## _wp16137

## Source details

**Canonical URL:** [_wp16137](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16137.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16137.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16137.pdf.json)

---

### Matching algorithm performance and sample construction
- Panel constructed from Quarterly Labor Force Survey (QLFS) of South Africa from 2008Q1 to 2014Q3; survey interviews approximately 30,000 households every quarter.
- Rotating panel: 25 percent of housing units rotated out every quarter; housing units surveyed for four consecutive quarters.
- Individual-level panel constructed by matching housing-unit and individual identifiers, checking sex and race consistency and that age does not differ by more than one year.
- Total observations: 2,355,379; matched to next quarter: 1,318,338; match rate: 56 percent.
- Comparison to official Stats SA individual-level panel (one-off panel between 2013 Q3 and 2013 Q4):
  - Out of total observations in 2013 Q3, both Stats SA and the algorithm matched 62.70 percent and failed to match 33.98 percent.
  - For 96.68 percent of the sample, the algorithm gave the same outcome as the official panel.
  - Algorithm generated 0.36 percent false matches.
- Matched sample restricted to individuals aged 15 to 64 for analysis.

### Representativeness and limitations
- Attrition and household moves can produce a matched sample that is not fully representative.
- Matched sample is quite similar to the full sample in observable characteristics (shares by age, sex, race, and education groups are similar).
- Survey weights used in regressions; standard errors clustered at the household level.

### Transition analysis: definitions and estimation
- Transition dummies:
  - Job-finding rate: 1 if unemployed in t and employed in t+1; 0 otherwise.
  - Job-exit rate: 1 if employed in t and unemployed in t+1; 0 otherwise.
  - Transition to formal employment: 1 if unemployed or informal in t and employed in formal sector in t+1; 0 otherwise.
- Two definitions of unemployment used: baseline (broad, includes discouraged workers) and robustness (narrow).
- Probit regressions estimated; reported results are marginal effects (dy/dx), not raw coefficients.
- Regressions include time dummies and province dummies.
- Individual categorical covariates:
  - Education: less than primary; primary but less than secondary; secondary but less than university; university or more.
  - Race: Black; Colored; Indian/Asian; White.
  - Age groups: 15 to 24; 25 to 34; 35 to 44; 45 to 54; 55 to 64.
  - Gender: male; female.
- Additional covariates:
  - Job-finding: Dummy for long-term unemployed (1 if unemployed for less than a year? — preserved from source wording: Dummy for long-term unemployed takes value 1 for those who have been unemployed for less than a year and 0 otherwise); Dummy for experience (1 if never employed before; 0 if has been employed before).
  - Job-exit: Sector of employment (formal, informal, private household); Contract duration (limited term, permanent, unspecified duration, not applicable).
- Two versions of some regressions include trade-union membership (question asked only after 2010Q3; smaller sample).

### Key micro findings — job‑finding rates (Appendix II levels and marginal patterns)
- Job‑finding probabilities (levels from Appendix II, Table 1):
  - Individuals with prior work experience: 0.131.
  - Individuals who have never worked before: 0.071.
  - Implied inexperienced job‑finding rate is about 46 percent lower than experienced (0.071 vs 0.131).
- Experience and demographics:
  - People with prior work-experience have almost 50 percent higher job-finding rate than those without experience; effect stronger for young job seekers.
  - Those unemployed for short periods have a finding probability 0.072 (45 percent) more on average than the long‑term unemployed.
  - Older individuals and men are more likely to transition into employment compared to 15 to 24 year olds and women.
- Education and job‑finding:
  - Job‑finding rate does not differ substantially across education groups and race.
  - Suggested explanations: higher reservation wages for more educated individuals; selection (lower unemployment among more educated); low quality of education weakens signal of productivity.
- Informal sector as stepping stone:
  - Probability that an unemployed individual (with prior work experience) finds a formal sector job: 0.072.
  - Probability that an informal sector worker finds a formal sector job: 0.112.
  - Informal sector workers have a 55 percent higher probability of transitioning into formal sector employment than the unemployed (0.112 vs 0.072).

