## _wp13146

## Source details

**Canonical URL:** [_wp13146](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13146.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13146.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13146.pdf.json)

---

### I. Introduction: research question and contribution
- Research question:
  - Does public hiring increase (“crowd in”) private employment or decrease (“crowd out”) private employment? If crowding out, is it “partial crowding out,” “full crowding out,” or “more than full crowding out”?
- Context and motivation:
  - Unemployment fell in much of the first decade of the 21st century but reversed after the great recession of 2008; employment rates are at their lowest levels in two decades.
  - Unemployment reduces output and income, increases inequality, erodes human capital, and has psychic and political costs.
  - Problem acute in the Middle East, North Africa, Caucasus, and Central Asia (MCD countries) with fast-growing labor forces and recent political transitions.
- Contribution:
  - Investigates effects of public hiring on unemployment and private employment, including crowding-out magnitude.
  - Assembles a dataset expanding non-OECD country coverage, with attention to MCD countries and data gaps.
  - Uses two measures of public employment: a “narrow” measure (public administration) and a “broad” measure (including other forms of public employment, including state-owned enterprises).

### II. Key mechanisms and literature findings
- Mechanisms discussed:
  - Product market channel: substitutability of public and private goods affects crowding-out magnitude.
  - Labor market channel: public-sector wages, job security, and probability of public employment alter private job search and acceptance.
  - Education channel: acquisition of qualifications aimed at public-sector jobs rather than private-sector productive skills.
- Selected empirical findings:
  - Algan et al. (2002): For 17 OECD countries (1960–2000), creation of 100 public jobs crowds out 150 private-sector jobs; unemployment equation implies 100 public jobs add 33 unemployed workers overall — “more than full crowding out.”
  - Malley and Moutos (1996): Evidence of full crowding out in Sweden.
  - Faggio and Overman (2012): Public employment increases local wages and can crowd out tradable-sector employment while crowding in nontradable-sector employment; short-run net effect in UK regional data was none, with crowding out over longer horizons.
  - Demekas and Kontolemis (2000): Public wage increases can lead to more-than-proportional private wage increases and higher unemployment (Greece).
  - Feldmann (2009a, 2009b): Larger public sector correlates with higher overall unemployment in developing countries; larger effects for women and youth and higher long-term unemployment shares.
- Gap addressed:
  - Most empirical work focuses on advanced countries; limited work extends to developing countries using direct public and private employment data.

### III. Data assembly and definitions
- Sample coverage:
  - Data collected for up to 194 countries over the period 1988–2011.
- Primary sources:
  - Key Indicators of the Labor Market (KILM), LABORSTA (ILO), and IMF World Economic Outlook (WEO); supplemental country reports and data provided to the IMF by country authorities.
- Public employment measures:
  - Narrow measure (“public administration”): ISIC category “Public administration and defense; compulsory social security” as reported to the ILO; excludes other public investment or business activities.
  - Broad measure: includes other forms of public employment, potentially including state-owned enterprises (SOEs).
- Other definitions and normalizations:
  - Public and private employment rates and the unemployment rate are expressed as a percentage of the total labor force.
  - Private-sector employment sometimes calculated by subtracting public employment from total employment when private employment not reported separately.
  - Controls: Real GDP growth, urbanization rate, trade openness (IMF WEO); Hiring and Firing Regulations index from Economic Freedom of the World (range 0 to 10, higher = freer labor markets).
  - Labor force primarily from the ILO; ILO is principal source of unemployment data.

### IV. Stylized facts and regional focus (MCD)
- Descriptive statistics (2006–2011 and related periods):
  - MCD unemployment rate averages about 9 percent, almost one and a half times that of the rest of the world (ROW).
  - MCD private-sector employment rate (broad definition) about 70 percent, almost 10 percentage points lower than in the ROW.
  - More than 20 percent of all MCD employees are in the public sector (broad definition), one and a half times as high as elsewhere.
  - MENAP region unemployment consistently around 10 percent.
  - CCA countries (Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyz Republic) show falling unemployment since 2000.
  - MENAP private-sector employment rates (narrow definition) lowest since at least the mid-1990s.
  - Government is a large employer in MENAP, CCA, and CEE relative to other regions; MENAP has the highest relative importance of public administration employment globally.
  - Using narrow definition, government employees comprise almost half the Kuwaiti labor force (proportion appears to have fallen substantially over time).
  - Country notes: Algeria and many Mashreq countries have high proportions in the public sector (broad measure); GCC countries show high proportions under the narrow definition.
- Time trends (Table 1 summary):
  - On balance, fall in ratio of public to private employment for the MCD as a whole.
  - MENAP: decline due to faster private-sector growth than public employment.
  - CCA: substantial fall in broadly defined public employment driven by downsizing and transition effects (SOE layoffs, privatization).
  - MENAP oil exporters (Kuwait, Oman, Qatar): falls in narrowly defined public employment; UAE pattern: fall in public administration employment offset by rise in broader government employment.
  - MENAP oil importers: broad-definition public employment proportions flat or rising; narrow-definition data limited but Morocco shows a short-period large rise in importance of public administration.
- Recent policy-context numbers:
  - Oman 2012 budget plans to hire more than 36,000 Omanis in the public sector (twice the historical rate).
  - Saudi 2012 budget provides for an additional 60,000 jobs in the Ministry of the Interior alone (well above historical rate for the entire government).
- Outlook implication:
  - GDP growth rates expected to be substantially lower than pre-Lehman Brothers levels post-2008; slower private-sector job creation expected while announced public hiring could accelerate public hiring and wage bills.

### V. Empirical approach and estimation strategies
- Four sets of regression specifications:
  - Regressions of unemployment on public employment.
  - Regressions of private-sector employment on public employment.
  - Use of both broad and narrow measures of public employment.
  - Variety of estimation approaches (FE, RE, FE with AR(1), RE with AR(1), Systems GMM, dynamic panels).
- Equations (compact notation from source):
  - Unemployment equation: Unemp_it = β Prv_it + γ Pub_it + ν_i + θ_t + ε_it
  - Employment equation: Prv_it = β Pub_it + γ X_it + ν_i + θ_t + ε_it
  - All rates are percentages of the total labor force.
- Sample, periods, and controls:
  - Six periods of data ranging from 1988 to 2011.
  - Coverage up to 139 countries and 454 observations depending on specification.
  - Controls X: GDP growth rate, urbanization rate, openness (trade/GDP), Hiring and Firing Regulations index.
- Identification and endogeneity mitigation:
  - Fixed effects control for time-invariant unobservables; GDP growth controls for cyclical private labor demand; hiring and firing index controls for legislative changes.
  - Systems GMM and dynamic panels use internal instruments (lags) due to data limitations; external instruments not used.
  - Formal bias discussion follows Wooldridge (2002) logic.

