## COVID-19

## Source details

**Canonical URL:** [COVID-19](https://www.imf.org/-/media/files/publications/reo/whd/2020/oct/english/covid-19.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/reo/whd/2020/oct/english/covid-19.pdf.md)
- [Structured JSON version](/-/media/files/publications/reo/whd/2020/oct/english/covid-19.pdf.json)

---

### Overview and health impact
- LAC region has 8.2 percent of the world population (640 million people) and accounted for 36 percent of all deaths (369 thousand) by early October.
- The region experienced the largest recession on record.
- LAC economic activity is expected to contract by 8.1 percent in the current year, compared with a global growth contraction of 4.4 percent and emerging countries contraction of 3.3 percent.

### Pandemic timeline and regional patterns
- The first case in Brazil was confirmed on February 25; by late March Brazil had had only 5 thousand cases.
- Peaks in cases and deaths in the region appeared to have been reached only recently, in late August.
- The region currently has a higher rate of new deaths per million people than the United States and the European Union (data as of October 1, 2020).
- Subregional differences:
  - South America and Mexico have been hit harder than Central America and the Caribbean.
  - The Caribbean has the lowest death toll; some islands eradicated the epidemic but some saw rebounds after reopening to travel.
- Country-level highest death tolls per million: Peru (highest), followed by Brazil, Bolivia, Chile, Ecuador, and Mexico.
- Official statistics likely understate cases and deaths due to low testing; “excess deaths” in a number of countries significantly exceed official COVID-19 deaths.

### Lockdowns, mobility, and observed pandemic dynamics
- Many Latin American countries implemented early, stringent, and protracted lockdowns—typically starting around the same time as in Europe but more stringent (Oxford stringency indicators) and longer-lived.
  - At the peak, mobility in Peru was 80 percent below normal.
- Early and stringent lockdowns prevented an explosion of daily cases and deaths that would have overwhelmed weak health systems, but did not contain total deaths.
- Distinct pandemic patterns identified:
  - “Forest fire”: explosion in daily deaths followed by rapid decline (observed in some Western European countries like France and Spain).
  - “Slow burn”: steady increase and plateaus over months (observed in much of Latin America, e.g., Brazil and Mexico).
  - “Put-out”: early reduction of contact rate below a threshold leading to eradication (observed in some Caribbean islands).
- SEIR model insights:
  - No action → forest fire (explosion then rapid decline).
  - Early reduction of contact rate below threshold → put-out.
  - Early reduction but above threshold → slow burn.
  - Late intervention → forest fire; the later the intervention, the more severe the forest fire and the higher the peak deaths.
- Timeliness evidence:
  - In Peru, the daily number of deaths two weeks after lockdown (proxy for spread at lockdown time) was still very low; in a number of European countries, it was much higher—implying Latin America avoided forest fires but did not eradicate the epidemic.
- R-effective outcomes:
  - In Peru, R-effective fell after lockdown but remained above 1.

### Structural factors and behavioral drivers limiting lockdown effectiveness
- Structural constraints reducing lockdown effectiveness include:
  - High degrees of economic informality and informal working conditions.
  - Densely populated poor living conditions and agglomerations in food markets and banks.
  - Low government effectiveness and weak institutional capacity.
  - Low testing leading to underestimation of cases and deaths.
- Empirical associations:
  - Higher population density and weak health systems may hamper effectiveness of containment policies.
  - Cross-country regressions show population size significant for deaths; population density and urban share are not significant in those regressions.
- Anecdotal evidence from Peru:
  - Crowded living conditions and agglomerations in food markets and banks may have contributed to spread.
  - Limited financial inclusion and informal work forced households to visit banks in person to receive cash transfers and to make cash food purchases.
- Lockdown fatigue and economic necessity:
  - Mobility in Latin America increased from April onwards; the rebound in mobility occurred while daily deaths were still rising, whereas in Europe the rebound occurred when daily deaths were in retreat.
  - Increased mobility from April may have contributed to further spread.

