## wp17257

## Source details

**Canonical URL:** [wp17257](https://www.imf.org/-/media/files/publications/wp/2017/wp17257.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp17257.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp17257.pdf.json)

---

### I. Introduction — purpose and scope
- Objective: Investigate the relationship between economic growth and job creation in developing countries, focusing on low and lower middle-income countries (LLMICs).
- Analytical dimensions:
  - Study growth patterns in output and total employment, and synchronicity of regime changes using the modified multiple structural breaks test by Berg, Ostry, and Zettelmeyer (2012).
  - Study the short-run relationship between output and labor market through the Okun’s Law framework, following Ball, Leigh, and Loungani (2017).
- Key findings:
  - Regime changes are common in both output growth and employment growth in LLMICs, but they are not synchronized.
  - Out of 17 up-breaks in GDP growth, only 3 are associated with up-breaks in employment growth; out of 9 down-breaks in GDP growth, only 2 are associated with down-breaks in employment growth.
  - The short-run relationship between output and total employment holds in half of the LLMICs; average Okun coefficient = 0.20, with substantial cross-country heterogeneity.
  - Okun responsiveness is negatively associated with poverty rate and skill-mismatch level.

### II. Data description and summary statistics
- Data source: IMF World Economic Outlook April 2017.
- Sample averages (LLMIC sample):
  - For 2001-2015: average economic growth rate = 5.2 percent; average employment growth rate = 1.9 percent; average unemployment rate change = -0.2 percent.
  - Correlations across countries for 2001-2015: correlation between average employment growth and average output growth = 0.2; correlation between average unemployment change and average output growth = -0.1.
- Selected country-period examples (preserved exactly):
  - Cabo Verde (1981-1990): GDP 5.79 EMP 3.57 UR 0.39; (1991-2000): GDP 6.15 EMP 3.89 UR -0.30; (2001-2015): GDP 4.35 EMP 1.33 UR -0.75.
  - Indonesia (1981-1990): GDP 6.75 EMP 1.24 UR 0.26; (1991-2000): GDP 4.84 EMP 1.96 UR 0.36; (2001-2015): GDP 5.41 EMP 1.73 UR -0.01.
- Sample averages by period:
  - 1981-1990 Average GDP 3.55 EMP 3.07 UR -0.19.
  - 1991-2000 Average GDP 2.13 EMP 2.24 UR -0.24.
  - 2001-2015 Average GDP 5.18 EMP 1.86 UR -0.17.
- Note: For many LLMICs employment growth was much lower than output growth during 2001-2015; unemployment rate decreases were limited in most countries.

### III. Long-run relationships (trend correlations)
- Method: Trends filtered with HP with a smooth parameter of 100.
- Summary counts:
  - Out of 25 LLMICs, all but 3 countries (Georgia, Moldova, and Ukraine) have positive long-run correlation between real GDP and total employment.
  - Of 19 countries with available unemployment data, 8 countries have positive long-run correlation between total employment and unemployment rate; 6 countries have positive correlation between real GDP and unemployment rate.
- Selected country correlations (exact values preserved):
  - Belize: GDP-EMP 1.00 EMP-UR -0.64 GDP-UR -0.68.
  - Bolivia: GDP-EMP 0.95 EMP-UR -0.98 GDP-UR -1.00.
  - Mongolia: GDP-EMP 0.95 EMP-UR -1.00 GDP-UR -1.00.
  - Morocco: GDP-EMP 0.98 EMP-UR -0.99 GDP-UR -0.95.
  - Pakistan: GDP-EMP 0.99 EMP-UR 0.47 GDP-UR 0.77.
  - Ukraine: GDP-EMP -0.11 EMP-UR 0.82 GDP-UR -0.83.
- Interpretation: Real GDP and total employment are closely linked in most countries, but mismatches between total employment and unemployment rate raise questions about the usefulness of unemployment rates as a labor-market indicator in these contexts.