### Key micro findings — job‑exit rates (Appendix II levels and marginal patterns)
- Job‑exit probabilities (levels from Appendix Table 2):
  - Less than primary: 0.055 (col 1 broad) and 0.033 (col 3 narrow).
  - University or more: 0.025 (col 1) and 0.019 (col 3).
  - Implied job-exit of university vs less than primary: about 55 percent lower (0.025 vs 0.055 in col 1).
- Education strongly reduces job-exit: university-educated have much more stable jobs.
- Race, age, sector effects:
  - Whites have much lower job‑exit rates compared to Blacks: Whites 0.020 (col 1) vs Blacks 0.051 (col 1) — difference 0.032 (almost 60 percent lower).
  - Age 55-64 job‑exit 0.022 (col 1) vs age 15-24 job‑exit 0.082 (col 1) — difference 0.06 (75 percent lower).
  - Formal sector workers less likely to be separated than informal sector workers (Formal 0.042 vs Informal 0.066, col 1).
- Trade-union effects:
  - Union members: job‑exit probability 0.031 (col 2); non-members: 0.049 (col 2) — not being a union member increases job‑exit probability by around 58 percent (0.031 vs 0.049).
  - Industry trade-union density has a negative effect on future job‑finding: raw coefficient -0.231, significant at 1 percent (clustered SEs at industry‑time level).

### Business cycle and aggregate interpretation
- During the crisis (sample includes the GFC), unemployment increased due to a fall in job‑finding rate rather than a rise in job‑exit rate.
  - Time‑dummy coefficients indicate job‑exit rate remained more or less unchanged while job‑finding rate declined substantially.
- Result accords with literature that fluctuations in unemployment are driven mainly by job‑finding rate.

### Aggregate counterfactuals (steady‑state stock‑flow exercises)
- Methodology: steady state of stock‑flow transition equations with quarterly transition rates estimated from micro panel; substituting sample averages gives steady state share of unemployed u = 32.2 percent; average unemployment rate for pooled sample = 31.8 percent.
- Counterfactuals (using reported levels exactly):
  - If all individuals had the job‑finding rate of experienced individuals (0.131 vs sample average 0.106), steady state unemployment = 28 percent (reduction of 4 percentage points from average).
  - If all individuals had the job‑finding rate of the inexperienced, steady state unemployment = 39 percent.
  - Difference implies aggregate unemployment rate difference of 11 percentage points (39 percent vs 28 percent).
  - If all had the job‑finding rate of the short‑term unemployed, aggregate unemployment = 25 percent.
  - If all had the job‑finding rate of the long‑term unemployed, aggregate unemployment = 35 percent.
  - Job‑exit (education) counterfactuals:
    - If everyone had one higher level of education (and corresponding lower job‑exit rate), counterfactual unemployment = 30 percent.
    - If everyone had the job‑exit rate of college graduates, counterfactual unemployment = 24 percent.
- Transition probability matrix (t to t+1) reported:
  - Employed → Employed: 0.931
  - Employed → Unemployed: 0.048
  - Employed → OLF: 0.020
  - Unemployed → Employed: 0.106
  - Unemployed → Unemployed: 0.755
  - Unemployed → OLF: 0.138
  - OLF → Employed: 0.025
  - OLF → Unemployed: 0.095
  - OLF → OLF: 0.880
- Example counterfactual unemployment rates (reported):
  - Original Unemployment: 0.32
  - If everyone has some experience before: 0.28
  - If no one has experience before: 0.39
  - If everyone moves to the next level of education: 0.32 (finding-rate panel)
  - If all are graduates: 0.31 (finding-rate panel)
  - If everyone was temporarily unemployed: 0.25 (exit-rate panel)
  - If everyone was long term unemployed: 0.35 (exit-rate panel)
  - If everyone moves to the next level of education: 0.30 (exit-rate panel)
  - If all are graduates: 0.24 (exit-rate panel)

### Cross‑country comparison
- Using ILO KLIM dataset (monthly outflow/inflow rates) for 2001–2012:
  - Ratio of job‑finding to exit rate estimated in this paper = 2.5 (Table 6).
  - Ratio in the KLIM data = 3.1.
  - South Africa’s job‑finding rate is one of the lowest among selected countries; only Romania and Turkey have lower job‑finding rates in the sample.
  - Countries with lower job‑finding rates also have much lower job‑exit rates.
  - South Africa has the lowest ratio of job‑finding to job‑exit rate in the sample.

