## Section V empirically tests this

## Source details

**Canonical URL:** [Section V empirically tests this](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020214-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020214-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020214-print-pdf.pdf.json)

---

### Summary of main empirical findings
- States with high tourist-exposure fare disproportionally worse, controlling for other factors.  
- There is some evidence that export-exposed states fared better, particularly in retail sales.  
- High- and low-manufacturing exposure states are taken from Chiquiar et al. (2017).  
- Economic activity measure: average of the percent change (July 2020 - January 2020/earlier; 2020 March - April year-over-year changes for debit/credit flows) of the following variables: debit and credit card flows, IMSS employees and employers, retail sales, Google workplaces and transit stations.  
- Data as of July 2020. Cases are in per capita terms. Source: National authorities and IMF staff calculations.

### Data, sample and stylized facts
- Individual-level dataset: more than 800K observations; includes age, gender, pre-existing conditions, smoker status, and residence (state and municipality).  
- Diseases and pre-existing conditions observed: obesity, diabetes, hypertension, immunocompromised, pneumonia, chronic obstructive pulmonary disease (COPD), cardiovascular disease.  
- Age and positivity shares:
  - About 68 percent of the people who tested are below 50 years; 62 percent of the people who tested positive are below 50 years.  
  - <25: among those tested, 34 percent tested positive.  
  - 25-50: among those tested, 45 percent tested positive.  
  - ≥50: 32 percent of tested; 38 percent of positives.  
  - 50-75: among those tested, 56 percent tested positive.  
  - >75: among those tested, 59 percent tested positive.  
- Gender and positivity:
  - Testing split almost equally; 53 percent of positive tests were male and 47 percent female.  
- Comorbidity positivity rates:
  - Among people tested with diabetes, 55 percent tested positive.  
  - Among those reporting pneumonia, 70 percent tested positive.

### Probit model: specification and dependent variables
- Two probit models estimated:
  - Probability of testing positive (1=positive; 0=negative).  
  - Probability of death (1=death; 0=positive).  
- Right-hand-side variables: age, gender, and a wide set of health conditions.  
- Estimation approach: start with an unrestricted specification including all health conditions and remove conditions not statistically significant or not of expected sign.  
- Functional form: Prob(testing positive=1|taking test) OR Prob(death=1|testing positive) = function(age, gender, variables representing health conditions).

### Probit model results — age, gender, and health conditions
- Age effects:
  - Predicted probability of testing positive:
    - 20-year old (others at mean): 40.5 percent.  
    - 60-year old (others at mean): 52.6 percent.  
  - Probability of death increases non-linearly with age, particularly after around 50 years:
    - Age increase 20 → 30: raises probability of death by 0.8 percentage points.  
    - Age increase 40 → 50: raises probability of death by 2.7 percentage points.  
    - Age increase 60 → 70: raises probability of death by 6.1 percentage points.
- Testing positive — health conditions and other factors:
  - Obesity increases probability of testing positive by 6.8 percentage points (other variables at mean).  
  - Diabetes increases probability of testing positive by 4.8 percentage points (excerpted).  
  - Pneumonia increases probability of testing positive by 20.7 percentage points (excerpted).  
  - Factors associated with higher probability of testing positive: gender, pneumonia, obesity, diabetes, hypertension.  
  - Factors associated with higher probability of death: gender, pneumonia, kidney-related disease, diabetes, immunocompromised status, obesity, hypertension, COPD, and other diseases.

