg and Ostry (2011) methodology to SACU to assess how income inequalities affect growth spell duration.
- Definition of “growth spells” (Berg, Ostry and Zettelmeyer, 2012): periods of real GDP per capita growth of at least 5 years, identified as beginning with an upbreak of per capita growth in excess of a minimum of 2 percent and ending wither with a downbreak followed by a period of an average growth of less than 2 percent, or simply the end of the sample.
- Empirical implication: countries with lower income inequalities experience on average longer growth spells.

### I.B. Comparative analysis and policy-relevant evidence
- Growth incidence estimates (preliminary BCWIS 2009/10 vs 2002/03 HIES) indicate middle percentiles (between percentiles 15 and 75) experienced faster real consumption per capita growth than the lowest 15 percent or highest 25 percent.
- Targeting performance:
  - Botswana spends about 3.2 percent of GDP on social safety nets but covers only about 20 percent of all poor households (World Bank 2010 Public Expenditure Review for Botswana).
  - In contrast, Brazil and Chile achieved significant inequality reduction with lower shares of GDP spent on social safety nets, implying potential efficiency gains from better targeting.
- International lessons (Box I.1) summarized:
  - Brazil: Bolsa Familia expanded and contributed to Gini decline from around 0.594 in 2001 to 0.539 in 2009; program targets poor households with school-age children via conditional cash transfers.
  - Chile: combination of fiscal policy rule, progressive redistribution through universal and targeted programs, and targeted information systems (e.g., Chile Solidario) helped sustain growth while decreasing inequality.
  - Indonesia: macroeconomic stability, fiscal discipline, trade liberalization, predictable agricultural policies, and investments in labor-intensive infrastructure supported diversification and poverty reduction.

### II. Closing the jobs gap in the SACU region — principal findings
- High unemployment in Botswana and other SACU countries stems from:
  - Government employment policies that affect reservation wages.
  - Skill mismatches between labor supply and labor demand.
  - A low effective cost of capital that encourages capital-intensive activity over labor-intensive sectors.
- Policy implications for reducing unemployment:
  - Adopt prudent public employment policies to avoid upward pressure on reservation wages and crowding out private employment.
  - Align university, tertiary, and vocational curricula with demand for skilled labor to reduce skill mismatches.
  - Address low effective cost of capital to shift incentives toward labor-intensive sectors and job creation.

### Empirical notes and data highlights used in the analysis
- Botswana long-run growth performance: average annual growth rate of 4.3 percent (real GDP per capita, 1980–2010).
- Poverty and inequality metrics: national headcount 30.6 percent (2002/03) to 20.7 percent (2009/10); rural 44.8 percent to 25.5 percent; urban 10.6 percent to 14.0 percent; Gini coefficients 0.54 (1985/86) and 0.61 (1993/94).
- Income composition: cash earnings approximately 73.5 percent of total income (BCWIS 2009/10).
- Social safety net spending: about 3.2 percent of GDP in Botswana (World Bank 2010 estimate); program coverage of poor households about 20 percent.

### Policy recommendations — actionable priorities
- Reduce inequalities through:
  - Early investments in human capital of the poor (education, health).
  - Improve targeting efficiency of existing welfare programs to reach the very poor.
  - Carefully designed fiscal redistributive measures that limit adverse growth effects.
- Reduce structural unemployment through:
  - Prudent, restrained public employment policies.
  - Aligning tertiary and vocational education curricula with labor market needs.
  - Tackling distortions from a low effective cost of capital to incentivize labor-intensive sectors.

*Source: Executive Summary, IMF selected issues paper — “Inequality and Growth in the Southern African Customs Union Region” and “Closing the Jobs Gap in the Southern Africa Customs Union Region.”*

### 13.      In order to determine the impact of inequality on growth, Berg et al (2012)

### _cr12235 - 13.      In order to determine the impact of inequality on growth, Berg et al (2012)

### Inequality and growth spell duration
- Berg et al (2012) use a proportional hazard model where the dependent variable is the duration of growth spells and estimate effects of economic and political variables on the probability that a growth spell will end.
- Higher income inequalities are associated with shorter growth durations and "appear to be a major contributing factor (Figure I.1)."
- "Improvements in income distribution, namely a reduction in the Gini coefficient from the 50th to the 60th percentile, would typically be associated with 50 percent longer growth period."
- Controlling for terms of trade, FDI received, price competitiveness, etc., "income inequalities, as measures by the Gini coefficient, do play a very significant impact on growth spells duration."
- Other statistically significant factors (to a lower degree than inequality) include:
  - investment in infrastructure;
  - external shocks (for instance, changes in terms of trade or nominal US interest rate);
  - quality of public institutions, notably as measured by the autocratic degree of political regimes;
  - financial sector development.
- Ethnic, linguistic and religious heterogeneity do not seem to have significant association with length of growth spells.
- Human capital measures (education, health) are associated with improved predicted duration of growth spells.
  - Education measure noted: "Improvement in primary education enrolment rate."
  - Health measure noted: "Child mortality rate level and change."

