## Annex I — COVID-19 and Labor Markets in Colombia

## Source details

**Canonical URL:** [Annex I — COVID-19 and Labor Markets in Colombia](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021235-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021235-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021235-print-pdf.pdf.json)

---

### I. Introduction: scope and main findings
- COVID-19 disrupted labor markets through mandatory lockdowns, restrictions in mobility, and shifts in consumption behavior.
- In Colombia, the pandemic induced "the deepest recession on record, with about a quarter of total employment being temporarily lost at the height of mandatory lockdowns in the spring of 2020."
- Paper documents using GEIH micro-data:
  - the magnitude, structure, and distributional impact of the COVID-19 shock on the labor market in Colombia;
  - main channels of impact;
  - the role of social protection policies; and
  - the informal nature of the subsequent recovery.
- Distributional impacts:
  - Women, the young, and those with low levels of education were the most adversely affected in income and employment.
  - Social policies only partly offset these losses.
- Role of informality:
  - Informal sector had greater employment losses initially, particularly among women.
  - Informal economy did not absorb the adverse employment shock to the degree observed in past recessions, contributing to record total employment losses.
  - Informal jobs rebounded toward pre-COVID employment levels faster than formal jobs.
  - Lack of regulation oversight and rigidities of formal employment appear to have helped the informal sector adapt labor prices and quantities quicker.
- Informality as double-edged:
  - Positive: Increased informality in lockdown-sensitive sectors during the second half of 2020 appears to have boosted adaptability—decreasing the correlation of employment losses with both stringency measures and workplace mobility in later lockdown waves.
  - Negative: Lags in recovery of formal employment and unemployment rates, and increases in informality rates—both overall and within sectors—remain at risk of becoming permanent, with potential medium-term losses in productivity and average incomes.

### II. Data and definitions
- Primary micro-data: Gran Encuesta Integrada de Hogares (GEIH), monthly samples; over fifteen thousand households sampled each month with a population about 60 thousand.
- Key GEIH variables: demographics; employment status (employed, unemployed, out of labor force); employer characteristics (sector, size, self-employment).
- Informality measure: constructed following the Colombian National Statistical Office (DANE) definition.
- Additional data: Google Community Mobility Reports; Oxford Government Response Tracker (OxGRT) Stringency Index; daily reported COVID-19 cases from Colombia’s Ministry of Health and Social Protection.
- Comparative micro-data: Chile (Encuesta Nacional de Empleo) and Peru (Encuesta Nacional de Hogares).

### III. Informality definitions and overlap
- DANE emphasizes precarious working conditions; ILO focuses on participation in social security and tax systems.
- Despite conceptual differences, DANE and ILO definitions present a large overlap in covered workers.
- Both definitions broadly capture greater vulnerability of informal workers to macroeconomic and idiosyncratic shocks.

### IV. Initial shock: March–April 2020
- Timeline and measures:
  - First COVID-19 case confirmed on March 6.
  - State of emergency declared on March 17.
  - First mandatory quarantine began on March 25; school closures remained during 2020 with a gradual hybrid return covering less than 15 percent of the student population starting in February of the following year.
- Labor market impacts:
  - Relative to pre-pandemic employment levels, about a quarter of employment had been temporarily lost by April 2020.
  - Pre-pandemic unemployment rate of 12 percent rose above 20 percent.
  - Labor force participation rate collapsed from 63 percent to a low of 52 percent.
  - Around 7 of every 10 workers who lost their jobs between February and April dropped out of the labor force.
- Mechanisms for exit: discouragement, household duties/childcare, sector shutdowns preventing search, and lockdown restrictions on job search.

### V. Regional comparison (LA5) and sectoral heterogeneity
- Net employment losses in Q2 2020 were higher than in Brazil, Mexico and Chile but lower than Peru.
- Colombian case distinguished by larger movements in the unemployment rate relative to regional peers.
- Sectoral dynamics:
  - Construction and manufacturing: significant losses in first half of 2020 and significant recovery in second half.
  - Education services and hospitality: prolonged disruptions and weaker gains.
  - By end-2020, most of the economy remained below pre-COVID employment except agriculture, trade services, and the public sector.
- Outstanding aggregate job gap at end-2020: 1.4 million.

