## WHAT DRIVES WAGE GROWTH IN POLAND?

## Source details

**Canonical URL:** [WHAT DRIVES WAGE GROWTH IN POLAND?](https://www.imf.org/-/media/files/publications/cr/2019/cr1938.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2019/cr1938.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2019/cr1938.pdf.json)

---

### Labor market overview and recent developments
- Nominal wages: about 4 percent (y/y) during 2010–16; accelerated since mid-2017, reaching 7.6 percent in Q3:2018.
- Real wages: increased from about 2½ percent during 2010-16 to 5½ percent in Q3:2018.
- Labor productivity (adjusted for changes in output prices): about 4½ percent during 2010–16; accelerated to 6½ percent in 2018.
- Real unit labor costs (RULCs): increased moderately despite wage acceleration.
- Unemployment rate: fell by more than 1 percentage point per year since 2014, reaching 3.8 percent in Q3:2018.
- Foreign workers (FWs) estimated stock: above 1 million in effective full-time equivalents (around 5 percent of total employment, compared with less than 1 percent in 2014); vast majority from Ukraine.
- Employment measurement: official LFS employment growth 1.4 percent (y/y) in 2017; staff estimate adjusting for FWs raises employment growth to 2.8 percent, implying adjusted real labor productivity growth declines by 1.7 percentage points to 1.8 percent.

### Empirical methodology
- Model specification: error-correction model (ECM):
  - dlogW_t = α + β_d * D_t-i + β_f * F_t-i + β_fw * FW_t + δ*(φ_1 logRW_t-i - φ_2 logTLP_t-i) + ε_t
- Dependent variable:
  - dlogW_t = change in nominal wages (total compensation of employees per hour worked).
- Error correction term:
  - δ*(φ_1 logRW_t-4 - φ_2 logTLP_t-4) captures long-run relationship between real wages and trend labor productivity.
- Short-run determinants included:
  - Domestic factors (D): inflation expectations; FW-adjusted changes in labor productivity; unemployment rate gap (HP-filtered); underemployment indicators (change in share of involuntary part-time employment, temporary employment and self-employed).
  - Foreign factors (F): labor market conditions in the euro area and Ukraine.
  - Foreign worker-related variables (FW): share of FWs in total employment; change in FW share.

### Long-run relationship between real wages and trend labor productivity
- Main finding:
  - Real wages track trend labor productivity; estimated coefficient significant and close to unity.
- Residuals:
  - Real wages exceeded fitted long-run level beginning in mid-2016, coinciding with an acceleration in labor productivity that is captured in trend productivity with a long delay.

### Short-run drivers of nominal wage growth (empirical results)
- Primary determinants: domestic factors and the error correction term.
- Unemployment rate gap: coefficient negative and significant across specifications—confirming inverse relationship between unemployment and wage growth.
- Labor productivity growth and error correction item: significant in all specifications.
- Expected inflation: not significant in any specification (robust to alternative measures including NBP and ECB surveys and actual CPI).
- Foreign spillovers: labor market indicators in the euro area and Ukraine insignificant.

### Role of foreign workers (FWs)
- Theoretical channels:
  - Substitution: FWs with similar skills can substitute for Polish workers, reducing bargaining power and dampening wages.
  - Complementarity and demand expansion: FWs in sectors with shortages can free Polish workers to reallocate to more productive tasks, support higher wages, and shift the production frontier outward.
- Empirical findings:
  - Significantly negative coefficients on the FW share and the change in the FW share imply a dampening effect on Polish wages.
  - Significantly positive coefficient on the interaction term implies FWs can support Polish wages.
  - Net impact depends on level of FW share: larger support when FW share is higher.
- Illustrative simulation:
  - From an initial FW share of 10 percent, a one percentage point increase in the FW share raises wages of Polish workers by 3.8 percent.
  - When FW share is small, the dampening effect tends to dominate.
- Current assessment:
  - The FW share around 5 percent is close to the neutral point where FWs have a neutral effect on wage growth; recent rises likely produced a net dampening effect.
- Caveats:
  - Results preliminary due to short time series and measurement issues with FW data.

### Decomposition of recent wage acceleration (2016–17)
- Main contributors to nominal wage acceleration relative to long-run average:
  - Error correction (correction of earlier “wage undershoot” from 2013–15).
  - Labor scarcity (tight labor market).
  - Offsetting factors: sluggish labor productivity (after adjusting for FWs) and net dampening effect of FWs.
- Outlook:
  - With real wages currently above the level consistent with trend productivity, correction of the overshoot would be expected to dampen future wage growth.
  - Actual future dynamics will depend on labor market tightness, FW share, and labor productivity dynamics.

