## 1. All Regions

## Source details

**Canonical URL:** [1. All Regions](https://www.imf.org/-/media/files/publications/wp/2018/wp18165.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp18165.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp18165.pdf.json)

---

### I. Introduction — demographic pressures and participation trends
- Dependency ratios are projected to reach "a striking 50 percent in more than three-quarters of countries by 2050."
- Median country statistics:
  - Population growth: "the median European economy’s population grew at only one fourth of a percent in 2016" and "more then half of countries will see negative growth rates by 2025."
  - Life expectancy: "the median country’s life expectancy increased by more than 12 years over the course of the past five and a half decades."
  - Healthy life expectancy: increased "by 3.6 years for men and by 2.8 years for women in the median European country between 1990 and 2015."
- Aggregate labor force participation patterns:
  - Participation rates increased in the majority of European countries and exhibited convergence, "especially from the bottom."
  - Contrast with the United States: "stark decline in participation observed in the United States" despite milder demographic pressures.
- Research aims:
  - Examine evolution and drivers of labor force participation rates in Europe, including heterogeneity across regions.
  - Tease out roles of aging, educational gains, cyclical conditions, and novel regional exposures to routinization and offshoring.
  - Construct employment-weighted regional measures of routinizability and offshorability following Das and Hilgenstock (2018) and related methodologies to allow comparability with U.S. analysis (Hilgenstock and Koczan 2018).

### II. Literature — policy, cohort, and technology effects
- Policy and institutional drivers of female participation emphasized: part-time incentives, maternity leave, childcare provision, paid parental leave, and family-friendly policies.
- Cipollone and others (2013): labor market reforms can account for "almost 25 percent" of the increase in participation of young women over two decades, and "more than 30 percent" for highly educated women.
- Cohort and lifecycle effects: cohort effects can account for a substantial part of recent increases in women’s participation.
- Technology and trade:
  - Gregory, Salomons, and Zierahn (2016): direct employment-reducing effects of automation on middle-skill employment offset by compensatory effects through product demand and local demand spillovers in European regions.
  - Graetz and Michaels (2015): industry-level adoption of industrial robots in EU countries "raised labor productivity, increased value-added, augmented worker wages, had no measurable effect on overall labor hours, and modestly shifted employment in favor of high-skilled workers."
- Contribution of this paper:
  - Use novel measures of initial local exposures to routinization and offshoring to study differential regional effects and comparisons to the United States.
  - Analyze participation at the subnational (roughly 250-region) level to exploit within-country heterogeneity.

### III. Data and empirical strategy
- Geographic and temporal scope:
  - Sample: roughly 250 NUTS regions (NUTS-2 preferred; NUTS-1 or country-level for some countries) with labor market indicators available from 2000.
  - Balanced panel includes 237 non-overlapping NUTS regions from 28 European economies for some statistics.
- Key variables and proxies:
  - Aging: old-age dependency ratios defined as "the population age 65 and older as a percentage of the population age 15 to 64."
  - Cyclical conditions: real GDP growth rates (Eurostat).
  - Education: share of the population enrolled in tertiary education (Eurostat).
  - Routinization and offshoring exposures: employment-weighted averages of occupational scores:
    - Routinability: scores from Autor and Dorn (2013) based on "routine task intensity" for 330 occupations using the DOT.
    - Offshorability: measures based on Goos, Manning, and Salomons (2014) converting Blinder and Krueger (2014) assessments into ISCO categories.
  - Employment-by-education data from the European Union Labour Force Survey (EULFS).
- Regression approach:
  - Regressions examine regional participation changes as functions of aging, cyclical conditions, education, and initial exposures to routinization/offshoring; controls include education and other regional characteristics.
  - Regression specification (cross-sectional regional-level, changes 2000–2016):
    - ∆yij = β1 ∆dep ratioij + β2 growthij + β3 ∆educij + β4 routij0 + β5 offshij0 + αj
    - ∆yij is the change in labor force participation between 2000 and 2016 in region i country j.
    - Initial exposures included one at a time due to high correlation; regressions control for time-invariant country characteristics (αj); standard errors adjusted for heteroskedasticity.