### Factors affecting the death toll (fundamentals)
- Box 3 findings (summarized):
  - Population size, low hospital capacity, aging population, and lack of routine BCG vaccination are on average associated with higher COVID-19 deaths.
  - LAC has a relatively young population and routine BCG vaccination, but health systems were poorly prepared.
  - Even controlling for “deep-determinants”, regressions residuals are excessively large for Peru, Brazil, Chile, Bolivia, and Mexico—indicating more deaths than explained by fundamentals alone.
- Additional micro-level findings:
  - Municipality-level data indicate total death tolls are higher in more densely populated municipalities and in locations further away from the equator (suggesting climate influences virus reproduction).

### Geographical spread within countries
- Subnational spread:
  - In late April about three quarters of Latin America regions had less than 20 deaths per million; by July it was less than a quarter.
  - Sharp regional differences: some regions now have more than 2,000 deaths per million while others have less than 20.
- Country examples:
  - Brazil: In mid-May, 20 percent of Brazil had a death toll of more than 100 per million; by mid-July this had increased to 100 percent.
  - Mexico (municipality level): In April almost all of Mexico had less than 20 deaths per million; by July less than 10 percent remained under 20 per million. A tenth of municipalities had between 1,000 and 2,000 deaths per million, while 7 percent had less than 100 deaths per million.

### Economic impact of lockdowns and behavioral changes — methods and data
- Two empirical approaches:
  - Cross-section using quarterly GDP (short sample of countries with quarterly data).
  - Panel of 17 Latin American countries using monthly economic activity indices (February to June).
- Estimators and identification:
  - Cross-section: OLS with controls including average past growth (2016Q1-2019Q4).
  - Panel: fixed-effects estimator with country dummies and time interactions to absorb time-invariant factors and capture perceptions, fatigue, and common trends.
- Mobility data used as high-frequency proxy for economic activity (daily frequency, subnational coverage).

### Cross-country regression findings (2020Q2 GDP growth)
- Model explains more than half of the variation in 2020Q2 year-over-year GDP growth (R-squared reported above 0.5 in table specifications).
- Key estimated elasticities and effects:
  - A tightening of the Oxford stringency index from 50 to 100 would in itself be expected to cause a drop of 6 percentage points in GDP growth.
  - An increase in the daily death toll from 2 to 5 per million reduces year-on-year growth by almost 2.5 percentage points (column (b) result).
- Heterogeneity by development:
  - Impact of deaths on activity varies with GDP per capita: the interaction coefficient is positive, implying that the richer a country the less the negative impact of deaths, ceteris paribus.
- Regional dummy:
  - The dummy for LA6 countries in this sample is not significant, suggesting the large drop in GDP in the region is not an idiosyncratic region-related phenomenon.

### LAC panel (monthly activity) findings and key statistics
- The fixed-effects panel model explains around 80 percent of the variation in monthly real activity growth rates in the region (R2 reported as 0.809 in preferred specifications).
- Both policy (stringency) and behavioral (deaths) factors had significant impacts on economic growth.
- Diminishing effects over time:
  - Using the point estimates in column (4), the behavioral proxy becomes indistinguishable from zero by June: -2.8*(T=6) - 16.9 = -0.1.
  - Impact of stringency in April (T=4), May (T=5) and June (T=6) are, respectively: -0.25, -0.22, -0.20.
  - Moving from a stringency index of 50 to 100 would cause a growth decline of approximately 10 percent (panel interpretation).
- Table 2 sample statistics (selected):
  - Stringency coefficients: -0.264 ***, -0.231 ***, -0.348 ***, -0.336 *** (across columns).
  - New Deaths coefficients: -18.881 ***, -16.586 ***, -16.884 *** (across columns).
  - New Deaths * T coefficients: 3.291 ***, 2.791 ***, 2.835 *** (across columns).
  - R2 values reported: 0.701, 0.779, 0.809, 0.809 (across columns).
  - Observations: 61, 75, 75, 75 (across columns).