### Inequality simulations and redistribution counterfactuals (NIDS wave 3, 2012)
- Data: NIDS wave 3 (2012) covered 8,040 households and 32,633 interviewed individuals.
- Baselines:
  - Baseline individual-level post‑tax income Gini (labor force sample) = 0.68.
  - Baseline household per‑capita income Gini = 0.665 (includes imputed rent and transfers).
- Simulation procedure (partial equilibrium, random selection):
  1. Compute baseline inequality in NIDS.
  2. Randomly select a subsample of unemployed to become employed.
  3. Assign incomes to the newly employed based on quantile regressions of log income on age, education, race and sex; predicted income assigned if higher than their unemployed income.
  4. Recompute inequality.
- Simulation results (reported):
  - Individual‑level Gini: starting at 0.68; decline in unemployment of 10 percentage points leads to Gini falling to 0.645 when using the median.
  - Household per‑capita Gini: baseline 0.665; decline in unemployment of 10 percentage points reduces Gini to 0.645 in median-based simulations.
- Redistribution counterfactual:
  - A shift in Gini from 0.665 to 0.645 would require an increase in transfers of about 40 percent, which was equivalent to 3.3 percent of GDP in 2012 or about 11.1 percent of government expenditure.
- Caveats:
  - Static exercise; dynamic effects (experience and skill accumulation) could lead to larger reductions in inequality.
  - Assigning a single predicted income per demographic cell understates within-cell dispersion and may overstate fall in inequality; robustness checks using full distribution of employed incomes for the newly employed produced larger declines in inequality.

### Stylized marginal-effects excerpts from transition regressions (selected reported figures)
- Job-Finding Rate (marginal effects, selected):
  - Female: -0.031*** (column 1) and -0.032*** (column 2); -0.036*** (columns 3 and 4).
  - Never employed: -0.061*** (column 1) and -0.052*** (column 2).
  - Short term unemployed (<1 year): 0.072*** (columns 1 and 2) and 0.078*** (columns 3 and 4).
  - Observations: 173,626 (columns 1 and 2); 105,887 (columns 3 and 4).
- Job-Exit Rate (marginal effects, selected):
  - University or more: -0.030*** (column 1) and -0.028*** (column 2); -0.015*** (columns 3 and 4).
  - Age 25-34: -0.025*** (col 1); Age 55-64: -0.060*** (col 1).
  - Informal sector: 0.024*** (column 1) and 0.017*** (column 2).
  - Permanent job type: -0.077*** (column 1) and -0.066*** (column 2).
  - Observations: 328,063 (columns 1 and 3); 174,683 (columns 2 and 4).
- Transition to formal employment (Unemployment/Informal → Formal, marginal effects, selected):
  - Work in Informal sector (if unemployed): 0.052*** (col 1) and 0.049*** (col 2).
  - Unemployed never worked: -0.039*** (col 1) and -0.040*** (col 2).
  - Education gradient (col 1 examples): Primary but not secondary 0.013***; Secondary but not university 0.036***; University or more 0.071***.
  - Female: -0.042*** (col 1) and -0.040*** (col 2).
  - Observations: 270,203 (cols 1 and 2); 196,021 (cols 3 and 4).

### Policy‑relevant implications and recommendations (as stated)
- Providing work experience (formal or informal) is key to lowering unemployment in the short run, especially for youth, women, and blacks.
- Targeted policy interventions such as the Youth Tax Incentive Scheme (Employment Tax Incentive Act cited in source) may help improve employability of young South Africans.
- Labor market reforms to increase the influence of outsiders may help create jobs (reduce insider-outsider frictions).
- Improving educational outcomes remains crucial for longer-term unemployment reduction, but education alone may not address short-run employability without experience.
- Reducing unemployment is a more viable and sustainable way of reducing income inequality compared to redistribution alone, given the high fiscal cost of matching the inequality gains via transfers.

*Source: _wp16137 — sections I–III, Appendix II, Appendix Tables 2–3, and V (as provided).*

### 1. Comparison of Matching Algorithm to Official Panel—2013Q3 to 2013Q4 .....................28

### 1. Comparison of Matching Algorithm to Official Panel—2013Q3 to 2013Q4 .....................28

### Matching algorithm performance and sample construction
- Panel constructed from Quarterly Labor Force Survey (QLFS) of South Africa from 2008Q1 to 2014Q3, survey interviews approximately 30,000 households every quarter.
- Rotating panel: 25 percent of housing units rotated out every quarter; housing units surveyed for four consecutive quarters.
- Individual-level panel constructed by matching housing-unit and individual identifiers, then checking sex and race consistency and that age does not differ by more than one year.
- Total observations: 2,355,379; matched to next quarter: 1,318,338; match rate: 56 percent.
- Comparison to official Stats SA individual-level panel (one-off panel between 2013 Q3 and 2013 Q4):
  - Out of total observations in 2013 Q3, both Stats SA and the algorithm matched 62.70 percent and failed to match 33.98 percent.
  - For 96.68 percent of the sample, the algorithm gave the same outcome as the official panel.
  - Algorithm generated 0.36 percent false matches.

### Representativeness and limitations of matched sample
- Attrition and household moves can produce matched sample that is not fully representative.
- Table 2 (referenced) indicates matched sample is quite similar to the full sample in observable characteristics: shares by age, sex, race, and education groups are similar for full sample and matched sub-sample.
- Analysis restricted to individuals aged 15 to 64 (working age population).