### VI. Key empirical findings — unemployment regressions
- Broad public-sector employment rate (Table 2, columns (1)–(3)):
  - Impact on unemployment rate close to zero; coefficients insignificantly different from 0 and consistent with full crowding out.
  - Reported coefficients for UnempRt:
    - Public-Sector Employment Rate: -0.050, -0.027, 0.066 (columns (1), (2), (3))
    - Public-Administration Employment Rate: -0.060, 0.242*, -0.167** (columns (1), (2), (3))
  - GDP Growth Rate coefficients (examples): -0.062, -0.079, -0.120, -0.428*** (various columns).
  - Observations reported: 419, 292, 292, 441, 314, 314 across columns (1)–(6).
  - Number of countries reported: 116, 82, 21, 33, 94, 94.
  - p-value (H0: b=-1) reported as 0.000 in all six columns in Table 2.
- Narrow (public administration) measure (Table 2, columns (4)–(6)):
  - Without controls (column (4)) results align with broad measure.
  - With controls (column (5)), public administration employment rate increases the unemployment rate significantly: a 1 percentage point increase in public administration employment rate leads to a ¼ percentage point higher unemployment rate.
  - GMM estimates (column (6)) imply significantly less than full crowding out for public administration employment, but the extent remains large.
- Systems GMM (Table A.4) highlights:
  - Short run coefficients indicate an initial positive effect of public sector hiring on labor markets but a subsequent negative effect, resulting in full long run crowding out.
  - Selected GMM estimates (examples):
    - Unemployment Rate (-1): 0.404***; 0.278*; 0.641***; 0.751***.
    - Public-Sector Employment Rate: 0.129; 0.197***.
    - LR coefficient examples: 0.217; -0.089; 0.404; -0.192.
    - p-value(H0: LR = -1): 0.000; 0.000; 0.000; 0.133.
  - Interpretation: some specifications point to more than full crowding out for public employment (significant in column (3)) and (insignificantly) less than full crowding out for public administration employment.

### VII. Key empirical findings — private-sector employment regressions
- Strong negative relationship between public- and private-sector employment rates; coefficients generally close to -1 and not significantly different from -1 (consistent with full crowding out).
- Examples from Table 3:
  - Public-Sector Employment Rate coefficients (PrvEmpRt columns): -0.849***, -0.979***, -1.271*** (columns (1), (2), (3)).
  - Public-Administration Employment Rate coefficients (PrvEmpRt columns): -1.007***, -1.059***, -0.795*** (columns (4), (5), (6)).
  - Example interpretation: "100 new public jobs crowd out 98 private jobs (column (2))."
- Additional statistics (Table 3 examples):
  - Observations: 396, 282, 282, 454, 315, 315 across columns (1)–(6).
  - R-squared examples: 0.488, 0.644, 0.139, 0.148.
  - p-value (H0: b=-1) examples: 0.014, 0.680, 0.027, 0.968, 0.772, 0.129.
- Systems GMM for private employment (Table A.8):
  - Short run coefficients indicate substantially less than full crowding out in some specifications; long run estimates often insignificantly different from -1 (consistent with full crowding out).
  - Selected LR coefficient examples: -0.78; -0.825; -0.945; -1.143.
  - p-value(H0: LR = -1): 0.337; 0.522; 0.982; 0.306.

### VIII. Regional heterogeneity — MCD (MENAP and CCA)
- Estimation approach:
  - Interaction terms with MENAP and CCA region dummies added; models often reported without controls due to limited coverage.
- Unemployment equation with interactions (Table 4, columns (1)–(2)):
  - Broad measure (column (1)): Public-Sector Employment Rate -0.096**; Public Employment X MENAP 0.156**; Public Employment X CCA 0.355***.
  - Narrow measure (column (2)): Public-Administration Employment Rate -0.078; Public Administration Employment X CCA 0.259***.
  - Aggregate tests:
    - MENAP countries: no clear relationship between unemployment and public employment.
    - CCA countries: positive effect of public employment on unemployment (worse and more-than-full crowding out in unemployment equations).
- Private employment equation with interactions (Table 4, columns (3)–(4)):
  - Broad measure (column (3)): Public-Sector Employment Rate -0.828***; Public Employment X MENAP -0.274*.
  - Narrow measure (column (4)): Public-Administration Employment Rate -0.891***; Public Administration Employment X CCA 0.291***.
  - Aggregate tests:
    - Crowding-out effects exist in both MENAP and CCA.
    - Full crowding out in MENAP; reject null of full crowding out in CCA.
- Net regional assessment:
  - MENAP countries may have stronger crowding-out effects than the rest of the world.
  - CCA countries: unemployment equations indicate more-than-full crowding out, while employment equations point to only partial crowding out.
- Appendix with controls (Table A.9 and Table A.10) notes:
  - Including controls sometimes increases standard errors and changes statistical significance of CCA coefficients.
  - Private-sector employment regressions (Table A.10) show MENAP crowding out larger than rest of world; CCA results less reliable due to coverage but sometimes imply crowding in in specific specifications.
- Selected numeric highlights from Table 4:
  - Observations: 419, 287, 396, 293 across columns (1)–(4).
  - Number of countries: 116, 88, 110, 89 across columns (1)–(4).
  - p-value(H0: b+bM=0) examples: 0.206, 0.128, 0.000, 0.000.
  - p-value(H0: b+bC=-1) examples: 0.000, 0.000, 0.019, 0.003.

### IX. Labor force participation implications
- Stylized facts:
  - Many MCD countries have low labor force participation rates and high unemployment, yielding low employment-to-population ratios.
- Inference from contrasting employment and unemployment results:
  - CCA: unemployment equations suggest more-than-full crowding out while employment equations suggest partial crowding out → suggests fall in government employment may have led to decreased labor force participation.
  - MENAP and full sample on average: similar effects in employment and unemployment specifications → generally no large net effect on labor force participation.
- Suggested follow-up:
  - Direct analysis of relationship between labor force participation and public hiring recommended as valuable corroboration.

### X. Robustness, alternative estimators, and selected appendix findings
- Estimation methods and robustness checks used:
  - FE with AR(1), RE, RE with AR(1), Systems GMM, dynamic panels, GMM/GIV in static framework, long lags, three-stage-least squares, cointegration analysis, time-series causality analysis.
  - Internal instruments (lags) used due to limited data; external instruments not used for panel of developing countries.
- Appendix I (Unemployment regressions) key points:
  - FE AR(1) (Table A.1): main finding—“All coefficients are insignificant and very close to our main results in Table 2, suggesting that the public employment rate has no clear relationship with unemployment rate.”
  - Selected FE AR(1) coefficients examples:
    - Public-Sector Employment Rate: 0.031; -0.004 (std. errors (0.076) (0.091)).
    - GDP Growth Rate: -0.309***; -0.386***.
    - Observations: 303; 210; 308; 220.
    - p-value(H0: b=-1): 0.000 in multiple columns.
  - RE and RE AR(1) (Tables A.2–A.3): coefficients close to zero; Hausman tests mixed.
  - Systems GMM (Table A.4): some specifications point to more than full crowding out; LR p-values often reject LR = -1.
- Appendix II (Private-sector regressions) key points:
  - FE AR(1) (Table A.5): “The crowding out effects reported here are lower than elsewhere in the paper.”
  - RE and RE AR(1) (Tables A.6–A.7): Public-Sector Employment Rate coefficients often around -0.896*** to -0.978***.
  - Systems GMM (Table A.8): LR coefficient examples: -0.78; -0.825; -0.945; -1.143; p-values for LR = -1 vary across specifications.
- Appendix III (MCD interactions with controls) highlights:
  - Adding controls can increase standard errors on regional interaction terms and change statistical significance.
  - Selected coefficients (examples from Table A.9 and Table A.10):
    - Public Employment X CCA: 0.249***; 1.366; -0.453 (various columns).
    - Public-Administration Employment X CCA: 0.206***; 1.273; -0.426.
    - Public-Sector Employment Rate (private equation): -0.891***; -0.963***; -0.980*** (columns (1)–(3) in Table A.10).
    - Public Employment X CCA in private equation: 0.125; 4.944***; 3.761***.

