## _cr16211

## Source details

**Canonical URL:** [_cr16211](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2016/_cr16211.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2016/_cr16211.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2016/_cr16211.pdf.json)

---

### Regional disparities and income convergence
- Poland closed over a quarter of its per capita income gap with the EU28 average during 1995–2014.
- GDP growth has averaged around 3 percent since 2008.
- Western regions bordering Germany have on average GDP per capita about 30 percent higher than eastern regions bordering Belarus and Ukraine.
- Mazowieckie (includes Warsaw) has a per capita income comparable to France; the poorest eastern-border regions have per capita incomes comparable to Bulgaria.
- Poverty rates and long-term unemployment are higher in the east.
- Regions have been growing at about the same rate during the past decade, limiting catch-up by poorer regions.
- Estimated potential gains from reducing regional gaps:
  - Poland’s GDP per capita could be increased by 7 percent if one-third of the gap between regions with GDP per capita below average and the four regions which are above average was closed.
  - This would imply an additional 5 percentage point convergence to the EU28 average income level.
- Labor productivity is strongly associated with regional income per capita; wealthier regions tend to have higher labor productivity.

### Determinants and correlates of regional productivity
- Positive correlates:
  - Foreign direct investment (FDI) intensity: higher regional FDI associated with higher productivity; western regions benefit from integration into the German supply chain.
  - Educational attainment: higher share of population 13+ with higher education correlates with higher productivity.
  - Transportation: major highways and higher railroad density concentrate in western regions, facilitating business location and labor mobility.
- Negative correlates:
  - Higher share of employment in agriculture associated with lower productivity.
  - Registered unemployment rates are higher in less productive regions.
- EU cohesion funds have been directed more to less productive eastern regions; value of EU co-financed projects per capita is correlated with productivity patterns.
- Wealthier regions show higher productivity across most sectors (agriculture, industry, construction, market services, public sector).

### Empirical model, data and estimation
- Estimation method:
  - Blundell and Bond system GMM estimator (Arellano and Bover (1995); Blundell and Bond (1998)), two-step with Windmeijer’s adjustment.
  - Time dummies included; all available lags used as GMM instruments.
- Sample and measurement:
  - Annual data between 2002 and 2012; unit of observation: region-year; Number of regions: 63.
  - Dependent variable: real labor productivity growth, measured as gross value added per employed (including self-employed), deflated by the GDP deflator.
- Structural variables included:
  - Labor quality: share of population with education below secondary school.
  - Structural transformation: change in the share of employment in agriculture, industry, and services.
  - Registered unemployment rate.
  - Patent applications per million employed (EPO) as innovation proxy.
  - Local government finances: investment spending share and non-investment spending to gross value added.
- Spillover variables:
  - Share of enterprises with foreign capital as FDI proxy.
  - Neighbors’ value added gap (labor productivity of neighboring regions).

### Main empirical findings (regional productivity growth)
- Foreign investment:
  - In the base model, a ten percent higher share of enterprises with foreign capital results in a 1.8 percentage point increase in labor productivity growth.
- Structural transformation:
  - A 1 percent decline of the share of agricultural employment in total employment adds about 0.5 percentage points to productivity growth.
  - Labor reallocation to industry results in higher overall productivity growth; reallocation to services does not show a significant impact on services productivity in the regressions.
- Educational attainment:
  - Decreasing the share of population with below secondary education by 10 percent boosts productivity growth by 5 percentage points.
- Labor market tightness:
  - Availability of idle workers has a negative impact on labor productivity by 0.4 percentage point per 1 percent higher unemployment rate.
- Local government expenditures:
  - Investment expenditures of local governments are an insignificant determinant of labor productivity growth.
  - Non-investment expenditures have a positive impact of 0.2–0.5 percentage point faster productivity growth per a 1 percentage point increase in the expenditure to gross value added ratio.
- Spatial spillovers and innovation:
  - Positive spillovers from proximity to regions with high productivity levels are almost absent.
  - Patent applications per million employed are not a significant determinant of labor productivity improvements.

### Structural and cross-regional reallocation (Box 2)
- Structural transformation contributed more to productivity growth in Poland than in regional peers.
- Structural transformation had a larger impact in poorer eastern regions due to higher initial agricultural shares and scope to move labor to more productive sectors.
- Cross-regional reallocation of labor contributed only 2.5 percentage points to total productivity gains in 2000–12.
- Sector decomposition (figure annotations in source): contribution of labor reallocation across sectors to productivity growth, 2000–12 (percentage points) plotted values include 5.7, 3.4, 4.2, -0.5 for Total, Industry, Services, Agriculture respectively (as presented in the source).

### Policy implications for regional convergence
- Increase productivity in lagging regions through:
  - Improving educational attainment and reducing skill mismatches in the east.
  - Scaling up public infrastructure (transportation network and connectivity).
  - Facilitating labor mobility across regions to allow reallocation of workers to higher-productivity sectors and regions.
  - Targeted use of EU cohesion funds and local government investment spending to support regional development and infrastructure that enhances productivity.
- Agricultural sector reforms to encourage consolidation and reduce incentives to remain on low-productivity farms (detailed in the Conclusions and policy directions section).

---

### Choice of model and robustness (system GMM)
- Estimator: Blundell and Bond system GMM; two-step with Windmeijer adjustment; time dummies included; all available lags used as GMM instruments.
- Panel dimensions and sample in reported specifications:
  - No. of Observations: 630 630 630 693 630 630 630 630 630 (as reported across specifications).
  - No. of Regions: 63 (in each specification reported).
  - No. of Instruments: reported values include 52 70 70 66 70 62 71 73 85.
- Specification tests (selected reported p-values):
  - AR1 (p-value): (0.000) across specifications.
  - AR2 (p-value): reported values include (0.156), (0.038), (0.012), (0.143), (0.100), (0.119), (0.082), (0.114), (0.100).
  - Hansen test (p-value): reported values include (0.173), (0.358), (0.240), (0.156), (0.219), (0.141), (0.278), (0.629), (0.219).
- Selected coefficient signs and statistical significance (examples as reported):
  - Real Productivity (t-1): negative and statistically significant (e.g., -0.258***, -0.304***, -0.353***).
  - Change in share of employment in agriculture (t): negative and significant (e.g., -0.587***, -0.492***, -0.455**, -0.517***, -0.427**, -0.490***, -0.538***).
  - Change in share of employment in industry (t): positive and significant where included (0.508***).
  - Share of enterprises with foreign capital (t-1): positive and significant in many specifications (e.g., 0.184*, 0.2046, 0.248**, 0.469***, 0.387**, 0.423**).
  - Share of population below secondary education (t-1): negative and significant (e.g., -0.439***, -0.495***, -0.553***, -0.619***, -1.055***).
  - Unemployment rate (t-1): negative and significant where included (e.g., -0.396***, -0.320*, -0.293**).