### SACU growth performance and contributing factors to spell endings
- Over the past decade SACU growth performance has been mixed relative to peers with similar income per capita; except Lesotho, all exhibit weaker GDP per capita growth.
- Despite proximity to a large emerging economy and natural resources, translating potential into improved GDP per capita growth remains difficult.
- Cross-country comparisons show growth-spell vulnerabilities for most SACU countries, except Botswana (and to some extent South Africa).
- South Africa’s growth spell started in 1998 (after the democratic reforms of 1994). Namibia showed capacity to generate high growth rates with a growth spell that started in 1995.
- Reported specific growth figures:
  - "average growth of GDP per capita of 2¼ percent (Namibia) and 3 percent (South Africa)."
- Hazard ratios:
  - A hazard ratio above 1 indicates a higher, country-specific, risk of a growth spell ending compared to the sample average.
  - "All countries have a hazard ratio significantly above 1, except Botswana, with a hazard ratio of 0.7."
- Table 1 and associated discussion identify country-specific contributing factors to spell endings; notable findings:
  - For smaller SACU members, insufficient trade liberalization contributed from 14 percent up to 28 percent to the end of growth spells.
  - Political institutions matter: using Polity IV autocracy measure, autocracy contributed substantially to the end of growth periods in Lesotho and Swaziland (examples: "Lesotho … autocracy was responsible for 24 percent of the end of the growth period," "in Swaziland autocracy contributed 26 percent and 36 percent to the end of the two spells, respectively").
  - The text stresses that concentrated political power can increase inequalities and raise risks of political crises or social unrest, which can lower investment and growth.

### Complexity of causality and country-specific factors
- While high income inequalities are associated with growth spell vulnerabilities, "the causal relation remains complex."
- "For almost all countries, income inequalities have remained high throughout the sample." (Income inequality measured by Gini coefficient has low variation across time.)
- Structural and sectoral features can affect both inequalities and growth; example: Botswana’s heavy dependence on mineral extraction (mostly diamonds) which is not labor intensive and thus less likely to reduce income inequalities.

### Quantified potential gains from reducing inequality
- SACU countries exhibit higher income inequalities than countries with similar GDP per capita; BLNS (Botswana, Lesotho, Namibia, Swaziland) exhibit much higher Gini coefficients than peers; South Africa’s Gini is only slightly higher than the peer average but showed a slight worsening (+1 percent) since 1994.
- Applying Berg and Ostry (2011) methodology to SACU, two experiments considered:
  - Set each SACU country’s inequality to its lowest historical level.
  - Set inequality to the average level encountered in countries of similar development.
- Key summarized outcome: "for all countries, the gains from this hypothetical improvement in inequality could be quite significant. For most spells, the average duration could have been increased from about 5–8 years, up to 15 years and above (Namibia, Botswana)."
- Table 2 (summary lines) reports Gini coefficients and associated changes in average duration; textual highlights include:
  - "The result for Namibia is largely driven by the high degree of income inequalities encountered during its growth spell."
  - "In Botswana, the unusual potential gain is largely driven by the length of the estimated growth spell."
  - "Thus, even marginal improvements in income distribution could result in a much prolonged growth spell."

### Policy recommendations: designing redistributive policies to strengthen growth
- Redistributive policies can potentially improve growth, but must be carefully crafted to avoid negative impacts on work and investment incentives.
- Two main considerations when implementing redistributive policies:
  - "Reducing inequalities in human capital should be at the core of policy intervention aimed at reducing future income inequalities and promoting growth."
  - "In parallel to promoting human capital investment, policies could also help private sector development, so that eventually new skills available are matched with corresponding vacancies. Otherwise the economy could well be trapped in structural imbalances between labor supply and demand."
- "Carefully crafted direct redistribution of income is also desirable, especially to alleviate extreme poverty."

