## Containment measures and COVID-19: empirical evidence (unit summary)

## Source details

**Canonical URL:** [Containment measures and COVID-19: empirical evidence (unit summary)](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020159-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020159-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020159-print-pdf.pdf.json)

---

### Introduction and scope
- Study period: January 1 to June 15, 2020 (daily data for main analysis); data cut-off for some sources: June 16, 2020; snapshot cited as of June 26, 2020.
- Global context cited: "close to 10 million cases" and "nearly 500 thousand deaths" as of June 26, 2020.
- Sample: balanced sample of 129 economies with at least 30 observation days after a significant outbreak (100 cases).
- Objective: quantify the causal effect of non-pharmaceutical containment measures (NPIs) on confirmed COVID-19 infections and deaths, and explain cross-country heterogeneity in effectiveness.

### Key containment measures analyzed
- Eight OxCGRT dimensions: (i) school closures; (ii) workplace closures; (iii) public event cancellations; (iv) gathering restrictions; (v) public transportation closures; (vi) stay-at-home orders; (vii) restrictions on internal movement; (viii) international travel bans.
- Aggregate Stringency Index: average of sub-indices normalized to range between 0 and 1.

### Data sources and coverage
- COVID-19 cases and deaths: Johns Hopkins Coronavirus Resource Center COVID-19 Dashboard; coverage begins January 22, 2020; data cut-off June 16, 2020 for that source.
- Containment measures: Oxford COVID-19 Government Response Tracker (OxCGRT); database starts January 1, 2020 and covers 151 economies.
- Additional NPI controls (testing policies, contact tracing, public information campaigns): OxCGRT, available for 176 countries from January 1, 2020.
- Temperature and humidity: Air Quality Open Data Platform; daily data for 95 countries from January 1, 2020.
- Tests conducted: Our World in Data; total tests and tests per thousand in 84 countries from January 1, 2020.
- Population and demographics: World Development Indicators (189 countries).
- Health robustness indicators: Global Health Security Index (195 countries) and Health Index (0–7 scale) from Global Competitiveness Report 2019 (114 economies).
- Mobility: Apple Maps Mobility Trends (walking and driving indices for 66 countries, starting January 13, 2020) and Google Mobility Reports (retail and transit-station mobility for 89 countries, starting February 15, 2020).

### Econometric approach
- Reduced-form "difference-in-difference" local-projection framework (Jordà (2005) style) to estimate dynamic cumulative effects over horizons h = 0,..,30.
- Main regression controls include: country-fixed effects, daily temperature and humidity, country-specific linear/quadratic/cubic time trends, testing/contact tracing/public information campaign variables, and lagged changes in infections (to address endogeneity from policy responses).
- Second specification: smooth-transition interaction (Granger and Teräsvirta-type) allowing containment effectiveness to vary with normalized country characteristics z (zero mean, unit variance); weighting function F(z) = exp(γz)/(1−exp(γz)) form as specified.
- Standard errors: robust, clustered at the country level; additional robustness with alternative lag structures and Driscoll-Kraay standard errors reported.

### Baseline findings (effects on infections)
- Containment measures have been, on average, "very effective" in flattening the pandemic curve and reducing confirmed infections.
- Country examples and magnitudes:
  - New Zealand: "almost 90 percent" reduction in the number of infections relative to a baseline with no containment measures (computed as (e^{θ_h}-1)*100).
- Weather effects: lagged daily temperature and humidity significantly affect cases and deaths; "cold and dry conditions" facilitate spread.
- Timing matters: Public Health Response Time (PHRT) — days from 100 confirmed cases to implementation of containment measures (excluding international travel restrictions) — is important; countries with low PHRT (faster implementation) "reduced the average number of infections and deaths by 95" while effects were not statistically significant for countries with relatively high PHRT.

### Robustness checks
- Excluding China and excluding the United States do not invalidate main findings.
- Alternative treatment definitions: using a containment dummy (start/end) rather than a severity index produces similar results.
- Controlling for mobility changes (anticipation of measures) does not quantitatively change baseline results (Figure 4a).
- Controlling for leads of the stringency index to account for serial correlation (per Teulings and Zubanov (2014)) preserves main results though magnitudes vary (Figure 4b).
- Results stable to alternative lag structures, projection horizons, and standard-error specifications.

