## COVID-19 Lockdowns and Exits in Asia: Some Lessons

## Source details

**Canonical URL:** [COVID-19 Lockdowns and Exits in Asia: Some Lessons](https://www.imf.org/-/media/files/oap/oap-home/2020/outreach-covid19lockdowns-pragyan-deb.pdf)

## Other formats

- [Markdown version](/-/media/files/oap/oap-home/2020/outreach-covid19lockdowns-pragyan-deb.pdf.md)
- [Structured JSON version](/-/media/files/oap/oap-home/2020/outreach-covid19lockdowns-pragyan-deb.pdf.json)

---

### Quantifying impact of containment measures
- Combine JHU’s data on COVID-19 spread with daily containment measures available from Oxford University.
- Use novel high-frequency indicators of economic activity, such as level of nitrogen oxide emissions.
- Assess impact at daily frequency without having to wait for lagged economic data such as GDP or IP.
- Help with econometric identification.

- Empirical approach: Local Projection methods by Jordà(2005)
  - ∆퐼퐼푖푖,푡푡+ℎ = 퐼퐼푖푖,푡푡+ℎ − I푖푖,푡푡+ℎ−1 and I푖푖,푡푡 is the log of the number of infections or economic indicator in country 푖푖 at date 푡푡.
  - 푐푐푖푖,푡푡 denotes the Containment Stringency Index.
  - 푋푋 is a vector of control variables which includes daily temperature and humidity levels, in addition to country-specific linear, quadratic and cubic time trends.
  - Impulse response specification:
    - ∆퐼퐼푖푖,푡푡+ℎ = 푢푢푖푖 + θℎ 푐푐푖푖,푡푡 + 푋푋푋푖푖,푡푡 Γℎ + �ℓ=1ℒ ψℎ,ℓ ∆퐼퐼푖푖,푡푡−ℓ + 휀휀푖푖,푡푡+ℎ.
  - Interaction specification to allow response to vary with country-specific characteristic z (normalized to zero mean and unit variance):
    - ∆퐼퐼푖푖,푡푡+ℎ = 푢푢푖푖 + 휃휃ℎ퐿퐿 F(z푖푖,푡푡) c푖푖,푡푡 + 휃ℎ퐻(1−F(z푖푖,푡푡)) c푖푖,푡푡 + 푋푋푋푖푖,푡푡 Γℎ + ∑ℓ=1ℒ F(z푖푖,푡푡) ψℎ,ℓ ∆퐼퐼푖푖,푡푡−ℓ + ∑ℓ=1ℒ (1−F(z푖푖,푡푡)) ψℎ,ℓ ∆퐼퐼푖푖,푡푡−ℓ + 휀휀푖푖,푡푡+ℎ.
    - with F(z푖푖푡푡) = e e e e e −γ z푖푖푖푖 /1− e e e e e −γ z푖푖푖푖, γ> 0.

### Containment effectiveness and disease dynamics
- Containment measures have been effective in reducing confirmed cases (reported as deviations from baseline in log percentage points over days after introduction of containment measures).
- Evidence distinguishes high mobility vs. low mobility contexts and shows mobility and number of cases co-movement.
- Public health response time matters:
  - Public Health Response Time measured in Number of days across regions: Asia and Pacific, Europe, Middle East and Central Asia, Americas, Africa.
  - Faster public health response associated with lower deviations in Confirmed Cases (Log percentage points).

### Short-term economic costs and high-frequency indicators
- Containment measures impose massive short-term economic costs, visible in:
  - GDP growth revisions (percentage points) plotted against Log deaths (per thousands).
  - NO2 emissions show large deviations from baseline in the days after containment introduction (Log percentage points).
- NO2 emissions and Industrial Production relationship:
  - NO2 emissions are strongly associated with level of economic activity.
  - Total man-made emissions of nitrogen oxides over the long term (1990-2018) are strongly correlated with conventional measures of economic activity such as GDP growth, manufacturing VA and IP.
  - At higher-frequency, over the period covered in the analysis of containment measures, estimated relationship between NO2 and IP statistically significant.
  - Implied finding: a 12 percent loss in Industrial Production due to containment measures.

