## background-paper-1-en

## Source details

**Canonical URL:** [background-paper-1-en](https://www.imf.org/-/media/files/publications/reo/whd/2023/october/english/background-paper-1-en.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/reo/whd/2023/october/english/background-paper-1-en.pdf.md)
- [Structured JSON version](/-/media/files/publications/reo/whd/2023/october/english/background-paper-1-en.pdf.json)

---

### 1. Introduction — Context, objectives, data, and methodology
- Context and objectives:
  - Latin America has aggregate income volatility twice as large as advanced economies (AEs) and higher than other Emerging Markets (EMs).
  - Large swings in income disrupt household consumption, affect investment, and push individuals into poverty, fueling social discontent.
  - Chapter aims to address:
    - How aggregate income and consumption volatility evolved in Latin America and comparisons with other regions.
    - Drivers of aggregate income volatility (external vs domestic).
    - Level and distribution of income volatility at the individual level across demographic groups.
    - Sensitivity of individual incomes to macroeconomic fluctuations.
    - Role of social transfers in mitigating individual income volatility.
- Definitions and data scope:
  - Income volatility: sample standard deviation of annual real GDP series (HP detrended, lambda = 100).
  - Individual-level analysis: micro data from labor force and household surveys for working-age population (age 14–64) in the labor force.
  - Regional groupings: AE, EM, LA, LA5 (Brazil, Chile, Colombia, Mexico, Peru), LA9 (Bolivia, Brazil, Chile, Colombia, Ecuador, Mexico, Paraguay, Peru, Uruguay), LA5* (Argentina, Brazil, Chile, Mexico, Peru).
- Research team: Juan Passadore (lead), Hyunmin Park, Ana Sofia Pessoa; supervision by Gustavo Adler and Anna Ivanova.
- Methodological notes:
  - Annex 2: decomposition of aggregate output variance into domestic and external components.
  - Annex 3: methodology for estimating response of consumption to exogenous income fluctuations (aggregate insurance).
  - Worker betas follow Guvenen et al. (2017); micro panels are short rotating panels tracking individuals over five or six quarters.

### Aggregate Income Volatility
- Evolution and comparisons:
  - Aggregate income volatility in Latin America has declined in recent decades.
  - Decline more pronounced in Latin America and other EMs, but Latin America remains more volatile than other EMs and AEs.
  - Decline in volatility occurred while GDP growth increased slightly in the past two decades relative to the previous two.
- Drivers of the decline:
  - More stable domestic conditions explain the decline over the past 2 decades.
  - LA9 decomposition: drop in aggregate income volatility over the past two decades due to lower volatility of domestic factors, more than offsetting an increase in volatility due to external factors.
  - External factors considered: US real interest rate, US GDP, China’s GDP, and terms of trade.
  - Variance contributions from external factors are driven mostly by changes in regression coefficients rather than factor volatilities, consistent with greater trade and financial integration.
- Consumption and aggregate insurance:
  - Consumption volatility has fallen in Latin America, consistent with the drop in income volatility.
  - Marked decrease in consumption volatility in LA5 countries.
  - Response of consumption to exogenous income fluctuations suggests improved consumption smoothing (see Annex 3).
- Key aggregate measurement framing:
  - Income volatility measured as sample standard deviation of annual real GDP (HP detrended).
  - Estimation periods compared include 1980–2000 and 2001–2019.
  - LA9 decomposition reports Total Variance, Domestic, and External components for those periods (figures in source).