### Heterogeneity: role of country characteristics
- Temperature: containment measures were more effective in countries with lower average temperatures; "stringent containment measures may have reduced the number of confirmed cases by more about 98 percent in countries with very low average temperature" and were not statistically significant in countries with very high average temperature.
- Humidity: similar pattern as temperature (cold/dry facilitates spread).
- Age structure: higher share of population above 65 increased effectiveness of containment measures; for countries with relatively high share of elderly, containment reduced cases "by about 98 percent 30 days after implementation"; negligible effect in very young populations.
- Population density: containment measures were more effective in low-density countries; example magnitude: "reduction by about 97 percent of COVID-19 cases 30 days after implementation" in low-density countries.
- Health preparedness: containment measures more effective in countries with higher Global Health Security Index and higher Health Index; "may have reduced the number of confirmed cases by more about 97 percent in countries with very strong health systems" and were not statistically significant in weak-health-capability countries.
- De facto social distancing: interactions with Apple/Google mobility indicate that containment measures are more effective where they produced larger mobility declines (greater social distancing).

### COVID-19 fatalities
- Effects on deaths are large but produce a delayed/aggregate response: average correlation between confirmed cases and deaths in the sample is above 0.9 (data caveat).
- Focusing on total impact after 30 days: containment measures "may have reduced the number of deaths by close to 97 percent after 30 days relative to a baseline of no containment."
- Country characteristics (population density, health infrastructure) are even more important for fatalities than for infections.

### Relative effectiveness across measure types
- Two estimation approaches used: (i) introduce each measure individually (subject to omitted-variable bias), (ii) include all measures together (subject to multicollinearity).
- Consistent result: all measures contributed significantly to reducing infections.
- Most effective measures (relatively): restrictions on international travel, restrictions on internal movement, and stay-at-home orders.

### Re-openings (easing containment)
- Sample for reopening analysis: balanced panel of 78 countries that lowered the stringency index after its peak.
- Relaxing containment measures: estimated dynamic response over 30 days shows an increase in confirmed cases "by more than 50 percent, relative to a baseline of stringent containment," with statistical significance at the 90 percent confidence level.

### Conclusions and policy implications
- In absence of a vaccine or effective treatments, containment measures are key to halting spread and limiting fatalities.
- Main empirical assessments indicate containment measures significantly reduce infections and deaths; stringent, swift, and well-enforced measures (leading to real mobility reductions) are most effective.
- Faster response (low PHRT), colder/drier climates, higher share of elderly, lower population density, and stronger health systems amplify the effectiveness of containment measures.
- International and internal travel restrictions and stay-at-home orders rank among the most effective individual measures.
- Easing measures leads to increased infections (more than 50 percent in the 30-day window), but the absolute effect is smaller than the effect of tightening measures.

*Source: IMF working paper content (wpiea2020159-print-pdf), analysis of daily data on COVID-19 cases, containment measures, and controls for 129 countries (January 1–June 15, 2020); canonical URL: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020159-print-pdf.pdf*

### References .............................................................................................................

### Containment measures and COVID-19: empirical evidence (unit summary)

### Introduction and scope
- Study period: January 1 to June 15, 2020 (daily data for main analysis); data cut-off for some sources: June 16, 2020; snapshot cited as of June 26, 2020.
- Global context cited: "close to 10 million cases" and "nearly 500 thousand deaths" as of June 26, 2020.
- Sample: balanced sample of 129 economies with at least 30 observation days after a significant outbreak (100 cases).
- Objective: quantify the causal effect of non-pharmaceutical containment measures (NPIs) on confirmed COVID-19 infections and deaths, and explain cross-country heterogeneity in effectiveness.

### Key containment measures analyzed
- Eight OxCGRT dimensions: (i) school closures; (ii) workplace closures; (iii) public event cancellations; (iv) gathering restrictions; (v) public transportation closures; (vi) stay-at-home orders; (vii) restrictions on internal movement; (viii) international travel bans.
- Aggregate Stringency Index: average of sub-indices normalized to range between 0 and 1.

### Data sources and coverage
- COVID-19 cases and deaths: Johns Hopkins Coronavirus Resource Center COVID-19 Dashboard; coverage begins January 22, 2020; data cut-off June 16, 2020 for that source.
- Containment measures: Oxford COVID-19 Government Response Tracker (OxCGRT); database starts January 1, 2020 and covers 151 economies.
- Additional NPI controls (testing policies, contact tracing, public information campaigns): OxCGRT, available for 176 countries from January 1, 2020.
- Temperature and humidity: Air Quality Open Data Platform; daily data for 95 countries from January 1, 2020.
- Tests conducted: Our World in Data; total tests and tests per thousand in 84 countries from January 1, 2020.
- Population and demographics: World Development Indicators (189 countries).
- Health robustness indicators: Global Health Security Index (195 countries) and Health Index (0–7 scale) from Global Competitiveness Report 2019 (114 economies).
- Mobility: Apple Maps Mobility Trends (walking and driving indices for 66 countries, starting January 13, 2020) and Google Mobility Reports (retail and transit-station mobility for 89 countries, starting February 15, 2020).