### Key sample statistics (Table 1 snapshot, N=847,163 observations; positives N=399,868; data as of 28th July 2020)
- Number of observations: 847,163; Number of positives: 399,868.  
- Age distribution of persons tested:
  - <25: 94,803 persons, 11 percent of all; positives in group: 32,279 (8 percent of positives); percent positive within group: 34.  
  - 25-50: 484,120 persons, 57 percent of all; positives in group: 215,862 (54 percent of positives); percent positive within group: 45.  
  - 50-75: 234,358 persons, 28 percent of all; positives in group: 131,826 (33 percent of positives); percent positive within group: 56.  
  - >=75: 33,882 persons, 4 percent of all; positives in group: 19,901 (5 percent of positives); percent positive within group: 59.  
- Gender:
  - Female: 424,887 persons (50 percent of all); positives among female: 186,897 (47 percent of positives); percent positive within female: 44.  
  - Male: 422,276 persons (50 percent of all); positives among male: 212,971 (53 percent of positives); percent positive within male: 50.  
- Health conditions (persons tested; percent of all tested; positives among those with condition; percent positive within condition):
  - Obesity: 137,147 persons, 16 percent of all; positives: 75,603, percent positive within group: 55.  
  - Diabetes: 105,839 persons, 12 percent of all; positives: 64,170, percent positive within group: 61.  
  - Asthma: 25,882 persons, 3 percent of all; positives: 10,793, percent positive within group: 42.  
  - Pneumonia: 119,698 persons, 14 percent of all; positives: 83,391, percent positive within group: 70.

### Health risks and individual-level findings (excerpted highlights)
- Increase in probability of testing positive:
  - 4.8 percentage points if the person has diabetes.  
  - 20.7 percentage points if the person has pneumonia.  
- Increase in probability of death:
  - Kidney-related diseases increase probability by 6 percentage points.  
- Likelihoods for both testing positive and death increase if the person is male.  
- Caveats:
  - Mexico’s fatalities could be underestimated by around three times.  
  - Analysis does not control for some relevant factors (e.g., individual’s income, living conditions, inclination to go for testing).  
- External supporting evidence noted (WHO March 2020; Apicella et al. 2020).

### Policy implication from health findings
- Protect vulnerable populations (the elderly and those with co-morbidities) during phased re-opening, especially given:
  - High informality and limited savings/safety net in Mexico.  
  - Potential delay in development of a widely available, cheap, and effective vaccine.

### State characteristics and Covid-19 case incidences (stylized facts)
- Income per capita (as of end-July 2020, cumulative per 1000):
  - Richer states had higher tests per capita (strong relationship; high R-square).  
  - Richer states had higher cases per capita and higher deaths per capita (R-squares lower than for tests per capita).  
  - Richer states had lower deaths per cases (possible better facilities or higher denominator due to higher testing).  
- Population density:
  - Densely populated states had more cases per capita by end-July 2020; result may be influenced by an outlier (Mexico City).  
  - Difference in average cases per capita between high-density states (top 75th percentile) and others lessened over time.  
- Health expenditure and capacity:
  - States below the 50th percentile in health expenditure/capacity had higher average deaths per 1000 cases over time.  
  - Indicators considered: health expenditure per capita; beds per capita (total and available); ICU beds available per capita.

### Regression evidence on state characteristics (summary of Table 2)
- Regression setup:
  - Panel regressions on monthly data (March–July 2020); controls include income per capita, population density, health expenditure per capita, containment measures, tests per capita; 1- or 2-lags of dependent variables; Driscoll-Kraay standard errors.
- Key estimated coefficients and signs (Table 2 highlights):
  - Test per capita: 0.3626*** (cases per capita), 0.0151*** (deaths per capita), -11.7767*** (deaths per cases).  
  - Measure (containment composite): 0.1511*** (cases per capita), 0.0543*** (deaths per capita).  
  - Health expenditure per capita: -0.0116** (deaths per capita), -20.6712* (deaths per cases).  
  - Population density: 0.0233** (deaths per cases).  
  - GDP per capita: 0.0845* (deaths per capita).
- Interpretation:
  - Higher health expenditure per capita associated with lower deaths per capita and lower deaths per cases (statistically significant).  
  - Richer states had more deaths per capita; highly populated states had more deaths per cases (the latter possibly driven by one high-density state).  
  - Results reflect correlations controlling for available factors; causation not claimed.