### XI. Main conclusions and policy implications
- Core empirical conclusion:
  - Regressions of unemployment on public employment and of private employment on public employment (two definitions of public employment) find robust evidence that public employment crowds out private employment; magnitude statistically indistinguishable from full crowding out for the complete sample of developing and advanced countries.
  - Therefore, a public job typically comes at the cost of a private-sector job and does not reduce overall unemployment.
- Specific policy implications:
  - Public-sector hiring:
    - (i) does not reduce unemployment,
    - (ii) increases the fiscal burden, and
    - (iii) inhibits long-term growth through reductions in private-sector employment.
  - Scarce public resources could be better spent on other social needs, including protecting the most vulnerable.
- Regional cautions and context:
  - Public sector is an important employer in MCD countries.
  - Data hint that public employment has fallen over time in MCD countries with a sizeable reduction in the broadly defined public sector in the CCA; econometric results imply this did not worsen unemployment.
  - Youth preference for public employment and recent public hiring initiatives (post-Arab Spring) could reverse trends; public hiring under weak private-sector employment growth could worsen unemployment and fiscal sustainability.
- Suggested further research:
  - Identify mechanisms (e.g., relationship between public and private sector wages, reservation wages).
  - Study relationship between government wage shares, tax rates, and private-sector employment.
  - Investigate efficacy of wage subsidy or salary top-up schemes for private-sector employment.

*Source: _wp13146 (PDF content provided).*

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

### _wp13146 - References

### I. Introduction: research question and contribution
- Research question: Does public hiring increase (“crowd in”) private employment or decrease (“crowd out”) private employment? If crowding out, is it “partial crowding out,” “full crowding out,” or “more than full crowding out”?
- Context and motivation:
  - Unemployment fell in much of the first decade of the 21st century but reversed after the great recession of 2008; employment rates are at their lowest levels in two decades.
  - Unemployment reduces output and income, increases inequality, erodes human capital, and has psychic and political costs (Layard, Nickell, and Jackman (2005); Kapstein and Converse (2008); World Bank (2012a); IMF (2010)).
  - The problem is especially acute in the Middle East, North Africa, Caucasus, and Central Asia (MCD countries), with fast-growing labor forces and recent political transitions exacerbating labor market pressures.
- Contribution of the paper:
  - Investigates effects of public hiring on unemployment and private employment, including whether public jobs crowd out private jobs and the implied effects on unemployment.
  - Assembles a dataset expanding non-OECD country coverage, with special attention to MCD countries and data gaps.
  - Uses two measures of public employment: a “narrow” measure (public administration) and a “broad” measure (including other forms of public employment, including state-owned enterprises).
- Core empirical result summary:
  - Overall results point to full crowding out: every public job comes at the cost of a private-sector job and does not reduce overall unemployment.
  - For the MCD countries, there is slight evidence of larger crowding-out effects.
- Policy implication highlighted:
  - Public-sector hiring at best offers no employment benefit and should be replaced with more socially beneficial fiscal expenditure.

### II. Key mechanisms and literature findings
- Mechanisms of crowding out discussed:
  - Product market channel: government activity can replace private-sector products, with substitutability of public and private goods determining crowding-out magnitude.
  - Labor market channel: public-sector wages, job security, and probability of public employment can alter private-sector job search and acceptance behavior.
  - Education channel: individuals may acquire qualifications aimed at public-sector jobs rather than private-sector productive skills.
- Selected empirical findings from literature:
  - Algan et al. (2002): For 17 OECD countries (1960–2000), creation of 100 public jobs crowds out 150 private-sector jobs; unemployment equation implies 100 public jobs add 33 unemployed workers overall — interpreted as more than full crowding out.
  - Malley and Moutos (1996): Evidence of full crowding out in Sweden using Vector Error Correction Model.
  - Faggio and Overman (2012): Public employment increases local wages and can crowd out tradable-sector employment while crowding in nontradable-sector employment; short-run net effect in UK regional data was none, with crowding out occurring over longer horizons.
  - Demekas and Kontolemis (2000): Public wage increases can lead to more-than-proportional private wage increases and higher unemployment via wage channels (evidence for Greece).
  - Feldmann (2009a, 2009b): Larger public sector correlates with higher overall unemployment in developing countries, with larger effects for women and youth and higher long-term unemployment shares; uses an index-based measure of government size rather than direct public employment data.
- Gaps in prior work:
  - Most empirical work focuses on advanced countries; very little has extended to developing countries using direct public and private employment data.

### III. Data assembly and definitions
- Sample coverage:
  - Data collected for up to 194 countries over the period 1988–2011.
- Primary sources:
  - Key Indicators of the Labor Market (KILM) and LABORSTA (ILO), and IMF World Economic Outlook (WEO).
  - Supplemental sources: individual country reports and data provided to the IMF by country authorities, especially for MCD countries where official series are limited.
- Public employment measures constructed:
  - Narrow measure (“public administration”): occupations classified under ISIC category “Public administration and defense; compulsory social security” as reported by national authorities to the ILO and other sources. This excludes other public investment or business activities.
  - Broad measure: includes other forms of public employment, potentially including state-owned enterprises (described elsewhere in the source text).

### IV. Stylized facts and regional focus (MCD)
- Stylized facts confirmed by the data (described in Section IV of the source):
  - Unemployment and private-sector employment outcomes in the MCD are among the worst in the world.
  - The relative size of government employment in the MCD is high by global standards.
  - The relative importance of government employment has been falling over time in the data series but may not capture recent accelerations in public-sector hiring.