### Robustness checks and extensions
- System GMM applied to productivity dynamics in each of three main sectors.
- Data transformed into 2-year non-overlapping periods.
- Cross-section and random effects models run on same variables.
- Total factor productivity used as dependent variable within a random effects specification.
- Results broadly consistent across robustness checks.

---

### Aging population and workforce projections
- Working-age population (people aged 15–64) of around 27 million is projected to shrink by about 3 million people or about 11 percent by 2030.
- By 2060, Eurostat projections suggest the working-age population will have shrunk by more than 30 percent.
- Old-age dependency ratio: today, for every one hundred persons of working age, there are about 20 people aged 65+; by 2060 this number is expected to increase to about 60 people.
- Poland’s fertility rate declined from close to three in 1960 to 1.3 in 2014.

### Female labor force participation: levels and potential
- As of 2014, women accounted for about 45 percent of the civilian labor force in Poland.
- For every 100 women of working age, only 61 were participating in the labor force (2014).
- Women’s life expectancy at birth: 81 years; men: 73 years.
- OECD estimate quoted: closing the gender participation gap in OECD countries by 2030 could increase the level of GDP by 12 percent on average across the OECD (OECD, 2012).
- Back-of-the-envelope for Poland (labor share 0.5): increasing women’s labor force participation rate to that of men could be associated with about 5 percent higher output.
- Female participation trends and gaps:
  - Female participation in Poland declined by 3½ percentage points during 2000–07, while NMS7 experienced an increase of about 1 percentage point.
  - In 2014, a female participation gap of more than 6 percentage points remained vis-à-vis NMS7 and EU15, where average participation rates exceeded 67 percent.
  - Prime working-age women (25–54): generally on par with EU15.
  - Youngest (15–24) and oldest (55–64) women: participation gap of more than 15 percentage points vis-à-vis EU15.
  - Gender gap relative to men: more than 13 percentage points lower than men in Poland; Nordic average gap is less than 5 percentage points; OECD average gap is close to 17 percentage points.
  - Part-time employment among women in Poland is 8 percent of employed women (2014), less than half the OECD average.

### Women in corporate leadership
- In 2013, in a sample of about 10 thousand non-financial corporations in Poland, less than a quarter of senior positions were held by women.
- Cribis survey: 22 percent of 242,000 commercial code companies in Poland report being managed by women.
- Polish Ministry of Treasury gender-representation plan for state-controlled companies aimed at 30 percent female representation at supervisory boards; publicly quoted state-controlled firms were close to that target in mid-2015.

### Drivers of female labor supply and policy levers
- Individual and preference factors: number of young children, life-cycle effects, education, intergenerational spillovers, and social norms affect participation.
- Policy factors:
  - Tax policy: joint taxation raises participation tax rate for second earner; Poland’s participation tax rate for second earner is well above the median European country and the OECD average (2013 data).
  - Childcare and pre-school services: Poland’s pre-school spending as of 2011 corresponds to around two thirds of the OECD average; coverage for 2-year olds corresponds to less than 20 percent of the OECD average.
  - Leave policy: generous leave can boost participation but excessive leave can have adverse effects.
- Policy recommendation summary:
  - Redirect resources toward pre-school education to boost female labor force participation.
  - Promote flexible work arrangements to combine parenting with work.
  - Address disincentives from high participation tax rates on the second earner.
  - Reconsider proposals to lower the retirement age that could reduce female participation.

---

### Family policy package: components and recent changes
- Components: tax breaks, services, and cash benefits.
- Tax breaks (2013):
  - Joint filing by married couples accounted for PLN3.2 billion (0.2 percent of GDP) in 2013.
  - Child tax allowances accounted for PLN5.5 billion (0.3 percent of GDP) in 2013.
  - Child tax allowance for the third child after 2014 changes is 80 percent above that for the first and second children; it goes up by an additional 35 percent for the fourth and subsequent children.
- Services (pre-school education):
  - Share of children aged 3–5 covered by pre-school education almost doubled during 2004–14, but an enrollment gap remains relative to the OECD average.
  - Early childhood education for 2-year olds in Poland corresponds to less than 20 percent of the OECD average.
- Cash benefits and ‘Family 500+’ (April 2016):
  - Program expected to cover about 2.7 million families, including 3.6 million children (about 10 percent of Poland’s population).
  - Benefits amount to about 40 percent of average disposable per-capita income in 2014.
  - Government estimates put total annual budgetary cost at around 1 percent of GDP during 2017–19, likely raising family policy spending to close to 3 percent of GDP per year—above the OECD average.
  - Program envisages PLN500 per month for the second and each subsequent child; for poorer families with monthly income of less than PLN800 per person, the first child will also be entitled.

### Leave and retirement policy changes
- Paid maternity leave increased from 20 to 26 weeks in 2013; an additional 26 weeks of paid parental leave added.
- Maternity benefits: 100 percent of salary when 26 weeks of leave are taken or 80 percent of salary when total of 52 weeks are taken.
- As of 2016, maternity benefits paid to uninsured women in amount of PLN 1,000 per month.
- Childcare leave with job protection up to three years is available; childcare leave is unpaid but counted in employment period and covered by social insurance (contributions paid by the state budget).
- Retirement policy:
  - Early retirement schemes phased out starting in 2009; statutory retirement age increased to 67 with 2013 pension changes, equalized for men and women via gradual process (target for men in 2020 and for women in 2040).
  - President submitted draft law to restore previous retirement ages (mentioned in source).

### Potential implications and recommendations
- Potential adverse effects:
  - ‘Family 500+’ may weaken women’s incentives to work; cash transfers often associated with lower female employment rates.
  - Lowering retirement age could reduce female labor force participation.
- Policy recommendations:
  - Shift resources toward childcare and pre-school education rather than primarily cash transfers.
  - Promote flexible work arrangements and address tax disincentives for second earners.
  - Avoid policies that lower effective retirement age for women.

---

### Banking sector performance and bank asset tax impacts
- Banking sector soundness:
  - Banks are well capitalized; ratio of non-performing loans to gross loans compares well regionally.
  - ROAA in Poland is below that in some comparable countries; ROAE close to regional average.
- Profitability trends:
  - ROAA trending down starting in 2014 Q3; low interest rates and increased costs in 2015 affected profitability.
  - Average tax outlays paid by banks relative to assets in 2015 were 20 percent; with the new asset tax this ratio is expected to double.
- Bank asset tax design:
  - Levied starting February 2016 on total value of assets less PLN 4 billion, own funds, and purchased sovereign debt at a monthly rate of 0.0366 percent (or 0.44 percent per year).
  - Exclusions: state-owned banks (relevant for BGK with legal status of state enterprise) and banks under recovery, in receivership, liquidation, or bankruptcy.
  - Tax may incentivize banks to increase holdings of exempted local government bonds and government securities.