### Mobility, behavior, and policy interaction
- Mobility is a timely proxy for economic activity; close link observed between decline in mobility and decline in economic activity in first half of 2020.
- Both lockdowns (policy stringency) and behavioral responses (fear of infection proxied by deaths) contributed to sharp mobility declines.
- Evidence of diminishing behavioral and policy effects on mobility over time:
  - Regions with higher daily deaths per million had lower mobility (column 1 of regional regressions).
  - The coefficient on New deaths (lag) * T is positive, implying the impact of new deaths on mobility declined over time.
  - The interaction of time with stringency is positive and highly significant, implying the impact of stringency on mobility also declined over time.
- Country/regional regressions (Peru and Argentina examples; regional fixed-effects):
  - Peru: New deaths (lag) effects cited as (-3.06 + 0.19 * T) * New deaths (lag) example—impact becomes less pronounced over time.
  - Table 3 selected coefficients:
    - New deaths (lag): -0.74***, -3.06***, -1.01***, -0.53*** (Peru models); -1.18**, -24.58***, -5.78***, -2.97*** (Argentina models).
    - New deaths (lag) * T: 0.19***, 0.09***, 0.02*, 1.25***, 0.37***, 0.05 (various columns).
    - Stringency: -0.84***, -0.88***, -0.70***, -0.86***.
    - Stringency * T: 0.01***, 0.02*** (columns reported).
    - Observations per model: 645–648 (Peru), 608–611 (Argentina).
    - R2 reported up to 0.92 in some specifications.

### Reopenings, recoveries, and risks
- Reopening patterns:
  - LAC countries have been gradually easing lockdowns but stringency remains elevated in most of the region; LA6 stay-at-home requirements were as stringent as France in April and May for the last five months referenced.
  - Example (Brazil): in mid-August the stringency index was 17 points lower than at the peak; deviation of mobility from baseline in mid-August was 35 percentage points less than at the peak.
- Mobility and activity recovery:
  - For a sample of 17 LAC countries, average mobility recovered from -60 percent (year-over-year) in April to -44 percent (year-over-year) in June.
  - Average economic contraction eased from -20 percent to -10 percent (year-over-year) over the same period.
- Trade-offs and resurgence risk:
  - The easing of restrictions and reopening have been associated with a pick-up in new COVID-19 deaths in some countries (Argentina, Colombia, Paraguay, Costa Rica, Suriname).
  - The region’s vulnerability is elevated due to reliance on stringency with weak testing capacity and limited tracing.
  - Limitations in testing and tracing increase vulnerability to second waves.

### Drivers of partial effectiveness and policy implications
- Structural and policy constraints that limited effectiveness of lockdowns:
  - High degree of informality, high poverty, crowded urban living conditions.
  - Weak health system capacity and low government effectiveness in some cases.
  - Inability to ramp up testing and tracing capacity; testing remains weak in several countries.
  - Limited fiscal resources in some cases hampered responses.
- Policy implications and recommendations:
  - Reopenings should be cautious given vulnerability to resurgence if reopening happens too quickly.
  - Lockdown periods should be used efficiently to prepare for safe reopening—specifically by strengthening testing and tracing capacity.
  - Recognize that both policy stringency and behavioral responses contributed to economic contractions, but their impacts wane over time as households face economic necessity and “fatigue.”
  - Given diminishing effectiveness of lockdowns over time, complementary measures (test-and-trace, targeted interventions) are critical to contain spread while limiting economic damage.

### Summary conclusions
- Early and stringent lockdowns in LAC initially reduced mobility sharply and slowed the early explosion of infections, helping avoid immediate overwhelm of medical capacity.
- Lockdowns slowed but did not stop the pandemic; effectiveness declined over time as mobility picked up while cases and deaths remained elevated, producing a “slow burn” death pattern in some cases.
- The significant economic impact in LAC reflects both government policies (lockdowns) and behavioral responses; empirical analysis cannot definitively rank their relative contributions.
- Structural factors and policy weaknesses (especially weak testing/tracing) left the region vulnerable to continued high death tolls and to resurgences as reopenings proceeded.