### Use of matched panel for transition analysis
- Panel allows following the same individual across quarters to analyze transition probabilities:
  - Job-finding rate: dummy = 1 if unemployed in quarter t and employed in quarter t+1; 0 otherwise.
  - Job-exit rate: dummy = 1 if employed in quarter t and unemployed in quarter t+1; 0 otherwise.
  - Probability of transitioning to formal employment: dummy = 1 if unemployed or informal in quarter t and employed in formal sector in t+1; 0 otherwise.
- Two versions of LHS variables constructed as robustness check: baseline uses broad definition of unemployment (includes discouraged workers), robustness uses narrow definition.
- Regressions include time dummies and province dummies; survey weights used; standard errors clustered at the household level.

### Key methodological details
- Probit regressions used; reported results are marginal effects (dy/dx), not raw coefficients.
- Individual-level characteristics included categorically:
  - Education: less than primary; primary but less than secondary; secondary but less than university; university or more.
  - Race: Black; Colored; Indian/Asian; White.
  - Age groups: 15 to 24; 25 to 34; 35 to 44; 45 to 54; 55 to 64.
  - Gender: male; female.
- Additional variables:
  - For job-finding regressions: Dummy for long-term unemployed (takes value 1 for those who have been unemployed for less than a year and 0 otherwise); Dummy for experience (takes value 1 if the individual has never been employed before and 0 if she has been employed before).
  - For job-exit regressions: Sector of employment (formal, informal, private household); Contract duration (limited term, permanent, unspecified duration, not applicable).

### Findings from transition analysis (overview points presented in source)
- Prior work-experience is an important determinant of job-finding rates; education has almost no effect in job-finding regressions.
- People with prior work-experience have almost 50 percent higher job-finding rate than those without experience; experience is even more important for young job seekers.
- Counterfactual exercise suggests difference in job-finding rates between experienced and those without experience can result in an aggregate unemployment rate difference of 11 percentage points.
- Long-term unemployment lowers future job-finding rates; women have lower job-finding rates compared to men.
- Education is an important determinant of job-exit rates: individuals with higher education have significantly lower job-exit rates.
- Young individuals and women have both low job-finding rates and high job-exit rates.
- Blacks have higher job-exit rates compared to other racial groups.
- Trade-union membership increases job-security for insiders, but higher union density in previous industry reduces job-finding prospects for outsiders (harmful effect on outsiders).
- Informal sector acts as a stepping stone to formal employment: unemployed individuals have a much lower probability of finding a formal sector job compared to those working in the informal sector.

### Inequality counterfactuals summarized in source
- Using data from the third wave of the National Income Dynamics Study (NIDS) conducted in 2012:
  - Reducing unemployment by 10 percentage points would lead to a fall in the Gini coefficient from 0.665 to 0.645.
  - Achieving a similar reduction in Gini solely through fiscal transfers would require an increase in transfers by about 40 percent, equivalent to 3.3 percent of GDP in 2012 or about 11.1 percent of government expenditure.
- Implication: without progress on reducing unemployment, reducing inequality through fiscal transfers alone is difficult.

### Policy-relevant implications highlighted in source
- Providing work experience (formal or informal) is key to lowering unemployment in the short run, especially for youth, women, and blacks.
- Targeted policy interventions such as the Youth Tax Incentive Scheme may help improve employability of young South Africans.
- Labor market reforms to increase the influence of outsiders may help create jobs.
- Improving educational outcomes remains crucial for longer-term unemployment reduction, but education alone may not address short-run employability without experience.

*Source: _wp16137 - 1. Comparison of Matching Algorithm to Official Panel—2013Q3 to 2013Q4, sections I–III (as provided).*