### Promoting human capital investment; health and education constraints
- Income inequalities are primarily related to disparities in human capital (health, education).
- On health: the HIV/AIDS epidemic is a major contributor; SACU countries share the highest prevalence rates in the world (see Table 3).
  - Higher HIV prevalence is associated with fewer years of education for affected groups, contributing to increased inequalities and intergenerational transmission (orphans less likely to attend school).
- On education: health issues negatively affect individual returns to education.
- Market failures explain underinvestment in human capital by the poor:
  - Two constraints: (i) limited resources prevent investment (Galor and Zeira, 1993; Piketty, 1997); (ii) low private returns because of externalities.
  - Public provision of education and health services can facilitate investment by the poor, financed by taxation of richer individuals; there is a trade-off between disincentives from higher taxation and gains from broader human capital provision.
  - By correcting market failures and externalities, public intervention can lead to higher overall investment and higher growth despite fiscal pressure.
- Cost and quality considerations:
  - Public systems (e.g., universal education programs) can be very costly in terms of buildings, teacher training, and overall budget.
  - Donor assistance could be desirable to mitigate fiscal risks and facilitate quality goals.
  - Quality of education is not solely a function of funding; factors like pupils-to-teachers ratio and teacher skills matter.
  - "It could thus be very well the case where growth and inequality reduction gains would be maximized with targeted improvements in the quality of the education system."

### Select numeric facts and tables cited in the chapter
- "Reduction in the Gini coefficient from the 50th to the 60th percentile … associated with 50 percent longer growth period."
- Hazard ratio: Botswana = 0.7; other SACU countries have hazard ratios "significantly above 1."
- Reported average GDP per capita growth during spells: Namibia = "2¼ percent"; South Africa = "3 percent."
- Table 3: Highest HIV/AIDS Prevalence Rates in the World, 2009 (prevalence rates among 15–49 age group; selected entries)
  - Swaziland: 125.9
  - Botswana: 224.8
  - Lesotho: 323.6
  - South Africa: 417.8
  - Namibia: 713.1
  - World average: 1.9
  - World median: 0.4
  - World st. dev.: 4.4
  - Note: Table 3 reports "Prevalence rates computed among the 15-49 age group." and includes population growth rates in a separate column.

*Source: IMF Staff estimates and computations.*

### 23.      SACU countries have made significant efforts to improve the access to

### _cr12235 - 23.      SACU countries have made significant efforts to improve the access to

### Education access and quality
- SACU countries exhibit among the best scores in primary school completion rates in sub-Saharan Africa (Lloyd and Hewett, 2009).
- Expansion of access has reduced inequalities in “quantitative parameters” (years of education).
- Qualitative differences in education contribute to income disparities (van der Berg, 2009; Keswel, 2009) and help explain why overall income inequalities have not been reduced yet in South Africa.
- Human capital investments yield positive growth impact largely when they complement technological efforts and private sector development (Benhabib and Spiegel, 1994; del Barrio, Lopez, and Serrano, 2002; Engelbrecht, 2002; Frantzen, 2002).
- Caution: results are sensitive to the quality of measurement of human capital (De la Fuente and Domenech, 2006).

### Employment opportunities and labor market characteristics
- Official unemployment rates in SACU range between 20 to 50 percent (Leigh and Flores, 2011).
- To provide jobs for current jobless and new entrants, SACU members would have to increase employment by at least 10 million full-time positions over the next decade 2012–21.
- Even achieving that increase would leave the ratio of employment to working age population below 50 percent.
- SACU labor markets are fairly segmented: formal vs. informal, urban vs. rural, and “good jobs sector” vs. “bad jobs sector.”
- Institutional factors (minimum wages, strong unions, centralized wage bargaining, insider-outsider dynamics) contribute to duality and rationing of good jobs.
- There is no single policy fix; a combination of carefully designed initiatives plus faster growth is required to close the jobs gap.

### Rural employment, land access, and poverty dynamics
- A large fraction of populations depend on subsistence agriculture: about ¾ of the population in Namibia and Swaziland, ½ in Botswana.
- Low productivity and low investment in rural areas limit individual profitability of human capital investments (World Bank, 2006).
- The poor in rural areas typically face lack of access to financial services and land ownership.
- Persistence of inequalities underscores the need for continued improvements in access to land and land-asset inequalities inherited from apartheid.
- Successful land reforms should provide fair, transparent, and long-lasting rights and foster the use of land for collateral (World Bank, 2006).
- Unequal access to land ownership can lead to lower education investment and preserve a low-skilled agricultural labor force; more equal land access can incentivize higher education investment and emergence of skilled-labor intensive industrial activities (Galor and others, 2009; Rajan, 2009).