### V. Empirical approach and main estimation strategies
- Four sets of regression specifications:
  - Regressions of unemployment on public employment.
  - Regressions of private-sector employment on public employment.
  - Use of both broad and narrow measures of public employment.
  - Variety of estimation approaches (detailed regression tables and appendices are provided in the source).
- Appendices and tables referenced (contents preserved from source):
  - Appendix I. Regressions of Unemployment Rate on Public Employment.
  - Appendix II. Regressions of Private-Sector Employment on Public Employment.
  - Appendix III. Regression with MCD Interaction Terms and Controls.
  - Table 1 : Public and Private Employment in the MCD Countries.
  - Table 2 : Regression of Unemployment Rate on Public Employment Rate.
  - Table 3 : Regression of Private-Sector Employment Rate on Public Employment Rate.
  - Table 4 : Regression with MCD Region Interaction Terms.
  - Figures include: overview of key labor statistics (2008–2011 average), unemployment rate weighted by labor force, private-sector employment rate, importance of public-sector employment (2008–2011 average), proportion of public employment in the MCD countries (latest year), unemployment and public employment (2006–11 average), and correlation between public- and private-sector employment (2006–11 average).

### VI. Findings, interpretation, and policy recommendations
- Main empirical finding:
  - Evidence points to full crowding out: each public-sector job is associated with the loss of a private-sector job, leaving overall unemployment unchanged.
- Implications for public hiring:
  - Public hiring does not reduce unemployment and therefore offers at best no employment benefit.
  - Public-sector hiring should be replaced with more socially beneficial fiscal expenditure.
- Regional nuance:
  - Slight evidence that crowding-out effects are larger in MCD countries.
- Suggested next steps:
  - Further research avenues are offered in Section VI of the source (details of specific suggestions appear in the full text).

*Source: _wp13146 - References (PDF content provided).*

### 2.      Meanwhile, in addition to the public employment data above, the ILO LABORSTA

### _wp13146 - 2.      Meanwhile, in addition to the public employment data above, the ILO LABORSTA

### Data sources and definitions
- Public-sector employment data:
  - ILO LABORSTA dataset “Public Sector Employment” includes governmental agencies and state-owned enterprises (SOEs); referred to as the “broad” measure of public employment.
  - Narrow measure excludes SOEs (public administration only).
- Private-sector employment:
  - Calculated where necessary by subtracting public employment from total employment when countries do not report private-sector employment separately.
  - Both resulting series are referred to as “private-sector employment” or “private employment” regardless of broad or narrow public measures.
- Labor force and unemployment:
  - Labor force primarily obtained from the ILO and supplemented with other sources.
  - ILO is the principal source of unemployment data.
- Control variables for regressions:
  - Real GDP growth, urbanization rate, and trade openness from the IMF WEO database.
  - Labor rigidity indicators (e.g., “Hiring and firing regulations” index) from the Economic Freedom of the World (EFW) index. The index ranges from 0 to 10, with higher ratings indicating freer labor markets.
- Normalization:
  - Public and private employment rates and the unemployment rate are expressed as a percentage of the total labor force.

### Descriptive analysis — key facts and statistics (2006–2011 and related periods)
- Regional patterns (Figure 1, 2008–2011 average):
  - MCD unemployment rate averages about 9 percent, almost one and a half times that of the rest of the world (ROW).
  - MCD private-sector employment rate (broad definition) about 70 percent, almost 10 percentage points lower than in the ROW.
  - More than 20 percent of all MCD employees are in the public sector (broad definition), one and a half times as high as elsewhere.
- Regional unemployment evolution (Figure 2):
  - MENAP region shows a high unemployment rate, consistently around 10 percent.
  - CCA countries (with sufficient data: Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyz Republic) show falling unemployment since 2000.
- Private-sector employment rates (Figure 3):
  - MCD private-sector employment rates are lower than all regions except CEE under the broad definition.
  - MENAP private-sector employment rates (narrow definition) have been the lowest since at least the mid-1990s.
  - CCA private-sector employment (broad definition including SOEs) is not materially lower than most regions and has risen slightly since 2000.
- Importance of public employment (Figure 4 and Figure 5):
  - Government is a large employer in MENAP, CCA, and CEE relative to other regions.
  - MENAP has the highest relative importance of public administration employment in the world.
  - Using narrow definition, government employees comprise almost half the Kuwaiti labor force (proportion appears to have fallen substantially over time).
  - Specific country observations: Algeria and many Mashreq countries have high proportions of employees in the public sector (broad measure); GCC countries show high proportions under the narrow definition.
- Time trends and Table 1 summary:
  - On balance, evidence suggests a fall in the ratio of public to private employment for the MCD as a whole.
  - MENAP: decline generally due to faster private-sector growth than public employment.
  - CCA: substantial fall in broadly defined public employment driven by downsizing and transition effects (e.g., layoffs from SOEs, privatization).
  - MENAP oil exporters (e.g., Kuwait, Oman, Qatar) show falls in narrowly defined public employment; UAE pattern suggests fall in public administration employment offset by rise in broader government employment.
  - MENAP oil importers: broad-definition public employment proportions flat or rising; narrow-definition data limited but show large rise in importance of public administration in Morocco over a short period.
- Recent policy/contextual numbers:
  - Oman 2012 budget plans to hire more than 36,000 Omanis in the public sector (twice the historical rate).
  - Saudi 2012 budget provides for an additional 60,000 jobs in the Ministry of the Interior alone (well above historical rate for the entire government).
- Macroeconomic outlook and implications:
  - Over the next few years (post-observed-data period up to 2008), GDP growth rates expected to be substantially lower than pre-Lehman Brothers levels because of global slowdown and political unrest (Arab Spring), despite high oil prices for hydrocarbon exporters.
  - Slower GDP growth implies slower private-sector job creation, while announced initiatives to increase government employment could accelerate public hiring and wage bills.
  - Recent budget information and IMF forecasts point to a substantial rise in government expenditure on public employment in both MENAP and CCA countries.

### Key empirical relationships and graphical findings
- Unemployment vs. public employment (Figure 6, 2006–11 average):
  - No clear relationship between unemployment and public employment using either broad or narrow public measures.
  - Variations in unemployment likely reflect multiple structural factors (demographic transition, skills mismatches, labor market rigidities, high reservation wages).
- Public vs. private employment (Figure 7, 2006–11 average):
  - Clear negative correlations between public employment rate and private-sector employment rate for both broad and narrow definitions.
  - Graphical evidence suggests crowding out: higher public employment rates associated with lower private-sector employment rates across the world and within the MCD region in particular.

### Econometric analysis — specification, estimation, and identification
- Equations estimated (four equations total):
  - Unemployment equation: Unemp_it = β Prv_it + γ Pub_it + ν_i + θ_t + ε_it (notation from source text)
  - Employment equation: Prv_it = β Pub_it + γ X_it + ν_i + θ_t + ε_it (source provides compact notation)
  - All rates are percentages of the total labor force.
- Sample and periods:
  - Six periods of data ranging from 1988 to 2011.
  - Coverage falls with data availability; up to 139 countries and 454 observations depending on specification.
- Control variables X:
  - GDP growth rate (controls for business cycle fluctuations, with time dummies).
  - Urbanization rate.
  - Openness (ratio of trade to GDP).
  - Hiring and firing regulations index (EFW).
- Estimation methods:
  - Several methods used; main focus on fixed effect (FE) within-groups estimation to control for country-specific time-invariant unobservables and slight measurement differences.
- Endogeneity concerns and mitigation:
  - Public hiring may respond to labor market slack (reverse causality). FE controls for time-invariant unobservables; GDP growth controls for cyclical private labor demand; hiring and firing index controls for legislative changes over time.
  - Residual time-varying country-specific unobservables may remain. The direction of potential bias depends on how government hires change relative to labor force and private hiring dynamics.
  - The source provides formal discussion of bias following Wooldridge (2002) logic (expressions shown in original text).