### IV. Growth-pattern regime changes in output and employment
- Method: Regime changes identified using Berg et al. (2012) modification of Bai and Perron (1998, 2003); minimum interstitiary period = 8 years.
- Aggregate counts:
  - Identified 17 up-breaks and 9 down-breaks in real GDP growth across the 25 LLMICs.
  - Identified 10 up-breaks and 17 down-breaks in employment growth across the 25 LLMICs.
  - Country coverage: 13 countries experienced at least one regime change in both real GDP and employment growth; 5 countries experienced at least one regime change in real GDP growth only; 4 countries experienced at least one regime change in employment growth only.
- Synchronization (exact cases preserved):
  - Out of the 17 up-breaks in GDP growth, only 3 cases (India 2004, Tajikistan 2000, and Uzbekistan 2001) are associated with up-breaks in employment growth.
  - Out of the 9 down-breaks in GDP growth, only 2 cases (Belize 2007 and Tajikistan 2008) are associated with down-breaks in employment growth.
- Country examples:
  - Belize: Output growth up-break in 1998 and down-break in 2006; employment growth had no corresponding up-break around 1998 and a down-break in 2007.
  - Tajikistan: Both output growth and employment growth experience an up-break in 2000 and a down-break in 2008 (synchronized).
  - Moldova: Output growth up-break in 2001; no corresponding labor-market up-break until employment up-break in 2010.
  - Morocco: No regime change in output growth, but total employment growth had a down-break in 2004.
- Stylized pattern:
  - Employment growth decelerations are associated with GDP growth down-breaks.
  - Employment growth accelerations are generally not associated with GDP growth up-breaks.

### V. Short-run relationship: Okun’s Law and cross-country heterogeneity
- Specifications estimated for 25 LLMICs:
  - Gaps specification: et − et* = γ(yt − yt*) + εt (HP filter smoothing parameter 100).
  - Changes specification: Δet = α + βΔyt + εt.
- Country-specific Belize (Table 3):
  - Gaps Specification: Coef. 0.23*; R-sq 0.14; Obs. 26.
  - Changes Specification: Coef. 0.34*; R-sq 0.15; Obs. 25.
- Summary of country-specific Okun results (25 countries):
  - Gaps version: positive coefficients in 19 countries, 13 statistically significant.
  - Changes version: positive coefficients in 19 countries, 9 statistically significant.
- Distributional statistics:
  - Gaps-version coefficients: average value 0.20; standard deviation 0.37; average R-squared 0.15.
  - Changes-version coefficients: mean value 0.16; standard deviation 0.12; average R-squared 0.12.
- Panel estimates (country & year fixed effects):
  - Gaps version: Output Gap coefficient 0.14***.
  - Changes version: Output Growth coefficient 0.17***.
  - Note: panel coefficient 0.14 is lower than averaged country-specific Okun coefficient (0.20); changes-panel coefficient 0.17 is close to averaged country-specific (0.16).

### VI. Determinants of cross-country heterogeneity in Okun responsiveness
- Hypothesized determinants examined: poverty rate, skill mismatch, size of informal sector, business regulations.
- Poverty rate:
  - Expected effect: higher poverty → weaker labor market responsiveness (negative relationship).
  - Correlations:
    - Gaps-version Okun coefficients vs. poverty rate: correlation -0.14.
    - Changes-version Okun coefficients vs. poverty rate: correlation -0.75.
  - Panel interaction results (Table 4):
    - Gaps: Interacted term with Poverty = -0.23** (standard error 0.12).
    - Changes: Interacted term with Poverty = -0.36*** (standard error 0.11).
- Skill mismatch:
  - Finding: higher skill mismatch associated with weaker labor market response.
  - Panel interaction (Table 4):
    - Gaps: Interacted term with Skill mismatch = -0.27* (standard error 0.16).
    - Changes: Interacted term with Skill mismatch = -0.17 (standard error 0.19) — not significant.
- Informal sector:
  - Expectation: larger informal sector → weaker employment response; empirical interaction estimates found positive but not significant effects.
  - Panel interaction (Table 4):
    - Gaps: Interacted term with Informality = 0.16 (standard error 0.16) — not significant.
    - Changes: Interacted term with Informality = 0.18 (standard error 0.19) — not significant.
- Business regulations:
  - Finding: higher level of business regulation associated with weaker labor market responses in expectation; interaction estimates not significant.
  - Panel interaction (Table 4):
    - Gaps: Interacted term with Business regulations = -0.12 (standard error 0.12) — not significant.
    - Changes: Interacted term with Business regulations = -0.16 (standard error 0.11) — not significant.
- Selected Table 4 panel numbers (exact values preserved):
  - Panel regression (column 1):
    - Gaps: Output Gap 0.14*** (0.03); R-squared 0.05; N. of Obs. 698.
    - Changes: Output Growth 0.17*** (0.05); R-squared 0.26; N. of Obs. 677.
  - Poverty interaction (column 2):
    - Gaps: Output Gap 0.15*** (0.06); Interacted term -0.23** (0.12); R-squared 0.19; N. of Obs. 326.
    - Changes: Output Growth 0.21*** (0.07); Interacted term -0.36*** (0.11); R-squared 0.34; N. of Obs. 316.
  - Skill mismatch interaction (column 3):
    - Gaps: Output Gap 0.22** (0.11); Interacted term -0.27* (0.16); R-squared 0.12; N. of Obs. 282.
    - Changes: Output Growth 0.22* (0.14); Interacted term -0.17 (0.19); R-squared 0.29; N. of Obs. 273.
  - Informality interaction (column 4):
    - Gaps: Output Gap 0.07 (0.07); Interacted term 0.16 (0.16); R-squared 0.11; N. of Obs. 620.
    - Changes: Output Growth 0.10 (0.09); Interacted term 0.18 (0.19); R-squared 0.27; N. of Obs. 601.
  - Business regulations interaction (column 5):
    - Gaps: Output Gap 0.17*** (0.07); Interacted term -0.12 (0.12); R-squared 0.16; N. of Obs. 350.
    - Changes: Output Growth 0.17*** (0.06); Interacted term -0.16 (0.11); R-squared 0.32; N. of Obs. 339.
- Note: Robust standard errors reported in parentheses. Significant at *10%, **5%, and ***1%.

