## 4.1  The effectiveness of Covid-19 containment measures

## Source details

**Canonical URL:** [4.1  The effectiveness of Covid-19 containment measures](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021099-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021099-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021099-print-pdf.pdf.json)

---

### Overview and scope
- Analysis covers containment measures implemented in over 150 countries.
- "Early" implementation: measures in place when the country reached 100 reported Covid-19 cases.
- Compares:
  - Stringency-based measures (lockdowns, closures, restrictions on mobility).
  - Public health measures (wide-scale testing, contact tracing, public information campaigns).

### Key empirical findings
- Successful containment combined:
  - Early targeted stringency measures (for example, monitoring international travel).
  - Strong public health measures (wide-scale testing, contact tracing, public information campaigns).
- Regional and historical patterns:
  - Asian economies with prior SARS experience (Hong Kong SAR, Taiwan Province of China, Singapore, Vietnam) acted quickly with targeted stringency and strong health measures and did not rely on quicker or stricter stay-at-home orders, workplace closures, transport closures, or internal mobility restrictions.
- Associations with deaths and growth:
  - Lower deaths and more successful containment in 2020 (measured by age-adjusted mortality rates) are associated with less stringent containment measures on average throughout the year, but with stronger public health measures.
  - Early stringency measures helped curb deaths during much of the year, although early stringency is not significant in explaining death rates recorded by the end of the year.
  - Countries with stricter containment measures, on average, experienced lower GDP growth in 2020.
  - Interaction effects: Covid-19 death rates are negatively related with economic growth, especially when average stringency measures were high.

### Evidence on economic and fiscal trade-offs
- External estimates cited:
  - "A 10 percent output loss from reduced economic activity and an equivalent fiscal cost (Gourinchas, 2020)."
  - Survey: "50 percent of survey participants incurred income losses averaging 5,293 U.S. dollars and wealth losses averaging 33,482 U.S. dollars (Coibion et al., 2020)."
  - Containment measures resulted in a loss of about "15 percent for industrial production over a 30-day horizon (Deb et al., 2020a)."
- This paper’s empirical results:
  - Lower deaths and more successful containment are associated with better GDP growth outcomes in 2020.
  - Lower death rates are associated with higher primary balances in 2020 and reduced debt levels in the medium run.
  - Results hold when controlling for age structure and Covid-19 fiscal support measures in sensitivity tests.

### Identification and robustness
- Addressing endogeneity via two-stage-least-squares:
  - Instrument: past experience in containing coronavirus epidemics (dummy for nine countries with past SARS/MERS experience).
  - First-stage: past experience associated with significantly lower death rates, controlling for countries’ income level; F-statistic = 17.6.
  - Second-stage: lower Covid-19 deaths, instrumented by past experience, have a positive impact on economic growth in 2020.
- Robustness checks:
  - Results robust to controls including population age structure (median age) and fiscal support measures.
  - Alternative instruments tested: including past Ebola experience and considering only SARS experience; both remain significantly associated with lower Covid-19 death rates and higher GDP growth when instrumenting deaths.

### Policy implications and recommended strategy
- No necessary trade-off between saving lives and saving the economy.
- Recommended "smart" containment strategy centered on:
  - Testing,
  - Contact tracing,
  - Public information campaigns,
  - Targeted stringency early on (for example, international travel controls, school closures, cancellation of public events).
- Countries with limited past epidemic experience often relied on stronger, longer-lasting stringency measures that reduced deaths but implied a trade-off for economic growth.
- Smart containment can mitigate both health and economic costs by emphasizing public health measures and early, targeted restrictions rather than broad, prolonged lockdowns.
- Authorities should prepare to adopt smart strategies to fight possible new waves; lessons for future pandemics: early travel controls plus strong health policies can reduce contagion while allowing relaxation of stringency measures later.

### Data, metrics, and empirical approach
- Sample and period:
  - Cross-sectional data for over 150 countries.
  - OxCGRT daily government measures spanning January 1- December 31, 2020.
- Main containment indicators (normalized 0-1):
  1. "Overall containment measures": containment and health index (8 containment/closure + first 3 health indicators).
  2. "Stringency measures": stringency index (8 containment/closure + public information campaigns).
  3. "Public health measures": principal component of public information campaigns, testing policy, contact tracing.
- Aggregation summary indicators:
  1. "Average strength": average value over the year (does not account for timing).
  2. "Early strength": measures in place when the 100th case was recorded (accounts for timing).
- Outcomes and controls:
  - Deaths and cases: Our World in Data, Johns Hopkins University.
  - Macroeconomic variables: IMF WEO (January 2021 vintage and other vintages for robustness).
  - Fiscal support: IMF policy tracker and Fiscal Monitor (above-the-line support as percent of GDP).
  - Mobility: Google Community Mobility Reports and Apple Mobility Trends summarized via principal components.
  - Health preparedness: hospital beds per thousand (Our World in Data).
  - Country classification: AEs, EMDEs, LIDCs.