### Summary conclusions from the presented sections
- MCD region exhibits higher unemployment, lower private-sector employment, and a larger role for the state as employer relative to other regions.
- Empirical graphs show:
  - No clear relationship between public employment and unemployment.
  - A strong negative correlation between public employment rates and private-sector employment rates consistent with crowding out.
- Policy environment and outlook:
  - Slower GDP growth post-2008 and political unrest could slow private-sector job creation.
  - Announced and budgeted increases in public employment and wages in MENAP and CCA could raise public employment and wage bills, potentially reinforcing crowding-out dynamics unless private-sector job creation accelerates.

*Source: _wp13146 - 2.      Meanwhile, in addition to the public employment data above, the ILO LABORSTA*

### 0. There is an underestimate of

### _wp13146 - 0. There is an underestimate of

### Methodology and identification
- Estimation methods:
  - Generalized Method of Moments (GMM)/Generalized Instrumental Variables (GIV) in a static framework, estimated as a system using multiple additional moment conditions (Cameron and Trivedi, 2006).
  - Use of potentially multiple lags of the endogenous variables (public employment and GDP growth) in levels as instruments for first-differenced equations and lags of differences as instruments for levels.
  - Fixed-effect (FE) estimation with autoregressive disturbances and random effects (RE) with and without autoregressive disturbances.
  - Dynamic panel regressions with one or two lags of the dependent variable and dynamic systems GMM to address endogeneity of public employment and GDP.
- Robustness checks:
  - Long lags of government size measures, three-stage-least squares, cointegration analysis, time-series causality analysis cited from literature.
  - Results from other estimation methods are reported in Appendices.
- Instrument strategy:
  - Internal instruments used due to data availability; external instruments not used for the panel of developing countries.

### Interpreting coefficients on public employment
- Unemployment equation interpretation (coefficient uβ):
  - uβ close to -1: additional public jobs are purely accounted for by a fall in unemployment → no crowding out.
  - uβ < -1: public employment generates private-sector jobs → crowding in.
  - uβ between 0 and -1: partial crowding out (some private jobs lost but fewer than public jobs created).
  - uβ close to 0: no change in unemployment because public job creation is completely cancelled by private job losses → full crowding out.
  - uβ > 0: overall unemployment rises → more than full crowding out.
- Private-sector employment equation interpretation (coefficient pβ):
  - pβ > 0: public employment crowds in private-sector employment.
  - pβ close to 0: no evidence of crowding out.
  - pβ between -1 and 0: partial crowding out.
  - pβ = -1: full crowding out (one private job substituted by one public job).
  - pβ < -1: more than full crowding out (private-sector job losses exceed public jobs created).

### Key empirical findings — unemployment regressions (high-level)
- Broad public-sector employment rate (columns (1)–(3)):
  - Impact on the unemployment rate is close to zero; p-values reject the hypothesis that there is no crowding out, implying at least partial crowding out.
  - Coefficients are insignificantly different from 0, consistent with full crowding out.
  - Specific reported coefficients for UnempRt:
    - Public-Sector Employment Rate: -0.050, -0.027, 0.066 (columns (1), (2), (3))
    - Public-Administration Employment Rate: -0.060, 0.242*, -0.167** (columns (1), (2), (3))
- Narrow (public administration) measure (columns (4)–(6)):
  - Without controls (column (4)) results align with broad measure.
  - With controls (column (5)), the public administration employment rate increases the unemployment rate significantly: a 1 percentage point increase in the public administration employment rate leads to a ¼ percentage point higher unemployment rate.
  - GMM estimates (column (6)) imply significantly less than full crowding out for public administration employment, but the extent remains large.
- Additional reported statistics in Table 2:
  - GDP Growth Rate coefficients in UnempRt columns include -0.062, -0.079, -0.120, -0.428*** (various columns).
  - Observations reported: 419, 292, 292, 441, 314, 314 across columns (1)–(6).
  - Number of countries reported: 116, 82, 21, 33, 94, 94 (as listed in Table 2).
  - p-value (H0: b=-1) reported as 0.000 in all six columns in Table 2.

### Key empirical findings — private-sector employment regressions (high-level)
- Strong negative relationship between public- and private-sector employment rates; coefficients generally close to -1 and not significantly different from -1.
- Examples from Table 3:
  - Public-Sector Employment Rate coefficients (PrvEmpRt columns): -0.849***, -0.979***, -1.271*** (columns (1), (2), (3))
  - Public-Administration Employment Rate coefficients (PrvEmpRt columns): -1.007***, -1.059***, -0.795*** (columns (4), (5), (6))
  - Interpretation example: "100 new public jobs crowd out 98 private jobs (column (2))."
- Additional reported statistics in Table 3:
  - Observations: 396, 282, 282, 454, 315, 315 across columns (1)–(6).
  - R-squared examples: 0.488, 0.644, 0.139, 0.148 across select columns.
  - p-value (H0: b=-1) examples: 0.014, 0.680, 0.027, 0.968, 0.772, 0.129 across columns.
- Summary inference:
  - Taken together with the unemployment results, public employment just about fully crowds out private-sector employment regardless of the definition; a rise in government hiring would be offset by decreases in private employment, resulting in no change in overall unemployment.
  - Symmetrically, reduced government hiring would be almost completely offset by additional private-sector jobs.