### Appendix — Key variables (selected entries and summary statistics)
- POL_wage gr: Obs 58, Mean 3.73, Std. Dev. 3.11, Min -1.16, Max 12.66.
- UE_GAP: Obs 82, Mean 0.00, Std. Dev. 1.28, Min -3.12, Max 2.10.
- CPI_exp1 year ahead expected inflation: Obs 66, Mean 2.65, Std. Dev. 0.95, Min 1.03, Max 5.96.
- LPgr: Obs 58, Mean 2.81, Std. Dev. 1.84, Min -0.86, Max 7.27.
- LPgr_migrant: Obs 58, Mean 2.34, Std. Dev. 2.13, Min -1.56, Max 7.16.
- Migrant_total: Obs 80, Mean 0.76, Std. Dev. 1.33, Min 0.00, Max 5.53.
- EA_wagegr: Obs 78, Mean 2.36, Std. Dev. 0.84, Min 0.05, Max 3.83.

*Prepared by Krzysztof Krogulski and Xin Cindy Xu; additional contributions for TFP section by Federico Joaquin Díez, Yevgeniya Korniyenko, Krzysztof Krogulski, and Robert Sierhej.*

---

### Appendix I — Measuring and Drivers of Firm-Level TFP

### A. Firm-Level TFP: Statistics Poland Data — sample and aggregate outcomes
- Sample period: 2005–16.
- Average sample size: more than 48 thousand nonfinancial firms per year.
- Employment covered: 4.6 million employees (annual).
- Annual sales in sample: PLN1.4 trillion.
- GVA grew by a cumulative 56 percent during 2005–16, with TFP accounting for more than a third of the increase.
- Sample-wide cumulative increase in TFP during 2005–16: just above 20 percentage points.
- Sectoral contributors:
  - Manufacturing accounts for the majority of the TFP increase; trade and construction also contributed positively; mining and utilities recorded negative productivity dynamics.
- Post-GFC TFP growth remained mostly positive but appreciably lower than the pre-crisis average.

### A. Firm-Level TFP: Statistics Poland Data — distributional patterns and firm attributes
- TFP distribution is highly skewed to the left; bulk of firms are in the lower half of the TFP distribution.
- Evidence of gradual convergence: diminishing inter-quartile distance and higher productivity growth in sectors starting at low TFP levels.
- Long right tail of TFP distribution indicates ample catch-up potential for less-productive companies.
- Firm size and ownership:
  - Very-large firms (more than 250 employees) were the most productive: their TFP level was 50 percent higher than for other firms in 2016.
  - The TFP gap between very-large and other firms halved during the sample period.
  - Foreign-owned firms have substantially higher TFP than SOEs or domestic private firms.
  - Initial TFP levels of foreign-owned firms and SOEs were roughly equal; a decade later foreign-owned firms were 33 percent more productive.
  - Domestic private firms were less productive than SOEs initially but caught up rapidly.
- Export status:
  - Export-oriented firms have much-higher TFP than firms serving only the domestic market and suffered less during the GFC.
- Manufacturing specifics:
  - Productivity convergence within manufacturing: divisions with highest TFP in 2005 saw more subdued gains.
  - Fastest TFP increases in high-tech manufacturing: production of computers and electronics, and electrical equipment.
  - Very-large manufacturing firms are both most productive and have strongest TFP growth.
  - In manufacturing, SOEs posted the most impressive TFP gains (contrary to whole-sample pattern).

### B. Firm-Level TFP: ORBIS Data — coverage and structural patterns
- Orbis data used: financials covering 2000–15 and ownership covering 2006–15.
- Orbis sample covers around 20 percent of total employment and about 40 percent of operational turnover revenue in the Polish economy.
- Ownership classification rules:
  - SOE if Polish government has a direct or ultimate stake of 25 percent or more.
  - Foreign firms if single foreign ultimate/direct owner of 10 percent or more.
  - Domestic private firms are all others.
- Key descriptive statistics (Poland, Orbis):
  - TFP level means:
    - Foreign Firms: 4.3423 (N obs 36,706).
    - SOEs: 3.4655 (N obs 9,177).
    - Domestic Private Firms: 4.1161 (N obs 252,963).
  - TFP growth means:
    - Foreign Firms: -0.0030 (N obs 27,961).
    - SOEs: 0.0050 (N obs 7,228).
    - Domestic Private Firms: -0.0188 (N obs 176,796).
- Structural patterns:
  - SOEs account for 3 percent of firms in the Orbis sample but about 20 percent share in total assets.
  - SMEs have lower TFP levels than large firms; largest firms by assets are the most productive.
  - SOEs show a bi-modal TFP distribution: most mass at the low-end but with a long right tail.
  - Heterogeneity across sectors: sectors with high SOE concentration see coexistence of both high and low TFP enterprises.