### Private sector development and demand for skills
- Human capital alone rarely has a significant impact on growth; its positive impact arises when it complements technological transfers and private sector development (Mankiw, Romer, Weil, 1992; Pritchett, 2001, 2006).
- Without demand for educated labor, the marginal return on education would decrease rapidly (Pritchett, 2001).
- Private sector development is essential to strengthen existing or develop new comparative advantages and to generate returns on human capital investments.

### Redistributive policies and social protection
- Redistributive policies (cash and in-kind transfers, progressive taxation) have potential benefits for addressing poverty and inequalities.
- Cash and in-kind transfers have been increasingly used by SACU countries and other sub-Saharan African countries (Garcia and Moore, 2012).
- Transfers are part of social protection for vulnerable groups (elderly, orphans) and can mitigate market failures by enabling investments otherwise unaffordable.
- Conditional cash transfers can reduce child labor and promote schooling (e.g., Bolsa Família in Brazil).
- The South African Child Support Grant (introduced 1998 as an unconditional cash transfer) has had some positive impact on health and education attainment by children (Coetzee, 2011).
- Measures to improve labor market flexibility and develop in-work tax credits could preserve work incentives and effective labor matching, but typically imply a trade-off between reducing inequalities and increasing employment.

### Empirical findings, inequality, and growth spells
- Estimates based on preliminary 2009/10 BCWIS data compared with 2002/03 HIES suggest a decrease in inequality (decline in the Gini coefficient) in Botswana over the intervening period.
- Welfare programs in Botswana have expanded but have been relatively less effective in targeting the very poor compared with middle-income countries like Chile, Brazil, and Indonesia.
- Botswana and SACU exhibit a high degree of income and non-income inequality; lowering income inequality has potential to extend the length of growth spells and offer durable solutions to poverty and long-run growth.
- Policy design must balance added fiscal pressures on businesses and individuals with benefits from greater investment in human capital and reduced poverty; fiscal sustainability must be preserved and timing is essential (short-term fiscal costs vs. longer-term benefits).

### Summary of regression coefficients (duration model of growth spells)
- Inequality (GINI) -0.05
- Income per capita at the begining of the growth spell -0.11
- Debt liabilities 0.00
- FDI liabilities 0.02
- Change of inflation within spell -0.01
- First lag of US interest rate change -0.24
- Overvaluation of exchange rate 0.00
- Polity IV autocracy measure -0.13
- Trade liberalization 0.66
- Terms of trade growth 0.01
- Constant 5.20
- Source: Berg and Ostry (2011).

### Regional unemployment context and comparative figures
- Official unemployment rates in SACU range between 20–50 percent and are largely a youth phenomenon.
- Selected countries outside Africa (unemployment or comparable rate as listed):
  - Chile: 9.99%
  - India: 3.26%
  - Thailand: 2.61%
  - Malaysia: 4.27%
  - Jamaica: 17.38%
  - Trinidad and Tobago: 13.51%

*Source: IMF staff report content (file _cr12235 - 23.      SACU countries have made significant efforts to improve the access to).*

### 5.      The unemployment data provides seven key observations (Figure II.2):

### _cr12235 - 5.      The unemployment data provides seven key observations (Figure II.2):

### Key observations on unemployment data quality and levels
- There are significant variations between the official unemployment rates and the implied unemployment rates for both SACU and other regions in SSA, highlighting weaknesses of the unemployment statistics across the region.
- Almost all data sources show that the unemployment rates in SACU are the highest in SSA, followed by those of natural resource rich economies in SSA. The non-natural resource rich economies in the region have on average lower unemployment rates.
- The high unemployment rate in SACU could reflect its better quality unemployment statistics.
- Youth unemployment rates in SACU are about twice the average official unemployment rates.
- Labor force surveys show that on average about 40 percent of the unemployed in SACU are long-term unemployed (unemployed for 6 months or longer).
- The share of people outside the labor force is a more reliable indicator of unemployment than the official unemployment rate; out of employment rates are significantly higher than official unemployment rates for all regions in SSA including SACU.
- In South Africa about 15 percent of the labor force are discouraged job seekers and in Botswana the proportion of discouraged job seekers is about 25 percent. In Botswana the number of discouraged job seekers is even higher than the number of individuals actively seeking work (the official unemployment rate). By contrast, only about 2 percent of people out of the labor force in Chile are discouraged workers.