### Containment measures — effectiveness analysis
- State-level local projection (Jorda 2005) results:
  - Change in containment measures associated with declines in new cases per capita.  
  - Negative coefficients become statistically significant only after about 30 days, indicating measures may take a long time to become effective.  
  - Results robust to alternative lag lengths, using raw new cases per capita, controlling for tests differently, and including lag of shock variable.
- Country-level regressions (G20 emerging markets and other emerging markets):
  - Both Mexico and other emerging markets show measures effective in reducing new cases per capita; effects statistically significant.  
  - For each percent change in measures, other emerging markets witnessed a higher drop in cases compared to Mexico — suggesting measures might have been less effective in Mexico for the percent-change magnitude tested.  
  - Caveats: estimates more reliable in panel cross-country settings than single-country time-series; significance can depend on lag choices. Mexico’s coefficients remain less negative across tested lags.

### Economic activity — role of state characteristics (stylized facts and regressions)
- Stylized correlations (May 2020 compared to January 2020; debit/credit cards use March–April year-over-year where noted):
  - Tourism exposure correlates strongly with activity drops:
    - IMSS-employment (R-square: 76 percent), debit card usage (35 percent), retail sales (28 percent), Google transit stations (33 percent).  
  - Export-exposed states experienced smaller drops in activity; evidence strongest for retail sales (R-square 21 percent) and credit card flows (24 percent).  
  - Lower mobility associated with higher activity drops; denser states fared worse though possibly driven by an outlier.
- Regression results (summary of Table 3):
  - Tourism (share of GDP) coefficients negative and often significant across indicators (debit/credit flows, retail, IMSS-employment, flight departures/arrivals, Google transit stations).  
  - Exports (share of GDP) coefficients often positive (e.g., 0.0003* for retail), suggesting higher export exposure associated with smaller declines in some indicators (retail particularly).  
  - Measure/mobility:
    - Measure coefficient strongly negative and statistically significant for several outcomes (e.g., retail: -0.4395***).  
  - Cases per 1000 persons enter with mixed signs across activity indicators.
- Economic significance of tourism:
  - For states at the 95th percentile of tourism exposure:
    - Around 50-60 percent month-over-month decline in flight arrivals and departures attributable to tourism exposure.  
    - 20 percent drop in year-over-year debit/credit card flows attributable to tourism exposure.  
    - 12 percent monthly decline in retail sales and 5 percent monthly decline in employment attributable to tourism exposure.  
  - For states at the 75th percentile of tourism exposure:
    - Around 10 percent drop in flight arrivals/departures.  
    - 3.5 percent drop in year-over-year debit/credit card flows.

### Role of exports — possible channels for resilience
- Channels identified for export-related resilience:
  - Higher remittances in higher export-exposed states:
    - States with high export-exposure had higher remittance growth in Q2 2020 compared to Q1 2020, potentially cushioning retail sales.  
  - Mexico’s exports declined less compared to many peers in initial months:
    - Mexico’s exports to the U.S. remained resilient initially; possible factors include trade diversion from China, proximity to the U.S., and pending USMCA ratification.  
  - Possible GVC resilience:
    - Countries with higher supply-chain integration showed lower nominal export decline in the initial phase; other forces (e.g., oil prices) could confound this link and require formal econometric study.

### Policy conclusions and recommended fiscal/health response magnitudes
- Health system investment:
  - States with lower per capita health indicators witnessed higher case fatality rates.  
  - Hannan et al. (2020) and Dudine et al. (2020) model-based suggestions: health spending could be increased by around 0.6-1.5 percent of GDP.
- Near-term fiscal support proposals (Hannan et al., 2020):
  - Near-term fiscal support of 2.5-3.5 percent of GDP total, including:
    - 1.2 percent of GDP for informal workers in hard-hit industries.  
    - 0.4 percent of GDP for wage subsidies for formal workers in hard-hit industries.
  - Combine near-term support with a credible medium-term tax reform to close fiscal gaps, finance public investment and social spending, and put debt on a downward trajectory.
- Structural policy recommendations:
  - Protect vulnerable populations (elderly and those with co-morbidities) during phased re-opening.  
  - Increase health spending and capacity to contain the pandemic and save lives.  
  - Provide larger emergency lifelines and fiscal support to vulnerable households and viable firms given high informality.  
  - Facilitate investment promoting broader integration into global or North American value chains and pursue complementary structural reforms to improve the business climate to reap benefits from USMCA.