### Bank-level regression and scenario results
- Sample: annual data for 55 Polish banks (39 commercial banks and 16 other depository corporations), covering 2005–15.
- Selected regression coefficients (as reported):
  - Lag of ROAA: 0.0745 (standard error 0.0969).
  - Lag of cost to income: -0.0152*** (0.0043).
  - Net interest margin: 0.1728** (0.0716).
  - Tax to pretax profits: -0.0091*** (0.0031).
  - Government securities to assets: 2.1937** (1.0585).
  - Observations: 328; Number of banks: 55; Number of instruments: 65.
  - P-value for Hansen statistics: 0.5505; P-value for AR2 statistics: 0.4699.
- Two forecast scenarios for 2016:
  - Scenario 1: interest rates and taxes remain at 2015 levels — annual ROAA could decline by 12 percent in 2016 relative to 2015.
  - Scenario 2: taxes increase corresponding to the new asset tax — the new bank asset tax could reduce profitability by an additional 18 percent (relative to Scenario 1).
  - Combined effect: total reduction in profitability of 30 percent relative to 2015.
  - If banks increase holdings of government securities by 10 percent relative to 2015, the difference in profitability relative to Scenario 1 would be 15 percent instead of 18 percent; a ten percent increase in debt holdings jointly with the new tax will reduce profits by 27 percent.

### Aggregate VAR and macro implications
- VAR baseline (no bank asset tax) projects average credit growth in 2016 is expected to decline by 1 percentage points relative to 2015.
- Two approaches to estimate tax impact on credit:
  - Approach 1 (uses bank-level tax impact on profitability): tax outlays increase would reduce credit growth by about 3 percentage points at end-2016 relative to the end-year baseline.
  - Approach 2 (includes tax payments directly in the VAR): tax outlays increase would reduce credit growth by about 4 percentage points at end-2016 relative to the end-year baseline.
- Impact on real GDP by end-2016 (via OLS regression of real GDP growth on lagged GDP growth and lagged credit growth):
  - Approach 1: real GDP growth about 0.3 percentage points below a no-tax scenario.
  - Approach 2: impact could be as large as 0.4 percentage points below a no-tax scenario.
- Oxford Economics estimate cited (source): the asset tax could reduce 2016 growth by 0.2 percentage points (Bloomberg, April 20, 2016).
- Observed short-term loan composition changes in early implementation:
  - Reduction in corporate and mortgage loans and an increase in consumer lending.
- Policy recommendation:
  - A Financial Activities Tax (FAT) based on profits and remuneration would be less distortionary than a tax on assets; example design: a 10–13 percent levy on bank profits and remunerations could generate the same revenue of PLN 4 billion expected from the current bank asset tax.
- Caveats:
  - Incentives to shift holdings from NBP bills to government bonds could complicate monetary policy.
  - Crowding out of private sector credit and potential adverse sovereign–bank feedback loops if public finances come under pressure.
  - Banks could shift lending to riskier consumer lending or less-regulated non-banks.

### VAR and OLS coefficient highlights (preserving reported values)
- Key VAR coefficients (Approach 1, Log(credits), column 1):
  - Lag 1: 1.1632*** (0.1434)
  - Lag 6: 0.2197** (0.1106)
  - Lag 8: -0.1868* (0.0969)
  - Lag of ROAA: 0.0131** (0.0054)
  - Lag of GDP growth rate: 0.4430*** (0.1676)
  - Observations: 43; R-squared: 0.99
- Key OLS (GDP growth rate, Table 3):
  - Lag 1: 0.8436*** (0.0882)
  - Lag 4: -0.3666*** (0.1016)
  - Credit growth rate Lag 1: 0.1775*** (0.0539)
  - Credit growth rate Lag 2: -0.1349** (0.0507)
  - Observations: 49; R2: 0.7242

*Prepared by IMF staff based on the supplied chapter text in _cr16211.*

### 1. Poland: Regions ______________________________________________________________________ 9

### 1. Poland: Regions

### Regional disparities and income convergence
- Poland closed over a quarter of its per capita income gap with the EU28 average during 1995–2014.
- GDP growth has averaged around 3 percent since 2008.
- Western regions bordering Germany have on average GDP per capita about 30 percent higher than eastern regions bordering Belarus and Ukraine.
- Mazowieckie (includes Warsaw) has a per capita income comparable to France; the poorest eastern-border regions have per capita incomes comparable to Bulgaria.
- Poverty rates and long-term unemployment are higher in the east.
- Regions have been growing at about the same rate during the past decade, limiting catch-up by poorer regions.
- Estimated potential gains from reducing regional gaps:
  - Poland’s GDP per capita could be increased by 7 percent if one-third of the gap between regions with GDP per capita below average and the four regions which are above average was closed.
  - This would imply an additional 5 percentage point convergence to the EU28 average income level.
- Labor productivity is strongly associated with regional income per capita; wealthier regions tend to have higher labor productivity.

### Stylized facts on determinants of regional productivity
- Correlates of higher regional productivity (illustrated in Figure 2):
  - Foreign direct investment (FDI) intensity: regions with higher FDI have higher productivity; western regions benefit from integration into the German supply chain.
  - Educational attainment: a higher share of population 13+ with higher education correlates with higher productivity.
  - Employment structure: regions with a higher share of employment in agriculture (the least productive sector) have lower productivity.
  - EU cohesion funds: EU funds have been directed more to less productive eastern regions; value of EU co-financed projects per capita is correlated with productivity patterns.
  - Transportation network: major highways and higher railroad density concentrate in western regions, linking them to Germany and facilitating business location and labor mobility.
  - Unemployment: registered unemployment rates are higher in less productive regions; higher unemployment is associated with lower productivity.
- Wealthier regions show higher productivity across most sectors (agriculture, industry, construction, market services, public sector), not explained by a single sectoral advantage.

### Empirical analysis: model variables and approach
- The empirical model examines determinants of labor productivity growth at regional level, including structural variables and spillover effects.
- Structural variables considered:
  - Labor quality: measured by share of population with education below secondary school.
  - Structural transformation: change in the share of employment in agriculture relative to industry and services; analysis includes changes in industry and services shares as well.
  - Unemployment: structural unemployment differences, search costs, labor mobility, and their effects on productivity and factor substitution.
  - Ability to innovate: proxy is number of patent applications to the European Patent Organization per million employed persons.
  - Local government finances: share of investment spending in total expenditures and share of non-investment spending to gross value added to control for local government size and spending composition.
- Spillover variables considered:
  - International spillovers: share of companies with foreign capital used as a proxy for FDI at regional level.
  - Regional spillovers: labor productivity of neighboring regions to account for spatial spillovers—both positive demand/investment spillovers and negative drainage of skilled labor or clustering effects.