### B. Firm-Level TFP: ORBIS Data — dynamics, frontier, and allocative efficiency
- Firm age and intangibles:
  - Older firms (>5 years) have higher TFP levels except around the GFC; younger firms (<5 years) exhibit faster TFP growth.
  - About 25 percent of firms registered any investment in intangible capital.
  - Firms investing in intangible capital have higher TFP and faster TFP growth than firms without such investments.
- Frontier dynamics:
  - Global frontier measured as average log TFP for top 5 percent of companies within each 2-digit industry and year; laggards capture average log productivity of all other firms.
  - Only 10 percent of firms identified as “frontier” in 2007 remained “frontier” in 2014.
  - Frontier firms’ TFP grew significantly; laggard firms were significantly impacted by the GFC with recovery visible only in the last few years of the sample.
  - In manufacturing and services, laggard firms’ TFP was adversely impacted by the GFC and has failed to recover (laggard manufacturing firms impacted significantly more).
  - Frontier firms in services saw a TFP increase about twice as large as frontier firms in manufacturing over the analyzed period.
- Allocative efficiency and cross-country comparison:
  - Following the GFC, resource allocation became more efficient: dispersion of TFP across firms (ratio of 75th to 25th percentiles) has been trending down since 2009.
  - Poland’s inter-quartile ratio (P75/P25) of TFP levels in 2015: 1.18, change 2006–15: -0.02.
  - Cross-country inter-quartile ratios (Level 2015; Change 2006–15): Czech R. 1.27 -0.01; Finland 1.22 0.01; France 1.19 0.01; Hungary 1.26 0.06; Italy 1.16 0.01; Spain 1.21 0.02; Slovakia 1.22 0.03; Slovenia 1.21 0.04; S. Korea 1.15 0.00.
  - Poland’s inter-quartile ratio in 2015 is below that of CEE4 peers and broadly in line with advanced countries in the analysis; Poland saw the largest decrease (narrowing) in TFP dispersion during 2006–15 among comparators.

### B. Empirical findings on structural drivers (Orbis regressions)
- Ownership effects:
  - Foreign-owned firms: associated with above-average TFP levels and growth rates (consistent across countries).
  - SOEs: associated with lower-than-average TFP levels and TFP growth rates (results hold across and within firms and across comparator countries).
- Within- and cross-firm regression highlights (selected coefficients):
  - TFP drivers — cross-sectional regression (TFP level):
    - Foreign coefficient (Poland): 0.094***.
    - SOEs coefficient (Poland): -0.421***.
    - N obs: 182,425.
  - TFP drivers — within-firm regression (TFP level):
    - Foreign coefficients: 0.019*** (Poland).
    - SOEs coefficients: -0.064*** (Poland).
    - N obs: 289,597.
  - TFP growth drivers — cross-sectional regression (TFP growth):
    - Foreign coefficients: 0.032*** (Poland).
    - SOEs coefficients: -0.059*** (Poland).
    - Lagged TFP: -0.178*** (Poland).
    - N obs: 182,425.
  - TFP growth drivers — within-firm regression (TFP growth):
    - Foreign coefficients: 0.006 (Poland; not significant).
    - SOEs coefficients: -0.051** (Poland).
    - Lagged TFP: -0.692*** (Poland).
    - N obs: 170,053.
- TFP growth and policy variables (by ownership, selected coefficients):
  - Foreign Firms (N obs 75,624): Regulation -0.006***; Lagged TFP -0.047***; Size 0.007***; Leverage 0.023***; Intangibles 0.070***; R2 0.059.
  - SOEs (N obs 12,011): Regulation -0.003 (not significant); Lagged TFP -0.023***; Size 0.004***; Leverage 0.037 (not significant); Intangibles 0.040***; R2 0.05.
  - Domestic Private Firms (N obs 1,580,189): Regulation -0.008***; Lagged TFP -0.099***; Size 0.015***; Leverage 0.039***; Intangibles 0.085***; R2 0.069.

### Quantitative counterfactual and robustness
- Back-of-the-envelope calculation:
  - If all SOEs were instead privately owned, the resulting aggregate TFP level would be, all else equal, almost nine percent higher (based on estimated coefficient on SOE dummy and accounting for number and size of SOEs).
- Robustness and estimation methods:
  - TFP estimated using Cobb-Douglas production function and control function methods (Olley-Pakes; Levinsohn-Petrin).
  - Baseline analyses used LP model with proxy = raw materials and energy on full sample.
  - Data cleaning, winsorization, and alternative estimation approaches applied to address endogeneity and selection.

### Conclusions and policy implications
- Ownership structure matters:
  - Foreign-owned firms are associated with strong TFP performance via above-average TFP levels and growth rates.
  - Prevalence of SOEs is found to be a drag on aggregate TFP outcomes, though heterogeneity exists with some SOEs performing as well as private firms.
- Policy recommendations:
  - Continue structural reforms to boost TFP growth:
    - Create an environment conducive to entrepreneurship by reducing barriers to entry.
    - Ensure a level playing field between state-owned and private firms.
    - Avoid barriers to scaling up businesses.
    - Encourage investments in innovation and R&D (investment in intangibles).
  - Monitor and mitigate macroeconomic costs associated with prevalence of SOEs and decreases in participation by foreign firms (e.g., lower FDI since the GFC).
  - Sustain or increase efforts toward structural reforms to boost TFP growth and sustain Poland’s future growth performance among regional and global peers.