### Appendix II reports the levels of the job-finding rate from the same regressions.

### _wp16137 - Appendix II reports the levels of the job-finding rate from the same regressions.

### Job‑finding rates: micro results and patterns
- Job‑finding rate does not differ substantially across different education groups and race.
- Possible explanations suggested:
  - Higher reservation wages for more educated individuals leading to longer search.
  - The educated unemployed are a selected sample because unemployment among the more educated is lower (due to lower job-exit rates).
  - Low quality of education may make educational attainment an unreliable signal of productivity and employability.
- Experience is an important determinant:
  - Estimated coefficients suggest the job finding rate of individuals with no experience is about 46 percent lower than that of individuals with prior work experience.
  - In Appendix II: the job‑finding probability of those who have worked before is 0.131, while those who have never worked before is 0.071 (Table 1 in Appendix II) which translates into a 46 percent lower job‑finding rate for the inexperienced workers.
- Result summaries:
  - Result II: Long‑term unemployment reduces future employability; the youth and women have lower probability of finding jobs.
    - Those unemployed for short periods have a finding probability which is 0.072 (45 percent) more on average than the long‑term unemployed.
    - Older individuals and men are more likely to transition into employment compared to 15 to 24 year olds and women.
  - Result III: Previous experience matters more for the youth.
    - 15 to 24 year olds with no experience have a much lower job‑finding probability compared to those who have worked before.
    - Even 25 to 34 year olds with no experience have a much lower job‑finding probability compared to 15 to 24 year olds with experience.
- Informal sector as stepping‑stone (Result VI):
  - Workers in the informal sector have a 55 percent higher probability of transitioning into formal sector employment than the unemployed.
  - Appendix II: The probability that an unemployed individual (with prior work experience) finds a formal sector job is 0.072 while that for an informal sector worker is 0.112 (Table 3 in Appendix II).

### Job‑exit rates: micro results and patterns
- Education matters strongly for job‑exit (Result IV):
  - Individuals with university education have much more stable jobs on average—the job‑exit rate of individuals with university education is about 55 percent less than that of individuals with less than primary education.
  - Appendix II: The job‑exit probability of individuals with less than primary education is 0.055, while that of university graduates is 0.025 (Table 2 in Appendix II).
- Race, age, and sector matter (Result V):
  - Whites and Indian/Asians have much lower job‑exit rates compared to Blacks.
  - On average, Whites have a job‑exit rate which is 0.032 less than that of Blacks (almost 60 percent lower).
  - 55 to 64 year olds have a job‑exit rate which is 0.06 less than 15 to 24 year olds (75 percent lower).
  - Formal sector workers are less likely to be separated from their job compared to informal sector workers.
- Trade unions and insider‑outsider dynamics (Result VIII):
  - Not being a trade union member increases job‑exit probability by around 58 percent (job‑exit probability of union members is 0.031 vs non‑members 0.049; Table 2 in Appendix II).
  - The trade union density of an individual’s previous industry has a negative effect on future job‑finding: raw coefficient on trade union density is -0.231 and statistically significant at the 1 percent level (clustered standard errors at industry‑time level). This implies previous work in a high trade union density industry reduces the probability of getting employed in the future.
  - Industries with high trade union density (e.g., mining; “electricity, gas and water”) show much lower job‑finding rates for unemployed who previously worked there compared to industries like agriculture, construction, or trade.

### Business cycle and aggregate interpretation
- Result VII: During the crisis (sample includes the GFC), unemployment increased due to a fall in job‑finding rate rather than a rise in job‑exit (job‑separation) rate.
  - Time‑dummy coefficients indicate job‑exit rate remained more or less unchanged while job‑finding rate declined substantially.
  - This accords with literature (Hall 2005; Shimer 2005) that fluctuations in unemployment are driven mainly by job‑finding rate.

### Aggregate counterfactuals (steady‑state stock‑flow exercises)
- Methodology: use steady state of stock‑flow transition equations (Cortes et al. (2015); Shimer (2012)) with quarterly transition rates estimated from the micro panel; Table 6 gives average transition values.
- Key steady‑state benchmarks:
  - Substituting sample averages into steady‑state equation gives steady state share of unemployed (u) = 32.2 percent.
  - Average unemployment rate for pooled sample = 31.8 percent.
- Counterfactuals using equation (2) (replace job‑finding or job‑exit rates with alternative group rates):
  - If all individuals had the job‑finding rate of experienced individuals (0.131 vs sample average 0.106), steady state unemployment would be 28 percent (a reduction of 4 percentage points from average).
  - If all individuals had the job‑finding rate of the inexperienced, steady state unemployment would be 39 percent.
  - Difference in job‑finding rates between experienced and inexperienced implies an aggregate unemployment rate difference of 11 percentage points (39 percent vs 28 percent).
  - Short‑term vs long‑term unemployed counterfactuals:
    - If all had the job‑finding rate of the short‑term unemployed, aggregate unemployment would be 25 percent.
    - If all had the job‑finding rate of the long‑term unemployed, aggregate unemployment would be 35 percent.
  - Job‑exit (education) counterfactuals:
    - If everyone had one higher level of education (and corresponding lower job‑exit rate), counterfactual unemployment = 30 percent.
    - If everyone had the job‑exit rate of college graduates, counterfactual unemployment = 24 percent.