### VII. Morocco: detailed case study and dynamic evidence
- Participation and education (Table 5):
  - Average participation rate 1999-2002 vs. 2011-2014 and change:
    - National: 52.2 → 48.6; Change -7%.
    - Without diploma: 51.2 → 49.0; Change -4%.
    - With diploma: 54.2 → 48.0; Change -11%.
    - Male: 78.1 → 73.4; Change -6%.
    - Female: 27.1 → 25.2; Change -7%.
    - 15 to 24 yr: 44.2 → 33.4; Change -25%.
    - 25 to 34 yr: 63.0 → 61.4; Change -3%.
    - 35 to 44 yr: 62.9 → 61.4; Change -2%.
    - 44 yr & over: 44.3 → 44.5; Change 0%.
    - Rural: 60.4 → 57.5; Change -5%.
    - Urbain: 46.4 → 42.7; Change -8%.
  - Working population with higher education: increased from 31.3 percent in 2000 to 41.4 percent in 2014.
- Employment by sector (annual average job creation and shares 2000–2014):
  - Total employment increased by 129 thousand jobs per year on average.
  - Service sector: 87 thousand jobs per year (67 percent of total); share increased from 34.7 percent in 2000 to 40.2 percent in 2014.
  - Construction: 31 thousand jobs per year (24 percent).
  - Agriculture, forest and fishing: 10 thousand jobs per year (8 percent).
  - Industrial sector (including handicrafts): 1 thousand jobs per year (1 percent).
- Informal sector importance (2013):
  - Employment in the informal sector around 2.4 million people, representing 36.3 percent of overall non-agricultural employment.
  - Survey: 233 thousand independent employers recruited 420 thousand employees and distributed 11.4 billion Dirham of wages in 2013.
  - That wage amount: almost 4 percent of national compensation and 11 percent of overall added value of the informal sector.
- Rolling-window dynamics (Morocco):
  - 10-year rolling-window estimates show the Okun coefficient becoming more negative and statistically significant over time; R-squared increases from around 0 to above 0.7 (gaps-version).
  - Rolling estimates for employment growth vs. real GDP growth show the coefficient becomes greater and more significant over time; R-squared increases from around 0 to above 0.6.
  - Interpretation: sectoral shift toward services and decreased participation (tightness of labor supply), despite the importance of informal market, are consistent with a strengthening Okun relationship in Morocco over time.

### VIII. Main conclusions and policy-relevant implications
- Empirical conclusions:
  - Regime changes are common in both output growth and employment growth across LLMICs, but are often unsynchronized.
  - Employment growth decelerations are associated with GDP down-breaks; employment accelerations are not reliably associated with GDP up-breaks.
  - Okun’s Law holds in roughly half of the 25 LLMICs; average Okun coefficient about 0.20 (gaps) with considerable cross-country heterogeneity.
  - Panel estimates indicate a positive and significant relationship between output and employment: gaps 0.14*** and changes 0.17***.
  - Cross-country heterogeneity in responsiveness of employment to output is negatively associated with poverty rate and skill mismatch; evidence on informality and business regulations is mixed or not significant.
- Policy-relevant implications:
  - Addressing poverty and skill mismatches may strengthen the employment response to growth.
  - Structural transformation (e.g., expansion of services) and labor force participation dynamics can alter the fit of standard Okun frameworks over time (Morocco example).
  - Deeper analysis and better data on informality are needed because informality can obscure the output–labor market relationship.