### Annex I — Data sources (selected)
- Covid-19 indicators:
  - Cases and deaths: State level: Ministry of Health (MoH); International: Johns Hopkins University (JHU).  
  - Tests: State level: IHME; International: IMF STA Database.  
  - Patient characteristics: State level: Ministry of Health (MoH).  
  - ICU availability: State level: IHME; International: IHME.  
  - Hospital beds, ventilator use: State level: Ministry of Health (MoH), IHME; International: IHME.  
  - Health expenditure per capita: State level: Ministry of Health (MoH); International: Healthcare Access and Quality (HAQ) Index.
- Economic activity:
  - Flight arrivals and departures: FlightRadar 24.  
  - Credit and debit card flows: Hernandez (2020).  
  - Employees and employers: IMSS.  
  - Population: INEGI.  
  - Tourism share of GDP: INEGI.  
  - Retail sales: INEGI.  
  - Remittances: INEGI.  
  - Google and Apple mobility: Haver.

*Source: wpiea2020214-print-pdf*

### Section V empirically tests this

### wpiea2020214-print-pdf - Section V empirically tests this

### Summary of main empirical findings
- States with high tourist-exposure fare disproportionally worse, controlling for other factors.  
- There is some evidence that export-exposed states fared better, particularly in retail sales.  
- High- and low-manufacturing exposure states are taken from Chiquiar et al. (2017).  
- Economic activity measure: average of the percent change (July 2020 - January 2020/earlier; 2020 March - April year-over-year changes for debit/credit flows) of the following variables: debit and credit card flows, IMSS employees and employers, retail sales, Google workplaces and transit stations.  
- Data as of July 2020. Cases are in per capita terms. Source: National authorities and IMF staff calculations.

### Data, sample and stylized facts
- Individual-level dataset contains more than 800K observations and includes age, gender, pre-existing conditions, smoker status, and residence (state and municipality).  
- Diseases and pre-existing conditions in the data include: obesity, diabetes, hypertension, immunocompromised, pneumonia, chronic obstructive pulmonary disease (COPD), cardiovascular disease.  
- About 68 percent of the people who tested are below 50 years, while 62 percent of the people who tested positive are in that age range.  
- Among people who tested and are below 25 years, 34 percent tested positive.  
- Among the age-group 25-50, 45 percent tested positive.  
- 32 percent of the people who tested are above or equal to 50 years of age, but 38 percent of the people who tested positive are in that age group.  
- Among the 50-75 age-group who took tests, 56 percent tested positive.  
- Among the age-group above 75 years who took tests, 59 percent tested positive.  
- Testing data are split almost equally between men and women; 53 percent of the positive tests were for male and 47 percent for female.  
- Among people who took tests and had diabetes, 55 percent tested positive.  
- Among people who took the test and reported pneumonia, 70 percent tested Covid-19 positive.

### Probit model: specification and dependent variables
- Two probit models are estimated:  
  - Probability of testing positive (1=positive; 0=negative).  
  - Probability of death (1=death; 0=positive).  
- Right-hand-side variables include age, gender, and a wide set of health conditions.  
- Approach: start with an unrestricted specification including all health conditions and then remove conditions that are not statistically significant or are not of the expected sign.  
- Functional form: Prob(testing positive=1|taking test) OR Prob(death=1|testing positive) = function(age, gender, variables representing health conditions).