### Selected quantitative results and statistics
- Instrument strength and IV results:
  - First-stage F-statistic = 17.6.
  - First-stage instrument coefficient on deaths: -0.43*** (standard error 0.12).
  - 2SLS second-stage: instrumented Covid-19 deaths coefficients on 2020 GDP Per Capita Growth (%) and 2020 GDP Growth (%) include -5.97* , -6.37** , -6.00* , -6.88** , -6.73** , -6.35** (standard errors 3.45 to 3.24).
- Containment measures and deaths (selected coefficients from tables):
  - Containment (average) coefficients in Table 2 up to 543.993 (Dec.) with standard errors such as (197.570).
  - Containment (at 100 cases) negative coefficients: -77.058* (Apr.), -161.650 (Dec.) with standard errors (40.723) and (134.740).
  - Stringency (average) for Deaths/pop: 278.327*** (Apr.) to 682.713*** (Dec.) with standard errors (93.063) to (176.121).
  - Health policies (average) for Deaths/pop: -172.993** (Apr.) to -513.122* (Dec.) with standard errors (81.060) to (302.976).
- Individual policy effects (Table 4, average-response coefficients for Deaths/pop):
  - School closures: 430.036*** (average), 529.049** (early response).
  - Workplace closures: 400.841*** (average), 553.320*** (early response).
  - Cancellations of public events: 407.690*** (average), 602.445** (early response).
  - Restrictions on gatherings: 368.835*** (average), 537.575** (early response).
  - International travel controls: -399.315*** (average), -508.694*** (early response).
  - Testing policy: -358.133** (average), -470.073*** (early response).
  - Contact tracing: -125.393 (average), -202.928 (early response).
  - Mask requirements: 207.184 (average), 102.368 (early response).
- Age and health capacity effects:
  - Median age coefficients include 3.684* (Apr.) up to 32.186*** (Dec.) with standard errors (1.987) to (6.790).
  - Hospital beds / 1,000 population coefficients include -9.127** (Apr.) to -19.613 (Dec.) with standard errors (4.531) to (15.373).
- Economic impacts of deaths (Tables 8–9):
  - Cumulated Covid-19 deaths coefficients for 2020 GDP Growth Per Capita (%): -2.73* (Apr 2020 WEO), -5.34*** (Jul 2020), -3.78*** (Oct 2020), -2.14** (Jan 2021 vs. 2019), -2.58***, -2.29** (Jan 2021 actual) with standard errors (1.38) to (0.96).
  - Similar coefficients for 2020 GDP Growth (%) in Table 9: -2.653* (Apr), -5.270*** (Jul), -3.711*** (Oct), -2.118** (Jan 2021 vs. 2019), -2.538***, -2.872*** with standard errors (1.376) to (0.949).
- Stringency, mobility, and fiscal outcomes:
  - Table 18: Containment (average) coefficient -2.32 (Actual vs. 2019); Stringency (average) -3.81* (Actual vs. 2019) with standard errors (2.15) and (2.03).
  - Table 19: Google mobility (average) positively associated with 2020 GDP Per Capita Growth (%): 1.13*** (Actual vs. 2019), 0.67* (Change vs. Jan 2020), 0.80** (Revision) with standard errors (0.42) to (0.40).
  - Table 22: Stringency (average) and Primary Balance examples: -0.96, -3.40* , -3.27** with standard errors (1.88) to (1.61).
  - Table 24: Covid-19 fiscal support in 2020 associated with Primary Balance: -0.57***, -0.31**, -0.17 with standard errors (0.11) to (0.10).
  - Tables 25–26: EMDEs and LIDCs coefficients for fiscal support negative and significant, e.g., EMDEs -2.85***, LIDCs -3.10** (Table 25, standard errors 0.74 and 1.26).

### Narrative evidence and case examples
- Success stories and patterns:
  - Countries with past SARS experience largely rank among the top-30 lowest age-adjusted death rates; Canada ranks 31st.
  - Low recorded deaths per million markedly lower in Asia and Africa; lower incidence in lower-income countries partly due to less testing and younger populations.
  - Vietnam example: by July 15 had only 380 cases and no deaths; early wide-ranging social distancing, large-scale testing, mobilized contact tracers, and strong public information campaign; Lowy Institute Covid Performance Index ranks Vietnam 2nd and the Philippines 79th in the 36 weeks following their hundredth confirmed case.
- Dynamics across waves:
  - Composition of successful countries changed with subsequent waves; some European countries successful until September 2020 (Lithuania, Slovak Republic, Austria, Georgia, Croatia, Slovenia) were unable to contain spread in late 2020.
  - Countries successful only in wave 1 relaxed stringency measures (international and domestic) relatively more than counterparts that remained successful, potentially contributing to later surges.
- Caveat: data underreporting and testing limitations noted throughout analysis.

*Source: wpiea2021099-print-pdf - 4.1  The effectiveness of Covid-19 containment measures — https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021099-print-pdf.pdf*

### 4.1  The effectiveness of Covid-19 containment measures. . . . . . . . . .  13

### 4.1  The effectiveness of Covid-19 containment measures.

### Overview and scope
- Analysis covers containment measures implemented in over 150 countries.
- "Early" implementation is defined as the measures in place at the time when the country reached 100 reported Covid-19 cases.
- The paper compares stringency-based measures (e.g., lockdowns, closures, restrictions on mobility) versus public health measures (e.g., wide-scale testing, contact tracing, public information campaigns).