### IV. Stylized facts and aggregate outcomes (2000–2016)
- Participation levels and changes:
  - 2016 participation levels range from "the high 30s" (regions in Southern Italy and Romania) to "the high 60s and above" (regions in Germany, the Netherlands, Sweden, and the United Kingdom).
  - Changes between 2000–2016 range from "6–9 percentage point increases (in parts of Spain and Germany) to 15–20 percentage point drops (in parts of Romania)."
- Convergence:
  - Regions with relatively higher participation rates in 2000 experienced larger declines (or smaller increases), while regions with relatively lower participation rates in 2000 experienced larger increases or smaller decreases.
  - Decreasing standard deviations of regions within countries over time indicate within-country convergence.
- Urban-rural pattern:
  - Initial population density is positively correlated with changes in participation: "Participation rates increased more (or declined less) in urban regions, while they declined more (or increased less) in rural regions."
  - Correlation is modest: regression fit reported R² = 0.0306 in the population density relation (Figure 7).
- Aggregate outcomes:
  - "Labor force participation rates increased in a majority (65 percent) of European regions over the period 2000–2016, with a median increase of 3 percentage points;"
  - "the remaining 35 percent of regions exhibited a median decline of 1.7 percentage points."
- Pre- and post-crisis dynamics:
  - After 2008, "increases appear to be somewhat smaller, and declines more widespread" than in 2000–2008.
  - About "a third of regions experienced a post-crisis decrease following a pre-crisis increase;"
  - "21 percent saw an increase in the post-crisis period after a decline in the pre-crisis period;"
  - "35 percent of regions saw increases in both periods and 11 percent saw decreases in both."

### V. Labor market adjustment patterns (2000–2016; pre- and post-crisis)
- Full period (2000–2016): rising employment was matched almost one-for-one by rising participation, with very little change in unemployment.
- Pre-crisis (2000–2008): flows mainly from unemployment and inactivity into employment.
- Post-crisis (2008–2016): employment started to decline, matched almost fully by a rise in unemployment, with still small increases in participation.
- Urban-rural differences:
  - Rural regions: larger swings in unemployment—larger decline pre-crisis and much larger increase post-crisis; post-crisis sharper employment drop translating into large increases in unemployment and some increases in inactivity.
  - Urban regions: pre-crisis flows mostly from inactivity to employment; post-crisis saw some rise in unemployment but small increases in employment and activity.
  - Over 2000–2016: urban regions saw large increases in employment and activity; rural areas saw small employment declines matched by moderate drops in activity.

### VI. Main regression results — drivers of change in labor force participation rates (selected coefficients from Table 1)
- Dependent variable: change in labor force participation rate (2000–2016).
- Average Real GDP Growth:
  - Column (1): -0.178 (0.240)
  - Column (2): -0.441 (0.275)
  - Column (3): -0.348 (0.277)
- Change in Old-Age-Dependency Ratio:
  - Column (1): -0.275*** (0.0627)
  - Column (2): -0.288*** (0.0824)
  - Column (3): -0.295*** (0.0844)
- Change in Postsecondary Share:
  - Column (1): 0.256*** (0.0557)
  - Column (2): 0.308*** (0.0703)
  - Column (3): 0.282*** (0.0716)
- Initial Exposure to Routinization:
  - Column (4): 6.544*** (1.484)
  - Column (5): 7.058*** (1.428)
- Initial Exposure to Offshoring:
  - Column (4): 6.770*** (1.686)
  - Column (5): 6.992*** (1.642)
- Observations and goodness of fit:
  - Observations: 185 (col 1), 185 (col 2), 262 (col 3), 177 (col 4), 176 (col 5)
  - R2: 0.698 (col 1), 0.693 (col 2), 0.680 (col 3), 0.751 (col 4), 0.741 (col 5)

### VII. Interpretation and mechanisms
- Negative associations:
  - Regions with lower real GDP growth experienced larger declines in participation (coefficients imprecisely estimated).
  - Regions with larger shifts toward population older than 65 experienced larger declines in participation; sensitivity to aging appears higher in Europe than in the United States.
- Positive associations:
  - Educational gains (increase in postsecondary/tertiary share) strongly correlated with increases in participation; effect somewhat weaker than in the United States.
  - Contrary to findings for the United States, initial exposures to automation (routinization) and offshoring are positively associated with subsequent increases in participation rates in Europe, robust to controls for population density and country-level policies (active labor market spending, union density, generosity of unemployment benefits).
- Possible explanations for positive trade/technology effects:
  - Smaller occupational-mix changes in Europe compared to US metropolitan areas over this period—fewer jobs automated or offshored.
  - An added second worker effect: households may add a second worker to make up for lost income of the first; consistent with large increases in female participation observed in Europe.