### Individual Income Volatility — incidence, heterogeneity, mechanisms, and betas
- Overall risk and incidence:
  - Individual labor-force income volatility is high in Latin America.
  - Probability of a large loss in real income (≥ 25 percent, year-over-year same quarter) has averaged about 22 percent over the past decade in LA5* (Argentina, Brazil, Chile, Mexico, Peru).
  - These probabilities have been stable over the last 10 years except during the pandemic years.
  - Within Latin America: share with large income losses averages about 30 percent in Peru and about 20 percent in Brazil.
- Demographic and employment heterogeneity:
  - Probability of large income losses larger for male, older, and less educated workers; pronounced differences across formality status.
  - 30 percent of informal workers face large labor income losses compared to 20 percent of formal workers.
  - Within informal workers: self-employed face 32 percent odds of large income swings; employed informal workers face 25 percent.
  - Employers, even in informal firms, tend to provide more income stability than self-employment.
- Employment transitions and mechanisms of loss:
  - Of the 22 percent experiencing large income losses:
    - 19 percent are employed in both periods.
    - Unemployed and those who become unemployed represent 3 percent.
  - Among employed workers who experience large income losses:
    - 48 percent remain informal.
    - 37 percent remain formal.
    - 10 percent transition from formal → informal.
    - 5 percent transition from informal → formal.
  - Among these employed workers with large income losses:
    - 31 percent experience a change in industry or occupation.
    - 19 percent reduce hours worked.
    - 18 percent experience both change in industry/occupation and reduction in hours.
- Poverty risk and intra-household insurance:
  - On average, 4.5 percent of the labor force crosses the poverty line each year due to a large loss of individual labor income in the same period.
  - Informal and less educated workers are more vulnerable to falling into poverty after large income losses.
  - Of the 22 percent experiencing large individual labor income swings, 16 percent belong to a household whose total labor income also drops by at least 25 percent.
  - Implied intra-household full mitigation: only 7 percent of the labor force (of those who experienced large individual losses) appear to have intra-household income pooling sufficient to fully mitigate large individual income shocks.
  - Poverty threshold: US$ 6.85 a day 2017 purchasing-power-parity.
- Sensitivity to aggregate fluctuations (worker betas):
  - Estimated average workers’ beta for Latin America is about 0.8 (elasticity of individual labor income to aggregate income).
  - Country range: 0.4 for Chile to 1.2 for Brazil.
  - Interpretation: a 1 percentage point reduction in aggregate income is associated with a 0.8 percentage point average fall in individual labor income.
  - Heterogeneity: informal, younger, and less educated workers are more sensitive to aggregate income volatility; differences likely related to sectoral employment composition.

### Betas by Country — measurement and demographic differences
- Definitions in betas analysis:
  - “Younger”: 14–39 years old; “Older”: 40–65 years old.
  - “Less educated”: those who did not complete secondary education.
  - “Formal” workers: those who pay social security contributions.
  - “Poor”: daily per capita household income < US$6.85 a day (2017 PPP).
  - LA5*: pooled regression of Argentina, Brazil, Chile, Mexico, Peru using 2019 PPP GDP weights.
- Estimation approach:
  - Country-level pooled OLS: Δy_{i,t} = α + β^Y ΔY_t + ε_{i,t}.
  - Cross-country pooled: Δy_{i,t,c} = β^Y ΔY_{t,c} + γ_c + ε_{i,t,c}.
  - Demographic exposure studied via interaction terms.
  - Individual weights rescaled and aggregated using 2019 PPP GDP weights.

### The Role of Social Transfers — coverage, replacement, poverty impact
- Social transfer categories:
  - Unemployment protection: severance pay, unemployment savings accounts, unemployment insurance (contributory pension excluded).
  - Other social transfers: noncontributory transfer programs and noncontributory pensions.
- Design notes and data limitations:
  - Severance pay included in unemployment protection where available (Argentina, Peru), converted to monthly equivalents.
  - Mexico excluded in many aggregates due to data limitations on other income sources, including social transfers.
  - Aggregates use 2019 PPP GDP weighted averages for LA5* or LA4* as noted.
- Coverage and targeting:
  - Other transfers coverage: 7.2 percent of the labor force.
  - Unemployment protection coverage: 1.2 percent of the labor force.
  - Probability of receiving unemployment protection higher among unemployed, formal, and more educated workers.
  - Share receiving other social transfers higher among poor, informal, and less educated workers.
- Replacement rates (ratio of transfers to previous total income):
  - For recipients across whole labor force:
    - Unemployment protection: transfers account on average for 70 percent of total previous labor income.
    - Other social transfers: transfers account on average for 42 percent of total previous income.
  - For unemployed who receive transfers:
    - Replacement rate of unemployment protection: 73 (percent).
    - Replacement rate of other social transfers: 57 (percent).
- Poverty impact of transfers:
  - After social transfers, share under US$6.85/person/day (2017 PPP) decreases by 3 percentage points on average.
  - Using US$2.15/person/day (2017 PPP), fraction in extreme poverty decreases on average from 6 to 4 percent after social transfers.