### VI. Uneven recovery through end-2020 and early-2021
- Labor force participation recovered 10 percentage points from May to December, reaching 62 percent.
- Over two thirds of those who joined the labor force became employed, contributing to a near 7 percentage point decrease in the unemployment rate to 13 percent by end-2020.
- Workers returning employed accounted for over 75 percent of total employment recovery on a year-on-year basis.
- By end-2020, total number of unemployed remained 2-3 percentage points above pre-COVID level.
- New lockdowns in January 2021 caused employment gains to stop, labor force participation to decline, and unemployment to rise again to 17 percent.

### VII. Distributional impacts: employment and income
- Job and labor income losses concentrated among lower-education groups, women, and the young—employment dropped by close to a third among these groups.
- Women: largest initial hit and significantly slower recovery due to childcare and home duties.
- Among those remaining employed, men, the least educated, and older workers experienced greater drops in monthly labor earnings.
- Incomes recovered but remained below pre-COVID levels as of end-2020.
- Poverty and household income distribution:
  - Mass of households reporting no income more than doubled from 2019 to 2020.
  - Poverty rates increased to levels not seen in the last decade.

### VIII. Social protection, transfers, and mitigation
- Over half of poorer households were covered by some form of transfer before COVID-19 (pension transfers, unemployment insurance, conditional and unconditional cash-transfer programs).
- Familias en Acción remained the most important transfer mechanism in coverage and magnitude.
- An expanded VAT-return transfer targeted poorest among Familias en Acción participants.
- Emergency transfers and employment protection measures worth around 1.3 percent of GDP in 2020 included:
  - Wage subsidies equivalent to 40 percent of the minimum wage for formal workers at firms experiencing over 20 percent in revenue losses.
  - Ingreso Solidario: a new unconditional transfer program aimed at 3 million at-risk households not covered under other programs.
- Expansion concentrated among poorer households but also reached households with earnings above the median by design; budgetary and technical constraints limited full coverage, particularly for informal workers.
- Majority of those reporting job loss due to COVID-19 were not enrolled in any government transfer program.
- GEIH self-reports indicate significant portions of transfers were allocated to households not among the poorest deciles, indicating scope for improved targeting.

### IX. Informality as a margin of adjustment: dynamics and magnitudes
- Informal sector acted as an important margin of adjustment due to flexibility in hiring/firing, hours worked, and labor prices.
- Informal employment comprised 51 percent of the aggregate contraction as of July 2020 relative to pre-pandemic levels (pre-COVID baseline informality rate: 48 percent).
- Gender composition and losses:
  - Women comprised about 45 percent of the employed but accounted for 52 percent of total job losses and 58 percent of informal-sector job losses.
  - Women's employment losses were mostly from informal workers; men's losses were more evenly split between formal and informal.
- Recovery contributions:
  - By December 2020, informal sector gains comprised 59 percent of total employment gains (58 percent for men and 59 percent for women).
  - Informality rate increased from 47 to 49 percent for both genders (pre-COVID total 48 percent; 46 percent for men; 50 percent for women).
- Worker-type elasticity:
  - Self-employment within informality was most elastic during downturn and upturn.
- Intensive margin:
  - Among informal workers, average weekly hours decreased by 28 percent by the height of lockdown (Q2 2020), rebounding close to pre-COVID levels by end-2020.
  - Among formal workers, maximum drop in hours was 18 percent.
  - Informal workers experienced severe drops in average monthly income not experienced by formal workers.