### VIII. Quantitative decompositions and magnitudes
- Decompositions (based on Table 1, columns 4 and 5) indicate:
  - Significant drags on participation from aging and cyclical factors, and from time-invariant country characteristics.
  - The most important positive contribution stems from changes in the share of the population with tertiary education.
    - A striking 12 percentage point increase in the tertiary education share on average would act to increase labor force participation by 3.3–3.6 percentage points, thereby more than offsetting downward pressures from aging and cyclical factors.
  - Overall quantitative contribution of trade and technology on participation was relatively small on average, but for some European regions effects are substantial.

### IX. Regional heterogeneity and subgroup patterns
- Regions with high initial exposures (above the 75th percentile) experienced subsequent increases in participation; examples include:
  - High routinization exposure: regions in France, Italy, Hungary.
  - High offshoring exposure: parts of the Czech Republic, Hungary.
  - Economically diverse urban regions: London, Paris, Catalonia, Madrid.
  - Some 'old industrial regions' in advanced Europe: parts of Northern Italy, Sweden, Switzerland.
- Regions with low initial exposures (below the 25th percentile) experienced declines in participation; examples include:
  - Regions in Poland, Romania, more agricultural regions.
  - Sparsely populated, agricultural regions: parts of Finland, Iceland, Norway, Romania, Scotland.
- Gender patterns:
  - Female participation increased more in regions initially more exposed to automation and offshoring; gender-specific regressions confirm large positive effects of exposures to routinization and offshoring on women’s participation.

### X. Policy implications and conclusions
- Spatial heterogeneity:
  - Striking within-country differences in participation evolution imply a spatial dimension to economic vulnerability: short- and medium-term costs from shocks are concentrated in particular places as well as sectors and occupations.
  - Findings call for explicit recognition of spatial heterogeneity in policy design.
- Research implications:
  - Further work using individual-level data is needed to examine the second earner effect in greater detail.
- Future risks:
  - Routinization and offshoring "may ... weigh on participation rates and challenge the sustainability of social insurance systems in the future as the forces of automation become stronger, especially in advanced Europe, and labor costs rise in emerging Europe, making these regions less attractive as destinations for offshoring."

### XI. Appendix — Figure 1 and Table 1 highlights
- Appendix Figure 1 notes:
  - Sources: Eurostat; and authors' calculations.
  - Employment rate, unemployment rate, and inactivity rate are defined as total employment, total unemployment, and total inactive population as a percentage of total population, respectively.
  - Horizontal axis tick labels: -1.5, -1, -0.5, 0, 0.5, 1, 1.5
- Regions displayed in Appendix Figure 1 (selected panels list orders of regions; panels include major cities and regions such as Berlin, Hamburg, Budapest, Bucharest, Warsaw, London, Munich, Rome, Madrid, Marseille, Brussels, Seville, Bratislava, Prague, Lyon, Amsterdam, Milan, Birmingham, Manchester, Torino, Barcelona, Paris, Stockholm, Vienna, Athens, Lisbon, Oslo).
- Appendix Table 1: mapping of country codes, region codes, region names, and NUTS level (selected entries shown for Austria, Belgium, Bulgaria, Switzerland, Czech Republic, Germany, Spain, Finland, France, and others).

*Source: IMF Working Paper wp18165 (excerpts provided).*

### References ________________________________________________________________21

### References (wp18165)

### I. Introduction — demographic pressures and participation trends
- Dependency ratios are projected to reach "a striking 50 percent in more than three-quarters of countries by 2050."
- Median country statistics:
  - Population growth: "the median European economy’s population grew at only one fourth of a percent in 2016" and "more then half of countries will see negative growth rates by 2025."
  - Life expectancy: "the median country’s life expectancy increased by more than 12 years over the course of the past five and a half decades."
  - Healthy life expectancy: increased "by 3.6 years for men and by 2.8 years for women in the median European country between 1990 and 2015."
- Aggregate labor force participation patterns:
  - Participation rates increased in the majority of European countries and exhibited convergence, "especially from the bottom."
  - Contrast with the United States: "stark decline in participation observed in the United States" despite milder demographic pressures.
- Research aims:
  - Examine evolution and drivers of labor force participation rates in Europe, including heterogeneity across regions.
  - Tease out roles of aging, educational gains, cyclical conditions, and novel regional exposures to routinization and offshoring.
  - Construct employment-weighted regional measures of routinizability and offshorability following Das and Hilgenstock (2018) and related methodologies to allow comparability with U.S. analysis (Hilgenstock and Koczan 2018).