### Key empirical findings
- Countries that successfully contained Covid-19 generally combined:
  - Early targeted stringency measures (for example, monitoring international travel).
  - Strong public health measures (wide-scale testing, contact tracing, public information campaigns).
- Asian economies with prior SARS experience (Hong Kong SAR, Taiwan Province of China, Singapore, Vietnam) acted quickly with targeted stringency and strong health measures, and did not rely on quicker or stricter stay-at-home orders, workplace closures, transport closures, or internal mobility restrictions.
- Lower deaths and more successful containment in 2020 (measured by age-adjusted mortality rates) are associated with less stringent containment measures on average throughout the year, but with stronger public health measures.
- Early stringency measures helped curb deaths during a large part of the year, although early stringency is not significant in explaining death rates recorded by the end of the year.
- Countries with stricter containment measures, on average, experienced lower GDP growth in 2020.
- Interaction effects: Covid-19 death rates are negatively related with economic growth, especially when average stringency measures were high.

### Evidence on economic and fiscal trade-offs
- Several external estimates referenced:
  - A 10 percent output loss from reduced economic activity and an equivalent fiscal cost (Gourinchas, 2020).
  - Based on survey data, 50 percent of survey participants incurred income losses averaging 5,293 U.S. dollars and wealth losses averaging 33,482 U.S. dollars (Coibion et al., 2020).
  - Containment measures resulted in a loss of about 15 percent for industrial production over a 30-day horizon (Deb et al., 2020a).
- This paper’s empirical results:
  - Lower deaths and more successful containment are associated with better GDP growth outcomes in 2020.
  - Lower death rates are associated with higher primary balances in 2020 and reduced debt levels in the medium run.
  - These results hold when controlling for age structure and Covid-19 fiscal support measures in sensitivity tests.

### Identification and robustness
- To address possible endogeneity, a two-stage-least-squares approach is employed:
  - Instrument: past experience in containing coronavirus epidemics.
  - First-stage: past experience is associated with significantly lower death rates, controlling for countries’ income level.
  - The F-statistic indicates the instrument is strong.
  - Second-stage: lower Covid-19 deaths, instrumented by past experience, have a positive impact on economic growth in 2020.
- Results are robust to controls including population age structure and fiscal support measures.

### Policy implications and recommended strategy
- There is not necessarily a trade-off between saving lives and saving the economy.
- A "smart" containment strategy is recommended, centered on:
  - Testing,
  - Contact tracing,
  - Public information campaigns,
  - Targeted stringency early on (for example, international travel controls, school closures, cancellation of public events).
- Countries with limited past epidemic experience often relied on stronger, longer-lasting stringency measures that reduced deaths but implied a trade-off for economic growth.
- Smart containment can mitigate both health and economic costs by emphasizing public health measures and early, targeted restrictions rather than broad, prolonged lockdowns.

*Source: wpiea2021099-print-pdf - 4.1  The effectiveness of Covid-19 containment measures. . . . . . . . .  13 — https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021099-print-pdf.pdf*