### Social Transfers as a Tool to Mitigate Income Volatility — effectiveness and limits
- Definitions and baseline measures:
  - Large income changes: absolute real income change > 25 percent.
  - Labor-force incidence before transfers:
    - Individual-level large income loss incidence: 22 percent.
    - Household-level large income loss incidence: 14 percent.
  - These measures remain unchanged after accounting for social transfers.
- Poverty transitions:
  - Before social transfers: 3.6 percent of the labor force falls below the poverty line following a large labor income loss.
  - After social transfers: probability decreases to 3.5 percent.
- Betas and exposure after transfers:
  - Worker betas do not decrease significantly after adding social transfers to individual income.
  - Individuals remain exposed to aggregate fluctuations even after receiving social transfers, including younger, less educated, informal, and poor workers.
- Pandemic-era transfers:
  - In 2020–21, temporary increases in social assistance reduced the share of the labor force with large income losses by 1.8 percentage points—larger than pre-pandemic effects.
  - Despite this, almost 25 percent of the labor force experienced large income losses during the pandemic, 3 percentage points higher than pre-pandemic levels.

### Reasons for limited insurance role of transfers — coverage gaps and responsiveness
- Unemployment protection coverage limited:
  - Only 14 percent of formal workers who become unemployed and experience large income losses receive an increase in transfers from unemployment protection.
- Other social transfers do not respond promptly:
  - Only 4 percent of the labor force experiencing large labor income losses receives an increase in other social transfers.
  - Only 12 percent of informal workers who become unemployed and experience large labor income losses receive an increase in other social transfers.
- Poor workers with large income losses:
  - Less than 2 percent receive a contemporaneous boost in unemployment protection.
  - 12 percent receive an increase in other social transfers.
- Transfers rarely adjust in periods of large income fluctuations, limiting their role as social insurance even for the poor or those who become poor after shocks.

### For the Labor Force — synthesis and policy recommendations
- Key findings:
  - Macroeconomic volatility in Latin America has decreased, driven by lower volatility of domestic factors and improved policies.
  - Sensitivity of aggregate consumption to aggregate income shocks has decreased, indicating improved aggregate insurance.
  - Income volatility remains high at the individual level, especially for informal workers; intra-household insurance is limited.
  - Social transfers, including unemployment protection, play a limited role in mitigating individual income volatility.
  - High labor income volatility poses significant challenges for households in or at risk of poverty.
  - Aggregates denote the average of Argentina, Brazil, Chile, and Peru in 2019 using PPP GDP weights; Mexico excluded due to data limitations on other income sources.
  - ILO (2021) figure: effective coverage for unemployment protection is 12.5 percent in LAC vs 18.6 percent in the world.
- Policy recommendations and design implications:
  - Revisit social protection frameworks to mitigate income volatility for the labor force, mindful of fiscal implications and incentive issues.
  - Increase coverage of social insurance programs such as unemployment savings accounts and unemployment insurance.
  - Shorten vesting periods where contribution periods for unemployment insurance are long so workers with fragmented formal employment histories can be insured.
  - Rethink design of social assistance programs to respond more quickly to income shocks.
  - Improve technical and administrative capacity to process social transfers; boost communication across government agencies to improve targeting and reduce program overlap.
  - Design social insurance to protect not only the poor but also those at risk of falling into poverty to preserve social cohesion amid larger and more frequent shocks (e.g., climate change, geoeconomic fragmentation).

