## 4.1 Regression discontinuity (wpiea2020282-print-pdf)

## Source details

**Canonical URL:** [4.1 Regression discontinuity (wpiea2020282-print-pdf)](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020282-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020282-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020282-print-pdf.pdf.json)

---

### Purpose, identification, and data
- Purpose and identification strategy:
  - Addresses endogeneity in adoption of lockdown measures by focusing on high-frequency changes in mobility around specific lockdown measures using a regression discontinuity (RD) design.
  - Strengthens identification by controlling for pre-existing mobility trends and epidemic severity in complementary local projection analysis and by excluding provinces with early or severe outbreaks.
- Data and scope:
  - Novel, confidential Vodafone mobility indicators at the provincial (NUTS3) level for Italy, Portugal, and Spain, disaggregated by gender and age groups.
  - Age groups: [18,24]; [25,44]; [45,64]; [65+].
  - Mobility indicator mob_{i,g,a,t} = percentage of people moving in province i, gender g, age group a, at time t.
  - Vodafone data validated against Apple and Google mobility indicators: correlations Italy and Portugal between 93 and 99 percent; Spain between 72 and 88 percent.

### RD estimation setup
- Running variable: time; treatment date: introduction of national stay-at-home orders.
- Baseline narrow bandwidth: 20 days around adoption of stay-at-home orders.
- Local linear regression with province fixed effects (α_p), day-of-week fixed effects (τ_dow), treatment indicator stay_{j,t}, women dummy women_{i,a}, date_{j,t}, interactions including stay × date and women × stay; standard errors clustered at the province level.
- Identification assumption: stay_{j,t} × date_{j,t} absorbs smooth relationship between date and the error term around lockdown introduction.

### Main RD findings (baseline, 20-day window)
- Overall mobility change at stay-at-home introduction:
  - The percentage of people leaving their homes in a day declined by about 15 percentage points.
- Age-specific stay-at-home coefficients (RD, Table 1 stay-at-home):
  - 18–24: -19.60*** (0.59)
  - 25–44: -12.95*** (0.46)
  - 45–64: -12.52*** (0.41)
  - 65+: -11.76*** (0.46)
- Women × stay-at-home interaction (RD, Table 1 women × stay-at-home):
  - 18–24: 2.31*** (0.50)
  - 25–44: -3.24*** (0.35)
  - 45–64: -1.74*** (0.34)
  - 65+: 1.55*** (0.57)
- Additional age-gender gaps at discontinuity (percentage points):
  - Age 25–44: women’s mobility declined by 3 percentage points more than men’s.
  - Age 18–24: gap of 2.3 percentage points.
  - Age 45–64: gap of 1.7 percentage points.
  - Age 65+: no disproportionate effect on women’s mobility.

### Northern Italy evidence on school closures
- Two discontinuities exploited: February 23 local school closures and March 9 national lockdown.
- Findings:
  - Divergence in mobility between men and women started at school closures.
  - Men’s mobility declined marginally at school closures; women’s mobility showed a clear discontinuity.
  - Heat map evidence: mobility of men and women similar before February 23; disproportionate reduction in women’s mobility across provinces after school closures; national stay-at-home orders further widened the mobility gap.
  - Lack of vertical pattern by province COVID-19 cases per population on March 9 suggests effects were not correlated with local intensity of the COVID-19 crisis.

### Interpretation from RD
- Largest mobility gender gap for people aged 25–44 suggests childcare needs induced by school closures largely explain the disproportionate impact on women’s mobility.
- Stay-at-home orders disproportionately reduced mobility of working-age people, especially those below 45 years of age.

---

### 4.2 Local projections (wpiea2020282-print-pdf)

### Methodology and identification
- Estimation approach: local projection regressions (Jordà, 2005) on province-level Vodafone mobility data; sample of 163 provinces between January 1 and June 29, 2020.
- Endogeneity mitigation:
  - Control for log of daily COVID-19 cases (ln∆cases) with weekly lag length.
  - Include lags of the dependent variable.
  - Exclude provinces that (i) registered the first 100 cumulative cases, (ii) had the highest number of COVID-19 cases by end-June 2020, and (iii) had more than five percent of the country’s total confirmed cases when the lockdown stringency index reached its maximum.
  - Province, gender (or age-group), and time fixed effects; standard errors clustered at the province level.
- Lockdown stringency index: reconstructed from Oxford Tracker as a simple average of nine sub-indicators, excluding public information campaigns and considering only national-level measures.
- Key interaction: lockdown stringency × gender dummy; lockdown stringency × age-group dummies.