### Box 1 — Lack of COVID-19 Testing, Underestimation of Cases and Excess Deaths
- Lack of testing:
  - The number of COVID-19 cases and of COVID-19 deaths may be significantly underestimated in Latin American countries, which may have implications for the assessment of the impact of the pandemic.
  - Limited testing capacity in LAC has been well below other regions.
  - Positive cases to tests ratio:
    - In the U.S., in Asian, MENA, and European countries, the ratio of the number of positive cases to the number of tests performed is below 20 percent.
    - In LAC, on the other hand, this ratio hovers between 50 and 100 percent.
  - Low testing seems to have been especially an issue in Mexico, Peru, Ecuador, Argentina, and Bolivia, while only Chile appears to have achieved an appropriate testing capacity.
- Underestimation of deaths:
  - The lack of testing capacity may also lead to an underestimation in the number of deaths.
  - Peru:
    - The number of “excess” deaths—i.e., the number of deaths in the current year in excess of the average of the previous few years—has been almost 3 times the number of the official deaths.
  - Chile:
    - Excess deaths are close to the official figure.
  - Economist-based observations:
    - Available data from the Economist suggest that in Mexico, Peru and Ecuador, there has been a significant undercounting of COVID-19 deaths.
    - Mexico: excess deaths as of end July were almost 3 times as high as the official number of COVID-19 deaths.
    - For Brazil and Chile, by contrast, the number of excess deaths is close to the official COVID-19 death count.
  - The combination of underestimated cases and deaths points towards an even more significant health impact of the pandemic in Latin America than the official figures would suggest.
- Key statistics from Box Table 1.1 (COVID-19 Excess Deaths, Per million):
  - Brazil — Dates: Mar 21st-Aug 14th — Excess deaths: 500 — Covid Deaths: 514 — Difference: -14
  - Mexico — Dates: Mar 28th-Jul 31st — Excess deaths: 1,000 — Covid Deaths: 369 — Difference: 631
  - Peru — Dates: Mar 31st-Aug 30th — Excess deaths: 2,110 — Covid Deaths: 882 — Difference: 1,228
  - Chile — Dates: Apr 7th-Sep 14th — Excess deaths: 620 — Covid Deaths: 678 — Difference: -58
  - Ecuador — Dates: Feb 29th-Aug 30th — Excess deaths: 1,830 — Covid Deaths: 381 — Difference: 1,449

*Source: IMF staff calculations and analysis from "COVID-19 in Latin America and the Caribbean" (October 2020).*

### Introduction

### Introduction

### Overview and health impact
- LAC region has 8.2 percent of the world population (640 million people) and accounted for 36 percent of all deaths (369 thousand) by early October.
- The region experienced the largest recession on record.
- LAC economic activity is expected to contract by 8.1 percent in the current year, compared with a global growth contraction of 4.4 percent and emerging countries contraction of 3.3 percent.

### Pandemic timeline and regional patterns
- The first case in Brazil was confirmed on February 25; by late March Brazil had had only 5 thousand cases.
- Peaks in cases and deaths in the region appeared to have been reached only recently, in late August.
- The region currently has a higher rate of new deaths per million people than the United States and the European Union (data as of October 1, 2020).
- Subregional differences:
  - South America and Mexico have been hit harder than Central America and the Caribbean.
  - The Caribbean has the lowest death toll; some islands eradicated the epidemic but some saw rebounds after reopening to travel.
- Country-level highest death tolls per million: Peru (highest), followed by Brazil, Bolivia, Chile, Ecuador, and Mexico.
- Official statistics likely understate cases and deaths due to low testing; “excess deaths” in a number of countries significantly exceed official COVID-19 deaths.

### Lockdowns, mobility, and observed pandemic dynamics
- Many Latin American countries implemented early, stringent, and protracted lockdowns—typically starting around the same time as in Europe but more stringent (Oxford stringency indicators) and longer-lived.
  - At the peak, mobility in Peru was 80 percent below normal.
- Early and stringent lockdowns prevented an explosion of daily cases and deaths that would have overwhelmed weak health systems, but did not contain total deaths.
- Distinct pandemic patterns identified:
  - “Forest fire”: explosion in daily deaths followed by rapid decline (observed in some Western European countries like France and Spain).
  - “Slow burn”: steady increase and plateaus over months (observed in much of Latin America, e.g., Brazil and Mexico).
  - “Put-out”: early reduction of contact rate below a threshold leading to eradication (observed in some Caribbean islands).
- SEIR model insights:
  - No action → forest fire (explosion then rapid decline).
  - Early reduction of contact rate below threshold → put-out.
  - Early reduction but above threshold → slow burn.
  - Late intervention → forest fire; the later the intervention, the more severe the forest fire and the higher the peak deaths.
- Timeliness evidence:
  - In Peru, the daily number of deaths two weeks after lockdown (proxy for spread at lockdown time) was still very low; in a number of European countries, it was much higher—implying Latin America avoided forest fires but did not eradicate the epidemic.
- R-effective outcomes:
  - In Peru, R-effective fell after lockdown but remained above 1.