### II. Literature — policy, cohort, and technology effects
- Policy and institutional drivers of female participation:
  - Part-time incentives, maternity leave, childcare provision, paid parental leave, and family-friendly policies are emphasized (Genre and others 2010; Thévenon 2013; Christiansen and others 2016; Miani and Hoorens 2014).
  - Cipollone and others (2013): labor market reforms can account for "almost 25 percent" of the increase in participation of young women over two decades, and "more than 30 percent" for highly educated women.
- Cohort and lifecycle effects:
  - Cohort effects can account for a substantial part of recent increases in women’s participation (Balleer and others 2014; Euwals and others 2007).
- Technology and trade:
  - Gregory, Salomons, and Zierahn (2016): in European regions, direct employment-reducing effects of automation on middle-skill employment have been offset by compensatory effects through product demand and local demand spillovers.
  - Graetz and Michaels (2015): industry-level adoption of industrial robots in EU countries "raised labor productivity, increased value-added, augmented worker wages, had no measurable effect on overall labor hours, and modestly shifted employment in favor of high-skilled workers."
- Contribution of this paper:
  - (1) Use novel measures of initial local exposures to routinization and offshoring to study differential regional effects and comparisons to the United States.
  - (2) Analyze participation at the subnational (roughly 250-region) level to exploit within-country heterogeneity.

### III. Data and empirical strategy
- Geographic and temporal scope:
  - Sample: roughly 250 NUTS regions (NUTS-2 preferred; NUTS-1 or country-level for some countries) with labor market indicators available from 2000.
  - Balanced panel includes 237 non-overlapping NUTS regions from 28 European economies for some statistics.
- Key variables and proxies:
  - Aging: old-age dependency ratios defined as "the population age 65 and older as a percentage of the population age 15 to 64."
  - Cyclical conditions: real GDP growth rates (Eurostat).
  - Education: share of the population enrolled in tertiary education (Eurostat).
  - Routinization and offshoring exposures: employment-weighted averages of occupational scores:
    - Routinability: scores from Autor and Dorn (2013) based on "routine task intensity" for 330 occupations using the DOT.
    - Offshorability: measures based on Goos, Manning, and Salomons (2014) converting Blinder and Krueger (2014) assessments into ISCO categories.
  - Employment-by-education data from the European Union Labour Force Survey (EULFS).
- Regression approach:
  - Regressions examine regional participation changes as functions of aging, cyclical conditions, education, and initial exposures to routinization/offshoring; controls include education and other regional characteristics.