*Source: Authors’ estimates based on the IMF World Economic Outlook April 2017.*

### References .............................................................................................................

### wp17257 - References .............................................................................................................

### I. Introduction — purpose and scope
- Objective: Investigate the relationship between economic growth and job creation in developing countries, focusing on low and lower middle-income countries (LLMICs).
- Two analytical dimensions:
  - Study growth patterns in output and total employment, and the synchronicity of regime changes across the two series using the modified multiple structural breaks test by Berg, Ostry, and Zettelmeyer (2012).
  - Study the short-run relationship between output and labor market through the Okun’s Law framework, following specifications in Ball, Leigh, and Loungani (2017).
- Key high-level findings:
  - Regime changes are common in both output growth and employment growth in LLMICs, but they are not synchronized.
  - Out of 17 up-breaks in GDP growth, only 3 are associated with up-breaks in employment growth; out of 9 down-breaks in GDP growth, only 2 are associated with down-breaks in employment growth.
  - The short-run relationship between output and total employment holds in half of the LLMICs; the Okun coefficient is 0.20 on average, with substantial cross-country heterogeneity.
  - Okun responsiveness is negatively associated with poverty rate and skill-mismatch level.

### II. Data description and summary statistics
- Data source: IMF World Economic Outlook April 2017.
- Periods analyzed and sample averages (for the LLMIC sample):
  - For 2001-2015: average economic growth rate = 5.2 percent; average employment growth rate = 1.9 percent; average unemployment rate change = -0.2 percent.
  - Correlations across countries for 2001-2015: correlation between average employment growth and average output growth = 0.2; correlation between average unemployment change and average output growth = -0.1.
- Table 1 reports average annual growth rates in real GDP (GDP), total employment (EMP), and average annual change in unemployment rate (UR) by country and period (1981-1990; 1991-2000; 2001-2015). Examples (preserved exactly as in source):
  - Cabo Verde (1981-1990): GDP 5.79 EMP 3.57 UR 0.39; (1991-2000): GDP 6.15 EMP 3.89 UR -0.30; (2001-2015): GDP 4.35 EMP 1.33 UR -0.75.
  - Indonesia (1981-1990): GDP 6.75 EMP 1.24 UR 0.26; (1991-2000): GDP 4.84 EMP 1.96 UR 0.36; (2001-2015): GDP 5.41 EMP 1.73 UR -0.01.
  - Sample averages by period: 1981-1990 Average GDP 3.55 EMP 3.07 UR -0.19; 1991-2000 Average GDP 2.13 EMP 2.24 UR -0.24; 2001-2015 Average GDP 5.18 EMP 1.86 UR -0.17.
  - Correlation rows in Table 1: 1981-1990 Corr. -0.38 0.42 0.67; 1991-2000 Corr. -0.28 0.16 -0.08.
- Note: For many LLMICs employment growth was much lower than output growth during 2001-2015; unemployment rate decreases were limited in most countries.

### III. Long-run relationships (trend correlations)
- Method: Trends filtered with HP with a smooth parameter of 100; Table 2 reports correlations among trend components of logged real GDP, logged total employment, and unemployment rate.
- Summary of long-run correlations (exact counts and examples preserved):
  - Out of 25 LLMICs, all but 3 countries (Georgia, Moldova, and Ukraine) have positive long-run correlation between real GDP and total employment.
  - Of 19 countries with available unemployment data, 8 countries have positive long-run correlation between total employment and unemployment rate; 6 countries have positive correlation between real GDP and unemployment rate.
- Selected country correlations from Table 2 (preserved exactly):
  - Belize: GDP-EMP 1.00 EMP-UR -0.64 GDP-UR -0.68.
  - Bolivia: GDP-EMP 0.95 EMP-UR -0.98 GDP-UR -1.00.
  - Mongolia: GDP-EMP 0.95 EMP-UR -1.00 GDP-UR -1.00.
  - Morocco: GDP-EMP 0.98 EMP-UR -0.99 GDP-UR -0.95.
  - Pakistan: GDP-EMP 0.99 EMP-UR 0.47 GDP-UR 0.77.
  - Ukraine: GDP-EMP -0.11 EMP-UR 0.82 GDP-UR -0.83.
- Interpretation: Real GDP and total employment are closely linked in most countries, but there are notable mismatches between total employment and unemployment rate in many LLMICs; this raises questions about the usefulness of unemployment rates as a labor-market indicator in these contexts.