*Source: Appendix I.*

### References ________________________________________________________________________________ 9

### WHAT DRIVES WAGE GROWTH IN POLAND?

### Labor market overview and recent developments
- Nominal wages: about 4 percent (y/y) during 2010–16; accelerated since mid-2017, reaching 7.6 percent in Q3:2018.
- Real wages: increased from about 2½ percent during 2010-16 to 5½ percent in Q3:2018.
- Labor productivity (adjusted for changes in output prices): about 4½ percent during 2010–16; accelerated to 6½ percent in 2018.
- Real unit labor costs (RULCs): increased moderately despite wage acceleration.
- Unemployment rate: fell by more than 1 percentage point per year since 2014, reaching 3.8 percent in Q3:2018.
- Foreign workers (FWs) estimated stock: above 1 million in effective full-time equivalents (around 5 percent of total employment, compared with less than 1 percent in 2014); vast majority from Ukraine.
- Employment measurement: official LFS employment growth 1.4 percent (y/y) in 2017; staff estimate adjusting for FWs raises employment growth to 2.8 percent, implying adjusted real labor productivity growth declines by 1.7 percentage points to 1.8 percent.

### Empirical methodology
- Model: an error-correction model (ECM) specified as:
  dlogW_t = α + β_d * D_t-i + β_f * F_t-i + β_fw * FW_t + δ*(φ_1 logRW_t-i - φ_2 logTLP_t-i) + ε_t
- Dependent variable: dlogW_t = change in nominal wages (total compensation of employees per hour worked).
- Error correction term: δ*(φ_1 logRW_t-4 - φ_2 logTLP_t-4) captures long-run relationship between real wages and trend labor productivity.
- Short-run determinants:
  - Domestic factors (D): inflation expectations; FW-adjusted changes in labor productivity; unemployment rate gap (HP-filtered); underemployment indicators (change in share of involuntary part-time employment, temporary employment and self-employed).
  - Foreign factors (F): labor market conditions in the euro area and Ukraine.
  - Foreign worker-related variables (FW): share of FWs in total employment; change in FW share.

### Long-run relationship between real wages and trend labor productivity
- Long-run finding: real wages track trend labor productivity; estimated coefficient significant and close to unity.
- Residuals: real wages exceeded fitted long-run level beginning in mid-2016, coinciding with an acceleration in labor productivity that is captured in trend productivity with a long delay.

### Short-run drivers of nominal wage growth (empirical results)
- Primary determinants: domestic factors and the error correction term.
- Unemployment rate gap: coefficient negative and significant across specifications—confirming inverse relationship between unemployment and wage growth.
- Labor productivity growth and error correction item: significant in all specifications.
- Expected inflation: not significant in any specification (robust to alternative measures including NBP and ECB surveys and actual CPI).
- Foreign spillovers: labor market indicators in the euro area and Ukraine insignificant.

### Role of foreign workers (FWs)
- Two potential channels:
  - Substitution: FWs with similar skills can substitute for Polish workers, reducing bargaining power and dampening wages.
  - Complementarity and demand expansion: FWs in sectors with shortages can free Polish workers to reallocate to more productive tasks, support higher wages, and shift the production frontier outward.
- Empirical findings:
  - Significantly negative coefficients on the FW share and the change in the FW share imply a dampening effect on Polish wages.
  - Significantly positive coefficient on the interaction term implies FWs can support Polish wages.
  - Net impact depends on level of FW share: larger support when FW share is higher.
- Illustrative simulation: from an initial FW share of 10 percent, a one percentage point increase in the FW share raises wages of Polish workers by 3.8 percent; when FW share is small, the dampening effect tends to dominate.
- Current assessment: the FW share around 5 percent is close to the neutral point where FWs have a neutral effect on wage growth; recent rises likely produced a net dampening effect.
- Caveats: results preliminary due to short time series and measurement issues with FW data.

### Decomposition of recent wage acceleration (2016–17)
- Main contributors to nominal wage acceleration relative to long-run average:
  - Error correction (correction of earlier “wage undershoot” from 2013–15).
  - Labor scarcity (tight labor market).
  - Offsetting factors: sluggish labor productivity (after adjusting for FWs) and net dampening effect of FWs.
- Outlook: with real wages currently above the level consistent with trend productivity, correction of the overshoot would be expected to dampen future wage growth. Actual future dynamics will depend on labor market tightness, FW share, and labor productivity dynamics.