### Probit model results — age, gender, and health conditions
- Age: important determinant of both testing positive and death.  
  - Predicted probability of testing positive for a 20-year old (keeping other variables at mean): 40.5 percent.  
  - Predicted probability of testing positive for a 60-year old (keeping other variables at mean): 52.6 percent.  
  - Increase in probability owing to age is statistically significant and of meaningful magnitude.  
  - Probability of death increases non-linearly with age, particularly after around 50 years:  
    - Age increase from 20 to 30 raises probability of death by 0.8 percentage points.  
    - Age increase from 40 to 50 raises probability of death by 2.7 percentage points.  
    - Age increase from 60 to 70 raises probability of death by 6.1 percentage points.  
- Testing positive — health conditions and other factors: keeping other variables at mean, the probability of testing positive increases by 6.8 percentage points if a person is obese.  
- Factors associated with a higher probability of testing positive: gender, pneumonia, obesity, diabetes, hypertension.  
- Factors associated with a higher probability of death: gender, pneumonia, kidney-related disease, diabetes, immunocompromised status, obesity, hypertension, COPD, and other diseases.

### Key sample statistics (Table 1 snapshot, N=847,163 observations; positives N=399,868; data as of 28th July 2020)
- Number of observations: 847,163; Number of positives: 399,868.  
- Age distribution of persons tested:  
  - <25: 94,803 persons, 11 percent of all; positives in group: 32,279 (8 percent of positives); percent positive within group: 34.  
  - 25-50: 484,120 persons, 57 percent of all; positives in group: 215,862 (54 percent of positives); percent positive within group: 45.  
  - 50-75: 234,358 persons, 28 percent of all; positives in group: 131,826 (33 percent of positives); percent positive within group: 56.  
  - >=75: 33,882 persons, 4 percent of all; positives in group: 19,901 (5 percent of positives); percent positive within group: 59.  
- Gender: Female 424,887 persons (50 percent of all); positives among female: 186,897 (47 percent of positives); percent positive within female: 44. Male 422,276 persons (50 percent of all); positives among male: 212,971 (53 percent of positives); percent positive within male: 50.  
- Health conditions (persons tested; percent of all tested; positives among those with condition; percent positive within condition):  
  - Obesity: 137,147 persons, 16 percent of all; positives: 75,603, percent positive within group: 55.  
  - Diabetes: 105,839 persons, 12 percent of all; positives: 64,170, percent positive within group: 61.  
  - Asthma: 25,882 persons, 3 percent of all; positives: 10,793, percent positive within group: 42.  
  - Pneumonia: 119,698 persons, 14 percent of all; positives: 83,391, percent positive within group: 70.

_Excerpted content as presented in the source PDF._

### 4.8 percentage points if the person has diabetes. The probability of testing positive increases

### wpiea2020214-print-pdf - 4.8 percentage points if the person has diabetes. The probability of testing positive increases

### Health risks and individual-level findings
- Increase in probability of testing positive:
  - 4.8 percentage points if the person has diabetes.
  - 20.7 percentage points if the person has pneumonia.
- Increase in probability of death:
  - Kidney-related diseases increase probability by 6 percentage points.
- Likelihoods for both testing positive and death increase if the person is male.
- Findings are indicative and subject to caveats:
  - Mexico’s fatalities could be underestimated by around three times.
  - Analysis does not control for some relevant factors (e.g., individual’s income, living conditions, inclination to go for testing).
- Supporting evidence from external sources:
  - WHO (March 2020) identified higher risk groups as (i) older people (people above 60 years; risk gradually increasing with age starting from 40 years) and (ii) those with underlying medical conditions such as cardiovascular disease, diabetes, respiratory disease, and cancer.
  - Apicella et al. (2020) note high fatalities in older patients where co-morbidities are common.

### Policy implication from health findings
- Important to protect vulnerable populations (the elderly and those with co-morbidities) during phased re-opening, especially given:
  - High informality and limited savings/safety net in Mexico.
  - Potential delay in development of a widely available, cheap, and effective vaccine.