### Labor market structure and sectoral composition
- With the exception of Lesotho, the average portion of the labor force in the agricultural sector in SACU is about 20 percent, reflecting prevalence of large scale commercial farming in SACU.
- By comparison, Labor Force Surveys report that 80 percent (Ethiopia) and 50 percent (Ghana) of employed workers are in the agriculture sector.
- Interpretation: SACU may have higher quality jobs and higher labor productivity that can coexist with higher unemployment (efficiency wage theory).

### Labor force composition and participation
- Female labor force participation (LFP) tends to be lower in SACU countries.
- Female LFP rate: South Africa is estimated at 46 percent; Ethiopia is estimated at 71 percent.
- High unemployment together with low LFP has resulted in very low ratios of employment to working-age population in SACU.

### Data quality recommendation
- Governments in SACU and more generally SSA need to invest more in improving the quality of the region’s unemployment data to better design and monitor job creation policies.

### Empirical analysis — Employment-Output Elasticity
- A commonly held view: to reduce unemployment substantially, investment must increase and SACU economic growth needs to be in the range of 6-10 percent.
- Estimated employment-output elasticity β averaged around 0.4; the constant term in the panel regression α was consistently negative in all estimated regressions.
- Adjusted R2 for the 4 respective regressions: 0.61, 0.58, 0.57 and 0.54.
- Estimated regression line (job creation vs employment-output elasticity): y = 0.0198x + 1.538; R² = 0.2066.
- Estimated regression line (effective cost of capital vs unemployment rate): y = －4.1461x + 30.153; R² = 0.0882.

### Cost of capital and structural distortions
- SACU members 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.
- The lower the cost of capital the higher the unemployment rate (Figure II.5).
- Interpretation: low effective cost of capital and significant regression intercepts suggest structural distortions in the SACU labor market that may contribute to persistently high unemployment and prevent market clearing.

### Wage policy, unions, and unemployment
- Public sector real wage growth in excess of productivity is closely correlated with the unemployment rate in SACU.
- SACU’s real wages in excess of productivity gains is significantly higher than other countries in the sample.
- Size of public sector and higher public sector wage awards influence private sector’s ability to create jobs; a bloated public sector can distort labor market outcomes and inflate wage expectations.
- Centralized collective bargaining contributes to weak link between pay and productivity and reduces real wage responsiveness to the business cycle.
- High real wage growth above productivity encourages firms to substitute capital for labor and increase informality.
- Union density is relatively high in SACU and shows a high degree of correlation with unemployment; panel regressions support a negative impact of union density on SACU’s overall unemployment rate.

### Skill mismatch and unemployment
- Skills Mismatch Index (SMI) constructed as sum over skill levels j of (Sijt − Mijt)^2 (using primary = low skilled, secondary = semi-skilled, tertiary = high skilled for supply; construction = low-skilled demand, manufacturing = semi-skilled demand, government and financial services = high-skilled demand).
- SACU shows substantial skill mismatch correlated with unemployment rates.
- Despite large spending on education, SACU has produced graduates whose skills are not in demand in the private sector.
- For Botswana unemployment rates are highest among college graduates; for South Africa they are highest among unskilled workers.
- Policy measures to reduce skill mismatch recommended: improve quality of education spending to support public-private partnerships for skills development, vocational and technical training, building ICT skills, and graduates’ internship programs (as in Botswana).

### Welfare benefits, labor market regulations, and unemployment
- Welfare benefits are not closely correlated with unemployment in SACU; welfare spending in SACU is on average lower than other regions in SSA and not associated with increased voluntary unemployment via replacement ratios.
- Panel regressions support the finding that welfare programs can help the unemployed and discouraged workers without raising unemployment.
- Hiring and firing costs show a low degree of correlation with unemployment rates in SACU; SACU countries have lower hiring and firing costs despite having the highest unemployment rates in SSA. Other regions with relatively high hiring and firing costs have lower unemployment rates.

*Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2012/_cr12235.pdf*

### 17.      Our estimated panel regressions (Appendix II.1) also show that hiring and firing

### Our estimated panel regressions (Appendix II.1) also show that hiring and firing costs is not a significant determinant of the overall level of unemployment in SACU.