### Section 3 describes the data used in the empirical analysis. Section 4 presents the

### wpiea2021099-print-pdf - Section 3 describes the data used in the empirical analysis. Section 4 presents the

### Narrative Discussion of Success Stories
- Recorded Covid-19 deaths per million inhabitants are markedly lower in Asia and Africa by the metric plotted in Figure 1; population age structure is noted as a likely significant factor.
- Recorded cases and deaths have been substantially lower in lower income countries (Figure 2), partly due to lack of widespread testing, younger populations, and the strong association between mortality and symptomatic cases and age.
- Three historically important epidemics since 2000 and their recorded impacts:
  - SARS (2003): 8,000 recorded cases, including 774 deaths; significant outbreaks in Mainland China, Hong Kong SAR, Taiwan Province of China, Canada, Singapore, and Vietnam (over 50 cases in each).
  - MERS (2013): over 2,500 cases and 881 recorded deaths; largest outbreaks affecting Saudi Arabia, South Korea, and United Arab Emirates (over 50 cases in each).
  - Ebola virus disease (EVD) (2014-2016 West Africa): estimated over 25,000 cases and 11,000 deaths; significant outbreaks in Guinea, Liberia, and Sierra Leone. 2018 outbreak in the Democratic Republic of Congo: over 2,000 confirmed and probable cases reported and 1,357 deaths.
- Countries with recent past pandemic experiences have been relatively successful in containing Covid-19 deaths:
  - Countries with past SARS experience saw low cases and deaths per capita, especially given their population age structure and widespread testing.
  - Countries with past MERS experience had low mortality but much larger outbreaks as measured by number of cases.
  - Lower-income countries previously hit by Ebola had less prevalent testing and younger populations that helped cushion impact.
- Ranking and incidence observations:
  - Table 1 ranks the top 30 countries with lowest age-adjusted death rates per capita; the six main countries with previous SARS experience all rank within the top-30, except Canada which ranks 31st.
  - Figure 3 shows markedly lower incidence of Covid-19 in the top-30 "high success" countries and the six countries with past SARS experience, compared to the rest of the world.
- Policy versus luck:
  - Factors like geographic position, connectivity, and population density may have reduced exposure for some countries, but the analysis argues policies played a much more important role.
  - Countries with past SARS outbreaks (excluding China in the SARS average) acted faster and implemented different strategies:
    - Implemented at least some stringency measures remarkably earlier and had stronger health policies throughout the year (Figures 4 and 5).
    - Quickly imposed restrictive measures such as international travel restrictions, school closures, and cancellation of public events.
    - Were proactive in implementing stronger health policies such as public information campaigns, testing policy, and contact tracing.
    - Relied relatively less on measures such as stay-at-home requirements, closures of public transport, restrictions on internal mobility.
    - Public information campaigns appear sufficient for people to ‘act cautiously’ and reduce mobility (Figure 6).
  - Early and targeted containment strategies (rapid implementation of targeted stringency, mass testing, contact tracing, public information campaigns) allowed these countries to curb spread despite proximity and interconnection with China.
- Vietnam case summary:
  - By July 15 had only 380 cases and no deaths related to Covid-19 despite large-scale testing.
  - Early response included wide-ranging social distancing and movement restrictions early on, mobilizing a large number of contact tracers (using a low-cost approach), and a strong public information campaign.
  - Lowy Institute’s Covid Performance Index ranks Vietnam 2nd and the Philippines 79th in the 36 weeks following their hundredth confirmed case; Vietnam began easing lockdown restrictions as early as April 23rd.
- Other "high success" countries (selected examples from Table 1): Finland, Estonia, Norway, Denmark, Iceland, Latvia, Japan, New Zealand, Australia, Fiji, Sri Lanka, Barbados, Jamaica, Trinidad and Tobago, Mauritius, Cyprus, Greece, Uruguay, Germany, Thailand, Malaysia, Canada.
- Dynamics over subsequent waves:
  - Composition of successful countries changed with subsequent waves; some European countries successful until September 2020 (Lithuania, Slovak Republic, Austria, Georgia, Croatia, Slovenia) were unable to contain spread in late 2020.
  - Figure 9 shows countries successful only in wave 1 relaxed several stringency measures (international and domestic) relatively more than counterparts that remained successful, potentially contributing to later surges.
- Caveat: data underreporting is noted as a caveat of the ranking.

### Data and Descriptive Statistics
- Sample and period:
  - Cross-sectional data for over 150 countries.
  - OxCGRT daily government measures spanning January 1- December 31, 2020, covering over 150 countries.
- Containment measures data (OxCGRT):
  - Overall government response index considers 19 indicators across: (i) containment and closure policies (8 indicators), (ii) health system policies (7 indicators), and (iii) economic policies (4 indicators).
  - Containment and closure sub-indicators: school closures, workplace closures, cancellation of public events, restrictions on gathering, public transport closures, stay-at-home requirements, restrictions on internal movement, international travel controls.
  - Health system sub-indicators: public information campaigns, testing policy, contact tracing, announced investment in healthcare, announced spending on vaccine development, facial covering requirements, vaccination policies.
  - Economic policy sub-indicators: cash payments to households, freezes on financial obligations for households, announced economic stimulus spending, international support to other countries.
- Main containment indicators used:
  1. ‘Overall containment measures’: overall ‘containment and health index’ (includes 8 containment/closure indicators and first 3 health indicators: public information campaigns, testing policy, contact tracing).
  2. ‘Stringency measures’: ‘stringency index’ (8 containment/closure indicators and public information campaigns), capturing policies that restrict movements.
  3. ‘Public health measures’: principal component of the 3 first health system indicators (public information campaigns, testing policy, contact tracing) focusing on ‘smart’ health measures.
- Aggregation summary indicators:
  1. ‘Average strength’ of containment measures: average value over the year (or sub-period); does not account for timing.
  2. ‘Early strength’ of containment measures: measures in place when the 100th case was recorded; accounts for timing.
  - All containment indicators normalized on a 0-1 scale.
- Covid-19 cases and deaths data:
  - Daily recorded Covid-19 cases and deaths from Our World in Data (OWID) database.
  - Complemented by Coronavirus Resource Center of Johns Hopkins University tracking official country announcements.
- Containment success metrics:
  - Lowy Institute’s Covid-19 Performance Index measures 98 countries’ relative success in the 36 weeks following their 100th confirmed case (based on confirmed cases and deaths, totals and per capita, cases as a proportion of tests, tests per thousand people).
  - Authors construct a ranking of containment success based on age-adjusted death rates for over 150 countries; age-adjusted death rates calculated as the residual when regressing deaths per capita on the share of population aged 70 and above.
- Macroeconomic variables (IMF WEO):
  - Variables from IMF’s World Economic Outlook (WEO) database: GDP per capita (in PPP USD terms), real GDP growth, real per capita GDP growth, primary balance (percent of GDP), gross public debt (percent of GDP).
  - Primary data source: January 2021 WEO; also consider projection revisions across vintages October 2019, January 2020, July 2020, October 2020.
- Fiscal support measures:
  - Data from IMF’s policy tracker on policy responses and IMF Fiscal Monitor database of Covid-19 Fiscal Response Measures published in June 2020, October 2020, and January 2021.
  - Use total above-the-line fiscal support announcements and amounts implemented in 2020, measured as percent of GDP.
- Mobility:
  - Google Community Mobility Report daily data by country across categories (retail and recreation, groceries and pharmacies, parks, transit stations, workplaces, residential); authors create a principal component summarizing categories into overall mobility indicator.
  - Apple Map’s Mobility Trends Report dataset also used.
- Health preparedness:
  - Hospital beds per thousand inhabitants from Our World In Data.
- Population metrics:
  - Population age, size, and density from Our World in Data.
- Country classification:
  - WEO country income groups: advanced economies (AEs), low-income developing countries (LIDCs), other emerging market and developing economies (EMDEs); differences in death rates, containment measures, and macroeconomic performance are noted and controlled for.
- Descriptive patterns highlighted:
  - Figure 2: higher incidence of Covid-19 cases and deaths reported by advanced economies, followed by emerging markets, and least by low-income developing countries, reflecting differences in reporting, testing policy, demographics, connectivity, and containment measures.
  - Figure 10: stringency measures on average stricter in emerging markets; more stringent measures in place earlier (at 100 cases) in developing countries; advanced economies had stronger average health policies, while early health policies were similar across groups.
  - Figure 11: GDP contractions in 2020 largest for emerging markets, followed by advanced economies, smallest for low-income developing countries.
  - Figure 12: primary deficits and increases in public debt largest for advanced economies, followed by emerging markets, smallest for low-income developing countries.
  - Figure 13: Covid-19 fiscal support measures substantially larger for advanced economies relative to developing countries (total above-the-line support as share of GDP).