### Econometric approach
- Reduced-form "difference-in-difference" local-projection framework (Jordà (2005) style) to estimate dynamic cumulative effects over horizons h = 0,..,30.
- Main regression controls include: country-fixed effects, daily temperature and humidity, country-specific linear/quadratic/cubic time trends, testing/contact tracing/public information campaign variables, and lagged changes in infections (to address endogeneity from policy responses).
- Second specification: smooth-transition interaction (Granger and Teräsvirta-type) allowing containment effectiveness to vary with normalized country characteristics z (zero mean, unit variance); weighting function F(z) = exp(γz)/(1−exp(γz)) form as specified.
- Standard errors: robust, clustered at the country level; additional robustness with alternative lag structures and Driscoll-Kraay standard errors reported.

### Baseline findings (effects on infections)
- Containment measures have been, on average, "very effective" in flattening the pandemic curve and reducing confirmed infections.
- Country examples and magnitudes:
  - New Zealand: "almost 90 percent" reduction in the number of infections relative to a baseline with no containment measures (computed as (e^{θ_h}-1)*100).
- Weather effects: lagged daily temperature and humidity significantly affect cases and deaths; "cold and dry conditions" facilitate spread.
- Timing matters: Public Health Response Time (PHRT) — days from 100 confirmed cases to implementation of containment measures (excluding international travel restrictions) — is important; countries with low PHRT (faster implementation) "reduced the average number of infections and deaths by 95" while effects were not statistically significant for countries with relatively high PHRT.

### Robustness checks
- Excluding China and excluding the United States do not invalidate main findings.
- Alternative treatment definitions: using a containment dummy (start/end) rather than a severity index produces similar results.
- Controlling for mobility changes (anticipation of measures) does not quantitatively change baseline results (Figure 4a).
- Controlling for leads of the stringency index to account for serial correlation (per Teulings and Zubanov (2014)) preserves main results though magnitudes vary (Figure 4b).
- Results stable to alternative lag structures, projection horizons, and standard-error specifications.

### Heterogeneity: role of country characteristics
- Temperature: containment measures were more effective in countries with lower average temperatures; "stringent containment measures may have reduced the number of confirmed cases by more about 98 percent in countries with very low average temperature" and were not statistically significant in countries with very high average temperature.
- Humidity: similar pattern as temperature (cold/dry facilitates spread).
- Age structure: higher share of population above 65 increased effectiveness of containment measures; for countries with relatively high share of elderly, containment reduced cases "by about 98 percent 30 days after implementation"; negligible effect in very young populations.
- Population density: containment measures were more effective in low-density countries; example magnitude: "reduction by about 97 percent of COVID-19 cases 30 days after implementation" in low-density countries.
- Health preparedness: containment measures more effective in countries with higher Global Health Security Index and higher Health Index; "may have reduced the number of confirmed cases by more about 97 percent in countries with very strong health systems" and were not statistically significant in weak-health-capability countries.
- De facto social distancing: interactions with Apple/Google mobility indicate that containment measures are more effective where they produced larger mobility declines (greater social distancing).

### COVID-19 fatalities
- Effects on deaths are large but produce a delayed/aggregate response: average correlation between confirmed cases and deaths in the sample is above 0.9 (data caveat).
- Focusing on total impact after 30 days: containment measures "may have reduced the number of deaths by close to 97 percent after 30 days relative to a baseline of no containment."
- Country characteristics (population density, health infrastructure) are even more important for fatalities than for infections.

### Relative effectiveness across measure types
- Two estimation approaches used: (i) introduce each measure individually (subject to omitted-variable bias), (ii) include all measures together (subject to multicollinearity).
- Consistent result: all measures contributed significantly to reducing infections.
- Most effective measures (relatively): restrictions on international travel, restrictions on internal movement, and stay-at-home orders.

### Re-openings (easing containment)
- Sample for reopening analysis: balanced panel of 78 countries that lowered the stringency index after its peak.
- Relaxing containment measures: estimated dynamic response over 30 days shows an increase in confirmed cases "by more than 50 percent, relative to a baseline of stringent containment," with statistical significance at the 90 percent confidence level.