### State characteristics and Covid-19 case incidences (stylized facts)
- Income per capita (as of end-July 2020, cumulative per 1000):
  - Richer states had higher tests per capita (strong relationship; high R-square).
  - Richer states had higher cases per capita and higher deaths per capita (R-squares lower than for tests per capita).
  - Richer states had lower deaths per cases (possible better facilities or higher denominator due to higher testing).
- Population density:
  - Densely populated states had more cases per capita by end-July 2020; result may be influenced by an outlier (Mexico City).
  - Difference in average cases per capita between high-density states (top 75th percentile) and others lessened over time.
- Health expenditure and capacity:
  - States below the 50th percentile in health expenditure/capacity had higher average deaths per 1000 cases over time.
  - Indicators considered: health expenditure per capita; beds per capita (total and available); ICU beds available per capita.
  - Evidence consistent with stronger health systems making containment measures more effective (stringent measures reduced cases by more than 97 percent in countries with strong health systems in cited literature).

### Regression evidence on state characteristics (summary of Table 2 and interpretation)
- Regression setup:
  - Panel regressions on monthly data (March–July 2020), controls include: income per capita, population density, health expenditure per capita, containment measures, tests per capita; 1- or 2-lags of dependent variables; Driscoll-Kraay standard errors.
- Key estimated coefficients and signs (Table 2 highlights):
  - Test per capita: 0.3626*** (cases per capita), 0.0151*** (deaths per capita), -11.7767*** (deaths per cases).
  - Measure (containment composite): 0.1511*** (cases per capita), 0.0543*** (deaths per capita).
  - Health expenditure per capita: -0.0116** (deaths per capita), -20.6712* (deaths per cases).
  - Population density: 0.0233** (deaths per cases).
  - GDP per capita: 0.0845* (deaths per capita).
- Interpretation:
  - Higher health expenditure per capita associated with lower deaths per capita and lower deaths per cases (statistically significant).
  - Richer states had more deaths per capita; highly populated states had more deaths per cases (the latter possibly driven by one high-density state).
  - Results reflect correlations controlling for available factors; causation not claimed.

### Containment measures — effectiveness analysis
- State-level local projection (Jorda 2005) results:
  - Change in containment measures associated with declines in new cases per capita.
  - Negative coefficients become statistically significant only after about 30 days, indicating measures may take a long time to become effective.
  - Results robust to alternative lag lengths, using raw new cases per capita, controlling for tests differently, and including lag of shock variable.
- Country-level regressions (G20 emerging markets and other emerging markets):
  - Both Mexico and other emerging markets show measures effective in reducing new cases per capita; effects statistically significant.
  - For each percent change in measures, other emerging markets witnessed a higher drop in cases compared to Mexico — suggesting measures might have been less effective in Mexico for the percent-change magnitude tested.
  - Caveats: estimates more reliable in panel cross-country settings than single-country time-series; significance can depend on lag choices. Mexico’s coefficients remain less negative across tested lags.

### Economic activity — role of state characteristics (stylized facts and regressions)
- Stylized correlations (May 2020 compared to January 2020; debit/credit cards use March–April year-over-year where noted):
  - Tourism exposure correlates strongly with activity drops: IMSS-employment (R-square: 76 percent), debit card usage (35 percent), retail sales (28 percent), Google transit stations (33 percent).
  - Export-exposed states experienced smaller drops in activity (positive slopes); evidence strongest for retail sales (R-square 21 percent) and credit card flows (24 percent).
  - Lower mobility associated with higher activity drops; denser states fared worse though possibly driven by an outlier.
- Regression results (summary of Table 3):
  - Tourism (share of GDP) coefficients (negative and often significant) indicate higher tourism exposure led to larger activity drops across indicators including debit/credit flows, retail, IMSS-employment, flight departures/arrivals, and Google transit stations.
  - Exports (share of GDP) coefficients often positive (e.g., 0.0003* for retail), suggesting higher export exposure associated with smaller declines in some indicators (retail particularly).
  - Measure/mobility:
    - Measure coefficient strongly negative and statistically significant for several outcomes (e.g., retail: -0.4395***; jobs/employers and other outcomes show significance in table).
  - Cases per 1000 persons enter with mixed signs across activity indicators.
- Economic significance of tourism:
  - For states at the 95th percentile of tourism exposure:
    - Around 50-60 percent month-over-month decline in flight arrivals and departures attributable to tourism exposure.
    - 20 percent drop in year-over-year debit/credit card flows attributable to tourism exposure.
    - 12 percent monthly decline in retail sales and 5 percent monthly decline in employment attributable to tourism exposure.
  - For states at the 75th percentile of tourism exposure:
    - Around 10 percent drop in flight arrivals/departures.
    - 3.5 percent drop in year-over-year debit/credit card flows.