*International Monetary Fund | October 2023 — Background paper: "Income Volatility and Social Insurance in Latin America" (background-paper-1-en)*

### 1. Introduction

### 1. Introduction

### Context and objectives
- Latin America has been historically volatile, displaying aggregate income volatility that is twice as large as that in advanced economies (AEs) and higher than that in other Emerging Markets (EMs).
- Large swings in income can disrupt household consumption, affect investment decisions, and push individuals into poverty, contributing to elevated social discontent and unrest in the region.
- There is a growing sense of economic uncertainty in the region, partly driven by recent global geopolitical developments and concerns about income instability going forward.
- This chapter aims to address:
  - How aggregate income and consumption volatility has evolved in Latin America and how it compares to other regions.
  - The drivers of aggregate income volatility (external vs domestic factors).
  - The level and distribution of income volatility at the individual level across demographic groups.
  - The sensitivity of individual incomes to macroeconomic fluctuations.
  - The role of social transfers in mitigating individual income volatility.

### Definitions and data scope
- Income volatility denotes the sample standard deviation of annual real GDP series (HP detrended, lambda = 100) for aggregate analyses.
- Individual-level analysis uses micro data from labor force and household surveys for the working-age population (age 14–64) who are in the labor force.
- Regional groupings used in figures and analysis include AE = advanced economies; EM = emerging markets; LA = Latin America; LA5 = Latin America 5 (Brazil, Chile, Colombia, Mexico, Peru); LA9 = Latin America 9 (Bolivia, Brazil, Chile, Colombia, Ecuador, Mexico, Paraguay, Peru, and Uruguay); LA5* = Argentina, Brazil, Chile, Mexico, and Peru.

### Research team
- This chapter was prepared by Juan Passadore (lead), Hyunmin Park, and Ana Sofia Pessoa, under the supervision of Gustavo Adler and Anna Ivanova.

### Methodological notes referenced
- Annex 2 details the methodology for decomposing aggregate output variance into domestic and external components.
- Annex 3 details the methodology for estimating the response of consumption to exogenous income fluctuations (aggregate insurance).
- Worker betas follow the approach in Guvenen et al. (2017); micro panel construction mainly uses short rotating panels tracking individuals over five or six quarters.

---

### Aggregate Income Volatility

- Evolution and comparisons:
  - Aggregate income volatility in Latin America has declined in recent decades.
  - The decline was more pronounced in Latin America and in other EMs, but Latin America continues to display higher volatility than other EMs and AEs.
  - The decline in volatility in Latin America was not accompanied by a decline in GDP growth; growth has actually increased slightly in the past two decades relative to the previous two.
- Drivers of the decline:
  - More stable domestic conditions explain the decline in aggregate volatility over the past 2 decades.
  - Decomposition for LA9 indicates the drop in aggregate income volatility over the past two decades was due to lower volatility of domestic factors, which more than offset an increase in volatility due to external factors.
  - External factors considered include the US real interest rate, US GDP, China’s GDP, and terms of trade.
  - The variance of output accounted for by each external factor equals the product of its variance and its regression coefficient; results suggest most changes in the variance of external components are due to changes in the coefficients rather than changes in factor volatilities, consistent with greater trade and financial integration.
- Consumption and aggregate insurance:
  - Consumption volatility has also fallen in Latin America, in line with the observed drop in income volatility.
  - There was a marked decrease in consumption volatility in LA5 countries.
  - The response of consumption to exogenous income fluctuations suggests improved consumption smoothing or aggregate insurance (see Annex 3).

Key aggregate statistics and measures preserved from source:
- Income volatility measured as sample standard deviation of annual real GDP (HP detrended).
- Estimation periods often compared: 1980–2000 and 2001–2019.
- LA9 decomposition shows Total Variance, Domestic, and External components for those periods (figures referenced).