### IV. Growth-pattern regime changes in output and employment
- Method: Identify regime changes in growth rates of output and total employment using Berg et al. (2012) modification of Bai and Perron (1998, 2003); minimum interstitiary period set to 8 years.
- Aggregate counts (Figure 1 and text):
  - Identified 17 up-breaks and 9 down-breaks in real GDP growth across the 25 LLMICs.
  - Identified 10 up-breaks and 17 down-breaks in employment growth across the 25 LLMICs.
  - Country coverage: 13 countries experienced at least one regime change in both real GDP and employment growth; 5 countries experienced at least one regime change in real GDP growth only; 4 countries experienced at least one regime change in employment growth only.
- Synchronization results (preserved exactly):
  - Out of the 17 up-breaks in GDP growth, only 3 cases (India 2004, Tajikistan 2000, and Uzbekistan 2001) are associated with up-breaks in employment growth.
  - Out of the 9 down-breaks in GDP growth, only 2 cases (Belize 2007 and Tajikistan 2008) are associated with down-breaks in employment growth.
- Case illustrations:
  - Tajikistan: both output growth and employment growth experience an up-break in 2000 and a down-break in 2008 (Figure 2 demonstrates synchronized regime changes).
  - Belize: synchronized down-break in 2007 (Figure 2 demonstrates synchronized regime changes).
- Broader pattern: Employment growth decelerations are generally associated with GDP growth down-breaks, while employment growth accelerations are not systematically associated with GDP growth up-breaks.

### V. Short-run relationship: Okun’s Law and cross-country heterogeneity
- Approach: Follow Ball, Leigh, and Loungani (2017) specifications to estimate short-run responses of total employment to real GDP; also estimate panel regressions and interaction-variable regressions to explore determinants of cross-country differences.
- Key quantitative results (preserved exactly):
  - The short-run relationship between output and total employment holds in half of the LLMICs.
  - Average Okun coefficient = 0.20.
- Cross-country determinants (findings):
  - The short-run responsiveness of total employment to output is negatively associated with poverty rate.
  - The short-run responsiveness of total employment to output is negatively associated with skill-mismatch level.
- Additional analyses referenced in figures and tables (listed in source):
  - Table 3: Estimates of Okun’s Law.
  - Table 4: Okun Coefficients from Panel Regression.
  - Figures 5–9: Histograms and scatter relationships linking Okun coefficients to poverty rate, skill mismatch, size of informal sector, and business regulation.

### VI. Country focus and supplementary material
- Morocco: Section V provides a closer look at the labor market in Morocco; figures and tables specific to Morocco include:
  - Table 5: Participation Rate in Morocco.
  - Figure 10: Evolution of Employment Share by Sector.
  - Figure 11: Rolling Estimation in Morocco, Employment.
- Appendix: Appendix Table lists up- and down-breaks in GDP growth and employment growth (Appendix Table 1).

*Source: Authors’ estimates based on the IMF World Economic Outlook April 2017.*

### 2008.  In  Belize,  output  growth  experiences  an  up-break  in  1998  and  a  down-break  in  2006;

### wp17257 - 2008. In Belize, output growth experiences an up-break in 1998 and a down-break in 2006;

### Regime changes in output and employment in LLMICs
- Belize:
  - Output growth: up-break in 1998 and down-break in 2006.
  - Employment growth: no corresponding up-break around 1998; down-break in 2007 following the output growth down-break.
- Examples of unsynchronized regime changes:
  - Moldova: output growth up-break in 2001 after a decade of constant negative growth; no corresponding up-break in the labor market until total employment up-break in 2010.
  - Morocco: no regime change in output growth, but total employment growth had a down-break in 2004.
- Aggregate counts (25 LLMICs):
  - 13 countries experienced at least one regime change in both output growth and employment growth.
  - 5 countries experienced regime change in output growth only.
  - 4 countries experienced regime change in employment growth only.
  - Out of the 17 up-breaks in GDP growth, only 3 cases are followed by up-breaks in employment growth.
  - Out of the 9 down-breaks in GDP growth, only 2 cases are followed by down-breaks in employment growth.
- Stylized pattern:
  - Employment growth decelerations are associated with GDP growth down-breaks.
  - Employment growth accelerations are generally not associated with GDP growth up-breaks.