### Role of exports — possible channels for resilience
- Channels identified for export-related resilience:
  - Higher remittances in higher export-exposed states:
    - States with high export-exposure had higher remittance growth in Q2 2020 compared to Q1 2020, potentially cushioning retail sales.
  - Mexico’s exports declined less compared to many peers in initial months:
    - Mexico’s exports to the U.S. remained resilient initially; possible factors include trade diversion from China, proximity to the U.S., and pending USMCA ratification.
  - Possible GVC (global value chain) resilience:
    - Countries with higher supply-chain integration showed lower nominal export decline in the initial phase, though other forces (e.g., oil prices) could confound this link and require formal econometric study.

### Policy conclusions and recommended fiscal/health response magnitudes
- Health system investment:
  - States with lower per capita health indicators witnessed higher case fatality rates.
  - Hannan et al. (2020) and Dudine et al. (2020) model-based suggestions: health spending could be increased by around 0.6-1.5 percent of GDP.
- Near-term fiscal support proposals (Hannan et al., 2020):
  - Near-term fiscal support of 2.5-3.5 percent of GDP total:
    - 1.2 percent of GDP for informal workers in hard-hit industries.
    - 0.4 percent of GDP for wage subsidies for formal workers in hard-hit industries.
  - Combine near-term support with a credible medium-term tax reform to close fiscal gaps, finance public investment and social spending, and put debt on a downward trajectory.
- Structural policy recommendations:
  - Protect vulnerable populations (elderly and those with co-morbidities) during phased re-opening.
  - Increase health spending and capacity to contain the pandemic and save lives.
  - Provide larger emergency lifelines and fiscal support to vulnerable households and viable firms given high informality.
  - Facilitate investment promoting broader integration into global or North American value chains and pursue complementary structural reforms to improve the business climate to reap benefits from USMCA.

*Source: wpiea2020214-print-pdf*

### ANNEX I. DATA SOURCES

### ANNEX I. DATA SOURCES

### Covid-19 indicators — data sources by coverage
- Cases and deaths
  - State level: Ministry of Health (MoH)
  - International: Johns Hopkins University (JHU)
- Tests
  - State level: IHME
  - International: IMF STA Database
- Patient characteristics
  - State level: Ministry of Health (MoH)
  - International: N/A
- UCI availability
  - State level: IHME
  - International: IHME
- Hospital beds
  - State level: Ministry of Health (MoH), IHME
  - International: IHME
- Ventilator use
  - State level: Ministry of Health (MoH), IHME
  - International: IHME
- Health expenditure per capita
  - State level: Ministry of Health (MoH)
  - International: Healthcare Access and Quality (HAQ) Index

### Economic activity — data sources by indicator
- Flight arrivals and departures
  - State level: FlightRadar 24
  - International: FlightRadar 24
- Credit and debit card flows
  - State level: Hernandez (2020)
  - International: (no international source listed)
- Employees
  - State level: IMSS
  - International: Haver analytics
- Employers
  - State level: IMSS
  - International: Haver analytics
- Population
  - State level: INEGI
  - International: UN
- FDI
  - State level: INEGI
  - International: WDI
- Tourism share of GDP
  - State level: INEGI
  - International: WDI
- Retail sales
  - State level: INEGI
  - International: Haver
- Remittances
  - State level: INEGI
  - International: Haver
- CO Air Quality Index (AQI)
  - State level: Air Quality Index (AQI)
  - International: Air Quality Index (AQI)
- NO2 Air Quality Index (AQI)
  - State level: Air Quality Index (AQI)
  - International: Air Quality Index (AQI)
- Google Mobility
  - State level: Haver
  - International: Haver
- Apple Mobility
  - State level: Haver
  - International: Haver

### Notes on charts in Annex II
- The charts show May 2020 (or the latest data available) as a percent of January 2020.
- The exception is debit and credit cards which show year-over-year March-April 2020 growth.

