## 1. Decomposition of GVA Growth by Labor Input and Sector

## Source details

**Canonical URL:** [1. Decomposition of GVA Growth by Labor Input and Sector](https://www.imf.org/-/media/files/publications/cr/2021/english/1estea2021002.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2021/english/1estea2021002.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2021/english/1estea2021002.pdf.json)

---

### A. Introduction
- The COVID-19 crisis poses risks of scarring to the labor market: job finding is lower and job separation is higher in recessions; switches in occupations after unemployment spells inflict earnings penalties.
- The paper studies vulnerable groups and policies to mitigate scarring by analyzing Estonia’s labor market adjustment from aggregate, sectoral, and social perspectives, and comparing with the Global Financial Crisis (GFC) and other Baltic and EU countries.

### B. Aggregate behavior — key findings
- GDP per capita and unemployment:
  - GDP per capita growth at the bottom in 2020Q2 was -7.3 percent in Estonia versus -14 percent for the EU average in 2020Q2 and -17.2 percent during the GFC.
  - Estonia’s unemployment rate increased from 4 percent in March to 8 percent in June 2020; the EU average increased from 6.5 to 7.6 percent in the same period.
  - Although Estonia’s unemployment rate remained below the GFC peak (19 percent), the increase in the percentage was nearly four times larger than the EU average.
- Participation and employment rates:
  - Participation rate (active persons over population above 15 years old) dropped from 63.7 in 2020Q1 to 62.6 percent in 2020Q2.
  - Employment rate (employed persons over population above 15 years old) dropped from 60.5 in 2020Q1 to 58.1 in 2020Q2; the EU average declined from 53.1 to 52.2 during the same period.
  - Estonia’s employment rate decline was 1.5 percentage points higher relative to the EU average.
  - Estonia’s employment rate recovered faster than the EU average by 2020Q4.

### C. Sectoral analysis — methodology and interpretation
- Decomposition components (notation preserved in source):
  - GVA per capita growth decomposed into sectoral contributions from:
    - growth of productivity per hours worked,
    - growth of hours worked per employment,
    - change in sectoral labor share (structural change),
    - growth of employment rate.
  - Notation and variables preserved from source: V_{s,t} (GVA), H_{s,t} (hours worked), N_{s,t} (number of workers), V_t (aggregate GVA), POP_t (population). Time-dependent weights ω_{s,VAR} defined in the source text.
- Interpretation:
  - Structural change contribution depends on whether sector productivity per worker is higher or lower than aggregate productivity; an increase in labor share of higher-productivity sectors yields positive structural change.

### D. Sectoral analysis — empirical findings
- Aggregate vs EU:
  - The contribution of employment to GVA per capita growth in 2020 was bigger in Estonia than the EU average.
- Timing and adjustment channels:
  - 2020H1: both Estonia and EU27 adjusted mainly by reducing hours worked per employment, with substantial employment loss already in 2020Q2.
  - 2020H2: Estonia’s drop in GVA per capita growth was mainly driven by lower employment, contrasting with the EU average.
  - Hourly productivity in Estonia increased in all quarters in 2020 (possible explanations: compositional effects from disproportionate lay-offs of low-productivity workers; worker adaptation and learning).
- Sector heterogeneity and patterns:
  - 2020Q1: the wholesale and retail trade, transport, accommodation and food services (labeled G-I whol sector) exhibited signs of slowdown.
  - 2020Q2: many sectors reduced hours per employment; G-I whol and M_N prof sectors showed employment loss.
  - 2020H2: after expiration of emergency/support measures, most sectors adjusted through reduced employment.
  - Productivity increases in G-I whol in 2020Q2 and end-2020 likely reflect compositional effects and worker learning.
  - 2020H2 recorded negative structural change numbers because employment declined in high-productivity sectors.