### X. Interaction of sectoral composition, informality, and the diminishing effect of lockdowns
- Sectors most affected at the height of the shock tended to have large levels of informal employment (negative correlation between employment losses and pre-COVID informality).
- Highly informal sectors experienced the largest employment losses but also bounced back more quickly; correlation between cumulative job destruction and informality disappears when accounting for recovered jobs.
- Decomposition of informality changes:
  - Decline in share of informal employment during lockdown driven mainly by contraction in high-informality sectors (between-sector effect).
  - Rise in informality in second half of 2020 driven by both recovery of those sectors and increases of informality within sectors (within-sector effect), indicating substitution away from formal job creation.
  - Net increase in informality could imply persistent shifts toward greater informality, potentially reversing pre-pandemic formalization trends.
- Lockdown sensitivity changed over time:
  - First lockdown: steep falls in mobility highly correlated with drop in employment; lockdowns were broad-based.
  - Over time: lockdowns became more differentiated and targeted; more industries allowed partial reopening.
  - Departmental-level correlation between mobility and employment weakened over time; during the second lockdown mobility fell slightly but employment stopped rebounding without falling in all departments.
- Sectoral exposure analysis:
  - Exposed sectors (binary indicator following Alfaro et al. (2020) and Morales et al. (2020)): by April 2020, median unexposed sector median contraction ~20 percent; median contraction in exposed sectors close to 40 percent.
  - Over following months, exposed sectors recovered faster; by November 2020 median gap comparable across groups.
  - Regression evidence (sector-department level, phases):
    - First lockdown: exposed sectors contracted by an extra 9 percent on average (Exposed Sector * First Lockdown = -0.0944***; standard error (0.0358)).
    - Reopening: gap reduces to 5 percent and is not statistically significant.
    - By the third lockdown the differential turns mildly positive but not significant in some specifications.
  - New Cases per 1000 residents and various constants and observations reported across specifications (examples: constants 9.552***, 9.542***, 9.537***, 8.732***, 8.681***, 8.660***; observations 3,828; 6,226; 7,659; 2,613; 4,804; 6,120).
- Mobility–employment regressions and informality:
  - Mobility * Unexposed Sector coefficients: 0.00280*** and 0.0178*** (standard errors (0.000566) and (0.00164)).
  - Mobility * Exposed Sector coefficients: 0.00539*** and 0.0152*** (standard errors (0.000529) and (0.00101)).
  - Phase interaction examples: Feb '20 * Mobility = 0.0104*** and 0.00698**; First Lockdown * Mobility = 0.00373*** and 0.0179***; First Reopening * Mobility = 0.00529*** and 0.0117***; Second Lockdown * Mobility = 0.000471 and 0.00810***.
  - Triple interactions show differential patterns by phase and exposure (e.g., Feb '20 * Mobility * Exposed Sector = 0.0140*** and 0.0101***).
  - Observations for these regressions: 6,699; Number of Sector-Departments: 501 (481 for informal employment regressions).
  - Columns focusing on informal employment show informal employment is more strongly associated with mobility changes across all sectors; the higher correlation of employment in exposed sectors with mobility does not hold when focusing only on informal workers, implying the observed difference is driven by formal employment dynamics.
  - Findings: exposed sectors shed more jobs in the first lockdown predominantly through formal jobs; recovery later occurred mostly through informal positions, raising informality rates within industries and nationally during reopening and the second lockdown.
  - The relationship between mobility and employment weakened over time, suggesting increased remote work or minimized personal interactions.

### XI. Policy-relevant implications and conclusion
- Aggregate impact: COVID-19 disrupted about a quarter of employment in Colombia with heterogeneous sectoral and demographic effects.
- Disproportionate burdens fell on women, the young, and the less educated due to concentration in lockdown-sensitive sectors and higher informality.
- Informality provided resilience during rebound, particularly in sectors exposed to the first lockdown, but may impose long-term costs:
  - Increased informality rates within several sectors;
  - Slower gains in formal salaried jobs;
  - Persistently elevated unemployment above pre-COVID levels;
  - Reduced female labor force participation due to school closures and caregiving burdens.
- Policy stance:
  - The flexibility of informal markets should not deter policies aimed at building a more formal economy in the medium-term to avoid persistent resource misallocations and lower aggregate productivity and potential output.

*Source: wpiea2021235-print-pdf — Annex I (IMF staff and authors’ calculations, GEIH, DANE, Haver Analytics and national authorities as referenced in the document).*