---

### Individual Income Volatility

- Overall risk and incidence:
  - Income volatility for individuals in the labor force is high in Latin America.
  - On average, the probability of experiencing a large loss in real income (of 25 percent or more, compared to the same quarter in the previous year) has hovered around 22 percent over the past decade in Argentina, Brazil, Chile, Mexico, and Peru (LA5*).
  - These figures have been remarkably stable over the last 10 years (except during the pandemic years).
  - Within Latin America, the share of individuals experiencing large income losses averages about 30 percent in Peru and about 20 percent in Brazil.
- Demographic and employment heterogeneity:
  - The probability of large income losses is larger for male, older, and less educated workers, though differences are especially pronounced across formality status.
  - 30 percent of informal workers face large labor income losses compared to 20 percent of formal workers.
  - Within informal workers: self-employed face the highest odds of large income swings (32 percent) while employed informal workers face 25 percent.
  - Employers, even in informal firms, tend to provide some degree of income stability versus self-employment.
- Employment transitions and mechanisms of loss:
  - Of the 22 percent of the labor force that face large income losses, 19 percent are employed in both periods, while the unemployed and workers who become unemployed represent only 3 percent of the cases.
  - Among workers who remain employed and experience large income losses:
    - 48 percent remain informal.
    - 37 percent remain formal.
    - 10 percent transition from formal → informal.
    - 5 percent transition from informal → formal.
  - Among these employed workers with large income losses:
    - 31 percent experience a change in industry or occupation.
    - 19 percent reduce hours worked.
    - 18 percent experience both change in industry/occupation and reduction in hours.
- Poverty risk and intra-household insurance:
  - On average, 4.5 percent of the labor force crosses the poverty line each year due to a large loss of individual labor income in the same period.
  - Informal and less educated workers are more vulnerable to falling into poverty following large income losses.
  - Of the 22 percent of the labor force experiencing large individual labor income swings, 16 percent belong to a household whose total labor income also drops by at least 25 percent.
  - These numbers imply that only 7 percent of the labor force (of those who experienced large individual losses) appear to have intra-household income pooling sufficient to fully mitigate large individual income shocks.
  - Poverty threshold used: US$ 6.85 a day 2017 purchasing-power-parity.
- Sensitivity to aggregate fluctuations (workers’ betas):
  - Estimated average workers’ beta for Latin America is about 0.8 (elasticity of individual labor income to aggregate income).
  - Country range in estimated betas: 0.4 for Chile to 1.2 for Brazil.
  - Interpretation: a 1 percentage point reduction in aggregate income is associated with a 0.8 percentage point average fall in individual labor income.
  - Heterogeneity: informal, younger, and less educated workers are more sensitive to aggregate income volatility; differences likely related to sectoral employment composition.

---

*International Monetary Fund | October 2023 — Background paper: "Income Volatility and Social Insurance in Latin America" (Chapter 1: Introduction)*

### 1.  Betas by Country

### 1.  Betas by Country

### Differences in Betas by Demographics
- Elasticity of individual labor income to aggregate income is estimated at the country level and via pooled regressions.
- “Younger” refers to 14–39 years old and “Older” refers to 40–65 years old.
- “Less educated” are those who did not complete secondary education.
- “Formal” workers are those who pay social security contributions.
- “Poor” refers to individuals whose daily per capita household income is lower than US$6.85 a day (2017 purchasing-power-parity).
- LA5* denotes the pooled regression of Argentina, Brazil, Chile, Mexico, and Peru using 2019 purchasing-power-parity GDP weights.

### Key statistical framing
- Country-level and pooled regression outputs are used to compute worker betas (elasticities).
- Differences in betas are reported relative to counterpart groups (e.g., male with respect to female).

---

### The Role of Social Transfers

### Social transfer categories analyzed
- Unemployment protection: includes severance pay, unemployment savings accounts, and unemployment insurance. Contributory pension excluded due to sample not including retirement-age population.
- Other social transfers: payouts from social safety nets such as noncontributory transfer programs and noncontributory pensions.