### Determinants of unemployment — empirical findings
- Hiring and firing costs is not a significant determinant of the overall level of unemployment in SACU (Appendix II.1).
- The high unemployment rate in SACU is not closely associated with labor market rigidities.
- Minimum wages (as a share of average wages in the economy) in SACU are relatively on the low side compared with some other countries (Figure II.6).
- The estimated panel regressions suggest that the high unemployment has little to do with the restrictiveness of the labor laws.
- Both the welfare benefits variable and the HIV/AIDs dummy were not significant in the estimated panel regressions. While HIV/AIDs reduced labor force growth in the early 1990s, government and donor efforts yielded significant progress on the rate of new infections especially among pregnant women.

### Correlations highlighted in figures (qualitative)
- Unemployment rate appears positively correlated with:
  - wage-productivity gap,
  - union density,
  - skill mismatch in the labor market.
- Unemployment rate appears to have little association with:
  - welfare benefits,
  - restrictiveness of labor laws.
- SACU countries have generally low minimum wages compared to peer countries (Figure II.6).

### Demographic factors
- Demographic pressures have not been a dominant factor explaining unemployment outcomes in SACU.
- Population growth is trending downwards (below 1 percent) across SACU countries (Figure II.7).
- This compares to an estimated 2½ percent annual population growth for the whole of SSA.
- The size of the working population as a ratio of total population is projected to increase as the impact of HIV/AIDs dissipates, but these ratios would generally remain low by standards in other regions.
- The demographic variable is not significant in the estimated panel regressions.

### Synthesis of causes (Summary of Results)
- The balance of evidence suggests SACU’s high unemployment rate is largely driven by its wage policy which distorts labor market outcomes, including through its impact on education choices and thus skill mismatch.
- Government hiring practices and the wage structure of the civil service have:
  - inflated wage expectations,
  - placed a premium on graduates with liberal art or social science degrees over actual skills in demand in the private sector,
  - influenced education choices and contributed to skills mismatch,
  - created voluntary unemployment as graduates wait for public sector jobs given their reservation wage,
  - exacerbated the misalignment between labor productivity and real wages, discouraging employment creation.
- A highly unionized government sector can also affect employment outcomes.
- The confluence of government hiring practices, public sector wage policy, and less flexible wage bargaining processes gives rise to distortions requiring fundamental change in SACU.
- Policy implication: SACU countries should change public sector wage policy to enhance fiscal sustainability and reduce distortions in labor markets.

### Skills, education, and sectoral policy implications
- At prevailing wage rates there is excess demand for skilled labor and in some countries excess supply of unskilled labor.
- Required actions:
  - Targeted interventions in key sectors combined with comprehensive education reform to create conditions for rapid growth with job creation.
  - Private sector-led improvements in technical/vocational training, with government standardizing curricula and accrediting programs.
- Relevant international experiences:
  - South Korea: reduced emphasis on university education and promoted tertiary and vocational training.
  - Japan: close links between training and industry, continuous curriculum development, and programs focused on job-market skills (Treichel, 2010).
- Governments in SACU have begun some of these initiatives.

### Tax policy and capital-labor bias
- The effective cost of capital variable is significant in the unemployment panel regressions, supporting the view that policies in SACU could be biased toward capital-intensive sectors at the expense of labor-intensive sectors.
- SACU’s wide-ranging tax incentives have resulted in a low effective tax rate on capital relative to labor (Figure II.8).
- Since the early 1980s, tax incentives have proliferated and produced an effective tax rate on capital which is low and favors capital-intensive activities.
- Policy recommendation: Streamlining tax incentives for capital will raise the effective cost of capital and reduce the distortionary impact on employment creation.

### Case study contrasts — Botswana and Chile (Box II.1) — policy lessons
- Despite similar fundamentals, Botswana has double-digit unemployment while Chile has generally kept unemployment below 10 percent.
- Factors associated with Chile’s better unemployment outcomes:
  - Sound fiscal policy, reduced size of government, and a composition of government spending that favors growth; public sector wage growth broadly in line with economy-wide productivity levels.
  - Better education outcomes and progress in addressing skill mismatch; greater labor-force occupational flexibility.
  - Chile has 30 percent enrollment in tertiary education compared with only 12.6 percent in South Africa (the highest in sub-Saharan Africa).
  - Progress in diversifying the economy away from primary commodity dependence, expanding services in value added and employment.

### Unemployment and income inequality
- High structural unemployment hinders inclusive growth.
- Analysis suggests sustained GDP growth alone cannot improve income inequality if not associated with reductions in long-term structural unemployment.
- Reductions in structural unemployment have a substantial positive impact on income distribution.
- Better education outcomes that reduce structural unemployment can reduce income inequality and make growth more inclusive.
- Policy implications: incentives to hire less-skilled workers and training programs for workers with stagnant wages or long unemployment spells.