### Structural transformation and cross-regional labor reallocation (Box 2)
- Overall productivity change decomposes into:
  - Within-sector changes, and
  - Changes due to reallocation of labor across sectors (structural transformation).
- Structural transformation has supported productivity growth in Poland and had a larger impact than in regional peers.
- Structural transformation contributed especially in poorer eastern regions, reflecting their higher share of agriculture and greater scope to shift labor from low-productivity farming to more productive sectors.
- Cross-regional reallocation of labor contributed only 2.5 percentage points to total productivity gains in 2000–12, indicating limited aggregate role for cross-regional labor movements.
- Sector decomposition indicates that cross-regional labor movements benefited industry and services but not agriculture, suggesting bottlenecks in cross-regional labor mobility and possible adverse policy incentives discouraging reallocation.

- (Figure annotations / contribution values presented in source)
  - Contribution of labor reallocation across sectors to productivity growth, 2000–12 (percentage points): Total, Industry, Services, Agriculture (figure shows values including 5.7, 3.4, 4.2, -0.5 as plotted).
  - Cross-regional reallocation of labor contribution to productivity growth, 2000–12: shown as percent of overall productivity growth (figure annotation in source).

### Key policy implications and recommendations (as identified in the chapter)
- Increase productivity in lagging regions to reduce regional income disparities and promote income convergence by:
  - Improving educational attainment and reducing skill mismatches in the east to raise labor quality and attract productive investment.
  - Scaling up public infrastructure (transportation network and connectivity) to attract investment to less productive regions and integrate them into supply chains.
  - Facilitating labor mobility across regions to allow reallocation of workers to higher-productivity sectors and regions and to remove bottlenecks limiting cross-regional labor movements.
- Ensure targeted use of EU cohesion funds and local government investment spending to support regional development and infrastructure that enhances productivity.

*Prepared by Krzysztof Krogulski, Robert Sierhej, and Aaron Thegeya. Source: Central Statistical Office, Eurostat, and IMF staff calculations.*

### 10.      Choice of model. The model is estimated using a system GMM estimator

### 10.      Choice of model. The model is estimated using a system GMM estimator

### Methodology and data
- Estimator and identification:
  - Blundell and Bond system generalized method of moments (system GMM) estimator (Arellano and Bover (1995); Blundell and Bond (1998)).
  - Two-step procedure applied; Windmeijer’s adjustment used to correct downward bias in standard errors.
  - Time dummies included to control for common shocks; individual effects removed in the GMM transformation.
  - All available lags used as GMM instruments; contemporary shifts of employment included alongside lagged levels of explanatory variables.
- Sample and measurement:
  - Annual data between 2002 and 2012.
  - Unit of observation: region-year.
  - Number of regions: 63.
  - Dependent variable: real labor productivity growth, measured as gross value added per employed (including self-employed), deflated by the GDP deflator.
  - Independent variables include: change in share of employment in agriculture, industry and services; share of population with education attainment below secondary level; registered unemployment rate; patent applications per million of employed persons; investment share of expenditures and non-investment expenditures of local governments as percent of gross value added; share of enterprises with foreign capital; neighbors’ value added gap.

### Main empirical findings (determinants of regional labor productivity growth)
- Foreign investment:
  - Foreign investment is a significant positive determinant of labor productivity growth.
  - In the base model, a ten percent higher share of enterprises with foreign capital results in a 1.8 percentage point increase in labor productivity growth.
- Structural transformation:
  - Structural transformation is a significant determinant of labor productivity growth.
  - A 1 percent decline of the share of agricultural employment in total employment adds about 0.5 percentage points to productivity growth.
  - Labor reallocation to industry results in higher overall productivity growth; reallocation to services does not show a significant impact on services productivity in the regressions.
- Educational attainment:
  - Level of educational attainment matters: decreasing the share of population with below secondary education by 10 percent boosts productivity growth by 5 percentage points.
- Labor market tightness:
  - Tighter labor market conditions support productivity enhancements: availability of idle workers has a negative impact on labor productivity by 0.4 percentage point per 1 percent higher unemployment rate.
- Local government expenditures:
  - Investment expenditures of local governments are an insignificant determinant of labor productivity growth.
  - Other (non-investment) expenditures have a positive impact of 0.2–0.5 percentage point faster productivity growth per a 1 percentage point increase in the expenditure to gross value added ratio.
  - Overall, local government outlays do not matter much for regional productivity dynamics.
- Spatial spillovers and innovation measures:
  - Positive spillovers from proximity to regions with high productivity levels are almost absent.
  - Patent applications per million employed are not a significant determinant of labor productivity improvements (not statistically significant; potential measurement limitations due to limited use of European patents in the analyzed period).

### Sectoral and cluster analysis
- Sectoral decomposition:
  - Structural transformation is particularly important in agriculture.
  - Foreign direct investment (FDI) is important in industry and services.
- Urban/rural/industrial clusters:
  - Analysis by urban and rural clusters shows that foreign direct investment and the level of education matter in each cluster: higher levels of FDI and higher levels of education are correlated with higher labor productivity growth.

### Regression output and specification diagnostics (summary from Appendix I)
- Sample size and panel dimensions:
  - No. of Observations: 630 630 630 693 630 630 630 630 630
  - No. of Regions: 63 63 63 63 63 63 63 63 63
- Instruments and tests:
  - No. of Instruments: 52 70 70 66 70 62 71 73 85
  - AR1 (p-value): (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
  - AR2 (p-value): (0.156) (0.038) (0.012) (0.143) (0.100) (0.119) (0.082) (0.114) (0.100)
  - Hansen test (p-value): (0.173) (0.358) (0.240) (0.156) (0.219) (0.141) (0.278) (0.629) (0.219)
- Selected coefficient signs and significance (as reported):
  - Real Productivity (t-1): negative and statistically significant across specifications (e.g., -0.258***, -0.304***, -0.353***, etc.).
  - Change in share of employment in agriculture (t): negative and significant (e.g., -0.587***, -0.492***, -0.455**, -0.517***, -0.427**, -0.490***, -0.538***).
  - Change in share of employment in industry (t): positive and significant in specifications where included (0.508***).
  - Share of enterprises with foreign capital (t-1): positive and significant in many specifications (e.g., 0.184*, 0.2046, 0.248**, 0.469***, 0.387**, 0.423**).
  - Share of population below secondary education (t-1): negative and significant (e.g., -0.439***, -0.495***, -0.553***, -0.619***, -1.055***).
  - Unemployment rate (t-1): negative and significant in specifications where included (e.g., -0.396***, -0.320*, -0.293**).
  - Local govt investment share of expenditures (t-1): insignificant (e.g., -0.0507).
  - Local govt non-investment expenditures to GVA (t-1): positive in some specs (0.204, 0.503*).
  - Neighbors GVA per capita gap (t-1): limited and mixed (0.0179, -0.00700).
- Specification tests interpretation:
  - Arellano-Bond test for second order autocorrelation confirms no significant autocorrelation among instruments (AR2 p-values reported).
  - Hansen tests for over-identification restrictions do not reject the null in reported specifications (Hansen p-values reported), suggesting instrument validity.