### Conclusions and policy implications
- In absence of a vaccine or effective treatments, containment measures are key to halting spread and limiting fatalities.
- Main empirical assessments indicate containment measures significantly reduce infections and deaths; stringent, swift, and well-enforced measures (leading to real mobility reductions) are most effective.
- Faster response (low PHRT), colder/drier climates, higher share of elderly, lower population density, and stronger health systems amplify the effectiveness of containment measures.
- International and internal travel restrictions and stay-at-home orders rank among the most effective individual measures.
- Easing measures leads to increased infections (more than 50 percent in the 30-day window), but the absolute effect is smaller than the effect of tightening measures.

*Source: IMF working paper content (wpiea2020159-print-pdf), analysis of daily data on COVID-19 cases, containment measures, and controls for 129 countries (January 1–June 15, 2020); canonical URL: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020159-print-pdf.pdf*

### References

### wpiea2020159-print-pdf - References

### Cited works
- Alesina, A.F., Furceri, D., Ostry, J.D., Papageorgiou, C. and Quinn, D.P., 2019. Structural Reforms and Elections: Evidence from a World-Wide New Dataset (No. w26720). National Bureau of Economic Research.
- Auerbach, A.J. and Gorodnichenko, Y., 2013. Output spillovers from fiscal policy. American Economic Review, 103(3), pp.141-46.
- Brotherhood L., Kircher P., Santos C., Tertilt M., 2020. An economic model of the COVID-19 epidemic: the importance of testing and age-specific policies. CEPR Discussion Paper DP14695, Centre for Economic Policy Research.
- Chinazzi et al., 2020. The effect of travel restrictions on the spread of the 2019 novel coronavirus (COVID-19) outbreak. Science, 368, 395-400.
- Cowling et al., 2020. Impact assessment of non-pharmaceutical interventions against coronavirus disease 2019 and influenza in Hong Kong: an observational study. Lancet Public Health 2020; 5: e279–88.
- Eichenbaum M., Rebelo S., Trabandt M., 2020. The Macroeconomics of Epidemics. NBER Working Papers 26882, National Bureau of Economic Research, Inc.
- Forslid R., Herzing M., 2020. Assessing the consequences of quarantines during a pandemic. CEPR Discussion Paper DP14699, Centre for Economic Policy Research.
- Granger, C.W.J. and Teräsvirta, T., 1993. Modelling Nonlinear Economic Relationships Oxford University Press. New York.
- Hsiang, S., Allen, D., Annan-Phan, S., Bell, K., Bolliger, I., Chong, T., Druckenmiller, H., Huang, L.Y., Hultgren, A., Krasovich, E. and Lau, P., 2020. The effect of large-scale anti-contagion policies on the COVID-19 pandemic. Nature, pp.1-9.
- Jordà, Ò., 2005. Estimation and inference of impulse responses by local projections. American economic review, 95(1), pp.161-182.
- Kraemer et al., 2020. The effect of human mobility and control measures on the COVID-19 epidemic in China. Science, 368, 493–497.
- H. Tian et al., 2020. An investigation of transmission control measures during the first 50 days of the COVID-19 epidemic in China. Science10.1126/science.abb6105 (2020).
- Ramey, V.A. and Zubairy, S., 2018. Government spending multipliers in good times and in bad: evidence from US historical data. Journal of Political Economy, 126(2), pp.850-901.
- Teulings, Coen N., Nikolay Zubanov. 2014. “Is Economic Recovery a Myth? Robust Estimates of Impulse Responses.” Journal of Applied Econometrics, Vol. 29: 497-514.