### IV. Stylized facts and empirical findings
- Cross-regional heterogeneity:
  - Participation levels in 2016 range from "the high 30s" (regions in Southern Italy and Romania) to "the high 60s and above" (regions in Germany, the Netherlands, Sweden, and the United Kingdom).
  - Changes between 2000–2016 range from "6–9 percentage point increases (in parts of Spain and Germany) to 15–20 percentage point drops (in parts of Romania)."
- Convergence:
  - Labor force participation rates have converged: "Regions with relatively higher participation rates in 2000 experienced larger declines (or smaller increases) ... while regions with relatively lower participation rates in 2000 experienced larger increases or smaller decreases."
  - Evidence of within-country convergence indicated by decreasing standard deviations of regions within countries over time.
- Urban-rural pattern:
  - Initial population density is positively correlated with changes in participation: "Participation rates increased more (or declined less) in urban regions, while they declined more (or increased less) in rural regions."
  - Correlation is modest: regression fit reported R² = 0.0306 in the population density relation (Figure 7).
- Aggregate regional outcomes, 2000–2016:
  - "Labor force participation rates increased in a majority (65 percent) of European regions over the period 2000–2016, with a median increase of 3 percentage points;"
  - "the remaining 35 percent of regions exhibited a median decline of 1.7 percentage points."
- Pre- and post-crisis dynamics:
  - After the Global Financial Crisis and the European sovereign debt crisis (2008–2016), "increases appear to be somewhat smaller, and declines more widespread" than in 2000–2008, with substantial regional heterogeneity:
    - About "a third of regions experienced a post-crisis decrease following a pre-crisis increase;"
    - "21 percent saw an increase in the post-crisis period after a decline in the pre-crisis period;"
    - "35 percent of regions saw increases in both periods and 11 percent saw decreases in both."
- Drivers decomposed (empirical results summary):
  - Aging and cyclical conditions exert downward pressure on participation rates.
  - Educational attainment produced a "striking rise" that more than offset drags from aging and cyclical conditions.
  - Routinization and offshoring:
    - Contrary to findings for the United States, "the forces of trade and technology did not result in detachment from the workforce in European regions on average, once labor markets had time to adjust."
    - Regions with high initial exposures to routinization and offshoring experienced "so-far larger increases in participation."
    - Types of regions with larger participation increases despite high exposures include:
      - Economically diverse urban regions (examples: London, Paris, Madrid, Catalonia).
      - Some 'old industrial regions' (examples: parts of Northern Italy, Sweden, Switzerland) where automation may have progressed more slowly or displaced workers found alternative employment.
      - Regions in Emerging Europe (parts of the Czech Republic and Hungary) that benefited from offshoring as receiving destinations integrated into regional supply chains.
    - Gender dimension: "This could be driven by an added second worker effect: higher initial exposures to routinization and offshoring are associated with larger subsequent increases in women’s participation."
- Future risks:
  - Routinization and offshoring "may ... weigh on participation rates and challenge the sustainability of social insurance systems in the future as the forces of automation become stronger, especially in advanced Europe, and labor costs rise in emerging Europe, making these regions less attractive as destinations for offshoring."

### V. Structure of the full paper (as provided)
- Section II: literature review.
- Section III: data description.
- Section IV: stylized facts (regional heterogeneity, convergence, urban-rural patterns).
- Section V: empirical strategy, regression results, regional contributions of factors.
- Section VI: conclusion.

_Italic: Content extracted from wp18165 - References (IMF working paper PDF)._

### 1. All Regions

### 1. All Regions

### Labor market adjustment over 2000–2016; pre- and post-Global Financial Crisis
- Over the full 2000–2016 period, rising employment was matched almost one-for-one by rising participation, with very little change in unemployment.
- Pre-crisis (2000–2008): flows were characterized by movements from unemployment and inactivity into employment.
- Post-crisis (2008–2016): employment started to decline, matched almost fully by a rise in unemployment, with still small increases in participation.
- Urban-rural differences:
  - Rural regions: larger swings in unemployment—larger decline pre-crisis and much larger increase post-crisis; pre-crisis flows mostly from unemployment to employment; post-crisis sharper employment drop translating into large increases in unemployment and some increases in inactivity.
  - Urban regions: pre-crisis flows mostly from inactivity to employment; post-crisis saw some rise in unemployment but in the context of (small) increases in employment and activity.
- Over 2000–2016: urban regions saw large increases in employment and activity; rural areas saw small employment declines matched by moderate drops in activity.

### Empirical strategy: drivers of changes in participation
- Aim: examine regional drivers of changes in labor force participation between 2000 and 2016, including effects of aging, cyclical factors, education, trade and technology.
- Two novel measures tested: initial local exposures to routinization (푟표푢푡푖푗0) and to offshoring (표푓푓푠ℎ푖푗0).
  - The two measures are highly correlated; most regions are relatively highly exposed to both automation and offshoring or neither.
  - Regional exposures could not be constructed for Germany and the Netherlands; initial values shown at country level and regions from these countries are excluded from the regression analysis.
- Regression specification (cross-sectional regional-level, changes 2000–2016):
  - ∆푦푖푗 = 훽1 ∆푑푒푝 푟푎푡푖표푖푗 + 훽2 푔푟표푤푡ℎ푖푗 + 훽3 ∆푒푑푢푐푖푗 + 훽4 푟표푢푡푖푗0 + 훽5 표푓푓푠ℎ푖푗0 + 훼푗
  - ∆푦푖푗 is the change in labor force participation between 2000 and 2016 in region i country j.
  - Controls: change in old-age dependency ratio (∆푑푒푝 푟푎푡푖표푖푗), average real GDP growth (푔푟표푤푡ℎ푖푗), change in educational attainment (∆푒푑푢푐푖푗).
  - Initial exposures included one at a time due to high correlation.
  - Regressions control for time-invariant country characteristics (훼푗); standard errors adjusted for heteroskedasticity.
  - Baseline results do not include measures of population growth and migration; but results for a smaller subsample are robust to their inclusion.