### Robustness and extensions
- Robustness checks performed:
  - System GMM applied to productivity dynamics in each of three main sectors.
  - Data transformed into 2-year non-overlapping periods.
  - Cross-section and random effects models run on same variables.
  - Total factor productivity used as dependent variable within a random effects specification.
  - Results were broadly consistent across robustness checks.

### Conclusions and policy recommendations (from the analysis)
- Key diagnosis:
  - Significant and enduring income disparities between western and eastern regions of Poland correlated with labor productivity differences.
  - Labor productivity growth in poorer eastern regions driven significantly by structural transformation; in wealthier western regions driven by higher investment and integration with the German supply chain.
  - Similar growth rates in labor productivity across regions have prevented eastern regions from catching up to western regions.
- Policy directions to support regional productivity convergence:
  - Support structural transformation and boost productivity in agriculture:
    - Review incentives pertaining to employment in agriculture to identify mechanisms encouraging people to stay in low-productivity farms.
    - Consider reforming the highly subsidized pension scheme for farmers to gradually align it with the regular system to discourage inefficient farming motivated by pension arbitrage.
    - Promote consolidation of agricultural production in poorer eastern regions to exploit economies of scale.
    - Move from agricultural taxation based on farm size and quality of land to an income-based tax to reduce disincentives to scale up farms and help define the base for social security contributions.
    - Accompany reforms with measures to address skill mismatches and bottlenecks in labor mobility.
  - Encourage labor mobility and reduce structural unemployment:
    - Address bottlenecks in labor market mobility (for example, improve functioning of the housing rental market, which is currently shallow).
    - Increase investment in active labor market policies to improve job searching efficiency across regions, upgrade skills, and reduce skill mismatches.
  - Attract investments to less productive regions:
    - Increase FDI presence in less productive regions to support regional productivity catch-up.
    - Strengthen transportation networks in poorer regions and better target EU funds to support investment attraction.
    - Invest in education and align vocational curricula to local industry needs to improve access to skilled labor for investors.
    - Improve efficiency of local tax administration, which correlates positively with regional productivity.

*Source: IMF staff calculations and analysis in _cr16211 - 10.      Choice of model. The model is estimated using a system GMM estimator*

### 1.      Poland is facing an aging population. By 2030, its working-age population

### _cr16211 - 1.      Poland is facing an aging population. By 2030, its working-age population

### Aging population and projections
- Working-age population (people aged 15–64) of around 27 million is projected to shrink by about 3 million people or about 11 percent by 2030.
- By 2060, Eurostat projections suggest the working-age population will have shrunk by more than 30 percent.
- Old-age dependency ratio: today, for every one hundred persons of working age, there are about 20 people aged 65+; by 2060 this number is expected to increase to about 60 people.
- Implication: dramatic increase in dependency ratio with adverse repercussions for social spending and potential growth.

### Fertility trends
- Poland’s fertility rate declined from close to three in 1960 to 1.3 in 2014.
- Result: Poland ranks among the countries in Europe with the lowest fertility rate.
- Declining fertility noted as a contributing factor to aging population challenges.

### Could increasing women’s labor force participation help?
- As of 2014, women accounted for about 45 percent of the civilian labor force in Poland.
- For every 100 women of working age, only 61 were participating in the labor force (2014).
- Women’s life expectancy at birth: 81 years; men: 73 years.
- OECD estimate quoted: closing the gender participation gap in OECD countries by 2030 could increase the level of GDP by 12 percent on average across the OECD (OECD, 2012).
- Back-of-the-envelope calculation for Poland (standard production function with labor share of 0.5): increasing women’s labor force participation rate to that of men could be associated with about 5 percent higher output.
- Closing the overall participation gap vis-à-vis the EU frontier would have a notably larger effect.

### Stylized facts on female labor force participation
- Female participation in Poland declined by 3½ percentage points during 2000–07, while other emerging European New EU Member States (NMS7) experienced an increase of about 1 percentage point.
- In 2014, a female participation gap of more than 6 percentage points remained vis-à-vis NMS7 and EU15, where average participation rates exceeded 67 percent.
- Age-group patterns:
  - Prime working-age women (25–54 years): generally on par with EU15.
  - Youngest (15–24) and oldest (55–64) working-age women: participation gap of more than 15 percentage points vis-à-vis EU15 persisted.
  - Women aged 55–59 almost doubled their participation rate after early retirement schemes were tightened in 2009.
  - Participation among 60–64 year-olds has remained broadly unchanged during the 2000s, while in NMS7 and EU15 this age group became markedly more active.
- Gender gap relative to men: women’s participation rate is more than 13 percentage points lower than men; Nordic average gap is less than 5 percentage points; OECD average gap is close to 17 percentage points.
- Life-cycle pattern: women of prime childbearing ages (20–39) have an average gender participation gap of close to 15 percent.
- Part-time employment: part-time employment among women in Poland is 8 percent of employed women (2014), less than half the OECD average; increases in participation would likely come from more women entering the labor force rather than increased part-time work.
- Education: more years of education associated with higher employment; women with tertiary education are more likely to be active.

### Women in the corporate sector
- In 2013, in a sample of about 10 thousand non-financial corporations in Poland, less than a quarter of senior positions were held by women.
- Cribis survey: 22 percent of 242,000 commercial code companies in Poland report being managed by women (concentrated in education, social assistance, healthcare services, and small retail trade).
- Polish Ministry of Treasury (MoT) introduced a gender-representation plan for state-controlled companies aiming at 30 percent female representation at supervisory boards; publicly quoted state-controlled firms were close to that target in mid-2015.
- Christiansen and others (2016a) found a strong positive association between firms’ gender diversity in senior positions and corporate financial performance, particularly in sectors with larger female shares and where creativity and innovative capacity are in demand.

### Drivers of female labor force participation (literature overview)
- Individual characteristics:
  - Larger number of young children in the household reduces likelihood of employment.
  - Labor supply varies over the life cycle.
  - More years of education generally associated with higher employment.
  - Growing up with a working mother increases likelihood of working (intergenerational spillovers).
- Preferences and attitudes:
  - Social norms affect women’s regard toward participating in the labor force.
  - ISSP-based evidence: women’s attitudes in Nordic countries are more favorable toward working than in several emerging European countries.
  - Attitudes have become more favorable toward women working during the past decade.
- Policies:
  - Tax policy: joint taxation (as in Poland) tends to have higher marginal tax rates on the second earner, usually a woman, creating disincentives.
  - Childcare and family benefits:
    - Availability of high-quality childcare facilitates working outside the home and supports female participation.
    - Family cash allowances can discourage participation due to income effects.
    - Leave policy allows temporary absence to care for children and can boost participation, but excessive/too generous leave can have the opposite effect and may lead employers to view women as high-risk workers.