### Empirical Analysis — The effectiveness of Covid-19 containment measures
- Two primary empirical objectives:
  1. Study the role of containment measures in reducing spread and severity of the pandemic, measured by deaths per capita, focusing on strategies of more successful countries.
  2. Assess the effect of containing Covid-19 on the macroeconomy and countries’ public finances.
- Literature context:
  - Several studies cited in the literature review show different government measures have been effective in containing spread (section begins empirical analysis building on this literature).
- Analytic approach elements (as described in Sections 3 and 4 setup):
  - Use of cross-country regression frameworks where age-adjusted death rates are an outcome (age-adjusted defined as residuals from regressing deaths per capita on share of population aged 70+).
  - Containment policy variables include aggregate indices ('overall containment measures', 'stringency measures', and 'public health measures'), summarized either as ‘average strength’ over the year or ‘early strength’ at the 100th case.
  - Controls and heterogeneity: WEO country income groups (AEs, EMDEs, LIDCs) and macro variables from WEO; testing and reporting limitations noted as data caveats.
- Key empirical patterns described qualitatively:
  - Countries with past SARS experience implemented earlier and stronger public health measures and targeted stringency (international travel controls, school closures, cancellation of public events) and relied less on broad mobility restrictions.
  - Early implementation of health measures (testing, contact tracing, public information) and targeted travel controls are associated with lower age-adjusted death rates in the cross-country comparisons and rankings.
  - Countries that strengthened testing policy and maintained stricter international travel controls while relaxing many domestic stringency measures were among those successful in containing age-adjusted death rates.
  - Countries that relaxed stringency measures relatively more after initial success were more likely to experience surges in subsequent waves.
- Note on limitations:
  - Testing data not available for as wide a sample; thus age-adjusted death rates are used rather than infection rates because infection rates would be even more dependent on countries’ testing policies.
  - Data underreporting acknowledged as a caveat of the ranking and cross-country comparisons.

*Italic: Source: wpiea2021099-print-pdf (content provided from sections 2–4 of the PDF).*

### 19. For example, Cowling et al. (2020) show that non-pharmaceutical interventions

### wpiea2021099-print-pdf - 19. For example, Cowling et al. (2020) show that non-pharmaceutical interventions

### Detecting “smart” measures
- Objective: identify containment measures taken by countries that successfully contained the pandemic and how they differed from other countries.
- Key distinction emphasized:
  - Stringency measures (e.g., lockdown restrictions).
  - Health policy measures (e.g., widespread testing and contact tracing).
- Importance of timing: imposing containment measures early, before the virus becomes widespread, is highlighted.