### Regional (MCD) heterogeneity — MENAP and CCA
- Approach:
  - Add interaction terms with MENAP and CCA region dummies; focus on model without control variables due to limited country coverage and identification concerns.
- Unemployment equation with interactions (Table 4, columns (1)–(2)):
  - Broad measure (column (1)): both MENAP and CCA interaction terms increase the effect of public-sector employment → suggests increases in public-sector employment in MCD region induce more unemployment.
  - Narrow measure (column (2)): no additional effect in MENAP, but a higher and strong positive effect in the CCA.
  - Aggregate tests:
    - MENAP countries show no clear relationship between unemployment and public employment.
    - CCA countries show a positive effect of public employment on unemployment (worse and more-than-full crowding out in unemployment equations).
- Private employment equation with interactions (Table 4, columns (3)–(4)):
  - Broad measure: MENAP shows stronger crowding-out effects than others; CCA shows no significant additional effect.
  - Narrow measure: crowding out in MENAP similar to others; crowding out significantly dampened in CCA.
  - Aggregate tests indicate:
    - Crowding-out effects exist in both MENAP and CCA.
    - Full crowding out in MENAP; reject null of full crowding out in CCA.
- Net regional assessment:
  - MENAP countries may have stronger crowding-out effects than the rest of the world.
  - CCA countries: unemployment equations indicate more-than-full crowding out, while employment equations point to only partial crowding out.
- Selected numeric values from Table 4:
  - Column (1) UnempRt: Public-Sector Employment Rate -0.096**; Public Employment X MENAP 0.156**; Public Employment X CCA 0.355***.
  - Column (2) UnempRt (narrow): Public-Administration Employment Rate -0.078; Public Administration Employment X CCA 0.259***.
  - Column (3) PrvEmpRt: Public-Sector Employment Rate -0.828***; Public Employment X MENAP -0.274*.
  - Column (4) PrvEmpRt (narrow): Public-Administration Employment Rate -0.891***; Public Administration Employment X CCA 0.291***.
  - Observations in Table 4: 419, 287, 396, 293 across columns (1)–(4).
  - Number of countries in Table 4: 116, 88, 110, 89 across columns (1)–(4).
  - p-value(H0: b+bM=0) examples: 0.206, 0.128, 0.000, 0.000.
  - p-value(H0: b+bC=-1) examples: 0.000, 0.000, 0.019, 0.003.

### Labor force participation implications
- Observed stylized facts:
  - Many MCD countries have low labor force participation rates and high unemployment, yielding particularly low employment-to-population ratios.
- Interpretation:
  - Comparing employment and unemployment results can infer effects of public hiring on labor force participation.
  - For CCA: unemployment equations suggest more-than-full crowding out while employment equations suggest partial crowding out → fall in government employment may have led to decreased labor force participation.
  - For MENAP and the full sample on average: similar effects in employment and unemployment specifications → generally no large net effect on labor force participation.
- Recommendation:
  - A more detailed direct analysis of the relationship between labor force participation and public hiring is suggested as a valuable corroboration.

### Main conclusions and policy implications
- Core empirical conclusion:
  - Regressions of unemployment on public employment and of private employment on public employment (two definitions of public employment) find robust evidence that public employment crowds out private employment; the magnitude is statistically indistinguishable from full crowding out for the complete sample of developing and advanced countries.
  - Therefore, a public job typically comes at the cost of a private-sector job and does not reduce overall unemployment.
- Specific policy implications summarized in the paper:
  - Public-sector hiring:
    - (i) does not reduce unemployment,
    - (ii) increases the fiscal burden, and
    - (iii) inhibits long-term growth through reductions in private-sector employment.
  - Together, these imply public hiring is detrimental to long term fiscal sustainability with limited benefit; scarce resources could be better spent on other social needs, including protecting the most vulnerable.
- Regional cautions:
  - Public sector is an important employer in MCD countries.
  - Data hint that public employment has fallen over time in MCD countries with a sizeable reduction in the broadly defined public sector in the CCA; econometric results imply this did not worsen unemployment.
  - However, youth preference for public employment and recent public hiring initiatives (in response to the Arab Spring) could reverse trends; public hiring under weak private-sector employment growth could worsen the problem.
- Suggested further research:
  - Identify mechanisms: e.g., relationship between public and private sector wages (reservation wages), relationship between government wage shares, tax rates, and private-sector employment.
  - Investigate efficacy of wage subsidy or salary top-up schemes for private-sector employment.

*Source: _wp13146 - 0. There is an underestimate of*

### APPENDIX I. REGRESSIONS OF UNEMPLOYMENT RATE ON PUBLIC EMPLOYMENT

### APPENDIX I. REGRESSIONS OF UNEMPLOYMENT RATE ON PUBLIC EMPLOYMENT

### Fixed Effects (FE) with AR(1) and serial correlation tests
- Table A.1 reports FE estimation of panel models with AR(1) disturbances by Baltagi and Wu (1999).
- Wooldridge (2002) serial correlation test p-value reported at bottom of table.
- Main finding: "All coefficients are insignificant and very close to our main results in Table 2, suggesting that the public employment rate has no clear relationship with unemployment rate."
- Selected coefficients and statistics (columns (1)–(4)):
  - Public-Sector Employment Rate: 0.031; -0.004 (standard errors (0.076) (0.091))
  - Public-Administration Employment Rate: -0.228; 0.079 (standard errors (0.167) (0.237))
  - GDP Growth Rate: -0.309***; -0.386*** (standard errors (0.081) (0.093))
  - Urbanization Rate: 0.244; 0.293 (standard errors (0.147) (0.204))
  - Hiring and Firing Regulations: -0.396; 0.162 (standard errors (0.364) (0.345))
  - Openness: 0.001*; 0.024* (standard errors (0.000) (0.014))
  - Observations: 303; 210; 308; 220
  - R-squared: 0.153; 0.219; 0.075; 0.323
  - Number of countries: 99; 74; 106; 81
  - p-value(H0: b=-1): 0.000; 0.000; 0.000; 0.000
  - Autocorr Test: 0.000; 0.000; 0.000; 0.000
- Robust standard errors in parentheses; significance: *** p<0.01, ** p<0.05, * p<0.1.
- All FE regressions include a constant term and are estimated with country-specific and period-specific fixed effects.

### Random Effects (RE) and RE with AR(1)
- Tables A.2 and A.3 present RE and RE with AR(1) disturbances.
- Results: coefficients close to zero; unlike Table 2 there is no significantly positive coefficient.
- Hausman Test p-values reported and "offers mixed results regarding the choice between FE and RE approaches."
- Selected RE coefficients (Table A.2, columns (1)–(4)):
  - Public-Sector Employment Rate: -0.029; -0.051 (std. errors (0.028) (0.048))
  - Public Administration Employment Rate: -0.034; 0.022 (std. errors (0.069) (0.097))
  - GDP Growth Rate: -0.092; -0.093 (std. errors (0.063) (0.072))
  - Hiring and Firing Regulations: -0.468**; -0.248 (std. errors (0.224) (0.222))
  - Openness: -0.000; -0.007 (std. errors (0.000) (0.009))
  - Observations: 419; 292; 441; 314
  - Number of countries: 116; 82; 133; 94
  - Hausman Test: 0.000; 0.001; 0.000; 1.000
  - p-value(H0: b=-1): 0.000; 0.000; 0.000; 0.000
  - Autocorr Test: 0.000; 0.000; 0.000; 0.000
- Selected RE with AR(1) coefficients (Table A.3, columns (1)–(4)):
  - Public-Sector Employment Rate: -0.035; -0.027 (std. errors (0.040) (0.054))
  - Public Administration Employment Rate: -0.022; 0.069 (std. errors (0.078) (0.132))
  - Hiring and Firing Regulations: -0.516**; -0.183 (std. errors (0.249) (0.235))
  - Openness: -0.000***; -0.003 (std. errors (0.000) (0.011))
  - Observations: 419; 292; 441; 314
  - Number of countries: 116; 82; 133; 94
  - Hausman Test: 0.861; 0.073; 0.039; 0.086
  - p-value(H0: b=-1): 0.000; 0.000; 0.000; 0.000
  - Autocorr Test: 0.000; 0.000; 0.000; 0.000
- Robust standard errors in parentheses; significance: *** p<0.01, ** p<0.05, * p<0.1.
- All RE regressions contain a constant term and are estimated with region-specific and period-specific fixed effects.