### Figure notes and empirical setup (methodology and data points preserved verbatim)
- Figure 1 Panel A: "Total cases (log scale)". Panel B: "Stringency of containment measure".
  - Source: Haver, OxCGRT Stringency Index and IMF Staff calculations.
  - Note: The index is scaled to vary between zero (least stringent) to 1 (most stringent) containment measures.
  - Index composed of the following categories: (i) School closing; (ii) Workplace closing; (iii) Cancel public events; (iv) Restrictions on gathering size; (v) Close public transport; (vi) Stay at home requirements; (vii) Restrictions on internal movement; (viii) Restrictions on international travel.
- Sample and timing across figures:
  - "Impulse response functions are estimated using a sample of 129 countries using daily data from the start of the outbreak."
  - "The analysis is restricted to countries with a significant outbreak that has lasted at least 30 days."
  - "t = 0 is the date when the outbreak becomes significant (100 cases) in each country."
  - "Results are based on June 15 data."
- Confidence intervals and axes:
  - Figures typically show the response and confidence bands at 90 and 95 percent (some panels specify 95 percent).
  - Horizontal axis: "the response x days after the containment measures" (or "after the easing of containment measures" where specified).
- Modeling notation and key variables (verbatim formulations preserved):
  - Estimates based on ∆푑_{t,i,h} = u_{i,t} + θ_h c_{i,t} + X′_{i,t} Γ_h + ∑_{ℓ=1}^{ℒ} ψ_{h,ℓ} ∆푑_{t-ℓ,i,h} + ε_{t,i,h} where ∆푑_{t,i,h} = d_{t,i,h} − d_{t−1,i,h} and d_{t,i} is the logarithm of the number of COVID-19 cases (or deaths, depending on specification) in country i observed at date t.
  - Model estimated at each horizon ℎ = 0, 1, ... H, with a lag structure ℓ = 1, 2 ... ℒ.
  - c_{i,t} is the index capturing the level of containment and mitigation measures; X is a matrix of time varying control variables and country specific linear, quadratic and cubic time trend.
  - For interaction specifications: ∆푑_{t,i,h} = u_i + u_t + θ_h [F(z_{i,t}) c_{i,t} + (1−F(z_{i,t})) c_{i,t}] + X′_{i,t} Γ_h + ... with F(z_{i,t}) defined and z the country-specific characteristics normalized to have zero mean and a unit variance.
  - Figures display log-difference changes whereas the text translates these into percent changes.
- Specific figure details:
  - Figure 2: "Effect of Containment Measures on Total Confirmed COVID-19 Cases" — response and confidence bands at 90 and 95 percent; model and definitions as above.
  - Figure 3: "Interaction with Public Health Response Time (deviation from baseline, log percentage points)" — interaction specification with F(z) and z normalized to zero mean and unit variance.
  - Figures 4a/4b: "Robustness Checks" — panels exclude China or US, include mobility controls, include time fixed effects, additional controls for serial correlation (leads), and combinations; last two panels add leads of the containment measure ∑(φ_ℓ c_{i,t+ℓ})_{ℓ=1}^ℒ to account for containment measures introduced during the response horizon.
  - Figure 5: "Summary of interactions with country characteristics (percent deviation from baseline 30 days after the introduction of containment measures)" — shows cumulative response after 30 days for baseline and each country characteristic.
  - Figures 6–8: "Interaction with Temperature and Elderly Population", "Interaction with Population Density and Health Infrastructure", "Interaction with Mobility" — each reports deviation from baseline, log percentage points, with 90 and 95 percent confidence bands; horizontal axis shows response x days after containment measures.
  - Figure 9: "Effect of Containment Measures on Total Confirmed COVID-19 Deaths (percent deviation from baseline 30 days after the introduction of containment measures)" — same sample and modeling as for cases; d is the logarithm of the number of COVID-19 deaths.
  - Figures 10–11: "Local projection response to different containment measures" (one at a time and together) — c_{i,t} indexes capture different types of containment and mitigation measures, introduced one at a time or simultaneously; figure 11 shows confidence bands at 95 percent.
  - Figure 12: "Effect of Containment Measures on Total Confirmed COVID-19 Cases" — horizontal axis shows response x days after the easing of containment measures.
- Annex figures:
  - Figure A1: "Effect of Containment Measures on Mobility (deviation from baseline, log percentage points)" — sample of 70 countries using daily data; Google mobility and retail indices reported as percent deviations for each day of the week to that corresponding day of the week in the baseline; smoothed over 7 days and then cumulated; graph shows response and confidence bands at 95 percent.
  - Figure A2: "Containment measures and mobility (mobility index, percent)" — Sources: Apple Mobility Indices, OxCGRT Stringency Index and IMF Staff calculations. "An index =100 suggest no decline in mobility compared to trends."
  - Figure A3: "Interaction with Health Security Index and Health Infrastructure (deviation from baseline, log percentage points)" — sample of 129 countries; interaction specification and normalization of z as above.
- Measurement and reporting conventions:
  - The index in Figure 1 is scaled between "zero (least stringent)" and "1 (most stringent)".
  - Mobility indices: Apple Mobility Indices and Google mobility/retail indices are used; Google indices are seasonally adjusted as percent deviations by weekday and smoothed over 7 days before cumulation.
  - All results cited are "based on June 15 data" where specified.

*Source: wpiea2020159-print-pdf - References (figures and notes as provided in the source PDF).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020159-print-pdf.pdf_