### Cross‑country comparison
- Using ILO KLIM dataset (monthly outflow/inflow rates) for 2001–2012 comparisons:
  - The ratio of job‑finding to exit rate estimated in this paper = 2.5 (Table 6).
  - The ratio in the KLIM data = 3.1.
  - South Africa’s job‑finding rate is one of the lowest among selected countries; only Romania and Turkey have lower job‑finding rates in the sample. Countries with lower job‑finding rates also have much lower job‑exit rates. South Africa has the lowest ratio of job‑finding to job‑exit rate in the sample.

### Inequality exercises using NIDS (2012, wave 3)
- Data and baseline:
  - NIDS wave 3 (2012) covered 8,040 households and 32,633 interviewed individuals.
  - Baseline individual‑level post‑tax income Gini (labor force sample) = 0.68.
  - Baseline household per‑capita income Gini = 0.665 (includes imputed rent and transfers).
- Simulation procedure (partial equilibrium, random selection):
  1. Compute baseline inequality in NIDS.
  2. Randomly select a subsample of unemployed to become employed.
  3. Assign incomes to the newly employed based on quantile regressions of log income on age, education, race and sex; predicted income assigned if higher than their unemployed income.
  4. Recompute inequality.
  - Random selection maintains steady state distribution of unemployment across demographic groups; reduces overall unemployment by an ‘x’ percent by reducing each group’s unemployment by the same ‘x’ percent.
  - Two quantile rules used: 25th percentile and median.
- Simulation results:
  - Individual‑level Gini:
    - Starting point: unemployment rate around 30 percent and Gini = 0.68.
    - A decline in unemployment of 10 percentage points leads to Gini falling from 0.68 to 0.645 when using the median.
  - Household per‑capita Gini:
    - Baseline = 0.665.
    - A decline in unemployment of 10 percentage points reduces the Gini to 0.645 in simulations when using median incomes.
  - Caveats:
    - Static exercise; dynamic effects (experience and skill accumulation) could lead to larger reductions in inequality.
    - Assigning a single predicted income per demographic cell understates within‑cell dispersion and may overstate fall in inequality; robustness checks using full distribution of employed incomes for the newly employed produced larger declines in inequality.
- Redistribution counterfactual (transfers):
  - To benchmark magnitude, compute percentage increase in all government transfer programs required to deliver the same fall in household per‑capita Gini as a decline in unemployment. (Figure 6 presents simulated Gini vs percent increase in transfers; baseline Gini with zero percent increase in transfers = 0.665.)

*Source: _wp16137 - Appendix II reports the levels of the job-finding rate from the same regressions.*

### 0.665 to 0.645 would require an increase in transfers of about 40 percent, which was

### _wp16137 - 0.665 to 0.645 would require an increase in transfers of about 40 percent, which was

### V. CONCLUSION — Key findings on unemployment and inequality
- Large skill mismatches, poor educational outcomes, and the apartheid legacies have hurt job growth and perpetuated inequality.
- Unemployment remains high, especially among youth, women, and blacks.
- Improving the quality of education is key to long-term unemployment reduction, but until education reliably signals productivity, policies that provide experience to young first-time entrants are important.
- Creating informality may not be ideal, but it can be a stepping stone into formal employment.
- Reducing unemployment is probably a more viable and sustainable way of reducing income inequality compared to redistribution alone, though redistribution has played a very important role in reducing poverty.

### Policy recommendations and implications
- Implement programs that provide on-the-job training and work experience to youth (example cited: Employment Tax Incentive Act) to improve employability of young first-time labor-market entrants.
- Focus on measures that increase job finding rates and reduce long-term unemployment as more effective tools for lowering inequality than relying solely on increased transfers.

### Quantitative comparisons and counterfactuals (static exercise and redistributive cost)
- A shift in Gini from 0.665 to 0.645 would require an increase in transfers of about 40 percent, which was equivalent to 3.3 percent of GDP in 2012 or about 11.1 percent of government expenditure.
- Therefore, redistributive policy alone to achieve the same fall in inequality that might arise from a fall in unemployment of 10 percentage points would be prohibitively expensive.