*Source: Authors’ estimates based on the IMF World Economic Outlook April 2017. Note: Up- and down-breaks identified using Berg et. al. (2012) methodology with minimum interstitiary period of 8 years.*

### Short-run relationship between output and total employment — Okun’s Law estimates
- Specifications estimated for 25 LLMICs:
  - Gaps specification: et − et* = γ(yt − yt*) + εt (HP filter smoothing parameter 100).
  - Changes specification: Δet = α + βΔyt + εt.
- Country-specific results for Belize (Table 3):
  - Gaps Specification: Coef. 0.23*; R-sq 0.14; Obs. 26.
  - Changes Specification: Coef. 0.34*; R-sq 0.15; Obs. 25.
- Summary of country-specific Okun results (25 countries):
  - Gaps version: positive coefficients in 19 countries, 13 statistically significant.
  - Changes version: positive coefficients in 19 countries, 9 statistically significant.
- Distributional statistics:
  - Gaps-version coefficients: average value 0.20; standard deviation 0.37; average R-squared 0.15 (R-squared mostly below 0.30).
  - Changes-version coefficients: mean value 0.16; standard deviation 0.12; average R-squared 0.12.
- Panel estimates (all countries, country & year fixed effects):
  - Gaps version: Output Gap coefficient 0.14*** (statistically significant).
  - Changes version: Output Growth coefficient 0.17*** (statistically significant).
  - Note: panel coefficient 0.14 is lower than averaged country-specific Okun coefficient (0.20); changes-panel coefficient 0.17 is close to averaged country-specific (0.16).

*Source: Authors’ estimates based on the IMF World Economic Outlook April 2017.*

### Determinants of cross-country heterogeneity in Okun responsiveness
- Hypothesized LLMIC particularities examined: poverty rate, skill mismatch, size of informal sector, business regulations.
- Poverty rate:
  - Expected effect: higher poverty → weaker labor market responsiveness (negative relationship).
  - Correlations reported:
    - Gaps-version Okun coefficients vs. poverty rate: correlation -0.14.
    - Changes-version Okun coefficients vs. poverty rate: correlation -0.75 (highly significant).
  - Panel interaction results (Table 4):
    - Gaps: Interacted term with Poverty = -0.23** (standard error 0.12).
    - Changes: Interacted term with Poverty = -0.36*** (standard error 0.11).
- Skill mismatch:
  - Finding: higher skill mismatch associated with weaker labor market response.
  - Panel interaction (Table 4):
    - Gaps: Interacted term with Skill mismatch = -0.27* (standard error 0.16).
    - Changes: Interacted term with Skill mismatch = -0.17 (standard error 0.19) — not significant.
- Informal sector:
  - Expectation: larger informal sector → weaker employment response; empirical relationship found positive (contrary to expectation) and not significant in interactions.
  - Panel interaction (Table 4):
    - Gaps: Interacted term with Informality = 0.16 (standard error 0.16) — not significant.
    - Changes: Interacted term with Informality = 0.18 (standard error 0.19) — not significant.
- Business regulations:
  - Finding: higher level of business regulation associated with weaker labor market responses.
  - Panel interaction (Table 4):
    - Gaps: Interacted term with Business regulations = -0.12 (standard error 0.12) — not significant.
    - Changes: Interacted term with Business regulations = -0.16 (standard error 0.11) — not significant.
- Table 4 key panel numbers (selected):
  - Panel regression (column 1):
    - Gaps: Output Gap 0.14*** (0.03); R-squared 0.05; N. of Obs. 698.
    - Changes: Output Growth 0.17*** (0.05); R-squared 0.26; N. of Obs. 677.
  - Poverty interaction (column 2):
    - Gaps: Output Gap 0.15*** (0.06); Interacted term -0.23** (0.12); R-squared 0.19; N. of Obs. 326.
    - Changes: Output Growth 0.21*** (0.07); Interacted term -0.36*** (0.11); R-squared 0.34; N. of Obs. 316.
  - Skill mismatch interaction (column 3):
    - Gaps: Output Gap 0.22** (0.11); Interacted term -0.27* (0.16); R-squared 0.12; N. of Obs. 282.
    - Changes: Output Growth 0.22* (0.14); Interacted term -0.17 (0.19); R-squared 0.29; N. of Obs. 273.
  - Informality interaction (column 4):
    - Gaps: Output Gap 0.07 (0.07); Interacted term 0.16 (0.16); R-squared 0.11; N. of Obs. 620.
    - Changes: Output Growth 0.10 (0.09); Interacted term 0.18 (0.19); R-squared 0.27; N. of Obs. 601.
  - Business regulations interaction (column 5):
    - Gaps: Output Gap 0.17*** (0.07); Interacted term -0.12 (0.12); R-squared 0.16; N. of Obs. 350.
    - Changes: Output Growth 0.17*** (0.06); Interacted term -0.16 (0.11); R-squared 0.32; N. of Obs. 339.