### Design notes and data limitations
- Severance pay included in unemployment protection when data are available (Argentina and Peru), converted to monthly equivalents.
- Mexico is excluded in many aggregates due to data limitations on other income sources, including social transfers.
- Aggregates use 2019 purchasing-power-parity GDP weighted averages for LA5* or LA4* as noted.

---

### What Do Social Transfers in Latin America Do?

### Coverage and targeting
- Other transfers have much larger coverage (7.2 percent of the labor force) than unemployment protection (1.2 percent of the labor force).
- Probability of receiving unemployment protection is higher among the unemployed, formal workers, and more educated.
- The share receiving other social transfers is higher among poor, informal, and less educated workers.
- By design, other social transfers are allocated toward groups vulnerable to poverty (less educated and informal workers).

### Replacement rates (ratio of transfers to previous total income)
- Unemployment protection: for the whole labor force who receive it, transfers account on average for 70 percent of total previous labor income.
- Other social transfers: for the whole labor force who receive them, transfers account on average for 42 percent of total previous income.
- For the unemployed who receive transfers:
  - Replacement rate of unemployment protection: 73 (percent).
  - Replacement rate of other social transfers: 57 (percent).

### Poverty impact of transfers
- After social transfers, share of population under the poverty threshold of US$6.85/person/day (2017 PPP) decreases by 3 percentage points on average.
- Using extreme poverty threshold US$2.15/person/day (2017 PPP), fraction in extreme poverty decreases on average from 6 to 4 percent after social transfers.

---

### Social Transfers as a Tool to Mitigate Income Volatility

### Individual and household-level effects
- Large income changes are defined as changes in real income greater than 25 percent in absolute value.
- The share of the labor force experiencing large income losses does not decrease materially after adding social transfers to individual income (shown across countries and groups).
- For the labor force:
  - Individual-level large income loss incidence: 22 percent (before transfers).
  - Household-level large income loss incidence: 14 percent (before transfers).
  - These measures remain unchanged after accounting for social transfers.
- Households falling into poverty after large labor income losses:
  - Before social transfers: 3.6 percent of the labor force falls below the poverty line following a large labor income loss.
  - After social transfers: probability decreases to 3.5 percent.

### Betas and exposure to aggregate fluctuations
- Worker betas do not decrease significantly after adding social transfers to individual income.
- Individuals remain exposed to aggregate fluctuations even after receiving social transfers.
- This holds for demographic groups with higher sensitivity: younger, less educated, informal, and poor workers.

### Pandemic-era transfers
- During 2020–21, temporary increases in social assistance spending reduced the share of the labor force with large income losses by 1.8 percentage points—larger than pre-pandemic effects.
- Despite this, almost 25 percent of the labor force experienced large income losses during the pandemic, 3 percentage points higher than pre-pandemic levels.

---

### Reasons for limited insurance role of transfers

### Coverage gaps and responsiveness
- Unemployment protection coverage is very limited even for the formal sector:
  - Only 14 percent of formal workers who become unemployed and experience large income losses receive an increase in transfers from unemployment protection.
- Other social transfers do not respond promptly to large income losses:
  - Only 4 percent of the labor force experiencing large labor income losses receives an increase in other social transfers.
  - Only 12 percent of informal workers who become unemployed and experience large labor income losses receive an increase in other social transfers.
- Poor workers with large income losses:
  - Less than 2 percent receive a contemporaneous boost in unemployment protection.
  - 12 percent receive an increase in other social transfers.
- Transfers rarely adjust in periods of large income fluctuations, limiting their role as social insurance even for the poor or those who become poor after shocks.