### Main regression results (Table 1): drivers of change in labor force participation rates
- Dependent variable: change in labor force participation rate (2000–2016).
- Selected coefficient estimates (standard errors in parentheses). Significance: * p < 0.1; ** p < 0.05; *** p < 0.01.
  - Average Real GDP Growth:
    - Column (1): -0.178 (0.240)
    - Column (2): -0.441 (0.275)
    - Column (3): -0.348 (0.277)
  - Change in Old-Age-Dependency Ratio:
    - Column (1): -0.275*** (0.0627)
    - Column (2): -0.288*** (0.0824)
    - Column (3): -0.295*** (0.0844)
  - Change in Postsecondary Share:
    - Column (1): 0.256*** (0.0557)
    - Column (2): 0.308*** (0.0703)
    - Column (3): 0.282*** (0.0716)
  - Initial Exposure to Routinization:
    - Column (4): 6.544*** (1.484)
    - Column (5): 7.058*** (1.428)
  - Initial Exposure to Offshoring:
    - Column (4): 6.770*** (1.686)
    - Column (5): 6.992*** (1.642)
- Observations and goodness of fit:
  - Observations: 185 (col 1), 185 (col 2), 262 (col 3), 177 (col 4), 176 (col 5)
  - R2: 0.698 (col 1), 0.693 (col 2), 0.680 (col 3), 0.751 (col 4), 0.741 (col 5)

### Interpretation of empirical findings
- Significant negative associations:
  - Regions with lower real GDP growth experienced larger declines in labor force participation, though coefficients are generally rather imprecisely estimated.
  - Regions with larger shifts toward population older than 65 experienced larger declines in participation; sensitivity to aging appears higher in Europe than in the United States.
- Significant positive associations:
  - Educational gains (increase in postsecondary/tertiary share) strongly correlated with increases in participation; effect somewhat weaker than in the United States.
  - Contrary to findings for the United States, initial exposures to automation (routinization) and offshoring are positively associated with subsequent increases in participation rates in Europe, robust to controls for population density and country-level policies (active labor market spending, union density, generosity of unemployment benefits).
- Possible explanations for positive trade/technology effects:
  - Smaller occupational-mix changes in Europe compared to US metropolitan areas over this period—fewer jobs automated or offshored.
  - An added second worker effect: households may add a second worker to make up for lost income of the first; consistent with large increases in female participation observed in Europe.

### Quantitative decompositions and magnitudes
- Decompositions based on Table 1, columns 4 and 5 show:
  - Significant drags on participation from aging and cyclical factors, and from time-invariant country characteristics.
  - The most important positive contribution stems from changes in the share of the population with tertiary education.
    - A striking 12 percentage point increase in the tertiary education share on average would act to increase labor force participation by 3.3–3.6 percentage points, thereby more than offsetting downward pressures from aging and cyclical factors.
- Overall quantitative contribution of trade and technology on participation was relatively small on average, but for some European regions effects are substantial.

### Regional heterogeneity and subgroup patterns
- Regions with high initial exposures (above the 75th percentile):
  - High routinization exposure examples: regions in France, Italy, Hungary—experienced subsequent increases in participation.
  - High offshoring exposure examples: parts of the Czech Republic, Hungary—experienced increases in participation as they integrated in regional supply chains.
  - Also includes diverse urban regions (for instance London, Paris, Catalonia, Madrid) and some ‘old industrial regions’ in advanced Europe (Northern Italy, Sweden, Switzerland).
- Regions with low initial exposures (below the 25th percentile):
  - Examples: regions in Poland, Romania, more agricultural regions—experienced declines in participation.
  - Sparsely populated, agricultural regions (parts of Finland, Iceland, Norway, Romania, Scotland) are lagging behind.
- Gender patterns:
  - Correlations suggest female participation increased more in regions initially more exposed to automation and offshoring; gender-specific regressions confirm large positive effects of exposures to routinization and offshoring on women’s participation.
  - The positive trade/technology-participation relationship may be driven by added second worker effect consistent with large increases in female participation.