### Structural factors and behavioral drivers limiting lockdown effectiveness
- Structural constraints reducing lockdown effectiveness include:
  - High degrees of economic informality and informal working conditions.
  - Densely populated poor living conditions and agglomerations in food markets and banks.
  - Low government effectiveness and weak institutional capacity.
  - Low testing leading to underestimation of cases and deaths.
- Empirical associations:
  - Higher population density and weak health systems may hamper effectiveness of containment policies.
  - Cross-country regressions show population size significant for deaths; population density and urban share are not significant in those regressions.
- Anecdotal evidence from Peru:
  - Crowded living conditions and agglomerations in food markets and banks may have contributed to spread.
  - Limited financial inclusion and informal work forced households to visit banks in person to receive cash transfers and to make cash food purchases.
- Lockdown fatigue and economic necessity:
  - Mobility in Latin America increased from April onwards; the rebound in mobility occurred while daily deaths were still rising, whereas in Europe the rebound occurred when daily deaths were in retreat.
  - Increased mobility from April may have contributed to further spread.

### Factors affecting the death toll (fundamentals)
- Box 3 findings (summarized):
  - Population size, low hospital capacity, aging population, and lack of routine BCG vaccination are on average associated with higher COVID-19 deaths.
  - LAC has a relatively young population and routine BCG vaccination, but health systems were poorly prepared.
  - Even controlling for “deep-determinants”, regressions residuals are excessively large for Peru, Brazil, Chile, Bolivia, and Mexico—indicating more deaths than explained by fundamentals alone.
- Additional micro-level findings:
  - Municipality-level data indicate total death tolls are higher in more densely populated municipalities and in locations further away from the equator (suggesting climate influences virus reproduction).

### Geographical spread within countries
- Subnational spread:
  - In late April about three quarters of Latin America regions had less than 20 deaths per million; by July it was less than a quarter.
  - Sharp regional differences: some regions now have more than 2,000 deaths per million while others have less than 20.
- Country examples:
  - Brazil: In mid-May, 20 percent of Brazil had a death toll of more than 100 per million; by mid-July this had increased to 100 percent.
  - Mexico (municipality level): In April almost all of Mexico had less than 20 deaths per million; by July less than 10 percent remained under 20 per million. A tenth of municipalities had between 1,000 and 2,000 deaths per million, while 7 percent had less than 100 deaths per million.

### Economic impact of lockdowns and behavioral changes
- The COVID-19 pandemic caused a major economic impact in LAC:
  - Year-on-year growth in 2020Q2 was well below previous quarters and variation in growth rates among countries increased notably.
  - The chapter aims to disentangle the roles of government policies (lockdowns) and voluntary behavioral responses in the sharp slowdown in growth using two approaches, including cross-country analysis controlling for deaths per million and stringency indicators.

*Prepared by a WHD team headed by Bas Bakker and Carlos Goncalves, including Pedro Rodriguez, Mauricio Vargas, Dmitry Vasilyev, Carlo Pizzinelli, Vibha Nanda, and Alain Brousseau.*

### 2020. This initial approach using quarterly data does not explore the time dimension and hence cannot

### COVID-19 in Latin America and the Caribbean — key findings (October 2020)

### Methods and data
- Two empirical approaches:
  - Cross-section using quarterly GDP (short sample of countries with quarterly data).
  - Panel of 17 Latin American countries using monthly economic activity indices (February to June).
- Estimators and identification:
  - Cross-section: OLS with controls including average past growth (2016Q1-2019Q4).
  - Panel: fixed-effects estimator with country dummies and time interactions to absorb time-invariant factors and capture perceptions, fatigue, and common trends.
- Mobility data used as high-frequency proxy for economic activity (daily frequency, subnational coverage).

### Cross-country regression findings
- Model explains more than half of the variation in 2020Q2 year-over-year GDP growth (R-squared reported above 0.5 in table specifications).
- Key estimated elasticities and effects:
  - A tightening of the Oxford stringency index from 50 to 100 would in itself be expected to cause a drop of 6 percentage points in GDP growth.
  - An increase in the daily death toll from 2 to 5 per million reduces year-on-year growth by almost 2.5 percentage points (column (b) result).
- Heterogeneity by development:
  - Impact of deaths on activity varies with GDP per capita: the interaction coefficient is positive, implying that the richer a country the less the negative impact of deaths, ceteris paribus.
- Regional dummy:
  - The dummy for LA6 countries in this sample is not significant, suggesting the large drop in GDP in the region is not an idiosyncratic region-related phenomenon.