### E. Decomposition of GVA by income recipients — findings
- Decomposition approach:
  - GVA growth decomposed into contribution from compensation of employees (workers) and the remainder (labeled business, mostly corporate profits and self-employed).
  - Wage subsidies are included in GVA and in compensation of employees where paid to workers.
- Aggregate pass-through:
  - Pass-through from output decline to compensation of employees was mild in aggregate but heterogeneous across sectors.
  - During the largest decline in 2020Q2, businesses absorbed nearly 75 percent of the shock.
  - Pre-COVID-19, GVA growth was mostly driven by compensation of employees; during the crisis most shock absorbed by businesses, contrasting with the GFC.
- Sectoral heterogeneity in pass-through:
  - G-I whol sector: more than half of the shock was absorbed by workers after 2020Q2.
  - J info and O–Q publ sectors (includes health services): exhibited strong growth in compensation of employees even when businesses recorded negative growth—indicating increased demand/shortage in ICT and health skills.

### F. Micro-data on job creation and destruction — findings
- Data and measurement:
  - Job creation and destruction measured from quarterly employment data (Estonia Tax and Customs Board) using firm-level changes in employment (formulas preserved in source).
  - Caveat: sectoral data (Eurostat) and micro-data (Tax and Customs Board) use different methodologies (e.g., full-time employee approach vs. Tax and Customs Board approach).
- Magnitudes and timing:
  - From 2020Q1 to 2020Q2, job destruction increased by nearly 15,000 and job creation decreased by around 5,000.
  - Job destruction decelerated in 2020Q3 with less stringent restrictions and resumed in 2020Q4 alongside worsening health conditions; job creation gradually recovered in 2020H2.
- Sectoral drivers:
  - Aggregate job destruction mirrored hospitality and high-touch sectors: accommodation and food services, administrative support, wholesale and retail trade, arts and entertainment.
  - 2020Q2 job destruction increase also driven by manufacturing, construction, and transportation.
- Sectors with net job creation:
  - Health services and ICT consistently recorded higher job creation than destruction before and during COVID-19, but magnitudes were insufficient to absorb losses from impacted sectors and re-entry constrained by specialized skills.

### G. Regional patterns
- Harju county (Tallinn) recorded the highest job destruction.
- Counties with prominent tourism/service sectors (Harju, Parnu, Saare) recorded high job destruction.
- Significant job destruction in Ida-Viru counties may reflect both epidemiological conditions and pre-crisis decline in the oil-shale industry.

### H. Gender perspective, age, and skills
- Gender unemployment timing:
  - Female unemployment increased faster initially in 2020Q2.
  - Male unemployment increased in 2020Q3.
  - Aggregate: "there is, however, no significant difference between males and females in aggregate."
- Gender pay gap:
  - Long-term: "the gender pay gap has declined in all sectors, although it remains among the highest in the EU."
- Age outcomes and scarring:
  - Younger cohorts were impacted disproportionately: "The unemployment rate for workers younger than 25 years old is more volatile due to the smaller population, higher on average historically, and increased more during the COVID-19 crisis as was the case in GFC."
  - Concern of scarring: "if young people lose jobs in their early careers, the impact tends to last long (e.g. Mroz and Savage, 2006)."
  - Granular 2020 findings:
    - Female unemployment was higher for the 30–35 years old group.
    - Male unemployment was higher for age groups 55–59 and 65–69 years old.
    - Participation rate declined for female 30–34 years old and male 65–60 years old groups.
- Skills and educational attainment:
  - "Although the unemployment rate increased for all the groups in 2020Q2, the group with the lowest educational attainment exhibited the largest increase."
  - "The group with the highest educational attainment exhibited a decline in unemployment in 2020Q4."
  - Risk: "This behavior also raises the concern of scarring risks related to a slower recovery (and associated lower pay) of the group with low educational attainment, which could expand inequality."