---

*Source: background-paper-1-en - 1. Betas by Country (PDF chapter).*

### 1.  For the Labor Force

### 1. For the Labor Force

### Key findings: macro and micro income volatility
- Although Latin America has a long history of high aggregate volatility, macroeconomic volatility has decreased in recent decades, driven by lower volatility of domestic factors, partly reflecting better policies.
- The sensitivity of aggregate consumption to aggregate income shocks has decreased, suggesting an improvement in aggregate insurance.
- Income volatility remains high at the individual level, especially for informal workers, and intra-household insurance is limited, implying there is room for social insurance to step in and protect the most vulnerable.
- Social transfers—including unemployment protection—play a limited role in mitigating individual income volatility, even for formal workers.
- High labor income volatility poses a significant challenge especially to households that are in or at risk of poverty.
- Aggregates denote the average of Argentina, Brazil, Chile, and Peru in 2019 using purchasing-power-parity GDP weights. Mexico is excluded due to data limitations on other income sources, including social transfers.
- “Poor” refers to people in poor households where daily per capita household income is lower than US$ 6.85 a day (2017 purchasing-power-parity).
- According to Figure 4.29 of ILO (2021), the effective coverage for unemployment protection is 12.5 percent in LAC, compared to 18.6 percent in the world.

### Policy recommendations and design implications
- Revisit social protection frameworks to mitigate income volatility for the labor force, especially for those who are in or at risk of falling into poverty, while being mindful of fiscal implications and incentive issues.
- Increase coverage of social insurance programs designed to mitigate income volatility such as unemployment savings accounts and unemployment insurance.
- In countries where the contribution period for unemployment insurance is long, shorten the vesting period so that more workers with fragmented formal employment histories could be insured.
- Rethink the design of social assistance programs so that they respond more quickly to income shocks.
- Improve technical and administrative capacity to process social transfers (Brollo et al., forthcoming).
  - Boost communication across government agencies to allow better targeting of vulnerable households and reduce overlap between various social transfer programs.
- Social insurance schemes should protect not only the poor but also those at risk of falling into poverty to maintain social cohesion in a world of larger and more frequent shocks (e.g., climate change, geoeconomic fragmentation).

### Data sources, sample definitions, and primary variables
- Aggregate data: World Economic Outlook database on real GDP, nominal GDP, CPI, real consumption, and nominal consumption for Advanced Economies, LA, and EMs at an annual frequency. External variables include WEO trade-weighted terms of trade, real interest rate (difference between realized inflation and the fed fund rate), and real GDP of Advanced Economies. Variables detrended with a Hodrick-Prescott filter equal to 100.
- Country groups: Advanced Economies (list provided); EM (list provided); LAC (Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Ecuador, Guatemala, Honduras, Mexico, Nicaragua, Panama, Peru, Paraguay, El Salvador, Uruguay).
- Micro datasets: short rotating panels with most individuals followed over five or six quarters; sample consists of individuals in the labor force for two consecutive years, age between 14 and 65 years old.
- Country-specific microdata sources:
  - Argentina: Encuesta Permanente de Hogares (EPH), quarterly, rotating panel 2-2-2. Unemployment protection includes unemployment insurance and severance pay (seguro de desempleo and indemnización por despido).
  - Brazil: Nova Pesquisa Mensal de Emprego (NPME) and PNADC (IBGE). NPME: monthly in six metro areas, panel 8 interviews over 16 months. PNADC: national, five consecutive quarters; data on severance pay and unemployment savings account not available, so unemployment protection includes only unemployment insurance (seguro-desemprego and seguro-defeso).
  - Chile: Encuesta Suplementaria de Ingresos (ESI) as complementary module of ENE; severance pay data not available, unemployment protection includes seguro de desempleo o cesantia.
  - Mexico: ENOE quarterly rotating panel; collects labor income from primary job only; other income sources not available.
  - Peru: ENAHO panel samples (2007–11, 2011–15, 2013–17, 2017–21); unemployment protection includes severance pay and Compensacion por Tiempo de Servicios.
- Main income variables:
  - Labor Income: sum of earnings from primary occupation and secondary jobs (except Mexico where secondary occupation income not available).
  - Income before social transfers: labor income and other income sources like rents, investment income, capital gains, donations, and pensions.
  - Income after social transfers: total income from labor, investment, rents, capital gains, pensions, donations together with transfers (including unemployment benefits, severance pay, and other government social transfers).
  - Note: income from other sources, including transfers, is not available for Mexico. Income from severance pay is not available for Brazil and Chile.
- Poverty identification: household income per capita before transfers used to identify people living in poverty, except Brazil and Mexico where total labor income is used due to data limitations.