### Trough effects of a full lockdown (local projection estimates at trough, days after shock = 7)
- Lockdown stringency (trough coefficients, Table 2):
  - 18–24: -27.07*** (1.80)
  - 25–44: -20.60*** (1.60)
  - 45–64: -19.75*** (1.57)
  - 65+: -15.19*** (1.77)
- Women × lockdown stringency (trough coefficients, Table 2):
  - 18–24: -3.58*** (0.95)
  - 25–44: -4.73*** (0.67)
  - 45–64: -3.69*** (0.73)
  - 65+: 2.33 (1.46)
- Observations and sample at trough:
  - Observations: 18,798; 18,798; 18,830; 17,872 (columns 1–4)
  - Provinces: 163; 163; 163; 157
  - R-squared: 0.94; 0.94; 0.94; 0.91

### Gender-differential local projection dynamics (age group 25–44)
- A full lockdown leads to a mobility trough after seven days.
- Point estimates (age 25–44):
  - Women’s mobility falls by 26 percentage points seven days after introduction of lockdowns.
  - Men’s mobility declines by about 21 percentage points seven days after introduction of lockdowns.
  - The women-minus-men mobility differential is statistically significant until the lockdown impulse weakens 15 days after initial tightening.

### Voluntary social distancing (ln∆cases) and gender
- For a given lockdown stringency, mobility responds to rising infections.
- Age group 25–44:
  - A doubling of daily COVID-19 cases leads to a contraction in mobility by about 0.3 percentage points 20 days after introduction of lockdowns.
  - The effect of rising infections on mobility is similar across men and women: no statistically significant gender difference in voluntary social distancing responses.

### Age heterogeneity (local projections and RD)
- RD: stay-at-home orders drastically reduced mobility across all age groups but drop stronger for younger cohorts; over 20 percentage points for people below 25; 11 percentage points for people aged 65 and above.
- Local projections:
  - People aged 18–24: trough of 30 percentage points nine days after introduction of lockdowns.
  - Differential relative to 65+: 18–24 vs 65+ largest and statistically significant from beginning of horizon; 25–44 and 45–64 become statistically larger relative to 65+ a few days after the lockdown impulse.
- Local projection trough gender-gap summary (20-day horizon; largest mobility gap between women and men):
  - Age 25–44: gender differential of 4.7 percentage points (largest).
  - Age 18–24: gender differential of 3.6 percentage points.
  - Age 45–64: gender differential of 3.7 percentage points.
  - Age 65+: fall in women’s mobility statistically indistinguishable from men’s.
- Voluntary social distancing by age:
  - Youngest cohorts react most forcefully to rising infections.
  - A doubling of COVID-19 cases leads to a fall in mobility by about one percentage point after 20 days for people aged 18–24.

---

### Robustness checks and sample splits
- Findings robust across:
  - Regression windows (20-day and 10-day).
  - Excluding Northern Italy regions with early measures.
  - Excluding Portugal and/or Spain (interpolated series); results closely mimic baseline.
  - Controlling for moving average of daily COVID-19 cases at province level.
  - Applying logistic transformation to mobility indicator.
  - Splitting samples by province characteristics: GDP per capita, population density, population size, rural vs not rural.
  - Replacing time fixed effects with day-of-the-week fixed effects: core findings persist; women-men differential appears more persistent without time fixed effects.
- Local projections excluding provinces with early and/or large outbreaks still confirm:
  - Lockdowns disproportionately reduce women’s mobility.
  - Lockdowns and voluntary social distancing took a larger toll on younger cohorts.
  - Estimated effects virtually identical in magnitude to baseline.