### Appendix — Key variables (selected entries and summary statistics)
- POL_wage gr: YoY growth rate of total labor compensation per hour worked in Poland — Obs 58, Mean 3.73, Std. Dev. 3.11, Min -1.16, Max 12.66.
- UE_GAP: HP-filtered gap of headline unemployment rate — Obs 82, Mean 0.00, Std. Dev. 1.28, Min -3.12, Max 2.10.
- CPI_exp1 year ahead expected inflation — Obs 66, Mean 2.65, Std. Dev. 0.95, Min 1.03, Max 5.96.
- LPgr: Growth rate of real labor productivity per hour — Obs 58, Mean 2.81, Std. Dev. 1.84, Min -0.86, Max 7.27.
- LPgr_migrant: Growth rate of real labor productivity per hour adjusted by migrant workers — Obs 58, Mean 2.34, Std. Dev. 2.13, Min -1.56, Max 7.16.
- Migrant_total: The share of temporary migrant workers (statement procedure) in total employment — Obs 80, Mean 0.76, Std. Dev. 1.33, Min 0.00, Max 5.53.
- EA_wagegr: YoY growth rate of total labor compensation per hour worked in EA — Obs 78, Mean 2.36, Std. Dev. 0.84, Min 0.05, Max 3.83.

---

*Prepared by Krzysztof Krogulski and Xin Cindy Xu; additional contributions for TFP section by Federico Joaquin Díez, Yevgeniya Korniyenko, Krzysztof Krogulski, and Robert Sierhej. References include IMF, 2017; IMF, 2018a; IMF, 2018b.*

### Appendix I.

### Appendix I

### A. Firm-Level TFP: Statistics Poland Data
- Sample and scope
  - Sample period: 2005–16.
  - Average sample size: more than 48 thousand nonfinancial firms per year.
  - Employment covered: 4.6 million employees (annual).
  - Annual sales in sample: PLN1.4 trillion.
  - Manufacturing and wholesale and retail trade are the two largest sectors by GVA.
  - Very-large and large firms dominate the sample in all sectors.
  - SOEs dominant in mining and energy and water supply (utilities); foreign-owned firms dominant in manufacturing, trade, and transport and ICT.

- Aggregate outcomes and contributions
  - GVA grew by a cumulative 56 percent during 2005–16, with TFP accounting for more than a third of the increase.
  - Sample-wide cumulative increase in TFP during 2005–16: just above 20 percentage points.
  - Manufacturing accounts for the majority of the TFP increase; trade and construction also contributed positively; mining and utilities recorded negative productivity dynamics.
  - Post-GFC TFP growth remained mostly positive but appreciably lower than the pre-crisis average.

- Distributional patterns and dynamics
  - TFP distribution is highly skewed to the left; the bulk of firms are in the lower half of the TFP distribution.
  - Evidence of gradual convergence: diminishing inter-quartile distance and higher productivity growth in sectors starting at low TFP levels.
  - Long right tail of TFP distribution indicates ample catch-up potential for less-productive companies.

- Firm size, ownership, and export status
  - Very-large firms (more than 250 employees) were the most productive: their TFP level was 50 percent higher than for other firms in 2016.
  - The TFP gap between very-large and other firms halved during the sample period.
  - Foreign-owned firms have substantially higher TFP than SOEs or domestic private firms.
  - Initial TFP levels of foreign-owned firms and SOEs were roughly equal; a decade later foreign-owned firms were 33 percent more productive.
  - Domestic private firms were less productive than SOEs initially but caught up rapidly.
  - Export-oriented firms have much-higher TFP than firms serving only the domestic market and suffered less during the GFC.

- Manufacturing specific findings
  - Productivity convergence within manufacturing: divisions with highest TFP in 2005 saw more subdued gains.
  - Fastest TFP increases in high-tech manufacturing: production of computers and electronics, and electrical equipment.
  - Very-large manufacturing firms are both most productive and have strongest TFP growth.
  - In manufacturing, SOEs posted the most impressive TFP gains (contrary to whole-sample pattern).

### B. Firm-Level TFP: ORBIS Data
- Data coverage and classification
  - Orbis financials covering 2000–15 and Orbis ownership covering 2006–15 used.
  - Orbis sample covers around 20 percent of total employment and about 40 percent of operational turnover revenue in the Polish economy (as reported by Eurostat and OECD).
  - Analysis includes firms in CEE4 (Czech Republic, Hungary, Slovak Republic, Slovenia), Italy, and Spain for comparison.
  - Ownership classification rules:
    - Firm is classified as owned/controlled by the State if the Polish government has a direct or ultimate stake of 25 percent or more.
    - Foreign firms are firms with single foreign ultimate/direct owner of 10 percent or more.
    - Domestic private firms are all others.

- Structural patterns
  - SOEs account for 3 percent of firms in the Orbis sample but about 20 percent share in total assets.
  - SMEs have lower TFP levels than large firms; largest firms by assets are the most productive.
  - TFP distribution by ownership:
    - Private firms’ (foreign and domestic) TFP distribution approximates normality.
    - SOEs show a bi-modal TFP distribution: one hump somewhat below private firms’ modes and a second hump at the low-end; most mass of SOEs distribution is at the low-end but with a long right tail.
  - Heterogeneity across sectors: sectors with high SOE concentration see coexistence of both high and low TFP enterprises, suggesting economic distortions.