### I. Government wage support and immediate impacts
- Design and timeline:
  - Wage compensation scheme initially approved to cover March–May 2020; extended to end-June 2020 under more conservative qualification criteria.
  - Duration employer must keep employees increased from one to three months.
  - Scheme was further deployed and targeted to regional epicenters at end-2020; re-introduced as part of March 2021 supplementary budget and used until end-May 2021.
- Revisions to qualification and rates (exact text from source):
  - "The turnover decline condition was revised up from the initial 30 to 50 percent decline."
  - "the employer was required to have reduced the staff or workload of workers by one-half (versus initially 30 percent)."
  - "The maximum compensation and replacement rate were reduced, respectively from EUR 1,000 to EUR 800 and from 70 to 50 percent."
- Coverage and targeting:
  - "The jobs benefiting from the government employment support measures reached 20 percent of the total in April 2020."
  - In April 2020, sectoral reach:
    - Accommodation and food services: nearly 80 percent of jobs benefited.
    - Arts and entertainment: nearly 40 percent of jobs benefited.
    - Financial services: lower number of jobs benefited (sector hit relatively less).
  - Comparison: Estonia supported more jobs during April 2020 than other Baltic countries and the average of 22 EU countries.
- Evidence on effectiveness and distributional impacts (reported findings):
  - Koppel and Laurimae (2021) reported:
    - The wage subsidy helped retain 65,000 jobs that would have otherwise been lost due to COVID-19.
    - Mitigated the impact on relative poverty: "the poverty rate of the population only increasing by 0.3 percentage points (ppts) compared to a 4 ppts increase in the counterfactual."
    - Reduced inequality relative to baseline, as lower-income workers benefited the most from wage subsidies.
- Adjustment channel and timing:
  - The use of government measures was associated with adjustment through the number of hours per worker in 2020H1; in 2020H2 the labor market started to adjust through the employment channel.
  - In the EU average, where wage support remained active in 2020H2, adjustment through hours worked continued and played a greater role.

### J. Policy implications and recommendations (as stated in source)
- Short and medium-term:
  - Wage support and social safety nets helped mitigate distributional impacts during the crisis.
  - Retention policies targeted to vulnerable groups should be maintained to ease unemployment dynamics until recovery is entrenched.
- Targeting and social protection:
  - Continue job retention measures until the pandemic abates markedly and guide them by pandemic monitoring and vaccine rollout.
  - Target support increasingly to identified vulnerable groups: examples: "increasing wage subsidies for youth or lower-skilled workers—to reduce the unequal impact of the shock."
  - "Social protection should remain easily accessible to vulnerable groups until job prospects are restored."
  - Advance agenda to enhance eligibility and flexibility of social safety nets, including coverage and cyclicality-dependency of unemployment benefits.
- Reallocation and activation as recovery proceeds:
  - As the economy recovers, more active use of reallocation policies is warranted to mitigate scarring and facilitate worker transition to expanding sectors.
  - Facilitate cross-sectoral reallocation to support shift to digital and greener economy.
  - Strengthen investments in human capital to boost productivity and ease occupational switching.
  - Use targeted active labor market policies (ALMPs), outreach and training measures for the lower-skilled, and hiring incentives where cost-effective.

### K. Key summary points (conclusion of the paper)
- Estonia navigated the crisis with a smaller drop in GDP per capita but a larger increase in the unemployment rate than the EU average.
- The channel of adjustment shifted from hours per worker to employment since 2020Q2; pass-through from output decline to compensation was limited and heterogeneous across sectors.
- Employment loss was salient in some sectors, including accommodation and food services; young cohorts and less educated suffered most, raising concerns of expanding inequality and scarring risks.
- Wage support and social safety nets helped mitigate the distributional impact; retention policies targeted to vulnerable groups should be maintained until recovery is entrenched.
- As the economy recovers, more active use of reallocation policies would facilitate reallocation.