### Annex I __________________________________________________________________27

### Annex I

### I. INTRODUCTION
- COVID-19 disrupted labor markets through mandatory lockdowns, restrictions in mobility, and shifts in consumption behavior.
- In Colombia, the pandemic induced "the deepest recession on record, with about a quarter of total employment being temporarily lost at the height of mandatory lockdowns in the spring of 2020."
- Using micro-data from a national household survey, the paper documents:
  - the magnitude, structure, and distributional impact of the COVID-19 shock on the labor market in Colombia;
  - main channels of impact;
  - the role of social protection policies; and
  - the informal nature of the subsequent recovery.
- Distributional and group-specific impacts:
  - Women, the young, and those with low levels of education were the most adversely affected groups in terms of income and employment losses.
  - Social policies only partly offset these losses.
- Role of informality:
  - The informal sector experienced greater employment losses when the pandemic first struck—particularly among women—because of intrinsic vulnerability of informal employment and because highly informal economic sectors were to a degree more sensitive to lockdown measures.
  - The informal economy did not absorb the adverse employment shock to the degree observed in past recessions, contributing to record total employment losses.
  - Informal jobs rebounded toward pre-COVID employment levels faster than formal jobs.
  - Lack of regulation oversight and rigidities of formal employment appear to have helped the informal sector adapt labor prices and quantities quicker, aiding the economic bounceback.
- Informality as a double-edged sword for the recovery:
  - Positive: Increased informality in lockdown-sensitive sectors during the second half of 2020 appears to have boosted adaptability during subsequent lockdowns—decreasing the correlation of employment losses with both stringency measures and workplace mobility in later lockdown waves.
  - Negative: Lags in recovery of formal employment and unemployment rates, and increases in informality rates—both overall and within sectors—remain at risk of becoming permanent, with potential medium-term losses in productivity and average incomes.
- Contribution to literature:
  - Extends early COVID-19 labor market studies by studying both the downturn and the upturn, and the impact of the subsequent lockdown in January 2021 using data up to March 2021.
  - Leverages high-frequency mobility indicators to track impact of different lockdown waves.
  - Closely linked to Eslava et al. (2020) and Morales et al. (2020) but uses a different empirical framework to study subsequent dynamics.

### II. DATA DESCRIPTION
- Primary micro-data source: Gran Encuesta Integrada de Hogares (GEIH), monthly samples representative of the national population.
  - Over fifteen thousand households are sampled each month with a population of about 60 thousand.
- Key variables available in GEIH:
  - demographic characteristics (e.g., age, gender, and education level);
  - employment status questions classifying workers as employed, unemployed, or out of the labor force; and
  - employer characteristics such as economic sector, size, and self-employment status.
- Informality measure:
  - Constructed following the definition from the Colombian National Statistical Office (DANE).
- Additional data sources:
  - Google Community Mobility Reports (mobility indicators);
  - Oxford Government Response Tracker (OxGRT) Stringency Index of COVID-19 national containment measures;
  - daily reported COVID-19 cases from Colombia’s Ministry of Health and Social Protection.
- Comparative analysis:
  - For other Latin American labor markets, the paper uses aggregate statistics from national statistical offices and micro-data for Chile (Encuesta Nacional de Empleo) and Peru (Encuesta Nacional de Hogares).

*Source: wpiea2021235-print-pdf - Annex I*

### Annex C reports the DANE definition of informality and compares it to that by the ILO. Overall, the DANE

### wpiea2021235-print-pdf - Annex C reports the DANE definition of informality and compares it to that by the ILO. Overall, the DANE

### Informality definitions and overlap
- DANE definition emphasizes the precarious nature of working conditions.
- ILO definition focuses on participation in the social security and tax systems of the workers and their employers.
- Despite differences, the two definitions present a large overlap in the workers they cover.
- Both definitions broadly capture a condition of greater vulnerability of informal workers to macroeconomic and idiosyncratic economic shocks.

### The initial COVID-19 impact (Colombia, March–April 2020)
- First COVID-19 case confirmed on March 6.
- Government declared a state of emergency on March 17.
- First mandatory quarantine began on March 25; school closures remained during 2020 with a gradual hybrid return covering less than 15 percent of the student population starting in February of the following year.
- Relative to pre-pandemic employment levels, about a quarter of employment had been temporarily lost by April 2020.
- Pre-pandemic unemployment rate of 12 percent rose above 20 percent—a historical record.
- Labor force participation rate collapsed from 63 percent to a low of 52 percent.
- Around 7 of every 10 workers who lost their jobs between February and April dropped out of the labor force.
- Mechanisms for exit from labor force included discouragement, increased household duties or childcare, sector-specific shutdowns preventing search, and lockdown regulations limiting job search.

### Regional comparative impact (LA5)
- Net employment losses in Q2 2020 (as a share of total employment) were higher than in Brazil, Mexico and Chile but lower than Peru.
- Colombian case distinguished by larger movements in the unemployment rate relative to regional peers.
- In other LA5 countries, the rise in unemployment between January and May 2020 was smaller than in Colombia both in percent terms and as a fraction of employment losses.