### Empirical approach and key regressions
- Cross-country regression to relate deaths to containment measures:
  - Death rate_i = β0 + β1 C_i + β2 X_i + u_i
  - Death rate_i denotes death rate per thousand inhabitants in country i.
  - C_i captures Covid-19 containment measures (average level since March 2020 and level implemented early on when country had 100 recorded cases).
  - X_i includes controls: median age, hospital beds per capita, GDP per capita, WEO country income groupings.
- Caveat: specification does not address endogeneity; results are correlations rather than causal relations.
- Outcome regression for economic impacts:
  - Y_i,h = β0 + β1 Death rate_i + β2 X_i + u_i
  - Y_i,h denotes outcome variable in country i at horizon h = 2020,...,2025.
  - Controls include log GDP per capita in U.S. dollar PPP terms and WEO country income groups.

### Evidence on containment measures and deaths
- Main empirical findings:
  - Strong stringency measures implemented early (at the time of 100 cases) are significantly associated with lower deaths up until October 2020.
  - Stronger overall public health measures (average during sample period) are associated with lower deaths per capita.
  - Stronger overall stringency (average during sample period) is positively associated with deaths per capita (likely due to endogenous relationship: stricter measures in places where pandemic was less contained).
  - Individual stringency measures are mostly positively related with country death rates (reverse causality), with the exception of international travel controls, which are strongly associated with reduced contagion.
  - Robust evidence in favor of strong health policies (e.g., widespread testing) being associated with lower death rates.
- Cross-checks and samples:
  - Table 5: t-test statistics comparing the 30 countries with lowest age-adjusted mortality rates to the rest of the sample.
    - These more-successful countries did not have stronger containment measures in place early on (at the time of 100 cases).
    - They had stronger health measures (testing policy and contact tracing) and stronger international travel controls throughout the year.
    - Their average stringency policies were less strict on average throughout the year.
  - Table 7: regressing Lowy Institute’s Covid Performance Index (measuring countries’ relative success over the 36 weeks following countries’ 100th confirmed case) on containment measures:
    - Less successful countries had more stringent measures on average (reverse causality).
    - Countries with stronger health measures and stronger international travel controls on average fared better.
    - Countries with more stringent measures in place early on (at 100 cases) also performed better.

### Does saving lives imply higher economic and fiscal costs?
- Summary of empirical results:
  - Lower deaths per capita are associated with higher economic growth, without significantly impacting countries’ primary balances and without being associated with significantly different fiscal support packages.
  - Countries with a higher success ranking (lower age-adjusted mortality rates) experienced higher GDP growth without significantly different fiscal costs.
  - Using past epidemic experience as an instrument confirms that containing Covid-19 deaths is associated with higher growth in 2020, without significantly impacting medium-term fiscal position.
- Specific econometric and data notes:
  - Outcomes from IMF WEO used: real GDP per capita growth rates, real GDP growth rates, primary balance as a share of GDP, public debt as a share of GDP.
  - WEO vintage: January 2021 used for actual 2020 levels and projections; results robust to other vintages (e.g., October 2019).
- Findings on the role of Covid-19 deaths:
  - Countries with higher Covid-19 death rates experienced:
    - Larger downward growth forecast revisions for each WEO publication in 2020.
    - A sharper fall in growth compared to 2019.
    - A significantly lower realized growth rate in 2020.
  - Higher deaths per capita were not associated with significantly different primary balances or with larger fiscal support packages implemented in 2020 or announced as of January 2021 or October 2020.
  - Interaction with income group: advanced economies that experienced higher death rates announced relatively smaller fiscal support packages as of late 2020/early 2021.
  - Medium-term projections: reduced deaths associated with higher GDP growth in 2020, lower GDP growth in 2021, and insignificantly different growth projections for future years; no significant differences for primary balances or public debt as a share of GDP throughout the forecast horizon.

### Role of past coronavirus experience (instrument)
- Instrument construction:
  - Dummy variable for nine countries with past coronavirus pandemic experiences from SARS and MERS: Taiwan Province of China, Singapore, Vietnam, Canada, Mainland China, Hong Kong SAR, Saudi Arabia, South Korea, United Arab Emirates.
- 2SLS evidence:
  - First stage: past experience is significantly associated with lower Covid-19 deaths per capita; F-statistic = 17.6 (above the Staiger-Stock rule of thumb of 10).
  - Second stage: lower deaths (instrumented by past experience) are associated with higher real GDP growth in 2020 and relatively better performance compared to 2019 and January 2020 WEO expectations.
  - Table 15: lower Covid-19 deaths associated with lower primary balances and higher fiscal support in 2020 (countries with higher success based on past experience offered larger fiscal support, which may partly explain better growth).
- Robustness:
  - Results robust to controlling for median age and to controlling for Covid-19 fiscal support measures implemented in 2020.
  - Alternative instruments tested:
    - Including countries with past Ebola experience in the instrument.
    - Considering only countries with past SARS experience.
  - Both alternative instruments remain significantly associated with lower Covid-19 death rates and with higher GDP growth in 2020 when instrumenting deaths.