- Firm age, intangibles, and frontier dynamics
  - Older firms (>5 years) have higher TFP levels except around the GFC; younger firms (<5 years) exhibit faster TFP growth.
  - About 25 percent of firms registered any investment in intangible capital.
  - Firms investing in intangible capital have higher TFP and faster TFP growth than firms without such investments.
  - Frontier vs laggard firms:
    - Global frontier measured as average log TFP for top 5 percent of companies within each 2-digit industry and year; laggards capture average log productivity of all other firms.
    - Only 10 percent of firms identified as “frontier” in 2007 remained “frontier” in 2014.
    - Frontier firms’ TFP grew significantly; laggard firms were significantly impacted by the GFC with recovery visible only in the last few years of the sample.
    - In manufacturing and services, laggard firms’ TFP was adversely impacted by the GFC and has failed to recover (laggard manufacturing firms impacted significantly more).
    - Frontier firms in services saw a TFP increase about twice as large as frontier firms in manufacturing over the analyzed period.

- Allocative efficiency and cross-country comparison
  - Following the GFC, resource allocation became more efficient: dispersion of TFP across firms (ratio of 75th to 25th percentiles) has been trending down since 2009.
  - Poland’s inter-quartile ratio (P75/P25) of TFP levels in 2015: 1.18, change 2006–15: -0.02.
  - Cross-country inter-quartile ratios (Level 2015; Change 2006–15):
    - Czech R. 1.27 -0.01
    - Finland 1.22 0.01
    - France 1.19 0.01
    - Hungary 1.26 0.06
    - Italy 1.16 0.01
    - Spain 1.21 0.02
    - Slovakia 1.22 0.03
    - Slovenia 1.21 0.04
    - S. Korea 1.15 0.00
  - Poland’s inter-quartile ratio in 2015 is below that of CEE4 peers and broadly in line with advanced countries in the analysis; Poland saw the largest decrease (narrowing) in TFP dispersion during 2006–15 among comparators.

- Empirical findings on structural drivers
  - SOEs are associated with lower-than-average TFP levels and TFP growth rates (results hold across and within firms and across comparator countries).
  - Foreign-owned firms have above-average TFP levels and growth rates (consistent across countries).
  - Ownership changes are associated with TFP differences when comparing the same firm before and after ownership change.
  - Domestic institutions matter: relaxing product market regulations has a positive and statistically significant effect on TFP growth for foreign and domestic private firms; restrictiveness of regulation is less important for SOEs.
  - Younger firms grow TFP faster than older firms, implying aggregate TFP could be raised by easing barriers to entry and scaling-up.
  - Investments in intangibles (proxy for R&D) are key for TFP growth; result holds across and within firms and across comparator countries.

- Quantitative counterfactual
  - Back-of-the-envelope calculation: if all SOEs were instead privately owned, the resulting aggregate TFP level would be, all else equal, almost nine percent higher (based on estimated coefficient on SOE dummy, accounting for number and size of SOEs).

### C. Conclusions and Implications
- Ownership structure matters
  - Foreign-owned firms are associated with strong TFP performance via above-average TFP levels and growth rates.
  - Prevalence of SOEs is found to be a drag on aggregate TFP outcomes, though heterogeneity exists with some SOEs performing as well as private firms.

- Policy implications and recommendations
  - Continue structural reforms to boost TFP growth:
    - Create an environment conducive to entrepreneurship by reducing barriers to entry.
    - Ensure a level playing field between state-owned and private firms.
    - Avoid barriers to scaling up businesses.
    - Encourage investments in innovation and R&D (investment in intangibles).
  - Monitor and mitigate macroeconomic costs associated with prevalence of SOEs and decreases in participation by foreign firms (e.g., lower FDI since the GFC).
  - Sustain or increase efforts toward structural reforms to boost TFP growth and sustain Poland’s future growth performance among regional and global peers.

*Source: Appendix I.*

### References

### References

### Key cited studies and working papers
- Andrews D., Criscuolo C. and P.N. Gal, 2016, “The Best Versus the Rest: The Global Productivity Slowdown, Divergence across Firms and the Role of Public Policy”, OECD Productivity Working Papers, No. 05.
- Budina, N., P. Deb, F. Díez, J. Fan, J. Luzi, F. Misch, A. Shabunina, 2018. “Drivers of Firm Productivity Growth and Misallocation,” Mimeo.
- Haltiwanger, J., R. Jarmin, and J. Miranda. 2013. “Who Creates Jobs? Small vs. Large vs. Young.” The Review of Economics and Statistics, Vol. 95, Issue 2, pp. 34–361.
- Hsieh, C., and P. J. Klenow, 2009, “Misallocation and Manufacturing TFP in China and India,” The Quarterly Journal of Economics, Vol. 124, No. 4.
- Kalemli-Ozcan, S., and others, 2016, “How to Construct Nationally Representative Firm-Level Data from the ORBIS Global Database,” NBER Working Paper, No. 21558.
- Levinsohn, J. and A. Petrin, 2003, “Estimating Production Functions Using Inputs to Control for Unobservables,” Review of Economic Studies, Vol. 70, No. 2.
- Olley, G. Steven and Ariel Pakes, 1996, “The Dynamics of Productivity in the Telecommunications Equipment Industry,” Econometrica, Vol. 64, No. 6, pp.:1263–1298.
- OECD, 2014, “Perspectives on Global Development 2014: Boosting Productivity to Meet the Middle-Income Challenge.”
- OECD, 2015, “The Future of Productivity.”
- Rovigatti, Gabriele, 2017, Production Function Estimation in R : The Prodest Package, Working Paper.
- Van Beveren, Ilke, 2012, “Total Factor Productivity Estimation: a Practical Review,” Journal of Economic Surveys, Vol 26, No. 1, pp. 98–128.
- IMF publications cited: 2016 Regional Economic Issues (May), 2017 Poland: Selected Issues, 2018 Regional Economic Outlook Chapter 3 (“Productivity Growth in Asia: Boosting Firm Dynamism and Weeding out the Zombies”).