---

### Policy-relevant implications and interpretation
- Aggregation bias:
  - Heterogeneous mobility responses across gender and age imply aggregate estimates of lockdown effects can suffer from aggregation bias and mismeasurement when groups are pooled.
- Gender and childcare:
  - Disproportionate mobility reduction among women, concentrated among ages 25–44, points to an uneven burden of unpaid care (for example, childcare when schools close), with implications for female labor supply and labor force participation.
  - Women’s employment concentration in contact-intensive sectors (hospitality, personal care, retail) amplifies economic vulnerability.
  - Policy consideration: targeted interventions to support women during the pandemic (for example, offering parental leave to both men and women to encourage equal burden sharing).
- Age and intergenerational equity:
  - Lockdowns and voluntary responses reduce mobility more for younger, working-age cohorts, implying larger economic costs for those relying on labor income and facing less stable contracts.
  - Younger cohorts’ stronger reactions to rising infections raise concerns about scarring effects and inter-generational fairness.
  - Policy consideration: targeted measures to protect higher-risk older people without unduly limiting younger people’s mobility and employment opportunities; social pact to at least partially compensate younger workers for economic losses attributable to lockdowns.
- Long-term concerns:
  - Greater mobility reduction among younger generations during lockdowns suggests potential for long-lasting scarring effects on cohorts entering the labor force during the crisis; warrants further investigation.

*Source: wpiea2020282-print-pdf — https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020282-print-pdf.pdf*

### 4.1    Regression discontinuity  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  

### 4.1    Regression discontinuity

### Purpose and identification strategy
- Addresses endogeneity in the adoption of lockdown measures by focusing on high-frequency changes in mobility around specific lockdown measures.
- Uses a regression discontinuity (RD) design to reduce the risk that other contemporaneous factors drive observed mobility changes.
- Strengthens identification by:
  - Controlling for pre-existing mobility trends and the severity of the country’s epidemic in complementary local projection analysis.
  - Excluding provinces that experienced early or particularly severe outbreaks so that national lockdown adoption is closer to exogenous for the sample considered.

### Data and scope
- Analysis is based on novel and confidential mobility indicators provided by Vodafone for Italy, Portugal, and Spain at the provincial level.
- The Vodafone data are disaggregated by gender and age groups, enabling the RD design to detect heterogeneous responses.

### Main empirical findings from the RD analysis
- Stay-at-home orders led to a sharper contraction in women’s mobility relative to men’s across Italy, Portugal, and Spain.
- In most provinces, stay-at-home orders coincided with school closures.
- In provinces in Northern Italy that imposed school closures before stay-at-home orders, school closures alone generated a considerable gender differential in mobility, indicating the uneven role of women in caring for children.
- Stay-at-home orders disproportionately reduced the mobility of working-age people, especially those below 45 years of age.

### Complementary evidence and comparisons
- Local projections corroborate the RD findings, showing disproportionate effects on women’s mobility:
  - Among people aged 24 to 45, a full lockdown—including all measures used by governments during the pandemic—reduces the number of women leaving home by almost 26 percent, against an impact on men of about 21 percent.
- Local projections also examine voluntary responses to rising COVID-19 cases (holding lockdown stringency constant):
  - Both men and women significantly reduce mobility when infections increase and do so with equal intensity.
  - Younger people reduce mobility more strongly when infections increase, contrary to expectations that older, higher-risk groups would isolate more; this is consistent with survey evidence that younger people are more alarmed by the risk of contracting COVID-19.

### Policy-relevant implications emphasized by the analysis
- Heterogeneous mobility responses across gender and age suggest that aggregate estimates of lockdown effects can suffer from aggregation bias and mismeasurement if groups are pooled.
- The disproportionate mobility reduction among women points to an uneven burden of unpaid care (for example, childcare when schools close), with implications for female labor supply and labor force participation.
- The disproportionate mobility reduction among younger, working-age cohorts implies larger economic costs for those who rely on labor income and face less stable contracts, raising concerns about scarring effects and inter-generational fairness.
- In the context of repeated infection waves and weak economic activity, containment measures should be nuanced and consider heterogeneous effects across groups; targeted measures to protect higher-risk older people without unduly limiting younger people’s mobility and employment opportunities are a relevant policy consideration.

*Source: wpiea2020282-print-pdf - 4.1    Regression discontinuity — https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020282-print-pdf.pdf*