### Measurement and methodological approaches
- Annual real income growth normalized by average income:
  - Δy_{i,t} = (y_{i,t} − y_{i,t−1}) / ((y_{i,t} + y_{i,t−1}) / 2)
  - y_{i,t} is the real income of individual i in year t and Δy_{i,t} is bounded between -2 and 2. Income variables are deflated using national CPI.
- Demographic group definitions used in analysis:
  - Gender: “female” and “male”.
  - Age: “younger” (14–39 years old) and “older” (40–65 years old).
  - Education: “more educated” and “less educated” (if hold a high school diploma or not).
  - Employment type: “formal” and “informal” (depending on whether workers pay social security contributions in t−1 or not).
  - Household income level: “poor” and “not poor” (average daily household income per capita lower than US$ 6.85 a day (2017 PPP)).
- Decomposition of aggregate income volatility (Raddatz 2007 approach):
  - Data generating process: y_{i,t} = φ y_{i,t−1} + β Z_t + ε_{i,t}; Z_t = ρ Z_{t−1} + ε^z_t.
  - Variance decomposition: V(y_{i,t}) = [β^2 × V(ε^z) / ((1−φ^2)(1−ρ^2))] + [V(ε_i) / (1−φ^2)].
  - First term measures external component; second term measures domestic shock variance adjusted by persistence.
  - Identifying assumption: domestic shocks and external variables are uncorrelated, E(ε_{i,t} × ε^z_t)=0.
- Contemporaneous consumption response to income shocks (Campbell and Mankiw 1994, IV approach):
  - C_{i,t} = β Y_{i,t} + ε^C_{i,t}; Y_{i,t} = γ Z_t + ε^Y_t, with Z_t exogenous world factors.
  - 2SLS estimation: first stage regress HP-detrended real output on external variables; second stage regress consumption on predicted output. LA9 sample: BRA, MEX, CHL, COL, PER, URY, PRY, ECU, BOL; GDP-weighted regression using purchasing-power-parity GDP weights.
- Worker betas methodology (Guvenen et al. 2017):
  - Country-level pooled OLS: Δy_{i,t} = α + β^Y ΔY_t + ε_{i,t}, where Δy_{i,t} is yearly individual real labor income and ΔY_t is real GDP growth.
  - Cross-country pooled: Δy_{i,t,c} = β^Y ΔY_{t,c} + γ_c + ε_{i,t,c}, with country fixed effects γ_c.
  - To study demographic exposure, interaction terms are added.
  - Individual weights rescaled using average total weights and 2019 PPP GDP: w^{New}_{it} = w_{it} / (∑∑ w_{it}_{N,T}) * PPP GDP_{2019}.

### Implications for social insurance coverage and targeting
- Unemployment protection coverage is on average lower in LAC than in the rest of the world (ILO 2021: 12.5 percent in LAC vs 18.6 percent in the world).
- Shortening vesting periods and expanding coverage (including for workers with fragmented formal employment histories) can increase insurance against income volatility.
- Faster-responding social assistance and improved administrative capacity and inter-agency communication can improve targeting of vulnerable households and reduce program overlap.
- Protecting those at risk of falling into poverty (not only the currently poor) is important for preserving social cohesion amid larger and more frequent shocks.

*Source: Regional Economic Outlook—Western Hemisphere, “Income Volatility and Social Insurance in Latin America,” International Monetary Fund | October 2023.*

---


_Source: https://www.imf.org/-/media/files/publications/reo/whd/2023/october/english/background-paper-1-en.pdf_