### Counterfactuals and transition probabilities (selected results)
- Transition probability matrix (t to t+1):
  - Employed → Employed: 0.931
  - Employed → Unemployed: 0.048
  - Employed → OLF: 0.020
  - Unemployed → Employed: 0.106
  - Unemployed → Unemployed: 0.755
  - Unemployed → OLF: 0.138
  - OLF → Employed: 0.025
  - OLF → Unemployed: 0.095
  - OLF → OLF: 0.880
- Counterfactual unemployment rate examples (Finding Rate / Exit Rate panels):
  - Original Unemployment: 0.32
  - If everyone has some experience before: 0.28
  - If no one has experience before: 0.39
  - If everyone moves to the next level of education: 0.32 (finding-rate panel)
  - If all are graduates: 0.31 (finding-rate panel)
  - If everyone was temporarily unemployed: 0.25 (exit-rate panel)
  - If everyone was long term unemployed: 0.35 (exit-rate panel)
  - If everyone moves to the next level of education: 0.30 (exit-rate panel)
  - If all are graduates: 0.24 (exit-rate panel)

### Empirical regression evidence — stylized results from transition regressions
- Job-Finding Rate (marginal effects, selected):
  - Female: -0.031*** (column 1) and -0.032*** (column 2); -0.036*** (columns 3 and 4)
  - Never employed: -0.061*** (column 1) and -0.052*** (column 2)
  - Short term unemployed (<1 year): 0.072*** (columns 1 and 2) and 0.078*** (columns 3 and 4)
  - Observations: 173,626 (columns 1 and 2); 105,887 (columns 3 and 4)
- Job-Exit Rate (marginal effects, selected):
  - University or more: -0.030*** (column 1) and -0.028*** (column 2); -0.015*** (columns 3 and 4)
  - Age effects: Age 25-34: -0.025*** (col 1) etc.; Age 55-64: -0.060*** (col 1) etc.
  - Informal sector (sector effect): 0.024*** (column 1) and 0.017*** (column 2)
  - Permanent job type: -0.077*** (column 1) and -0.066*** (column 2)
  - Observations: 328,063 (columns 1 and 3); 174,683 (columns 2 and 4)
- Transition to formal employment (Unemployment/Informal → Formal, marginal effects, selected):
  - Work in Informal sector (if unemployed): 0.052*** (col 1) and 0.049*** (col 2)
  - Unemployed never worked: -0.039*** (col 1) and -0.040*** (col 2)
  - Education gradient (col 1 examples): Primary but not secondary 0.013***; Secondary but not university 0.036***; University or more 0.071***
  - Female: -0.042*** (col 1) and -0.040*** (col 2)
  - Observations: 270,203 (cols 1 and 2); 196,021 (cols 3 and 4)

### Additional empirical notes
- Data sources and periods used in analyses:
  - QLFS 2008Q1 to 2014Q3 (transition regressions, transition probabilities, and counterfactuals)
  - NIDS wave 3 (inequality simulations in Figures 5 and 6)
  - ILO-KILM data averaged over 2001-2012 for cross-country comparisons (Figure 4)
- Matching algorithm comparisons and sample composition:
  - Table comparing algorithm matching to StatsSA official panel for 2013Q3 to 2013Q4 reported.
  - Full sample vs. matched sample shares reported (examples preserved exactly in source tables, e.g., Male 48.9 / 48.3; Female 51.1 / 51.69; African/Black 78.34 / 78.6; Employed 43.01 / 43.39; Unemployed 13.87 / 13.42).

*Italicized source attribution provided by original content unit.*

### Appendix Table 2. Transition Regression: Job-Exit Rate (Levels)

### Appendix Table 2. Transition Regression: Job-Exit Rate (Levels)

### Regression setup and sample
- Data: QLFS 2008Q1 to 2014Q3.
- Dependent variable: dummy = 1 if individual transitions from employment to unemployment (d_i,t).
- Reported figures: levels of the transition probability from the corresponding probit regression (levels counterpart of Table 4 in the main text).
- Columns:
  - (1) Broad
  - (2) Broad — Trade Union (membership included; question asked only after 2010Q3, smaller sample)
  - (3) Narrow
  - (4) Narrow — Trade Union
- Standard errors clustered at the household level. Statistical significance indicated as: *** p<0.01, ** p<0.05, * p<0.1.
- Observations: 328,063 (columns 1 and 3); 174,683 (columns 2 and 4).