### Systems GMM estimates for unemployment (Table A.4)
- Table A.4 reports Systems GMM estimates; long run coefficients robust to instrument set chosen to reduce standard errors subject to Sargan and Arellano and Bond (1991) autocorrelation tests.
- Notes:
  - Specifications with one lag of the dependent variable (columns (1)–(2)) indicate significant second-order autocorrelation.
  - Given second-order autocorrelation, columns (3)–(4) introduce an additional lag of the dependent variable.
  - Short run coefficients tended to indicate an initial short run positive effect of public sector hiring on labor markets (higher employment or lower unemployment) but a subsequent negative effect, resulting in full long run crowding out (footnote: specifications with lags of explanatory variables yielded similar long run results but with much larger standard errors).
- Selected estimates (columns (1)–(4)):
  - Unemployment Rate (-1): 0.404***; 0.278*; 0.641***; 0.751*** [GMM standard errors in brackets [0.126][0.143][0.119][0.141]]
  - Unemployment Rate (-2): -0.130*; -0.076 [0.073][0.049]
  - Public-Sector Employment Rate: 0.129; 0.197*** [0.091][0.069]
  - Public Administration Employment Rate: -0.064; -0.062 [0.148][0.179]
  - GDP Growth Rate: -0.187; -0.232; -0.411**; -0.659*** [0.126][0.145][0.198][0.200]
  - Openness: -0.000**; -0.029*; -0.006; 0.016 [0.000][0.016][0.057][0.049]
  - Urbanization Rate: 0.01; -0.077; -0.187; 0.24 [0.070][0.051][0.251][0.213]
  - Hiring and Firing Regulations: 0.249; 0.306; 0; -0.021 [0.263][0.237][0.000][0.027]
  - Observations: 260; 285; 216; 241
  - Number of countries: 79; 88; 75; 84
  - LR coefficient: 0.217; -0.089; 0.404; -0.192
  - p-value(H0: LR = 0): 0.164; 0.662; 0.023; 0.721
  - p-value(H0: LR = -1): 0.000; 0.000; 0.000; 0.133
  - In instruments: 43; 46; 39; 39
  - GMM lag structure and AR, Sargan p-values reported per column:
    - AR(1) p-value: 0.128; 0.748; 0.076; 0.197
    - AR(2) p-value: 0.022; 0.079; 0.309; 0.416
    - AR(3) p-value: 0.452; 0.155
    - Sargan p-value: 0.258; 0.218; 0.216; 0.515
- Interpretation: results point to more than full crowding out for public employment (significantly so in column (3)) and (insignificantly) less than full crowding out for public administration employment.

---

### APPENDIX II. REGRESSIONS OF PRIVATE-SECTOR EMPLOYMENT ON PUBLIC EMPLOYMENT

### FE with AR(1)
- Table A.5 reports FE AR(1) results for private-sector employment (PrvEmpRt).
- Main note: "The crowding out effects reported here are lower than elsewhere in the paper."
- Selected coefficients (columns (1)–(4)):
  - Public-Sector Employment Rate: -0.343**; -0.930*** (std. errors (0.142) (0.125))
  - Public Administration Employment Rate: -0.243; -0.175 (std. errors (0.254) (0.354))
  - GDP Growth Rate: 0.187; 0.382*** (std. errors (0.115) (0.134))
  - Urbanization Rate: -0.330; -0.264 (std. errors (0.223) (0.313))
  - Hiring and Firing Regulations: 0.118; -0.637 (std. errors (0.509) (0.480))
  - Openness: -0.001; 0.012 (std. errors (0.000) (0.022))
  - Observations: 286; 202; 315; 221
  - R-squared: 0.134; 0.473; 0.291; 0.112
  - Number of countries: 91; 71; 105; 82
  - p-value(H0: b=-1): 0.000; 0.575; 0.003; 0.021
  - Autocorr Test: 0.000; 0.000; 0.000; 0.000

### RE and RE with AR(1)
- Table A.6 (RE) selected coefficients (columns (1)–(4)):
  - Public-Sector Employment Rate: -0.896***; -0.978*** (std. errors (0.063) (0.048))
  - Public Administration Employment Rate: -0.971***; -0.897*** (std. errors (0.109) (0.150))
  - GDP Growth Rate: -0.084; -0.000 (std. errors (0.108) (0.150))
  - Openness: 0.000***; 0.027 (std. errors (0.000) (0.018))
  - Observations: 396; 282; 454; 315
  - Number of countries: 110; 80; 139; 94
  - Hausman Test: 0.010; 0.740; 0.952; 0.070
  - p-value(H0: b=-1): 0.098; 0.646; 0.787; 0.493
- Table A.7 (RE with AR(1)) selected coefficients (columns (1)–(4)):
  - Public-Sector Employment Rate: -0.839***; -0.959*** (std. errors (0.061) (0.069))
  - Public Administration Employment Rate: -0.874***; -0.778*** (std. errors (0.108) (0.181))
  - Observations: 396; 282; 454; 315
  - Number of countries: 110; 80; 139; 94
  - Hausman Test: 0.000; 0.000; 0.000; 0.164
  - p-value(H0: b=-1): 0.008; 0.551; 0.245; 0.221
  - Autocorr Test: 0.000; 0.000; 0.000; 0.000

### Systems GMM for private-sector employment (Table A.8)
- Table A.8 reports Systems GMM estimation for PrvEmpRt; short run coefficients indicate substantially less than full crowding out in some specifications, while long run estimates are often insignificantly different from -1 (consistent with full crowding out).
- Example short run interpretation: column (2) indicates a 1 percentage point rise in public administration employment would reduce private employment by 1/2 percentage point in that period.
- Selected estimates (columns (1)–(4)):
  - Private-Employment Rate (-1): 0.406*; 0.386; 0.415; 0.427*** (std. errors (0.23)(0.28)(1.82)(0.02))
  - Private-Employment Rate (-2): -0.048; 0.031 (std. errors (0.05)(0.03))
  - Public-Sector Employment Rate: -0.463; -0.599 (std. errors (0.30)(3.27))
  - Public Administration Employment Rate: -0.506; -0.619*** (std. errors (0.32)(0.07))
  - GDP Growth Rate: 0.185; 0.059; 0.21; 0.710*** (std. errors (0.33)(0.19)(0.51)(0.08))
  - Openness: 0; 0.027; 0; 0.030*** (reported formatting includes 0.00(0.05) 0.00(0.01))
  - Urbanization Rate: 0.056; 0.01; -0.057; -0.076** (std. errors (0.10)(0.07)(0.19)(0.03))
  - Hiring and Firing Regulations: -0.009; -0.439; 0.04; -2.145*** (std. errors (0.47)(0.37)(4.99)(0.18))
  - Observations: 223; 244; 158; 171
  - Number of countries: 73; 83; 60; 71
  - LR coefficient: -0.78; -0.825; -0.945; -1.143
  - p-value(H0: LR = 0): 0.001; 0.003; 0.698; 0.000
  - p-value(H0: LR = -1): 0.337; 0.522; 0.982; 0.306
  - Instruments: 32; 32; 45; 45
  - GMM lag structure and AR, Sargan p-values reported per column:
    - AR(1) p-value: 0.285; 0.405; 0.801; 0.171
    - AR(2) p-value: 0.037; 0.041; 0.182; 0.238
    - AR(3) p-value: 0.457; 0.19
    - Sargan p-value: 0.349; 0.155; 0.326; 0.168
- All Systems GMM regressions contain a constant term and are estimated with period-specific effects using systems GMM with instruments for difference and levels equations.