*References compiled from the source content.*

### Appendix I — Measuring Total Factor Productivity (TFP) of Non-Financial Enterprises in Poland: Data and Methods
- Data source and scope:
  - Annual reports of non-financial firms employing more than 9 persons for 2005–16 (reported to Statistics Poland, GUS, on SP statistical form).
  - Final database: over 585 thousand statistical units, 67 percent of all observations registered in the SP reports for the years 2005–16.
  - Time-varying sample: on average more than 48 thousand firms per year, with 4.6 million employees and annual sales at PLN1.4 trillion.
  - Note: sampling was not based on a representative method.
- Data cleaning and preparation rules:
  - Removed firms with no positive net revenue from sales and firms with zero cost of materials and energy, external services and travel expenses.
  - Removed companies meeting at least one of: value of fixed assets at the beginning and end of the year was zero; number of full-time equivalent employees was zero; labor cost (wages and social contributions) was zero.
  - Missing values replaced with zeros for: business travel expenses, intangible assets, costs of production for own use, value of goods and materials sold, excise tax, value of semi-finished products and production in progress, stock of finished products.
  - Transition from PKD-2004 to PKD-2007 classification of activities in 2008.
- TFP definition and measurement:
  - Production function assumption: Cobb-Douglas: Y_it = A_it K_it^{β_k} L_it^{β_l}.
  - Y_it and K_it were deflated to 2010 prices.
  - Log-linear form used: y_it = ω_it + β_k k_it + β_l d_it + u_it.
  - TFP estimator: ω̂_it = β̂_0 + v̂_it = y_it − β̂_k k_it − β̂_l d_it.
  - Individual TFP: TFTP̂_it = e^{ω̂_it}.
  - Fixed capital defined as average annual level: K̄ = (TK_{12} + TK_0)/2.
  - Firm-level TFP winsorized by removing the top and bottom percentile from TFTP_it distribution.
  - Sector aggregation: TFTP_{S,t} = Σ_{i∈S} w_{i,t} TFTP_{i,t} with weights w_{i,t} = K_{i,t}^{β_k} L_{i,t}^{β_l} / Σ_{i∈S} K_{i,t}^{β_k} L_{i,t}^{β_l}.
- Estimation challenges and methods:
  - Issues: endogeneity, sample selection bias, omitted variables; classical OLS can bias β_l upward and β_k downward.
  - Robust methods referenced: instrumental variables, generalized method of moments, control function methods (Olley-Pakes 1996; Levinsohn-Petrin 2003).
  - OP model uses investment as proxy; LP model uses outlays on materials and energy as proxy.
  - Models estimated in a 3-step procedure with bootstrapped standard errors.
- Estimation approaches applied:
  - Pooled OLS (classical linear regression).
  - Panel regressions with fixed and random individual effects.
  - Control function methods including OP and LP models.
  - LP model re-estimated on 3-year rolling sample window to produce time-variant estimates of β_l and β_k.
  - Statistically significant linear trend in logarithm of gross value added confirmed.
  - Baseline analyses used LP model (proxy = raw materials and energy) identified on full sample, including all firms.

*Prepared by M. Błażej, M. Górajski, D. Kotlewski, A. Rynio (Statistics Poland). All estimates performed in R; R packages used include prodest, estprod, dplyr, plm (Rovigatti, 2017).*