### LAC panel (monthly activity) findings
- The fixed-effects panel model explains around 80 percent of the variation in monthly real activity growth rates in the region (R2 reported as 0.809 in preferred specifications).
- Both policy (stringency) and behavioral (deaths) factors had significant impacts on economic growth.
- Diminishing effects over time:
  - Using the point estimates in column (4), the behavioral proxy becomes indistinguishable from zero by June: -2.8*(T=6) - 16.9 = -0.1.
  - Impact of stringency in April (T=4), May (T=5) and June (T=6) are, respectively: -0.25, -0.22, -0.20.
  - Moving from a stringency index of 50 to 100 would cause a growth decline of approximately 10 percent (panel interpretation).
- Table 2 sample statistics (selected):
  - Stringency coefficients: -0.264 ***, -0.231 ***, -0.348 ***, -0.336 *** (across columns).
  - New Deaths coefficients: -18.881 ***, -16.586 ***, -16.884 *** (across columns).
  - New Deaths * T coefficients: 3.291 ***, 2.791 ***, 2.835 *** (across columns).
  - R2 values reported: 0.701, 0.779, 0.809, 0.809 (across columns).
  - Observations: 61, 75, 75, 75 (across columns).

### Mobility, behavior, and policy interaction
- Mobility is a timely proxy for economic activity; close link observed between decline in mobility and decline in economic activity in first half of 2020.
- Both lockdowns (policy stringency) and behavioral responses (fear of infection proxied by deaths) contributed to sharp mobility declines.
- Evidence of diminishing behavioral and policy effects on mobility over time:
  - Regions with higher daily deaths per million had lower mobility (column 1 of regional regressions).
  - The coefficient on New deaths (lag) * T is positive, implying the impact of new deaths on mobility declined over time.
  - The interaction of time with stringency is positive and highly significant, implying the impact of stringency on mobility also declined over time.
- Country/regional regressions (Peru and Argentina examples; regional fixed-effects):
  - Peru: New deaths (lag) effects cited as (-3.06 + 0.19 * T) * New deaths (lag) example—impact becomes less pronounced over time.
  - Table 3 selected coefficients:
    - New deaths (lag): -0.74***, -3.06***, -1.01***, -0.53*** (Peru models); -1.18**, -24.58***, -5.78***, -2.97*** (Argentina models).
    - New deaths (lag) * T: 0.19***, 0.09***, 0.02*, 1.25***, 0.37***, 0.05 (various columns).
    - Stringency: -0.84***, -0.88***, -0.70***, -0.86***.
    - Stringency * T: 0.01***, 0.02*** (columns reported).
    - Observations per model: 645–648 (Peru), 608–611 (Argentina).
    - R2 reported up to 0.92 in some specifications.

### Reopenings, recoveries, and risks
- Reopening patterns:
  - LAC countries have been gradually easing lockdowns but stringency remains elevated in most of the region; LA6 stay-at-home requirements were as stringent as France in April and May for the last five months referenced.
  - Example (Brazil): in mid-August the stringency index was 17 points lower than at the peak; deviation of mobility from baseline in mid-August was 35 percentage points less than at the peak.
- Mobility and activity recovery:
  - For a sample of 17 LAC countries, average mobility recovered from -60 percent (year-over-year) in April to -44 percent (year-over-year) in June.
  - Average economic contraction eased from -20 percent to -10 percent (year-over-year) over the same period.
- Trade-offs and resurgence risk:
  - The easing of restrictions and reopening have been associated with a pick-up in new COVID-19 deaths in some countries (Argentina, Colombia, Paraguay, Costa Rica, Suriname).
  - The region’s vulnerability is elevated due to reliance on stringency with weak testing capacity and limited tracing.
  - Limitations in testing and tracing increase vulnerability to second waves.