### Policy implications and conclusions
- Striking within-country differences in participation evolution imply a spatial dimension to economic vulnerability: short- and medium-term costs from shocks are concentrated in particular places as well as sectors and occupations.
- Findings call for explicit recognition of spatial heterogeneity in policy design and for further work using individual-level data to examine the second earner effect in greater detail.

*Source: IMF Working Paper wp18165, "1. All Regions" (excerpts as provided).*

### Appendix Figure 1. Changes in Labor Market Dynamics, Individual Regions

### Appendix Figure 1. Changes in Labor Market Dynamics, Individual Regions

### Note and axes
- Sources: Eurostat; and authors' calculations.
- Note: Employment rate, unemployment rate, and inactivity rate are defined as total employment, total unemployment, and total inactive population as a percentage of total population, respectively.
- Horizontal axis tick labels (as shown): -1.5, -1, -0.5, 0, 0.5, 1, 1.5

### Panel 1. 2000-2016 — Regions displayed (order as in figure)
- Berlin
- Hamburg
- Budapest
- Bucharest
- Warsaw
- London
- Munich
- Rome
- Madrid
- Marseille
- Brussels
- Seville
- Bratislava
- Prague
- Lyon
- Amsterdam
- Milan
- Birmingham
- Manchester
- Torino
- Barcelona
- Paris
- Stockholm
- Vienna
- Athens
- Lisbon
- Oslo

### Panel 2. 2000-2008 — Regions displayed (order as in figure)
- Madrid
- Seville
- Barcelona
- Athens
- Rome
- Amsterdam
- Marseille
- Warsaw
- Milan
- Brussels
- Bratislava
- Bucharest
- Torino
- Paris
- Hamburg
- Budapest
- Lyon
- Prague
- Munich
- Oslo
- Berlin
- Birmingham
- London
- Manchester
- Stockholm
- Lisbon
- Vienna

### Panel 3. 2008-2016 — Regions displayed (order as in figure)
- Berlin
- Budapest
- Hamburg
- Munich
- London
- Bucharest
- Manchester
- Birmingham
- Warsaw
- Prague
- Brussels
- Stockholm
- Lyon
- Vienna
- Bratislava
- Marseille
- Rome
- Torino
- Milan
- Paris
- Lisbon
- Amsterdam
- Barcelona
- Seville
- Madrid
- Oslo
- Athens