### Annex III — State-level regression results (high-level mapping and key details)
- Regression focus: "The Role of State Characteristics on Economic Activity," with alternative mobility measures (IHME’s Composite Mobility and Apple Drive) used instead of containment measures.
- Common model features:
  - Regressions include lag of dependent variables.
  - Population density is habitat per km2; GDP per capita is million pesos per capita.
  - Standard errors in parentheses; significance: *** p<0.01, ** p<0.05, * p<0.1.

- Table AIII.1 (using IHME’s Composite Mobility)
  - Dependent variables across columns 1–9: debit/credit, retail, jobs, employers, flight_dep, flight_arr, g_work, g_transit, no2.
  - Selected coefficient examples (coefficient (standard error)):
    - GDP per capita: 13.0464 (9.4875) for debit/credit; -0.0376 (0.0352) for retail; -0.0464* (0.0119) for jobs.
    - Tourism (share of GDP): -0.8841*** (0.2046) for debit/credit; -0.0052** (0.0010) for retail.
    - Population density: 0.0007 (0.0012) for debit/credit; 0.0000*** (0.0000) for jobs.
    - Exports (share of GDP): 0.0032 (0.0307) for debit/credit; 0.0004** (0.0001) for retail.
    - Measure/mobility-app (mom change): 53.5039*** (15.8033) for debit/credit; 0.0008 (0.0011) for retail.
    - Cases (per 1000 persons): 9.8252 (6.3582) for debit/credit; -0.1830 (0.1440) for retail.
  - Model fit and sample:
    - Observations: 329, 69, 69, 68, 78, 99, 63, 5,  (as presented across columns)
    - R-squared: 0.7596, 0.6405, 0.6279, 0.5279, 0.5398, 0.5796, 0.9269, 0.9182, 0.3790
    - Number of groups: 32, 32, 32, 29, 30, 32, 32, 31, 2

- Table AIII.2 (using Apple Drive)
  - Dependent variables across columns 1–9: debit/credit, retail, jobs, employers, flight_dep, flight_arr, g_work, g_transit, no2.
  - Selected coefficient examples (coefficient (standard error)):
    - GDP per capita: 16.8814 (10.8577) for debit/credit; 0.0231 (0.0125) for retail; -0.0579** (0.0068) for jobs.
    - Tourism (share of GDP): -1.0033*** (0.2216) for debit/credit; -0.0035 (0.0013) for retail.
    - Population density: -0.0001 (0.0013) for debit/credit; -0.0000* (0.0000) for retail.
    - Exports (share of GDP): 0.0287 (0.0338) for debit/credit; 0.0005 (0.0002) for retail.
    - Measure/mobility Composite mobility: 0.4060** (0.1927) for debit/credit; 0.0054*** (0.0004) for retail; 0.0006*** (0.0000) for jobs.
    - Cases (per 1000 persons): 10.9017 (7.2132) for debit/credit; 0.0620*** (0.0058) for retail; 0.0051 (0.0037) for jobs.
  - Model fit and sample:
    - Observations: 32, 96, 96, 96, 87, 89, 96, 96, 35
    - R-squared: 0.7022, 0.8482, 0.6885, 0.6656, 0.7174, 0.7010, 0.9289, 0.9391, 0.2770
    - Number of groups: 32, 32, 32, 29, 30, 32, 32, 12

*Source: ANNEX I. DATA SOURCES (as provided in the content unit).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020214-print-pdf.pdf_