---

### APPENDIX III. REGRESSION WITH MCD INTERACTION TERMS AND CONTROLS

### Unemployment equation with MCD region interaction terms (Table A.9)
- Main note: "Our main results with the MCD region interaction terms in Table do not use control variables due to the loss of observations. This section presents both FE and RE estimation results with control variables."
- All estimated coefficients of the unemployment equation reported in Table A.9.
- Summary findings:
  - Columns (1)–(2) use RE estimation without control variables. For the MENAP countries, crowding out is slightly weaker than with FE in Table 4, but there is more than full crowding out for the CCA.
  - Last four columns include control variables. MENA region have more than full crowding out for public sector employment but mixed results for public administration employment.
  - Standard errors on coefficients of CCA dummies increase substantially when adding control variables, making additional effect of CCA countries statistically insignificant.
- Selected coefficients and test statistics (columns (1)–(6)):
  - Public-Sector Employment Rate: -0.056; -0.112; -0.080 (std. errors (0.040)(0.075)(0.062))
  - Public Employment X MENAP: 0.066; 0.237**; 0.189** (std. errors (0.071)(0.099)(0.083))
  - Public Employment X CCA: 0.249***; 1.366; -0.453 (std. errors (0.065)(1.204)(0.500))
  - Public Administration Employment Rate: -0.042; -0.192; -0.024 (std. errors (0.066)(0.213)(0.168))
  - Public Administration Employment X MENAP: -0.186; 0.013; -0.178 (std. errors (0.117)(0.235)(0.184))
  - Public Administration Employment X CCA: 0.206***; 1.273; -0.426 (std. errors (0.050)(0.994)(0.476))
  - GDP Growth Rate: -0.087; -0.176**; -0.071; -0.178* (std. errors (0.078)(0.081)(0.086)(0.093))
  - Hiring and Firing Regulations: -0.454; -0.152; -0.597**; -0.494* (std. errors (0.278)(0.351)(0.248)(0.286))
  - Openness: -0.000***; 0.017; -0.000***; 0.002 (std. errors (0.000)(0.013)(0.000)(0.013))
  - Observations: 419; 287; 292; 224; 292; 224
  - Number of countries: 116; 88; 82; 73; 82; 73
  - p-value(H0: b+bM=0): 0.861; 0.019; 0.001; 0.005; 0.002; 0.017
  - p-value(H0: b+bM=-1): 0.000; 0.000; 0.000; 0.000; 0.000; 0.000
  - p-value(H0: b+bC=0): 0.000; 0.048; 0.300; 0.294; 0.278; 0.334
  - p-value(H0: b+bC=-1): 0.000; 0.000; 0.064; 0.045; 0.343; 0.239
  - Hausman Test: 0.014; 1.000; 1.000; 1.000
- Robust standard errors in parentheses; significance: *** p<0.01, ** p<0.05, * p<0.1.
- All regressions contain a constant term and are estimated with country-specific, region-specific and period-specific fixed effects.

### Private-sector employment equation with MCD interaction terms (Table A.10)
- Employment equation results reported in Table A.10.
- Summary findings:
  - For MENAP countries, crowding out effects are larger than in the rest of the world; results point either to full crowding out or more than full crowding out.
  - For CCA, some evidence of less than full crowding out in column (2). Other CCA results less reliable due to loss of observations and country coverage; coefficients imply crowding in effects in some specifications.
- Selected coefficients and test statistics (columns (1)–(6)):
  - Public-Sector Employment Rate: -0.891***; -0.963***; -0.980*** (std. errors (0.074)(0.077)(0.070))
  - Public Employment X MENAP: -0.213; -0.070; -0.025 (std. errors (0.135)(0.117)(0.102))
  - Public Employment X CCA: 0.125; 4.944***; 3.761*** (std. errors (0.097)(1.338)(1.361))
  - Public-Administration Employment Rate: -0.897***; -0.843***; -0.778*** (std. errors (0.114)(0.284)(0.284))
  - Public-Administration Employment X MENAP: 0.079; -0.413; -0.449 (std. errors (0.212)(0.317)(0.297))
  - Public-Administration Employment X CCA: 0.315***; 5.165***; 3.514*** (std. errors (0.075)(1.394)(1.284))
  - GDP Growth Rate: -0.027; 0.007; -0.078; -0.092 (std. errors (0.091)(0.110)(0.092)(0.125))
  - Urbanization Rate: -0.248*; -0.148; -0.020; -0.050 (std. errors (0.133)(0.225)(0.090)(0.087))
  - Hiring and Firing Regulations: 0.279; 0.073; 0.420; 0.345 (std. errors (0.360)(0.408)(0.362)(0.385))
  - Openness: 0.000***; 0.006; 0.000***; 0.014 (std. errors (0.000)(0.021)(0.000)(0.017))
  - Observations: 396; 293; 282; 225; 282; 225
  - Number of countries: 110; 89; 80; 73; 80; 73
  - R-squared: 0.670; 0.233
  - p-value(H0: b+bM=0): 0.000; 0.000; 0.000; 0.000; 0.000; 0.000
  - p-value(H0: b+bM=-1): 0.288; 0.320; 0.532; 0.001; 0.919; 0.009
  - p-value(H0: b+bC=0): 0.000; 0.000; 0.004; 0.003; 0.040; 0.033
  - p-value(H0: b+bC=-1): 0.011; 0.003; 0.000; 0.000; 0.005; 0.004
  - Hausman Test: 0.024; 0.001; 1.000; 1.000
- Robust standard errors in parentheses; significance: *** p<0.01, ** p<0.05, * p<0.1.
- All regressions contain a constant term and are estimated with country-specific, region-specific and period-specific fixed effects.

*Source: _wp13146 - APPENDIX I. REGRESSIONS OF UNEMPLOYMENT RATE ON PUBLIC EMPLOYMENT*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13146.pdf_