*Source: Authors’ estimates based on the IMF World Economic Outlook April 2017. Note: Robust standard errors reported in parentheses. Significant at *10%, **5%, and ***1%.*

### Morocco: detailed case study and dynamic evidence
- Participation and education (Table 5):
  - Average participation rate 1999-2002 vs. 2011-2014 and change:
    - National: 52.2 → 48.6; Change -7%.
    - Without diploma: 51.2 → 49.0; Change -4%.
    - With diploma: 54.2 → 48.0; Change -11%.
    - Male: 78.1 → 73.4; Change -6%.
    - Female: 27.1 → 25.2; Change -7%.
    - 15 to 24 yr: 44.2 → 33.4; Change -25%.
    - 25 to 34 yr: 63.0 → 61.4; Change -3%.
    - 35 to 44 yr: 62.9 → 61.4; Change -2%.
    - 44 yr & over: 44.3 → 44.5; Change 0%.
    - Rural: 60.4 → 57.5; Change -5%.
    - Urbain: 46.4 → 42.7; Change -8%.
  - Working population with higher education: increased from 31.3 percent in 2000 to 41.4 percent in 2014.
- Employment by sector (annual average job creation and shares 2000–2014):
  - Total employment increased by 129 thousand jobs per year on average.
  - Service sector: 87 thousand jobs per year (67 percent of total); share increased from 34.7 percent in 2000 to 40.2 percent in 2014.
  - Construction: 31 thousand jobs per year (24 percent).
  - Agriculture, forest and fishing: 10 thousand jobs per year (8 percent).
  - Industrial sector (including handicrafts): 1 thousand jobs per year (1 percent).
- Informal sector importance (2013):
  - Employment in the informal sector around 2.4 million people, representing 36.3 percent of overall non-agricultural employment.
  - Survey: 233 thousand independent employers recruited 420 thousand employees and distributed 11.4 billion Dirham of wages in 2013.
  - That wage amount: almost 4 percent of national compensation and 11 percent of overall added value of the informal sector.
- Rolling-window dynamics (Morocco):
  - 10-year rolling-window estimates show the Okun coefficient becoming more negative and statistically significant over time; R-squared increases from around 0 to above 0.7 (gaps-version).
  - Rolling estimates for employment growth vs. real GDP growth show the coefficient becomes greater and more significant over time; R-squared increases from around 0 to above 0.6.
  - Interpretation: sectoral shift toward services and decreased participation (tightness of labor supply), despite the importance of informal market, are consistent with a strengthening Okun relationship in Morocco over time.

*Source: Haut-commissariat au Plan, Morocco, 2016; Authors’ estimates based on the IMF World Economic Outlook April 2017.*

### Main conclusions and policy-relevant implications
- Main empirical findings:
  - Regime changes are common in both output growth and employment growth across LLMICs, but are often unsynchronized.
  - Employment growth decelerations are associated with GDP down-breaks; employment accelerations are not reliably associated with GDP up-breaks.
  - Okun’s Law holds in roughly half of the 25 LLMICs; average Okun coefficient about 0.20 (gaps) with considerable cross-country heterogeneity.
  - Panel estimates indicate a positive and significant relationship between output and employment: gaps 0.14*** and changes 0.17***.
  - Cross-country heterogeneity in responsiveness of employment to output is negatively associated with poverty rate and skill mismatch; evidence on informality and business regulations is mixed or not significant in interactions.
- Policy implications (inferred from empirical patterns in the text):
  - Addressing poverty and skill mismatches may strengthen the employment response to growth.
  - Structural transformation (e.g., expansion of services) and labor force participation dynamics can alter the fit of standard Okun frameworks over time (Morocco example).
  - Deeper analysis and better data on informality are needed because informality can obscure the output–labor market relationship.