### Containment measures and economic outcomes
- Relationship between containment measures and economic performance:
  - Regression: Y_i = β0 + β1 C_i + β2 X_i + u_i, where Y_i is 2020 outcome and C_i average containment measures across the year.
  - Table 18: stronger average Covid-19 containment measures associated with lower economic growth, driven by stringency measures; health policies are insignificant in this regression.
  - Accounting for deaths (Table 19):
    - After controlling for deaths, evidence still suggests higher average stringency may have contributed to slowing economic activity.
    - Higher deaths, controlling for stringency, are associated with lower activity (possibly due to voluntary social distancing).
  - Interaction effects:
    - The marginal effect of the Covid-19 death rate on economic growth is lower (negative and significant) when it comes at the cost of higher average stringency measures.
  - Mobility:
    - Controlling for mobility fully explains any effects captured by stringency measures.
  - Fiscal outcomes:
    - Table 22 (columns 3 and 6): countries with stronger average measures (explained by stronger stringency) saw larger downward revisions in primary balances; robust to controls for pandemic severity and reduced mobility.
    - No general evidence that fiscal support packages depended on containment measures, pandemic severity, or mobility reduction (Table 26), with an exception:
      - Advanced economies provided larger fiscal support in 2020 in countries that applied more stringent containment measures (Table 25, column 4).
    - Fiscal space likely constrained fiscal support in many countries, especially lower-income countries; thus more stringent containment measures were not, on average, accompanied by more fiscal support.

### Concluding remarks and policy implications
- Synthesis of results:
  - Countries that imposed stronger containment measures on average experienced higher output losses.
  - Countries that contained the spread through smart containment measures fared better in economic growth.
  - Successful cases (including those with past epidemic experience) combined:
    - Early restrictions (including travel controls).
    - Smart containment: large-scale testing, contact tracing, public information campaigns.
- Policy recommendations:
  - Authorities should prepare to adopt smart strategies to fight possible new waves of infections.
  - Lessons for future pandemics: early travel controls plus strong health policies (testing, tracing, information) can reduce contagion while allowing relaxation of stringency measures later.

*Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021099-print-pdf.pdf*

### References

### References and Appendix (selected empirical results)

### Key empirical findings from tables and figures
- Containment measures and health policies have statistically significant associations with Covid-19 deaths per capita and country performance:
  - Containment (average) coefficients in Table 2 range up to 543.993 (Dec.) with standard errors such as (197.570); Containment (at 100 cases) shows negative coefficients like -77.058* (Apr.) and -161.650 (Dec.) with standard errors (40.723) and (134.740).
  - Stringency (average) in Table 3 shows large positive coefficients for Deaths/pop: 278.327*** (Apr.) to 682.713*** (Dec.) with standard errors (93.063) to (176.121).
  - Health policies (average) in Table 3 show negative coefficients for Deaths/pop: -172.993** (Apr.) to -513.122* (Dec.) with standard errors (81.060) to (302.976).
- Age and health system capacity matter:
  - Median age coefficients (Table 2) include 3.684* (Apr.) up to 32.186*** (Dec.) with standard errors (1.987) to (6.790).
  - Hospital beds / 1,000 population coefficients (Table 2) include -9.127** (Apr.) to -19.613 (Dec.) with standard errors (4.531) to (15.373).
- Economic classification differences:
  - EMDEs coefficients (Table 2) include -87.784*** (Apr.) and -24.388 (Dec.) with standard errors (26.245) to (90.340).
  - LIDCs coefficients (Table 2) range from -67.461 to -206.126 with standard errors (49.323) to (164.812).

### Containment measures — specific policy effects (Table 4, Table 3)
- Individual containment measures (one at a time) — average-response coefficients for Deaths/pop (Table 4):
  - School closures: 430.036*** (average), 529.049** (early response)
  - Workplace closures: 400.841*** (average), 553.320*** (early response)
  - Cancellations of public events: 407.690*** (average), 602.445** (early response)
  - Restrictions on gatherings: 368.835*** (average), 537.575** (early response)
  - International travel controls: -399.315*** (average), -508.694*** (early response)
  - Testing policy: -358.133** (average), -470.073*** (early response)
  - Contact tracing: -125.393 (average), -202.928 (early response)
  - Mask requirements: 207.184 (average), 102.368 (early response)
- Table 3 contrasts Stringency vs Health policies:
  - Stringency (at 100 cases) shows negative coefficients: -135.801*** (Apr.) to -92.057 (Dec.) with standard errors (42.304) to (125.434).
  - Health policies (at 100 cases) show mixed coefficients, e.g., 68.694 (Apr.), 80.589 (May), 28.796 (Jun.), and negative later months.

### Covid-19 containment success and performance (Tables 6–7, Table 12)
- Containment success rank is negatively associated with deaths and positively associated with GDP outcomes:
  - Containment success rank coefficient in Table 6: -6.07*** (DeathsSuccess) with standard error (0.36).
  - In Table 7, Containment (2020 average) coefficient for Score: -62.750*** (standard error 13.046); Containment (at 100 cases) coefficients: 31.924*** and 28.356*** (standard errors 8.574 and 7.900).
- Table 12: Containment success rank positively associated with 2020 GDP Per Capita Growth (%) and 2020 GDP Growth (%):
  - Coefficients include 0.019***, 0.015**, 0.013*, 0.017**, 0.015**, 0.014** with standard errors (0.007).