### Appendix Table 1. NUTS Regions (selected entries and structure)
- The appendix provides a tabulated mapping of country codes, region codes, region names, and NUTS level. Examples from the table as presented:
  - Austria (AUT)
    - AT 1 Ostösterreich 1
    - AT 11 Burgenland (AT) 2
    - AT 12 Niederösterreich 2
    - AT 13 Wien 2
    - AT 2 Südostereich 1
    - AT 21 Kärnten 2
    - AT 22 Steiermark 2
    - AT 3 Westösterreich 1
    - AT 31 Oberösterreich 2
    - AT 32 Salzburg 2
    - AT 33 Tirol 2
    - AT 34 Vorarlberg 2
  - Belgium (BEL)
    - BE1 Région de Bruxelles-Capitale 1
    - BE2 Vlaams Gewest 1
    - BE21 Prov. Antwerpen 2
    - BE22 Prov. Limburg (BE) 2
    - BE23 Prov. Oost-Vlaanderen 2
    - BE24 Prov. Vlaams-Brabant 2
    - BE25 Prov. West-Vlaanderen 2
    - BE3 Région wallonne 1
    - BE31 Prov. Brabant Wallon 2
    - BE32 Prov. Hainaut 2
    - BE33 Prov. Liège 2
    - BE34 Prov. Luxembourg (BE) 2
    - BE35 Prov. Namur 2
  - Bulgaria (BGD)
    - BG3 Severna i yugoiztochna Bulgaria 1
    - BG31 Severozapaden 2
    - BG32 Severen tsentralen 2
    - BG33 Severoiztochen 2
    - BG34 Yugoiztochen 2
    - BG4 Yugozapadna i yuzhna tsentralna Bulgaria 1
    - BG41 Yugozapaden 2
    - BG42 Yuzhen tsentralen 2
  - Switzerland (CHE)
    - CH1 Région lémanique 2
    - CH2 Espace Mittelland 2
    - CH3 Nordwest schweiz 2
    - CH4 Zürich 2
    - CH5 Ost schweiz 2
    - CH6 Zentralschweiz 2
    - CH7 Ticino 2
  - Czech Republic (CZE)
    - CZ1 Praha 2
    - CZ2 Střední Čechy 2
    - CZ3 Jihozápad 2
    - CZ4 Severozápad 2
    - CZ5 Severovýchod 2
    - CZ6 Jihovýchod 2
    - CZ7 Střední Morava 2
    - CZ8 Moravskoslezsko 2
  - Germany (DEU) (selected entries)
    - DE1 Baden-Württemberg 1
    - DE11 Stuttgart 2
    - DE12 Karlsruhe 2
    - DE13 Freiburg 2
    - DE14 Tübingen 2
    - DE2 Bayern 1
    - DE21 Oberbayern 2
    - DE22 Niederbayern 2
    - DE23 Oberpfalz 2
    - DE24 Oberfranken 2
    - DE25 Mittelfranken 2
    - DE26 Unterfranken 2
    - DE27 Schwaben 2
  - Spain (ESP) (selected entries)
    - ES1 Noroeste (ES) 1
    - ES11 Galicia 2
    - ES12 Principado de Asturias 2
    - ES13 Cantabria 2
    - ES2 Noreste (ES) 1
    - ES21 País Vasco 2
    - ES22 Comunidad Foral de Navarra 2
    - ES23 La Rioja 2
    - ES24 Aragón 2
    - ES3 Comunidad de Madrid 1
    - ES4 Centro (ES) 1
    - ES41 Castilla y León 2
    - ES42 Castilla-la Mancha 2
    - ES43 Extremadura 2
    - ES5 Este (ES) 1
    - ES51 Cataluña 2
    - ES52 Comunidad Valenciana 2
    - ES53 Illes Balears 2
    - ES6 Sur (ES) 1
    - ES61 Andalucía 2
    - ES62 Región de Murcia 2
    - ES63 Ciudad Autónoma de Ceuta (ES) 2
    - ES64 Ciudad Autónoma de Melilla (ES) 2
    - ES7 Canarias (ES) 1
  - Finland (FIN) (selected entries)
    - FI1 Manner-Suomi 1
    - FI13 Pohjois- ja Itä-Suomi 2
    - FI18 Etelä-Suomi 2
    - FI181 Helsinki-Uusimaa 3
    - FI19 Länsi-Suomi 2
    - FI2 Åland 1
  - France (FRA) (selected entries)
    - FR1 Île de France 2
    - FR2 Bassin Parisien 1
    - FR21 Champagne-Ardenne 2
    - FR22 Picardie 2
    - FR23 Haute-Normandie 2
    - FR24 Centre (FR) 2
    - FR25 Basse-Normandie 2
    - FR26 Bourgogne 2
    - FR3 Nord - Pas-de-Calais 2
    - FR4 Est (FR) 1
    - FR41 Lorraine 2
    - FR42 Alsace 2
    - FR43 Franche-Comté 2
    - FR5 Ouest (FR) 1
    - FR51 Pays de la Loire 2
    - FR52 Bretagne 2
    - FR53 Poitou-Charentes 2
    - FR6 Sud-Ouest (FR) 1
    - FR61 Aquitaine 2
    - FR62 Midi-Pyrénées 2
    - FR63 Limousin 2
    - FR7 Centre-Est (FR) 1
    - FR71 Rhône-Alpes 2
    - FR72 Auvergne 2
    - FR8 Méditerranée 1
    - FR81 Languedoc-Roussillon 2
    - FR82 Provence-Alpes-Côte d'Azur 2
    - FR83 Corse 2
    - FR9 Départements d'outre-mer 1
    - FR91 Guadeloupe 2
    - FR92 Martinique 2
    - FR93 Guyane 2
    - FR94 La Réunion 2
    - FRA50 Mayotte 3
- The table continues with entries for Greece (GRC), Croatia (HRV), Hungary (HUN), Ireland (IRL), Italy (ITA), Netherlands (NLD), Norway (NOR), Poland (POL), Portugal (PRT), Romania (ROU), Sweden (SWE), Slovenia (SVN), Slovak Republic (SVK), United Kingdom (GBR), and other subnational region codes as printed in the appendix.

*Source: Appendix Figure 1 and Appendix Table 1, as provided in the content unit.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18165.pdf_