### Uneven recovery (second half of 2020 and early 2021)
- Labor force participation rate recovered 10 percentage points from May to December, reaching 62 percent.
- Over two thirds of those who joined the labor force became employed, contributing to a near 7 percentage point decrease in the unemployment rate to 13 percent by end-2020.
- Workers returning to the labor force into employment accounted for over 75 percent of total employment recovery on a year-on-year basis.
- By end-2020, the total number of unemployed remained 2-3 percentage points above pre-COVID level.
- New lockdowns in January 2021 caused employment gains to stop, labor force participation to decline, and unemployment to rise again to 17 percent.
- Sectoral heterogeneity:
  - Construction and manufacturing experienced significant employment losses in the first half of 2020 and significant recovery in the second half.
  - Education services and hospitality experienced more prolonged disruptions and weaker gains.
  - By end-2020, most of the economy remained below pre-COVID employment, except agriculture, trade services, and the public sector.
- Outstanding aggregate job gap relative to pre-COVID employment at end-2020: 1.4 million.

### Distributional impact: employment and income losses
- Job and labor income losses concentrated among those with lower education, women, and the young—employment dropped by close to a third among these groups.
- Women experienced both the largest initial hit and significantly slower recovery due to childcare and home duties.
- Among those who retained employment, men, the least educated, and older workers experienced greater drops in monthly labor earnings.
- Incomes recovered but remained below pre-COVID levels as of end-2020.
- Pandemic worsened household income distribution:
  - Mass of households reporting no income more than doubled from 2019 to 2020.
  - Poverty rates increased to levels not seen in the last decade.

### Offsetting policies and transfers
- Over half of poorer households were covered by some form of transfer in Colombia before COVID-19 (including pension transfers, unemployment insurance, conditional and unconditional cash-transfer programs).
- Familias en Acción remained the most important transfer mechanism in coverage and magnitude among existing programs.
- An expanded VAT-return transfer program targeted poorest among Familias en Acción participants.
- Self-reported GEIH data indicates a significant portion of transfers were allocated to households not among the poorest deciles, indicating scope for improved targeting.
- Emergency transfers and employment protection measures worth around 1.3 percent of GDP in 2020 included:
  - Wage subsidies equivalent to 40 percent of the minimum wage for formal workers at firms experiencing over 20 percent in revenue losses.
  - A new unconditional transfer program (Ingreso Solidario) aimed at 3 million at-risk households not covered under other programs.
- Expansion concentrated among poorer households but also reached households with earnings above the median by design; budgetary and technical constraints limited full coverage, particularly for informal workers.
- Majority of those reporting job loss due to COVID-19 were not enrolled in any government transfer program.

### Informality as a margin of adjustment
- Informal sector acted as an important margin of adjustment due to greater flexibility in hiring/firing, hours worked, and labor prices.
- Informal employment comprised 51 percent of the aggregate contraction as of July 2020 relative to pre-pandemic levels (pre-COVID baseline informality rate: 48 percent).
- Gender composition and losses:
  - Women comprised about 45 percent of the employed but accounted for 52 percent of total job losses and 58 percent of informal-sector job losses.
  - Employment losses among women were mostly from informal workers; men's losses were more evenly split between formal and informal sectors.
- Recovery dynamics:
  - By December 2020, informal sector gains comprised 59 percent of total employment gains (58 percent of total gains for men and 59 percent for women).
  - Informality rate increased from 47 to 49 percent for both genders (pre-COVID total 48 percent; 46 percent for men; 50 percent for women).
- Worker-type elasticity:
  - Self-employment within informality was most elastic during downturn and upturn.
- Intensive margin (hours and earnings):
  - Among informal workers, average weekly hours decreased by 28 percent by the height of lockdown (Q2 2020), rebounding close to pre-COVID levels by end-2020.
  - Among formal workers, maximum drop in hours was 18 percent.
  - Informal workers experienced severe drops in average monthly income not experienced by formal workers.
- Many informal workers are independent and dependent on demand for services/products, which was depressed during lockdown.