*Source: 1estea2021002 - 1. Decomposition of GVA Growth by Labor Input and Sector (June 30, 2021).*

### 1. Decomposition of GVA Growth by Labor Input and Sector ___________________________ 5

### 1. Decomposition of GVA Growth by Labor Input and Sector ___________________________ 5

### A. Introduction
- The COVID-19 crisis poses risks of scarring to the labor market: job finding is lower and job separation is higher in recessions; switches in occupations after unemployment spells inflict earnings penalties.
- The paper studies vulnerable groups and policies to mitigate scarring by analyzing Estonia’s labor market adjustment from aggregate, sectoral, and social perspectives, and comparing with the Global Financial Crisis (GFC) and other Baltic and EU countries.

### B. Aggregate behavior — key findings
- GDP per capita growth at the bottom in 2020Q2 was -7.3 percent in Estonia versus -14 percent for the EU average in 2020Q2 and -17.2 percent during the GFC.
- Estonia’s unemployment rate increased from 4 percent in March to 8 percent in June 2020; the EU average increased from 6.5 to 7.6 percent in the same period.
- Although Estonia’s unemployment rate remained below the GFC peak (19 percent), the increase in the percentage was nearly four times larger than the EU average.
- Participation rate (active persons over population above 15 years old) dropped from 63.7 in 2020Q1 to 62.6 percent in 2020Q2.
- Employment rate (employed persons over population above 15 years old) dropped from 60.5 in 2020Q1 to 58.1 in 2020Q2; the EU average declined from 53.1 to 52.2 during the same period. Estonia’s employment rate decline was 1.5 percentage points higher relative to the EU average.
- Estonia’s employment rate recovered faster than the EU average by 2020Q4.

### C. Sectoral analysis — methodology and interpretation
- GVA per capita growth is decomposed into sectoral contributions from:
  - growth of productivity per hours worked,
  - growth of hours worked per employment,
  - change in sectoral labor share (structural change),
  - growth of employment rate.
- Notation and variables preserved from source: V_{s,t} (GVA), H_{s,t} (hours worked), N_{s,t} (number of workers), V_t (aggregate GVA), POP_t (population). Time-dependent weights ω_{s,VAR} defined in the source text.
- Interpretation: structural change contribution depends on whether sector productivity per worker is higher or lower than aggregate productivity; an increase in labor share of higher-productivity sectors yields positive structural change.

### C. Sectoral analysis — empirical findings
- In 2020:
  - The contribution of employment to GVA per capita growth in 2020 was bigger in Estonia than the EU average.
  - 2020H1: both Estonia and EU27 adjusted mainly by reducing hours worked per employment, with substantial employment loss already in 2020Q2.
  - 2020H2: Estonia’s drop in GVA per capita growth was mainly driven by lower employment, contrasting with the EU average.
  - Hourly productivity in Estonia increased in all quarters in 2020 (possible explanations: compositional effects from disproportionate lay-offs of low-productivity workers; worker adaptation and learning).
- Sector heterogeneity:
  - In 2020Q1, the wholesale and retail trade, transport, accommodation and food services (labeled G-I whol sector) exhibited signs of slowdown.
  - In 2020Q2, many sectors reduced hours per employment; G-I whol and M_N prof sectors showed employment loss.
  - In 2020H2, after expiration of emergency/support measures, most sectors adjusted through reduced employment.
  - Productivity increases in G-I whol in 2020Q2 and end-2020 likely reflect compositional effects and worker learning.
  - 2020H2 recorded negative structural change numbers because employment declined in high-productivity sectors.