### Appendix II — Orbis Data and Regression Results (selected tables and findings)
- Orbis database descriptive statistics (Poland):
  - TFP level (mean, p50, sd, min, max, N obs)
    - Foreign Firms: 4.3423, 4.4014, 0.8618, 1.9650, 6.3501, 36706
    - SOEs: 3.4655, 3.4848, 0.8347, 1.9652, 6.3435, 9177
    - Domestic Private Firms: 4.1161, 4.1412, 0.8099, 1.9651, 6.3525, 252963
  - TFP growth (mean, p50, sd, min, max, N obs)
    - Foreign Firms: -0.0030, 0.0029, 0.2147, -0.8976, 0.8046, 27,961
    - SOEs: 0.0050, 0.0031, 0.1688, -0.8691, 0.7970, 7,228
    - Domestic Private Firms: -0.0188, -0.0127, 0.2378, -0.8988, 0.8045, 176,796
- Ownership and firm characteristics (Table 2; Dependent Variable: TFP level):
  - Foreign Firms: Assets coefficient 0.071*** [7.184]; Revenue 0.082*** [8.200]; Age 0.031*** [7.061]; N obs 289,588.
  - SOEs: Assets coefficient 0.096*** [2.812]; Revenue 0.012 [0.319]; Age 0.066*** [5.303]; N obs 289,588.
  - Notes: regressions include firm and industry-year fixed effects. Robust t-statistics in parenthesis. *** 1 percent, ** 5 percent, * 10 percent.
- Ownership and TFP (Table 3; Multinomial regression marginal effects):
  - Dependent Variable: Ownership (control = foreign ownership)
  - TFP marginal effects:
    - dummy SOEs: -0.0375*** (0.00161)
    - Domestic Private Firms: -0.00658** (0.00312)
  - N obs 211,378. Controls: age, size, industry and year fixed effects. Standard errors clustered at the firm level.
- TFP drivers — cross-sectional regression (Table 4; Dependent Variable: TFP level):
  - Poland, CEE5, CEE5+IT+ES results:
    - Foreign coefficient: 0.094*** (Poland, col 1); 0.091*** (CEE5, col 2); 0.159*** (CEE5+IT+ES, col 3).
    - SOEs coefficients (cols 4–6): -0.421***, -0.471***, -0.264*** respectively.
  - N obs: 182,425; 721,979; 3,972,410. R2: 0.733; 0.666; 0.822 (and 0.737; 0.667; 0.821 in alternative columns).
  - Notes: regressions include firm level characteristics (total assets, company age, share of intangibles to total assets, variable to proxy for leverage), all regressors lagged by one year; include sector-year and country-year fixed effects.
- TFP drivers — within-firm regression (Table 5; Dependent Variable: TFP level):
  - Foreign coefficients: 0.019*** (Poland), 0.009** (CEE5), 0.005** (CEE5+IT+ES).
  - SOEs coefficients: -0.064***, -0.054***, -0.044*** respectively.
  - N obs: 289,597; 1,052,826; 5,286,057. R2: 0.928; 0.947; 0.972.
- TFP growth drivers — cross-sectional regression (Table 6; Dependent Variable: TFP growth):
  - Foreign coefficients: 0.032*** (Poland), 0.030*** (CEE5), 0.033*** (CEE5+IT+ES).
  - SOEs coefficients: -0.059***, -0.030***, -0.006*** respectively.
  - Lagged TFP: -0.178***, -0.105***, -0.097*** (and -0.180***, -0.105***, -0.096*** in alternative columns).
  - N obs: 182,425; 721,979; 3,972,410. R2: 0.733; 0.666; 0.822 (and 0.737; 0.667; 0.821).
- TFP growth drivers — within-firm regression (Table 7; Dependent Variable: TFP growth):
  - Foreign coefficients: 0.006, 0.005, 0.00 (not significant) with t-stats [0.995], [1.387], [-0.147].
  - SOEs coefficients: -0.051**, -0.031**, -0.025 (t-stats [-2.083], [-1.758], [-1.616]).
  - Lagged TFP: -0.692***, -0.723***, -0.680*** (t-stats [-174.99], [-341.74], [-758.68]).
  - N obs: 170,053; 677,169; 3,785,237. R2: 0.501; 0.492; 0.482.
- TFP growth and policy variables (Table 8; Dependent Variable: TFP growth by ownership):
  - Foreign Firms (col 1): Regulation -0.006*** [-4.692]; Lagged TFP -0.047*** [-33.433]; Size 0.007*** [10.716]; Leverage 0.023*** [2.902]; Intangibles 0.070*** [6.525]; N obs 75,624; R2 0.059.
  - SOEs (col 2): Regulation -0.003 [-1.090]; Lagged TFP -0.023*** [-11.678]; Size 0.004*** [3.614]; Leverage 0.037 [1.210]; Intangibles 0.040*** [2.469]; N obs 12,011; R2 0.05.
  - Domestic Private Firms (col 3): Regulation -0.008*** [-18.492]; Lagged TFP -0.099*** [-208.115]; Size 0.015*** [77.228]; Leverage 0.039*** [24.021]; Intangibles 0.085*** [30.106]; N obs 1,580,189; R2 0.069.
  - Notes: Regressions include firm and industry-year fixed effects. Robust t-statistics in parenthesis. *** 1 percent, ** 5 percent, * 10 percent.

*Prepared by Federico Joaquin Díez and Yevgeniya Korniyenko. Orbis database excludes firms operating outside the market economy (excluding sectors 83–99 based on NACE Rev.2).*

*Content derived from the supplied IMF document.*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2019/cr1938.pdf_