### Policies in Poland: current setting and scope for reform
- Participation tax rate for the second earner in a family in Poland is well above the median European country and the OECD average (2013 data), suggesting tax-induced disincentives to work.
- Public spending on family policy (tax breaks, services, cash transfers) in Poland as of 2011: 1¾ percent of GDP—about two-thirds of the OECD average.
- Observations:
  - Both tax and expenditure policy suggest available room to support women’s decision to work outside the home.
  - Allocation of spending (not just amount) matters when assessing policy needs; international practice can guide policy choices for Poland.
- Policy levers discussed in prior sections are relevant for Poland: tax reform to reduce second-earner disincentives, expansion and targeting of childcare and pre-school services, careful design of family cash transfers and leave policies to avoid adverse labor supply effects.

*Prepared by Lone Christiansen and Robert Sierhej, drawing on Christiansen and others (2016).*

### 15.      The family policy package consist of three main components: tax breaks, services, and

### _cr16211 - 15.      The family policy package consist of three main components: tax breaks, services, and

### Family policy package: components and baseline statistics
- The family policy package consists of three main components: tax breaks, services, and cash benefits.
- Tax breaks (personal income tax related):
  - Joint filing by married couples (family tax credits) accounted for PLN3.2 billion (0.2 percent of GDP) in 2013.
  - Child tax allowances accounted for PLN5.5 billion (0.3 percent of GDP) in 2013.
  - Child tax allowances are means-tested only for the first child and were increased for the third and any subsequent children in 2014.
  - After the 2014 changes, the child tax allowance for the third child is 80 percent above that for the first and second children, and it goes up by an additional 35 percent for the fourth and subsequent children.
  - Families can deduct child tax allowances against PIT, social security, and health contributions.
- Services (pre-school education):
  - Spending in this category is mainly related to pre-school education managed by local governments.
  - As of 2011, such spending in percent of GDP corresponded to around two thirds of the OECD average—though broadly on par with regional peers.
  - The share of children aged 3–5 covered by pre-school education almost doubled during 2004–14, but an enrollment gap remains relative to the OECD average.
  - The enrollment gap is particularly large for early childhood education: the ratio for 2-year olds in Poland corresponds to less than 20 percent of the OECD average.
  - Despite broadened coverage of early childhood facilities, availability remains scant by international standards.
- Cash benefits:
  - Largely means-tested child birth grants and family allowances for children aged 0–18.
  - In 2011, family-related cash allowances in Poland accounted for more than one third of family policy spending.
  - Family allowances have traditionally been low by OECD standards—about half the OECD average relative to GDP and 10 percentage points below the OECD average as a share of total family policy spending.
  - Ministry of Finance (MoF) noted family allowances covered some 40 percent of the population in the first quintile and 20 percent in the second quintile, but only half of the population below the absolute poverty line received family allowances.

### Selected international experience (Nordic perspective)
- Sweden (2014):
  - Labor force participation among women at almost 80 percent, 4¼ percentage points below men.
  - Women had on average about 1.9 children.
  - Close to 90 percent of Swedish women choose to work at full time.
  - Participation tax on the second earner around 25 percent.
  - Total public spending on family policy as of 2011 more than 3½ percent of GDP; more than half allocated to services such as childcare and pre-school.
  - Paid maternity and parental leave total 38 full-rate equivalent weeks.
  - Average effective retirement age during 2009–14 among women in Sweden was 64 years.
- Other Nordics:
  - Female labor force participation at 84 percent in Iceland coincides with fertility above 2.
  - Public spending on preschool and childcare in Iceland and Denmark as of 2011 exceeds 1½ percent of GDP.
  - Family cash allowances in Iceland at 1.2 percent of GDP rank below Sweden.
  - Paid leave entitlements vary: about 45 full-rate equivalent weeks in Norway to 27 in Denmark.
  - Effective retirement ages of 68 and 64 in Iceland and Norway, respectively, during 2009–14.

### Recent policy changes in Poland (2013–2016)
- Spending increase from new child benefits:
  - The ‘Family 500+’ benefits program took effect in April 2016 and markedly increased family cash allowances.
  - Program expected to cover about 2.7 million families, including 3.6 million children (about 10 percent of Poland’s population).
  - These benefits amount to about 40 percent of average disposable per-capita income in 2014.
  - Government estimates put total annual budgetary cost at around 1 percent of GDP during 2017–19, likely raising family policy spending to close to 3 percent of GDP per year—above the OECD average.
  - The program envisages non-means-tested benefits of PLN500 per month for the second and each subsequent child. For poorer families with monthly income of less than PLN800 per person, the first child will also be entitled to this benefit.
- Leave policy changes:
  - Paid maternity leave increased from 20 to 26 weeks in 2013; 20 weeks are obligatory (14 for the mother and 6 transferable to the father).
  - An additional 26 weeks of paid parental leave added, allowing extension of total leave period.
  - Maternity benefits amount to 100 percent of the salary when 26 weeks of leave are taken or to 80 percent of the salary when a total of 52 weeks of leave are taken, including 26 weeks of parental leave.
  - As of 2016, maternity benefits are paid also to uninsured women (students, unemployed, farmers) in the amount of PLN 1,000 per month, about ¼ of the economy-wide average wage.
  - Childcare (parental) leave with job protection of up to three years is available; childcare leave is unpaid but counted in employment period and covered by social insurance (contributions paid by the state budget). A person on childcare leave can request part-time work from the employer, which must be accepted.
- Retirement regulations:
  - Until 2009, early retirement allowed women to retire five years before the statutory age of 60 with 30 years of social security insurance; sector-specific early retirement schemes applied.
  - Early retirement schemes were phased out starting in 2009, increasing effective retirement age.
  - Statutory retirement age later increased to 67 with 2013 pension changes, equalizing for men and women via a gradual process by 3 months per year (target for men in 2020 and for women in 2040).
  - As of 2014, effective female retirement age remained well below the OECD average.
  - The President submitted to parliament a draft law to restore the previous retirement age of 60 and 65 for men and women, respectively.

### Potential implications of recent changes
- Recent measures could weaken women’s incentives to work:
  - ‘Family 500+’:
    - Lump-sum transfers will be a dominant factor in family policy.
    - Cash transfers are often found to have limited impact on fertility and tend to be associated with lower female employment rates.
    - In contrast, spending on childcare facilities appears to be more positively associated with fertility.
  - Potential lowering of the retirement age:
    - If lowered to 2013 levels, prolonged attachment to the labor market among women would likely not materialize, adversely impacting the effective old-age dependency ratio.
    - The increase in participation rates among 55–60 year-old women after early retirement schemes were phased out in 2009 highlights the effectiveness of measures that retain women in the workforce.