### D. Decomposition of GVA by income recipients — findings
- GVA growth decomposed into contribution from compensation of employees (workers) and the remainder (labeled business, mostly corporate profits and self-employed).
- Wage subsidies are included in GVA and in compensation of employees where paid to workers.
- Aggregate pass-through:
  - Pass-through from output decline to compensation of employees was mild in aggregate but heterogeneous across sectors.
  - During the largest decline in 2020Q2, businesses absorbed nearly 75 percent of the shock.
  - Pre-COVID-19, GVA growth was mostly driven by compensation of employees; during the crisis most shock absorbed by businesses, contrasting with the GFC.
- Sectoral heterogeneity in pass-through:
  - G-I whol sector: more than half of the shock was absorbed by workers after 2020Q2.
  - J info and O–Q publ sectors (includes health services): exhibited strong growth in compensation of employees even when businesses recorded negative growth—indicating increased demand/shortage in ICT and health skills.

### E. Micro-data on job creation and destruction — findings
- Job creation and destruction measured from quarterly employment data (Estonia Tax and Customs Board) using firm-level changes in employment (formulas preserved in source).
- Caveat: sectoral data (Eurostat) and micro-data (Tax and Customs Board) use different methodologies (e.g., full-time employee approach vs. Tax and Customs Board approach).
- COVID-19 impact:
  - From 2020Q1 to 2020Q2, job destruction increased by nearly 15,000 and job creation decreased by around 5,000.
  - Job destruction decelerated in 2020Q3 with less stringent restrictions and resumed in 2020Q4 alongside worsening health conditions; job creation gradually recovered in 2020H2.
- Sectoral drivers of job destruction:
  - Aggregate job destruction mirrored hospitality and high-touch sectors: accommodation and food services, administrative support, wholesale and retail trade, arts and entertainment.
  - 2020Q2 job destruction increase also driven by manufacturing, construction, and transportation.
- Sectors with net job creation:
  - Health services and ICT consistently recorded higher job creation than destruction before and during COVID-19, but magnitudes were insufficient to absorb losses from impacted sectors and re-entry constrained by specialized skills.

### F. Regional patterns
- Harju county (Tallinn) recorded the highest job destruction.
- Counties with prominent tourism/service sectors (Harju, Parnu, Saare) recorded high job destruction.
- Significant job destruction in Ida-Viru counties may reflect both epidemiological conditions and pre-crisis decline in the oil-shale industry.

### G. Policy implications and recommendations (as stated in source)
- Wage support and social safety nets helped mitigate distributional impacts during the crisis.
- Retention policies targeted to vulnerable groups should be maintained to ease unemployment dynamics until recovery is entrenched.
- As the economy recovers, more active use of reallocation policies is warranted to mitigate scarring and facilitate worker transition to expanding sectors.

*Source: 1estea2021002 - 1. Decomposition of GVA Growth by Labor Input and Sector (June 30, 2021).*

### 17.      From the gender perspective, male

### 17.      From the gender perspective, male

### Gender outcomes
- Male and female unemployment increased at different timing in 2020:
  - Female unemployment increased faster initially in 2020Q2.
  - Male unemployment increased in 2020Q3.
- Aggregate: "there is, however, no significant difference between males and females in aggregate."
- Long-term: "the gender pay gap has declined in all sectors, although it remains among the highest in the EU."
- Estonia—Gender wage gap progress since 2010 (figures shown in source): GPG in 2019 (Percent) and Change in GPG since 2010 (Percent) presented by sector and country in the source.

### Age outcomes
- Younger cohorts were impacted disproportionately:
  - "The unemployment rate for workers younger than 25 years old is more volatile due to the smaller population, higher on average historically, and increased more during the COVID-19 crisis as was the case in GFC."
  - Concern of scarring: "if young people lose jobs in their early careers, the impact tends to last long (e.g. Mroz and Savage, 2006)."
- Granular gender-age findings for 2020:
  - Female unemployment was higher for the 30–35 years old group.
  - Male unemployment was higher for age groups 55–59 and 65–69 years old.
  - Participation rate declined for female 30–34 years old and male 65–60 years old groups.