### Drivers of partial effectiveness and policy implications
- Structural and policy constraints that limited effectiveness of lockdowns:
  - High degree of informality, high poverty, crowded urban living conditions.
  - Weak health system capacity and low government effectiveness in some cases.
  - Inability to ramp up testing and tracing capacity; testing remains weak in several countries.
  - Limited fiscal resources in some cases hampered responses.
- Policy implications and recommendations highlighted by the analysis:
  - Reopenings should be cautious given vulnerability to resurgence if reopening happens too quickly.
  - Lockdown periods should be used efficiently to prepare for safe reopening—specifically by strengthening testing and tracing capacity.
  - Recognize that both policy stringency and behavioral responses contributed to economic contractions, but their impacts wane over time as households face economic necessity and “fatigue.”
  - Given diminishing effectiveness of lockdowns over time, complementary measures (test-and-trace, targeted interventions) are critical to contain spread while limiting economic damage.

### Summary conclusions
- Early and stringent lockdowns in LAC initially reduced mobility sharply and slowed the early explosion of infections, helping avoid immediate overwhelm of medical capacity.
- Lockdowns slowed but did not stop the pandemic; effectiveness declined over time as mobility picked up while cases and deaths remained elevated, producing a “slow burn” death pattern in some cases.
- The significant economic impact in LAC reflects both government policies (lockdowns) and behavioral responses; empirical analysis cannot definitively rank their relative contributions.
- Structural factors and policy weaknesses (especially weak testing/tracing) left the region vulnerable to continued high death tolls and to resurgences as reopenings proceeded.

*Source: IMF staff calculations and analysis from "COVID-19 in Latin America and the Caribbean" (October 2020).*

### Box 1. Lack of COVID-19 Testing, Underestimation of Cases and Excess Deaths

### Box 1. Lack of COVID-19 Testing, Underestimation of Cases and Excess Deaths

### Lack of testing
- The number of COVID-19 cases and of COVID-19 deaths may be significantly underestimated in Latin American countries, which may have implications for the assessment of the impact of the pandemic.
- This underestimation is likely related to the limited testing capacity, which in LAC has been well below other regions.
- The lack of testing is evident in the positive cases to tests ratio:
  - In the U.S., in Asian, MENA, and European countries, the ratio of the number of positive cases to the number of tests performed is below 20 percent.
  - In LAC, on the other hand, this ratio hovers between 50 and 100 percent.
- Low testing seems to have been especially an issue in Mexico, Peru, Ecuador, Argentina, and Bolivia, while only Chile appears to have achieved an appropriate testing capacity.

### Underestimation of deaths
- The lack of testing capacity may also lead to an underestimation in the number of deaths.
- Peru:
  - The number of “excess” deaths—i.e., the number of deaths in the current year in excess of the average of the previous few years—has been almost 3 times the number of the official deaths.
- Chile:
  - Excess deaths are close to the official figure.
- Economist-based observations:
  - Available data from the Economist suggest that in Mexico, Peru and Ecuador, there has been a significant undercounting of COVID-19 deaths.
  - Mexico: excess deaths as of end July were almost 3 times as high as the official number of COVID-19 deaths.
  - For Brazil and Chile, by contrast, the number of excess deaths is close to the official COVID-19 death count.
- The combination of underestimated cases and deaths points towards an even more significant health impact of the pandemic in Latin America than the official figures would suggest.

### Key statistics from Box Table 1.1 (COVID-19 Excess Deaths, Per million)
- Brazil — Dates: Mar 21st-Aug 14th — Excess deaths: 500 — Covid Deaths: 514 — Difference: -14
- Mexico — Dates: Mar 28th-Jul 31st — Excess deaths: 1,000 — Covid Deaths: 369 — Difference: 631
- Peru — Dates: Mar 31st-Aug 30th — Excess deaths: 2,110 — Covid Deaths: 882 — Difference: 1,228
- Chile — Dates: Apr 7th-Sep 14th — Excess deaths: 620 — Covid Deaths: 678 — Difference: -58
- Ecuador — Dates: Feb 29th-Aug 30th — Excess deaths: 1,830 — Covid Deaths: 381 — Difference: 1,449

*This box was prepared by Bas Bakker and Mauricio Vargas.*

---


_Source: https://www.imf.org/-/media/files/publications/reo/whd/2020/oct/english/covid-19.pdf_