- Regression/summary statistics (as reported):
  - Observations: 421
  - R-Squared: 0.0160 and 0.005 (reported across specifications).
  - Number of countries: 38 and 38.
  - Coefficient examples:
    - NO2 emissions (percent) 0.015** (0.006).
    - Industrial Production (percent) 0.27* (0.151).
    - Constant 0.004*** (0.0003) and 0.023*** (0.001).
  - Standard errors clustered at the country level in parentheses.
  - Significance notation: *** ρ<0.01, ** ρ<0.05, *** ρ<0.1.

### Lessons from Asia — empirical findings and recommendations
- Framing: Asian countries were first to be hit and many locked down their economies; quick and decisive action allowed some to gradually reopen. The presentation draws lessons from their experience.

- Lesson 1: Act fast.
  - Faster public health responses reduced confirmed cases (deviation from baseline in log percentage points).
  - Delays can significantly increase costs: extra infections per million relative to optimal containment grow with weeks since start of outbreak; comparison of "Optimal" vs. "Delayed" scenarios shows higher infections and larger GDP losses over time.
  - Public Health Response Time visualized by region (Number of days): Asia and Pacific, Europe, Middle East and Central Asia, Americas, Africa.

- Lesson 2: Put in place strong testing and tracing policies to avoid a second wave.
  - Testing and Tracing Policies at Time of Exit measured on an Index (7-day moving average) by region: Asia and Pacific, Middle East and Central Asia, Europe, Africa, Americas.
  - Strong testing and tracing at the time of exit associated with smaller deviations in Confirmed Cases (Log percentage points, 30 days after relaxation of containment measures).
  - Note on distribution: Median and 75th percentile overlaps for Americas.

- Lesson 3: Get the timing of exit right.
  - New Cases at Time of Exit (Per million people, 7-day moving average) differ by region: Asia and Pacific, Africa, Middle East and Central Asia, Americas, Europe.
  - Low new cases at exit vs. High new cases at exit:
    - Transit Mobility, Deviation from Baseline (Percent points) shows different mobility rebounds.
    - Premature exits can make the situation worse: Industrial Production, Deviation from Baseline (Percent, implied impact on industrial production 30 days after reopening) and GDP Recovery from Trough (Percent growth from trough) show worse epidemiological and economic outcomes for early exits.
    - Comparison of "Optimal" vs. "Early" reopening scenarios in GDP recovery from trough.

- Lesson 4: Use macroeconomic policies as cushion.
  - Fiscal and monetary policy cushioning matters for industrial production outcomes:
    - Scenarios presented include Low fiscal stimulus vs. High fiscal stimulus; Small policy rate cut vs. Large policy rate cut.
    - Impact of Fiscal Shocks on Industrial Production (Betas, percent) reported across environments: Baseline, Fixed exchange rate, Closed economy, Low debt to GDP, Low NO2 emissions, Low mobility, Low containment (controlling for mobility).
  - Policy combinations influence the implied impact on Industrial Production after 30 days.

### Overall conclusions
- Lessons Learnt from Asia summarized:
  - Lesson 1: Act fast.
  - Lesson 2: Put in place strong testing and tracing policies.
  - Lesson 3: Get the timing of exit right.
  - Lesson 4: Use macroeconomic policies as cushion.
- Objective: Limit scarring and lay the foundation for a stronger recovery.
  - Industrial Production, Deviation from Baseline (Percent, implied impact on industrial production 30 days after containment/reopening) shows containment and reopening effects spanning negative and positive deviations (axis from -14 to 8 percent in the reported figure).

---


_Source: https://www.imf.org/-/media/files/oap/oap-home/2020/outreach-covid19lockdowns-pragyan-deb.pdf_