### Skills and educational attainment
- Unemployment by education level:
  - "Although the unemployment rate increased for all the groups in 2020Q2, the group with the lowest educational attainment exhibited the largest increase."
  - "The group with the highest educational attainment exhibited a decline in unemployment in 2020Q4."
- Risk: "This behavior also raises the concern of scarring risks related to a slower recovery (and associated lower pay) of the group with low educational attainment, which could expand inequality."

### Government wage support and immediate impacts
- Design and timeline:
  - Wage compensation scheme initially approved to cover March–May 2020; extended to end-June 2020 under more conservative qualification criteria.
  - Duration employer must keep employees increased from one to three months.
  - Scheme was further deployed and targeted to regional epicenters at end-2020; re-introduced as part of March 2021 supplementary budget and used until end-May 2021.
- Revisions to qualification and rates (exact text from source):
  - "The turnover decline condition was revised up from the initial 30 to 50 percent decline."
  - "the employer was required to have reduced the staff or workload of workers by one-half (versus initially 30 percent)."
  - "The maximum compensation and replacement rate were reduced, respectively from EUR 1,000 to EUR 800 and from 70 to 50 percent."
- Coverage and targeting:
  - "The jobs benefiting from the government employment support measures reached 20 percent of the total in April 2020."
  - In April 2020, sectoral reach:
    - Accommodation and food services: nearly 80 percent of jobs benefited.
    - Arts and entertainment: nearly 40 percent of jobs benefited.
    - Financial services: lower number of jobs benefited (sector hit relatively less).
  - Comparison: Estonia supported more jobs during April 2020 than other Baltic countries and the average of 22 EU countries.

### Evidence on effectiveness and distributional impacts
- Findings from Koppel and Laurimae (2021) as reported:
  - The wage subsidy helped retain 65,000 jobs that would have otherwise been lost due to COVID-19.
  - Mitigated the impact on relative poverty: "the poverty rate of the population only increasing by 0.3 percentage points (ppts) compared to a 4 ppts increase in the counterfactual."
  - Reduced inequality relative to baseline, as lower-income workers benefited the most from wage subsidies.
- Adjustment channel:
  - The use of government measures was associated with adjustment through the number of hours per worker in 2020H1; in 2020H2 the labor market started to adjust through the employment channel.
  - In the EU average, where wage support remained active in 2020H2, adjustment through hours worked continued and played a greater role.

### Policy recommendations to mitigate labor market scarring risks
- Continue job retention measures until the pandemic abates markedly and guide them by pandemic monitoring and vaccine rollout.
- Target support increasingly to identified vulnerable groups:
  - Examples: "increasing wage subsidies for youth or lower-skilled workers—to reduce the unequal impact of the shock."
- Social protection:
  - "Social protection should remain easily accessible to vulnerable groups until job prospects are restored."
  - Advance agenda to enhance eligibility and flexibility of social safety nets, including coverage and cyclicality-dependency of unemployment benefits.
- Reallocation policies as recovery proceeds:
  - Facilitate cross-sectoral reallocation to support shift to digital and greener economy.
  - Strengthen investments in human capital to boost productivity and ease occupational switching.
  - Use targeted active labor market policies (ALMPs), outreach and training measures for the lower-skilled, and hiring incentives where cost-effective.

### Key summary points (conclusion of the paper)
- Estonia navigated the crisis with a smaller drop in GDP per capita but a larger increase in the unemployment rate than the EU average.
- The channel of adjustment shifted from hours per worker to employment since 2020Q2; pass-through from output decline to compensation was limited and heterogeneous across sectors.
- Employment loss was salient in some sectors, including accommodation and food services; young cohorts and less educated suffered most, raising concerns of expanding inequality and scarring risks.
- Wage support and social safety nets helped mitigate the distributional impact; retention policies targeted to vulnerable groups should be maintained until recovery is entrenched.
- As the economy recovers, more active use of reallocation policies would facilitate reallocation.

*REPUBLIC OF ESTONIA  INTERNATIONAL MONETARY FUND*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2021/english/1estea2021002.pdf_