### Estimated levels of job-exit probabilities (by characteristic)
- Education
  - Less than primary: 0.055***, 0.053***, 0.033***, 0.032***
  - Primary but not secondary: 0.055***, 0.053***, 0.037***, 0.036***
  - Secondary but not university: 0.043***, 0.041***, 0.031***, 0.030***
  - University or more: 0.025***, 0.025***, 0.019***, 0.019***
  - Other: 0.049***, 0.046***, 0.032***, 0.030***
- Race
  - African/Black: 0.051***, 0.048***, 0.035***, 0.033***
  - Colored: 0.049***, 0.051***, 0.032***, 0.033***
  - Indian/Asian: 0.024***, 0.022***, 0.019***, 0.018***
  - White: 0.020***, 0.020***, 0.013***, 0.014***
- Gender
  - Male: 0.048***, 0.047***, 0.035***, 0.034***
  - Female: 0.047***, 0.045***, 0.030***, 0.028***
- Age groups
  - Age 15-24: 0.082***, 0.074***, 0.054***, 0.050***
  - Age 25-34: 0.057***, 0.054***, 0.040***, 0.038***
  - Age 35-44: 0.043***, 0.040***, 0.029***, 0.027***
  - Age 45-54: 0.032***, 0.030***, 0.021***, 0.020***
  - Age 55-64: 0.022***, 0.018***, 0.013***, 0.011***
- Sector
  - Formal sector: 0.042***, 0.044***, 0.029***, 0.030***
  - Informal sector: 0.066***, 0.061***, 0.044***, 0.040***
  - Private households: 0.045***, 0.044***, 0.032***, 0.030***
- Type of Job
  - Limited Duration: 0.098***, 0.087***, 0.070***, 0.061***
  - Permanent: 0.021***, 0.022***, 0.015***, 0.016***
  - Unspecified Duration: 0.072***, 0.065***, 0.050***, 0.044***
  - Not Applicable: 0.049***, 0.031***
- Trade Union (columns 2 and 4 only)
  - Member: 0.031***, 0.021***
  - Not a member: 0.049***, 0.033***
  - Don't know: 0.047***, 0.033***

### Notes (from table)
- Columns 1 and 2 use the broad definition of unemployment for the LHS variable; columns 3 and 4 use the narrow definition.
- Columns 2 and 4 include whether the individual is a trade union member; these columns use a smaller sample because the trade union membership question was only asked after 2010Q3.

*Source: Appendix Table 2 from the provided IMF content unit.*

### Appendix Table 3. Unemployment/Informal to Formal Employment (Levels)

### Regression setup and sample
- Data: QLFS 2008Q1 to 2014Q3.
- Dependent variable: dummy = 1 if individual transitions from unemployment or the informal sector to formal employment (i_i,t).
- Reported figures: levels of the transition probability from the corresponding probit regression (levels counterpart of Table 5 in the main text).
- Columns:
  - (1) Broad — Unemployed/Informal
  - (2) Broad — Unemployed/Informal/Experience
  - (3) Narrow — Unemployed/Informal
  - (4) Narrow — Unemployed/Informal/Experience
- Standard errors clustered at the household level. Statistical significance indicated as: *** p<0.01, ** p<0.05, * p<0.1.
- Observations: 270,203 (columns 1 and 2); 196,021 (columns 3 and 4).

### Estimated levels of transition to formal employment (by characteristic)
- Unemployed / Informal indicators
  - Unemployed: 0.056***, 0.064***
  - Work in Informal sector: 0.108***, 0.113***
- Unemployed / Informal / Experience indicators (columns 2 and 4)
  - Unemployed with experience: 0.072***, 0.080***
  - Unemployed never worked: 0.033***, 0.040***
  - Work in Informal sector: 0.112***, 0.116***
- Education
  - Less than primary: 0.053***, 0.053***, 0.060***, 0.060***
  - Primary but not secondary: 0.066***, 0.066***, 0.078***, 0.077***
  - Secondary but not university: 0.089***, 0.091***, 0.104***, 0.106***
  - University or more: 0.124***, 0.124***, 0.143***, 0.144***
  - Other: 0.078***, 0.078***, 0.090***, 0.089***
- Race
  - African/Black: 0.067***, 0.068***, 0.079***, 0.079***
  - Colored: 0.095***, 0.091***, 0.104***, 0.101***
  - Indian/Asian: 0.111***, 0.107***, 0.127***, 0.123***
  - White: 0.147***, 0.142***, 0.178***, 0.172***
- Gender
  - Male: 0.095***, 0.094***, 0.111***, 0.110***
  - Female: 0.053***, 0.054***, 0.061***, 0.062***
- Age groups
  - Age 15-24: 0.065***, 0.078***, 0.081***, 0.095***
  - Age 25-34: 0.082***, 0.083***, 0.093***, 0.094***
  - Age 35-44: 0.076***, 0.071***, 0.087***, 0.083***
  - Age 45-54: 0.064***, 0.058***, 0.074***, 0.070***
  - Age 55-64: 0.051***, 0.048***, 0.063***, 0.060***

### Notes (from table)
- Columns 1 and 3 include a RHS variable for whether the individual was unemployed or in the informal sector.
- Columns 2 and 4 include RHS categories: unemployed with no experience, unemployed with experience, or in the informal sector.

*Source: Appendix Table 3 from the provided IMF content unit.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16137.pdf_