### Interaction with sectoral composition and informality dynamics
- Negative correlation between employment losses at the height of the shock and pre-COVID-19 informality level: sectors most affected had large levels of informal employment.
- Highly informal sectors experienced the largest employment losses but also bounced back more quickly; correlation between cumulative job destruction and informality disappears when accounting for recovered jobs.
- Decomposition of informality share during loss and recovery phases (between- and within-sector):
  - Most decline in share of informal employment during lockdown driven by contraction in high-informality sectors (between-sector effect).
  - Rise in informality in second half of 2020 driven by both recovery of those sectors and increase of informality within sectors (within-sector effect), indicating substitution away from formal job creation.
  - Net increase in informality could imply more persistent shifts toward greater informality, potentially reversing pre-pandemic formalization trends.

### The diminishing effects of lockdowns (intro)
- As sectoral composition and informality shifted, sensitivity of the economy to lockdown measures changed.
- Analysis combines high-frequency mobility data with employment changes at sector-region level to document a decrease in sensitivity.
- Colombia experienced two national lockdowns between the pandemic’s outset and March [text ends here].

*Source: wpiea2021235-print-pdf (IMF staff and authors’ calculations, GEIH, DANE, Haver Analytics and national authorities as referenced in the document).*

### 2021. Figure 12 (left panel) shows the timing of the containment measures at the national

### Figure 12 (left panel) shows the timing of the containment measures at the national

### Timeline of containment measures, cases, and stringency
- The Stringency Index is compiled by the Oxford COVID-19 Government Response Tracker (OxCGRT).
- The first lockdown: steep rise of the index in late March 2020 following Decree 457 of 2020 announced on the 25th of March of 2020; measures were implemented pre-emptively as the virus had not yet spread widely by that date.
- Cases rose progressively and reached a peak in August 2020.
- Restrictions were gradually lifted in the following months while cases subsided; school closures remained in place.
- A second wave of lockdown measures was implemented in early January 2021 after cases rose following the December holidays; social and employment support policies remained in place at similar intensity.
- The Stringency Index rose again during the second lockdown and cases fell steeply; the second round of tightening appears to have been more effective in reducing the spread within a short time frame.
- Visualization notes: the left panel reports the 7-day moving average of new reported cases (solid blue line) and the Stringency Index from OxGRT (dashed red line).

### Mobility response to lockdowns (department level)
- Workplace mobility index sourced from Google Community Mobility Reports, reported relative to the beginning of February 2020.
- At the outset of the first lockdown, mobility fell sharply and progressively rebounded, returning very close to pre-pandemic levels by the end of 2020 in most departments.
- During the second lockdown, despite stringency and cases being as high or higher than in the first lockdown, mobility only reached a trough of -25 percent compared to February 2020.
- Interpretation: the impact of the second lockdown on mobility was more muted — possibly because restrictions were different, the economy adapted, or both.
- Finding consistent with Bakker and Goncalves (2021): a decreasing correlation between stringency and mobility across Latin America.
- Visualization notes: right panel reports mean and minimum-maximum range of the 7-day moving average of workplace mobility relative to February 2020 across Colombia’s departments.

### The diminishing effect of lockdowns on the labor market
- First national lockdown was broad-based; few sectors allowed to operate and teleworking unfeasible in many sectors; fall in mobility highly correlated with drop in employment.
- Over time lockdowns became more differentiated by local conditions and targeted critical sectors; more industries were allowed to partially reopen.
- Departmental-level correlation between mobility and employment weakened over time as mobility and employment recovered.
- During the second lockdown (January 2021), mobility fell slightly but employment stopped rebounding without falling in all departments; the mobility–employment relationship continued to flatten.