### Conclusions and policy recommendations
- Job creation is a key challenge; no single measure will suffice — a combination of carefully designed initiatives and faster growth is required.
- Closing the jobs gap would require:
  - Faster economic growth — the analysis suggests economic growth to around 6-10 percent on average to put a significant dent in the unemployment rate in SACU.
  - A fundamental change in public sector wage policy to enhance fiscal sustainability and reduce labor-market distortions.
  - Aligning education policies to private-sector skill needs via public-private partnerships, vocational and technical training, and ICT skills development.
- Other policy initiatives to generate faster job creation:
  - Economic diversification to create labor-intensive sectors, limit Ballassa-Samuelson effects on non-tradables, and improve the investment climate and reduce the costs of doing business.
  - Targeted government intervention in key non-tradable sectors with high employment multipliers.
- Data and monitoring:
  - Urgent need for governments in SACU to invest in strengthening labor market/unemployment statistics.
  - Significant variations between official unemployment rates and implied unemployment rates highlight severe weaknesses in the unemployment data.
  - Better quality statistics are required to monitor policy effectiveness for job creation.

*Source: IMF staff analysis extracted from the provided content unit.*

### APPENDIX II.1

### APPENDIX II.1

### Table 1 — Estimated Panel Regressions (baseline sample: 33 countries, 1990-2009)
- Estimation techniques: pooled-regression, fixed effects estimator, Arellano-Bond GMM panel regression.
- Key coefficient signs and significance (lags indicated):
  - ∆Unemployment(-2): −.6.2*** (t-statistic (−5.14)) — reported under dynamic estimation.
  - Constant: 0.019 (0.85) ; 0.013 (0.63) ; 0.016 (0.49).
  - ∆Wages in excess of productivity(-1): 0.84*** (4.45); 0.76*** (3.93); 0.72*** (3.77).
  - ∆Union density(-1): 0.22** (2.23); 0.17** (2.37); 0.15** (2.19).
  - ∆Labor law restrictiveness(-2): 0.023 (1.25); 0.015 (0.86); 0.008 (0.62) — coefficient sign as expected but insignificant in all regressions.
  - ∆Skills mismatch index(-1): 0.65*** (4.22); 0.59*** (4.36); 0.55*** (4.11).
  - ∆Demographic(-2): 0.098 (1.83); 0.082 (1.76); 0.077 (1.50).
  - ∆Effective cost of capital(-3): −0.34** (−2.056); −0.28** (−2.340); −0.33** (−2.182).
  - Error-correction mechanism(-1): −0.056*** (−5.78); −0.049*** (−4.77); −0.037*** (−3.92).
- Goodness of fit and sample:
  - Adj. R-squared: 0.61 ; 0.54 ; 0.59.
  - Number of observations: 220 ; 220 ; 220.
- Notes on interpretation from table text:
  - A rising wage-productivity gap leads to an increase in the unemployment rate after a 1-year lag.
  - Union density raises unemployment after a 1-year lag.
  - Increase in skill mismatch raises unemployment.
  - Lowering of the effective cost of capital raises unemployment within a 3-year lag.
  - Labor law restrictiveness has the expected sign but is insignificant in all estimated panel regressions.
- Statistical notation:
  - Entries in parentheses are the calculated t-statistics.
  - * significant at 10%; ** significant at 5%; *** significant at 1 %.

### Estimation of the Link between Structural Unemployment and Income Inequality — methodology and rationale
- Purpose:
  - Examine link between structural unemployment and income inequality within the SACU context using a panel including three SACU countries and selected middle-income countries.
  - Decompose unemployment into structural (trend) and cyclical (deviation) components and investigate impact on income distribution, controlling for inflation.
- Hypotheses:
  - Temporary increases in unemployment (cyclical) may worsen income inequality if marginal, low-skill workers (bottom of income distribution) are laid off first.
  - Transitory unemployment income loss may be offset by unemployment insurance and welfare benefits, especially with growing incidence of dual earners.
- Structural unemployment estimation approaches:
  - Benchmark: regress unemployment on a constant, and linear and quadratic trend terms; fitted values = structural unemployment, trend deviations = cyclical unemployment.
  - Robustness checks:
    - Hodrick-Prescott (HP) filter.
    - Kalman filter (estimates trend using all observations).
  - Structural unemployment series from HP filter and fitted trends are broadly similar; benchmark uses fitted linear and quadratic trend series.
- Sample and estimation:
  - Sample: 11 middle-income countries including 3 SACU members (Botswana, Namibia and South Africa); Lesotho and Swaziland data not available.
  - Sample period: 1990-2009.
  - Estimation method: Arellano-Bond’s GMM panel regression.
  - Unit-root and cointegration tests to be reported in forthcoming working paper (Augmented-Dickey Fuller, Phillips-Perron, Co-integrating Durbin-Watson Statistics (CDWS)).