### Economic and fiscal impacts associated with Covid-19 deaths (Tables 8–11, 18–26)
- GDP growth revisions and deaths (Tables 8–9):
  - Cumulated Covid-19 deaths coefficients for 2020 GDP Growth Per Capita (%): -2.73* (Apr 2020 WEO), -5.34*** (Jul 2020), -3.78*** (Oct 2020), -2.14** (Jan 2021 vs. 2019), -2.58***, -2.29** (Jan 2021 actual), with standard errors (1.38) to (0.96).
  - Similar patterns in Table 9 for 2020 GDP Growth (%): -2.653* (Apr), -5.270*** (Jul), -3.711*** (Oct), -2.118** (Jan 2021 vs. 2019), -2.538***, -2.872*** with standard errors (1.376) to (0.949).
- Primary balance revisions (Table 10):
  - Cumulated Covid-19 deaths coefficients vary and are generally not significant in later columns: 1.245, 0.657, -1.217, -0.352, -0.555, -0.390 with standard errors (2.103) to (0.951).
  - Log GDP per capita shows negative and significant coefficients in several specifications, e.g., -1.204** and -1.321*** (standard errors 0.468, 0.459).
- Fiscal support packages and deaths (Table 11):
  - Covid-19 deaths coefficients in some specifications: 0.54, 0.30, 0.14, 0.98, 1.63, 2.01** with standard errors (0.65) to (0.85).
  - Interactions with AE dummy show negative coefficients: Covid-19 deaths x AE dummy -1.03, -3.08* , -4.30*** (standard errors 1.32 to 1.79).
- Relation of stringency, mobility, and fiscal outcomes (Tables 18–26):
  - Table 18: Containment (average) coefficient -2.32 (Actual vs. 2019) and Stringency (average) -3.81* (Actual vs. 2019) with standard errors (2.15) and (2.03).
  - Table 19: Google mobility (average) positively associated with 2020 GDP Per Capita Growth (%): 1.13*** (Actual vs. 2019), 0.67* (Change vs. Jan 2020), 0.80** (Revision) with standard errors (0.42) to (0.40).
  - Table 22: Stringency (average) and Primary Balance: e.g., Stringency (average) -0.96, -3.40* , -3.27** with standard errors (1.88) to (1.61).
  - Table 24: Covid-19 fiscal support in 2020 associated with Primary Balance: -0.57***, -0.31**, -0.17 (Actual, Change, Revision) with standard errors (0.11) to (0.10).
  - Tables 25–26: Fiscal support regressions show Stringency (average) coefficients near zero and EMDEs / LIDCs negative and significant coefficients for fiscal support, e.g., EMDEs -2.85***, LIDCs -3.10** (Table 25, standard errors 0.74 and 1.26).

### Instrumental-variable and 2SLS results (Tables 14–17, Figures 18–19)
- First-stage instrument strength:
  - Coronavirus experience instrument: coefficient -0.43*** (standard error 0.12) in 1st stage across IV specifications (Tables 14–17).
  - F-stat reported: 17.6 (Table 14).
- 2SLS second-stage estimates (Table 14, Table 15):
  - Table 14: Covid-19 deaths (instrumented) coefficients on 2020 GDP Per Capita Growth (%) and 2020 GDP Growth (%) include -5.97* , -6.37** , -6.00* , -6.88** , -6.73** , -6.35** with standard errors (3.45) to (3.24).
  - Table 15: Instrumented Covid-19 deaths coefficients on fiscal outcomes include 9.79* (Primary Balance actual) and other mixed coefficients, standard errors such as (5.49).
- Robustness and instrument checks (Tables 16–17):
  - Robustness tests show instrumented Covid-19 deaths coefficients such as -6.81** (Deaths second-stage) and 9.37* (Prim. bal.), with additional controls (Median age, Fiscal support) included.
  - Instrument robustness with alternative epidemic experience instruments: e.g., Epidemic experience -0.29***; SARS experience -0.41***; instrumented Covid-19 deaths coefficients like -7.04* (Deaths) and -7.44* (Fiscal) with standard errors (4.15) to (7.07).

### Figures and tables inventory (Appendix A.1)
- Figures listed include:
  - Figure 1: Global Waves of Covid-19 in 2020
  - Figure 2: Covid-19 Incidence by Income Group
  - Figures 3–20: Various timelines, containment measures comparisons, mobility trends, OLS and 2SLS effects on WEO growth and fiscal projections, marginal effects.
- Tables listed include:
  - Table 1: 2020 Success Stories in Covid-19 Containment (top-30 countries)
  - Tables 2–26: Empirical regressions on effects of containment measures, individual measures, Covid-19 success rankings, GDP and fiscal revisions, 2SLS regressions, robustness checks, and relations with stringency, mobility, and fiscal packages.
- Notes on statistical reporting:
  - Standard errors are reported in parentheses throughout tables.
  - Significance codes: *** p<0.01, ** p<0.05, * p<0.1.

*Source: References and Appendix tables and figures from the PDF content unit.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021099-print-pdf.pdf_