*Source: Authors’ estimates based on the IMF World Economic Outlook April 2017; Haut-commissariat au Plan, Morocco, 2016.*

### Appendix Table. Up- and Down-breaks in GDP Growth and Employment Growth

### Appendix Table. Up- and Down-breaks in GDP Growth and Employment Growth

### Up- and Down-breaks — Real GDP Growth
- Belize: Up-breaks 1998; Down-breaks 2006
- Bolivia: Up-breaks 2006; Down-breaks 1998
- Cabo Verde: Up-breaks 1992; Down-breaks 2008
- Cote d'Ivoire: Up-breaks 2011; Down-breaks
- Egypt: Up-breaks; Down-breaks
- El Salvador: Up-breaks; Down-breaks 1999, 2006
- Georgia: Up-breaks; Down-breaks
- Honduras: Up-breaks 1999; Down-breaks 2007
- India: Up-breaks 2002; Down-breaks
- Indonesia: Up-breaks 2006; Down-breaks 1998
- Kenya: Up-breaks 2009; Down-breaks
- Kyrgyzstan: Up-breaks; Down-breaks
- Moldova: Up-breaks 2001; Down-breaks
- Mongolia: Up-breaks; Down-breaks 1998
- Morocco: Up-breaks; Down-breaks
- Nicaragua: Up-breaks 1993; Down-breaks
- Pakistan: Up-breaks; Down-breaks 1996, 2004
- Philippines: Up-breaks 2009; Down-breaks
- Sao Tome and P.: Up-breaks 1988, 2002; Down-breaks
- Tajikistan: Up-breaks 2000; Down-breaks 2008
- Tanzania: Up-breaks 2003; Down-breaks
- Turkmenistan: Up-breaks; Down-breaks 2003
- Ukraine: Up-breaks 1999; Down-breaks 2007
- Uzbekistan: Up-breaks 1999; Down-breaks
- Vietnam: Up-breaks; Down-breaks 1997

### Up- and Down-breaks — Total Employment Growth
- Belize: Up-breaks ; Down-breaks 2007
- Bolivia: Up-breaks ; Down-breaks 1997, 2006
- Cabo Verde: Up-breaks ; Down-breaks 2000
- Cote d'Ivoire: Up-breaks 1988; Down-breaks 1996
- Egypt: Up-breaks ; Down-breaks
- El Salvador: Up-breaks 2006; Down-breaks 1998
- Georgia: Up-breaks ; Down-breaks
- Honduras: Up-breaks 2007; Down-breaks 2006
- India: Up-breaks 2004; Down-breaks 1987, 1996
- Indonesia: Up-breaks ; Down-breaks
- Kenya: Up-breaks ; Down-breaks 1996
- Kyrgyzstan: Up-breaks ; Down-breaks
- Moldova: Up-breaks 2010; Down-breaks
- Mongolia: Up-breaks 1990, 2006; Down-breaks
- Morocco: Up-breaks 2004; Down-breaks
- Nicaragua: Up-breaks ; Down-breaks
- Pakistan: Up-breaks ; Down-breaks
- Philippines: Up-breaks ; Down-breaks 1996
- Sao Tome and P.: Up-breaks ; Down-breaks
- Tajikistan: Up-breaks 2000; Down-breaks 2008
- Tanzania: Up-breaks 2011; Down-breaks
- Turkmenistan: Up-breaks 2003; Down-breaks
- Ukraine: Up-breaks ; Down-breaks
- Uzbekistan: Up-breaks 2001, 2010; Down-breaks
- Vietnam: Up-breaks ; Down-breaks

### Methodology and Notes
- Source: Authors’ estimates based on the IMF World Economic Outlook April 2017.
- Note: Up- and down-breaks in GDP growth and employment growth are identified using Berg et. al. (2012) methodology with minimum interstitiary period of 8 years.

*Source: Appendix Table. Up- and Down-breaks in GDP Growth and Employment Growth, wp17257.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp17257.pdf_