### Key findings and policy recommendations
- Findings:
  - Female labor force participation in Poland is low, particularly among seniors.
  - Labor force participation among women of prime working age (25–54) is below that of men.
  - Women tend to withdraw from the labor market earlier than men; effective retirement age for women in Poland at about 60 is notably below the OECD average of about 63.
  - Polish family-related policies historically focused on leave and cash transfers rather than childcare provision.
  - The ‘Family 500+’ program significantly increased spending on family cash transfers.
  - Cross-country empirical evidence tends to show a negative association between cash transfers and labor market participation, but a positive association with childcare provision.
- Policy recommendations:
  - Redirect resources toward pre-school education to have a larger positive impact on female labor force participation.
  - Such a shift toward childcare need not come at the expense of fertility.
  - Promote flexible work arrangements to help combine parenting with work, noting the low prevalence of part-time work.
  - Consider addressing disincentives to working arising from high participation tax rates on the second earner in a family.
  - Reconsider proposals to lower the retirement age, as lowering would likely reduce female labor force participation.

*International Monetary Fund — Republic of Poland, Section on Family Policy and Gender/Labor Implications*

### 1.      The Polish banking sector performs well on a number of financial soundness

### _cr16211 - 1.      The Polish banking sector performs well on a number of financial soundness

### Overview and context
- Banks are well capitalized and the ratio of non-performing loans to gross loans, though still elevated, compares well within the region.
- Return on average assets (ROAA) in Poland is below that in some comparable countries (e.g., the Czech Republic), while returns on average equity (ROAE) are close to the regional average.
- Recent events affecting profitability include low interest rates and increased costs in 2015, notably the bankruptcy of SK bank and the Bank Guarantee Fund (BFG) payment of about PLN 2 billion.

### Recent profitability trends and drivers
- ROAA has been trending down starting in 2014 Q3, with low interest rates being the main driving factor.
- Costs for banks grew at a much higher rate than income in 2015, resulting in declining efficiency as measured by the cost-to-income ratio.
- Average tax outlays paid by banks relative to assets in 2015 were 20 percent; with the new asset tax this ratio is expected to double.

### Bank asset tax: design and parameters
- The bank asset tax is levied starting February 2016 on the total value of assets less PLN 4 billion, own funds, and purchased sovereign debt at a monthly rate of 0.0366 percent (or 0.44 percent per year).
- State-owned banks (relevant for BGK with legal status of state enterprise) and banks under recovery proceedings, in receivership, in liquidation, or those who filed for bankruptcy are excluded from the tax.
- The tax may incentivize banks to increase holdings of exempted local government bonds and government securities.

### Bank-level analysis: methodology and sample
- Sample: annual data for 55 Polish banks and other depository corporations (39 commercial banks and 16 other depository corporations), covering 2005–15.
- Method: panel regressions using system GMM; instruments are internal (lags of instrumented variables, collapsed).
- Main regression specification includes lagged ROAA, lagged net loans/ assets, lagged equity/ assets, lagged cost/ income, lagged log(assets), net interest margin (NIM), GDP growth, tax/ pretax income, government securities/ assets, and year dummies.

### Key regression coefficients and diagnostics (as reported)
- Lag of ROAA: 0.0745 (standard error 0.0969)
- Lag of net loans to total assets: 0.0119 (0.0089)
- Lag of equity to total assets: 0.0418 (0.0257)
- Lag of cost to income: -0.0152*** (0.0043)
- Lag of log(assets): 0.0193 (0.0896)
- Net interest margin: 0.1728** (0.0716)
- GDP growth rate: 0.0721 (0.2079)
- Tax to pretax profits: -0.0091*** (0.0031)
- Government securities to assets: 2.1937** (1.0585)
- Observations: 328
- Number of banks: 55
- Number of instruments: 65
- P-value for Hansen statistics: 0.5505
- P-value for AR2 statistics: 0.4699

### Bank-level scenario results and quantified impacts on profitability
- Two forecast scenarios for 2016:
  - Scenario 1: interest rates and taxes remain at 2015 levels.
  - Scenario 2: taxes increase corresponding to the new asset tax.
- Scenario 1 outcome: annual ROAA could decline by 12 percent in 2016 relative to 2015, largely owing to higher costs in 2015.
- Scenario 2 outcome: the new bank asset tax could reduce profitability by an additional 18 percent (relative to Scenario 1).
- Combined effect: total reduction in profitability of 30 percent relative to 2015.
- If banks increase holdings of government securities by 10 percent relative to 2015 to reduce the tax base, the difference in profitability relative to Scenario 1 would be 15 percent instead of 18 percent (Scenario 2), and a ten percent increase in debt holdings jointly with the new tax will reduce profits by 27 percent.

### Aggregate analysis: VAR and macro implications
- An aggregate VAR model projects a baseline (no bank asset tax) where average credit growth in 2016 is expected to decline by 1 percentage points relative to 2015.
- Two approaches to estimate tax impact on credit growth:
  - Approach 1: uses bank-level estimate of tax impact on profitability to infer credit impact.
  - Approach 2: includes tax payments directly in the VAR.
- Tax outlays increase would reduce credit growth by between 3 (Approach 1) and 4 (Approach 2) percentage points at end-2016 relative to the end-year baseline.
- Using an OLS regression of real GDP growth on lagged GDP growth and lagged credit growth, the impact on real GDP by end-2016:
  - Approach 1: real GDP growth about 0.3 percentage points below a no-tax scenario.
  - Approach 2: impact could be as large as 0.4 percentage points below a no-tax scenario.
- Oxford Economics estimate cited: the asset tax could reduce 2016 growth by 0.2 percentage points (Bloomberg, April 20, 2016).

### Summary of findings
- The bank asset tax can significantly reduce bank profits and slow credit and economic growth.
- Quantified impacts:
  - Bank profitability: additional reduction of 18 percent from the asset tax; total reduction of 30 percent relative to 2015 when combined with other factors.
  - Credit growth: contraction by between 3 and 4 percentage points at end-2016 relative to a no-tax projection.
  - Real GDP growth: reduction of up to 0.4 percentage points by end-2016 attributable to the bank tax.

### Policy recommendations and caveats
- A Financial Activities Tax (FAT) based on profits and remuneration would be less distortionary than a tax on assets. Example designs referenced:
  - A 10–13 percent levy on bank profits and remunerations could generate the same revenue of PLN 4 billion expected from the current bank asset tax.
  - Country examples: Israel levies an FAT at its standard VAT rate of 17 percent; Iceland levies a 5.5 percent tax on wages and a 6 percent tax on profits above ISK 1 billion; Denmark levies a 10.5 percent tax on wages.
- Potential unintended effects and caveats:
  - Incentives to shift holdings from NBP bills to government bonds could complicate monetary policy.
  - Crowding out of private sector credit and potential adverse sovereign–bank feedback loops if public finances come under pressure.
  - Policy uncertainty from additional sectoral taxes may sour investor sentiment, raise financing costs, and weigh on public finances and growth.
  - Banks could respond by shifting lending toward riskier consumer lending or to less-regulated non-banks, with implications for credit composition and risk.