### Sectoral exposure and employment dynamics
- Sectoral exposure to restrictions measured as a binary variable following Alfaro et al. (2020) and Morales et al. (2020); exposed = 1, unexposed = 0 (Annex B details).
- Figure 14 (sector-level employment, Feb. 2020=100):
  - By April 2020, the median unexposed sector had experienced a 20 percent contraction while the median contraction in exposed sectors was close to 40 percent.
  - Over the following months, exposed sectors recovered faster; by November 2020 the median gap with respect to pre-pandemic levels was comparable across the two groups.
  - When the second lockdown occurred, the contraction in employment was very small in both sector groups.
- Regression evidence (sector-department level, phases: first lockdown April–July 2020; reopening August–December 2020; second lockdown January 2021 onwards):
  - Specification: Log(Employment_djt) = β * Cases_t + Σ_i δ_i * Phase_i * Exposed_j + γ_t + η_dj + ε_djt.
  - First lockdown: exposed sectors contracted by an extra 9 percent on average (table columns: Exposed Sector * First Lockdown = -0.0944***; standard error (0.0358)).
  - Reopening phase: gap reduces to 5 percent and is not statistically significant (reported: Exposed Sector * First Reopening = -0.0224 and -0.0353 in different samples; not significant).
  - By the third lockdown the differential turns mildly positive but still not significant (Exposed Sector * Second Lockdown = 0.0520 and 0.382*** in different specifications).
  - New Cases per 1000 residents coefficients reported variably across specifications, e.g., -0.164, 0.176, 0.0433, 0.224, 0.343*, 0.204 (standard errors reported in parentheses).
  - Constants reported as 9.552***, 9.542***, 9.537***, 8.732***, 8.681***, 8.660*** with standard errors noted.
  - Observations vary by specification: 3,828; 6,226; 7,659; 2,613; 4,804; 6,120 (as reported in the regression table).
  - Robust standard errors in parentheses; significance markers: *** p<0.01, ** p<0.05, * p<0.1.

### Mobility–employment interactions and informality
- Regression of log employment on workplace mobility, exposure, and phase (Table 3):
  - Mobility * Unexposed Sector coefficients: 0.00280*** and 0.0178*** in different columns (standard errors (0.000566) and (0.00164)).
  - Mobility * Exposed Sector coefficients: 0.00539*** and 0.0152*** (standard errors (0.000529) and (0.00101)).
  - Phase interactions (examples): Feb '20 * Mobility = 0.0104*** and 0.00698**; First Lockdown * Mobility = 0.00373*** and 0.0179***; First Reopening * Mobility = 0.00529*** and 0.0117***; Second Lockdown * Mobility = 0.000471 and 0.00810*** (with standard errors reported).
  - Triple interactions show differential patterns by phase and exposure (e.g., Feb '20 * Mobility * Exposed Sector = 0.0140*** and 0.0101***).
  - Observations: 6,699 across several specifications; Number of Sector-Departments: 501 (or 481 for informal employment regressions).
  - Columns 4–6 focus on informal employment: informal employment is more strongly associated with mobility changes across all sectors; the higher correlation of employment in exposed sectors with mobility does not hold when focusing only on informal workers, implying the observed difference is driven by formal employment dynamics.
- Informal employment dynamics:
  - Informal employment fell during the first lockdown, but exposed sectors experienced a substantially milder fall in informal employment in the first months.
  - The interacted coefficient for the reopening phase has a very similar value to the first lockdown, indicating persistence of differential informal labor dynamics through the second half of 2020.
  - Differential may have widened during the second lockdown, as the interacted coefficient is larger than for the first lockdown.
  - When exposed sectors shed more jobs in the first lockdown, they did so predominantly through formal jobs; their recovery later occurred mostly through informal positions, raising informality rates within these industries and nationally during the reopening and second lockdown.
  - The relationship between mobility and employment weakened over time, suggesting increased remote work or minimized personal interactions.

### Conclusion and policy-relevant implications
- Aggregate labor impact: COVID-19 disrupted about a quarter of employment in Colombia with heterogeneous sectoral and demographic effects.
- Disproportionate impacts: Women, the young, and the less educated experienced the largest losses in employment and labor income, reflecting concentration in lockdown-sensitive sectors and higher prevalence of job informality.
- Informality as margin of adjustment:
  - Informality increased resilience during the rebound, particularly in sectors directly exposed to the first lockdown.
  - However, informal adaptation may carry long-term costs: increased informality rates within several sectors, slower gains in formal salaried jobs, persistently elevated unemployment above pre-COVID levels, and reduced female labor force participation due to school closures and caregiving burdens.
- Policy stance: The flexibility of informal markets should not deter policies aimed at building a more formal economy in the medium-term to avoid persistent resource misallocations and lower aggregate productivity and potential output.

*From: wpiea2021235-print-pdf (2021) — Figure 12 and accompanying analysis.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021235-print-pdf.pdf_