### Table 2 — Structural Unemployment and Income Inequality (Arellano-Bond dynamic panel results)
- General result summary across specifications (Structural unemployment measured by fitted trend, HP filter, and Kalman filter):
  - An increase in structural unemployment is associated with:
    - An increase in the income share of the fourth highest and highest quintiles.
    - A decrease in the income share of the lowest three quintiles.
  - Changes in cyclical unemployment have no consistent impact on the income share of the richest forty percent.
- Panel A: Structural Unemployment from Fitted Trend (coefficients with t-statistics)
  - Constant: −0.201 (−1.609) ; 0.295* (−1.914) ; 0.393** (−2.277) ; 0.285 (−1.54) ; −1.165 (−1.914).
  - Structural Unemployment: −0.038 ** (−2.194) ; −0.058 ** (−2.002) ; −0.073 ** (−2.180) ; 0.049** (2.476) ; 0.216** (2.216).
  - Cyclical Unemployment: −0.026 (−1.516) ; 0.041** (−2.039) ; −0.014 (−0.545) ; 0.002 (0.09) ; 0.068 (−1.043).
  - Inflation: 0.051** (4.301) ; 0.031** (−2.234) ; 0.030* (−1.931) ; −0.007 (−0.422) ; −0.107** (−2.372).
  - R-squared: 0.55 ; 0.48 ; 0.60 ; 0.41 ; 0.58.
  - Durbin-Watson: 2.34 ; 2.26 ; 2.47 ; 2.25 ; 2.55.
- Panel B: Structural Unemployment from Hodrick-Prescott Filter
  - Constant: 0.218* (1.730) ; 0.240 (1.605) ; 0.320* (−1.877) ; 0.227 (1.254) ; −0.987 (−2.030).
  - Structural Unemployment: −0.040* (−1.992) ; −0.049** (−2.785) ; −0.061 * (−1.991) ; 0.039 (−2.226) ; 0.186** (2.042).
  - Cyclical Unemployment: −0.024 (−1.288) ; 0.047** (−2.138) ; −0.017 (−0.692) ; −0.0004 (−0.0009) ; 0.077 (−1.077).
  - Inflation: 0.051** (−4.337) ; 0.029** (−2.089) ; 0.028* (−1.759) ; −0.009 (−0.522) ; −0.102** (−2.226).
  - R-squared: 0.43 ; 0.35 ; 0.37 ; 0.41 ; 0.37.
  - Durbin-Watson: 2.35 ; 2.22 ; 2.42 ; 2.25 ; 2.49.
- Panel C: Structural Unemployment from Kalman Filter
  - Constant: 0.156 (1.660) ; 0.170 (1.545) ; 0.193 (−1.507) ; 0.126 (0.925) ; −0.607 (−1.654).
  - Structural Unemployment: −0.029 ** (−2.680) ; −0.037** (−2.603) ; −0.039** (−2.703) ; 0.022 (−2.226) ; 0.120* (2.765).
  - Cyclical Unemployment: −0.036 (−1.214) ; 0.082 (−2.367) ; −0.027 (−0.667) ; −0.003 (−0.070) ; 0.135 (−1.174).
  - Inflation: 0.049** (4.012) ; 0.024* (−1.669) ; 0.024* (−1.469) ; −0.011 (−0.647) ; −0.102** (−1.838).
  - R-squared: 0.45 ; 0.34 ; 0.18 ; 0.46 ; 0.39.
  - Durbin-Watson: 2.31 ; 2.22 ; 2.33 ; 2.17 ; 2.40.
- Source and statistical notes:
  - Source: IMF staff calculations.
  - Entries in parentheses are the calculated t-statistics.
  - * significant at 10%; ** significant at 5%; *** significant at 1 %.
  - Estimation Method: Dynamic Panel Data Modelling using the Arellano-Bond Estimator.

*Source: _cr12235 - APPENDIX II.1 (IMF staff calculations).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2012/_cr12235.pdf_