*Prepared by Yorbol Yakhshilikov. Content based solely on the supplied IMF chapter text.*

### 2. Aggregate analysis

### 2. Aggregate analysis

### Credit flow
- VAR framework chosen to capture dynamics between credit flows and deposit flows; bank characteristics (profitability, leverage, cost to income ratio, cost components) included as exogenous variables.
- Two approaches to estimate impact of the bank asset tax on credit, differing by exogenous variables:
  - Approach 1: exogenous variables include lagged real GDP growth and bank profitability (ROAA).
  - Approach 2: exogenous variables include real GDP growth, lagged average banking sector NIM, and tax outlays relative to pretax income.
- Data and modeling assumptions:
  - Approach 1 uses quarterly ROAA from Bankscope. For baseline forecast of four quarters in 2016, ROAA was adjusted downward by 12 percent year-to-year. In the impact scenario, ROAA in each quarter was reduced by 30 percent relative to the same quarter in 2015.
  - Approach 2 uses NIM and tax outlays from the Polish Financial Supervision Authority (KNF). Data for 2016Q1 reflect recent changes in tax outlays. For remaining 2016 quarters it was assumed tax outlays will be adjusted in the same proportion as in Q1 2016 relative to Q1 2015. No change in net interest margin was assumed relative to observed levels in 2016Q1.
  - Acknowledgement that because tax outlays only started in February, the estimate of the impact is likely underestimated.
- Stationarity and lag selection:
  - VAR estimated in log levels of credit and deposits.
  - Dickey-Fuller test on log credit rejects the null of a unit root; test for log deposits was inconclusive.
  - Lag structure chosen to match Akaike Information Criteria (modeled with up to 5 lags) and to approximate true coefficient estimates from OLS regressions of log(credits) on lags and the indicator of interest.
- Summary result statements:
  - Using staff projections for quarterly real GDP growth, the effect of the asset tax through reduced bank profitability on credit is reported in Table 2 (Approach 1, column 1).
  - Tax outlays are negatively associated with credit as shown in Table 2 (Approach 2, column 3).
  - When credit growth and deposit growth were modeled (differences rather than levels), the effect of bank profitability and tax was larger than in the levels VAR.
- Changes in loan composition observed in early implementation:
  - Empirical evidence from the first three months of the new asset tax suggests a reduction in corporate and mortgage loans and an increase in consumer lending.
  - Overall, these changes in loan structure are not detrimental to the direction of credit growth in 2016 as a result of the bank asset tax.
- Caveats:
  - Short data coverage and use of levels rather than differences in the VAR.
  - Series are highly persistent; as more reliable data become available the model could be revisited.

- Key coefficient estimates from Table 2 (selected, preserving values and significance as reported):
  - Dependent variable: Log(credits) — Approach 1 (column 1)
    - Lag 1: 1.1632*** (0.1434)
    - Lag 2: -0.2319 (0.1648)
    - Lag 6: 0.2197** (0.1106)
    - Lag 8: -0.1868* (0.0969)
    - Log(deposits) Lag 1: -0.1052 (0.1865)
    - Log(deposits) Lag 2: 0.2198 (0.1950)
    - Lag of ROAA: 0.0131** (0.0054)
    - GDP growth rate: 0.2820 (0.2206)
    - Lag of GDP growth rate: 0.4430*** (0.1676)
    - Constant: 0.0568 (0.2913)
    - Observations: 43
    - R-squared: 0.99
  - Dependent variable: Log(deposits) — Approach 1 (column 2)
    - Lag 1: 0.2507** (0.1243)
    - Lag 2: -0.0211 (0.1428)
    - Lag 6: 0.1853* (0.0958)
    - Lag 8: -0.3503*** (0.0840)
    - Log(deposits) Lag 1: 0.1381 (0.1616)
    - Log(deposits) Lag 2: 0.2114 (0.1690)
    - Log(deposits) Lag 8: 0.6332*** (0.1133)
    - Lag of ROAA: -0.0069 (0.0047)
    - Lag of net interest margin: 1.6404* (0.9708)
    - GDP growth rate: 0.4188** (0.1693)
    - Constant: 0.2911 (0.2524)
    - Observations: 43
    - R-squared: 0.99
  - Dependent variable: Log(credits) — Approach 2 (column 3)
    - Lag 1: 1.2993*** (0.1638)
    - Lag 2: -0.2974 (0.1885)
    - Lag 6: 0.1159 (0.1952)
    - Lag 8: -0.1802* (0.1066)
    - Log(deposits) Lag 1: -0.3153 (0.2350)
    - Log(deposits) Lag 2: 0.3538 (0.2530)
    - Log(deposits) Lag 8: 0.1555 (0.1525)
    - Lag of net interest margin: -0.0873 (0.7449)
    - Lag of tax to pretax income: -0.4052 (0.6401)
    - Lag of GDP growth rate: 0.1609 (0.1452)
    - Constant: -0.0649 (0.2391)
    - Observations: 43
    - R-squared: 0.99
  - Dependent variable: Log(deposits) — Approach 2 (column 4)
    - Lag 1: 0.1362 (0.1257)
    - Lag 2: 0.0975 (0.1447)
    - Lag 6: 0.0632 (0.1498)
    - Lag 8: -0.2905*** (0.0818)
    - Log(deposits) Lag 1: 0.4366** (0.1804)
    - Log(deposits) Lag 2: -0.0122 (0.1941)
    - Log(deposits) Lag 8: 0.6950*** (0.1170)
    - Lag of net interest margin: -0.0873 (0.7449)
    - Lag of tax to pretax income: -0.8012 (0.4912)
    - GDP growth rate: 0.4188** (0.1693)
    - Constant: 0.2423 (0.1835)
    - Observations: 43
    - R-squared: 0.99

### Real GDP growth
- Transmission mechanism considered: bank asset tax → credit growth → GDP growth.
- OLS specification: autoregressive process of GDP growth with lagged credit growth rates included.
- Core empirical result (Table 3):
  - Cumulative effect of the growth rate of credit over two consecutive quarters is significantly and positively associated with GDP growth.
- Key coefficient estimates from Table 3 (preserving values and significance):
  - Dependent variable: GDP growth rate
    - Lag 1: 0.8436*** (0.0882)
    - Lag 4: -0.3666*** (0.1016)
    - Credit growth rate Lag 1: 0.1775*** (0.0539)
    - Credit growth rate Lag 2: -0.1349** (0.0507)
    - Constant: 0.0153*** (0.0038)
    - Observations: 49
    - R2: 0.7242
- Additional note: In a VAR of GDP, investment, and consumption growth with credit growth as exogenous, the cumulative two-quarter effect of credit growth is of the same magnitude as in the OLS regression; credit growth is also positively associated with investment and consumption growth. Considering suppressed credit growth effects on investment and consumption, the total effect on the real economy in the OLS setup may be underestimated.

*Source: _cr16211 - 2. Aggregate analysis*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/scr/2016/_cr16211.pdf_