### Section 6 examines the robustness of the results.  Section 7 concludes.

### The COVID-19 crises in Italy, Portugal, and Spain

### Outbreak timelines and lockdown measures
- Italy
  - January 31: government declared a state of emergency and stopped flights from and to China.
  - February 21: first confirmed domestic case (Codogno, Lombardy).
  - February 22: decree quarantined more than 50,000 people from 11 municipalities (so called zone rosse).
  - February 23: local school closures in five regions in Northern Italy (see later analysis).
  - March 4: all schools and universities across Italy were closed for two weeks; sporting events behind closed doors until April 3.
  - March 8: lockdown of all 12 provinces in Lombardy and 14 provinces in Piedmont, Emilia-Romagna, Veneto, and Marche.
  - March 10: lockdown extended to the whole country; steep penalties including possibility of three months imprisonment for violators.
  - March 11: almost all commercial activities prohibited except supermarkets and pharmacies.
  - March 21: all non-essential businesses and industries closed; movement restricted.
  - May: many restrictions progressively eased.
  - June 3: freedom of movement across regions and other European countries restored.
- Portugal
  - March 2: first cases recorded.
  - March 18: entire Portuguese territory entered in a State of Emergency, which lasted until May 2.
  - April 9 to 13 (Easter week): special measures restricting movements between municipalities (concelhos); airports closed to civil transportation.
  - May 4: restrictions started to be eased and small stores reopened.
  - May 18: nurseries, the last two years of secondary school, restaurants, cafes, medium-sized stores and some museums reopened.
- Spain
  - January 31: first case (foreign tourist).
  - By mid-February: broader diffusion; all 50 provinces had confirmed cases by mid-March.
  - March 7: first local lockdown for a small municipality.
  - March 12: lockdown extended to four municipalities in Catalunya (70,000 people affected).
  - March 14: national lockdown and state of emergency; many nonessential activities closed, citizens permitted to travel to work and buy essential items.
  - March 25: official death toll surpassed that of China; most cases concentrated in Madrid.
  - March 28: order that all non-essential workers stay home from March 30 to April 9.
  - Early April: number of deaths peaked and then progressively declined until June 1 (first day without COVID-19 related deaths).
  - Beginning of May: progressive easing; May 11 allowed opening of small shops, terraces at half capacity, and places of worship at one-third capacity in 26 provinces/territories (about half of the population).

### A unique dataset (Vodafone mobility data)
- Data source and construction
  - Anonymized and aggregated mobility data provided by Vodafone through a confidential agreement.
  - Mobility indexed by analysis of connections of mobile phones to cell towers; home location identified by top 3 cells a phone connects to between 10pm and 5am; a customer is recorded as leaving home if the phone connects to a cell different from the home cells.
  - Gender and age differentiation possible using information customers provide when signing up for post-paid contracts.
  - Age groups: people aged between 18 and 24, between 25 and 44, between 45 and 64, and 65 and above.
  - For Spain and Portugal some daily observations are missing for oldest and youngest age groups; series are linearly interpolated to obtain a balanced panel.
- Privacy and controls
  - Data provided in anonymized form, reporting average mobility for a given gender and age group at the provincial (NUTS3) level when a minimum of 50 customers are available.
  - Data sharing protocol included technical and organizational controls and an ethical assessment prior to analysis.
  - In a few cases age information is inferred (examples: Spain separates 18–24 in family contracts based on factors including data usage; Portugal infers age from sequential number of personal identification cards with an error of five years at most).
- Advantages and validation
  - Key advantage: ability to differentiate mobility across gender and age groups to assess heterogeneous effects of lockdowns.
  - Vodafone mobility patterns broadly align with Apple and Google mobility data.
  - Correlations between Vodafone and Apple/Google indicators:
    - Italy and Portugal: range between 93 and 99 percent.
    - Spain: 72 and 88 percent.
  - Geographical disaggregation shows interdecile range of the mobility indicator as large as 20 percentage points across provinces; dispersion remains broadly constant over time.

### Mobility differentials by gender and age — preliminary evidence
- Lockdowns associated with larger drops in women’s mobility relative to men’s across all three countries (Figure 2 summary).
- Lockdowns reduce mobility of working-age people more strongly relative to people aged 65 and above.
- Panel evidence:
  - Binned scatter plots show mobility differentials (women-men) and age-group differentials relative to 65+; mobility indicator residualized with respect to day-of-week fixed effects.

### Empirical strategies to identify heterogeneous effects
- Identification challenges
  - Lockdown adoption is endogenous (e.g., more likely when health risks rise), and voluntary behavior changes could confound estimated effects.
  - Policymakers may impose lockdowns when mobility is high, potentially biasing associations.
- Two empirical strategies used
  - Regression discontinuity (RD) designs focusing on high-frequency changes in mobility around specific lockdown measures (narrow time window reduces confounding).
  - Local projections controlling for lagged mobility and severity of the epidemic (new infections); local projections estimated using provinces without severe outbreaks to treat national lockdown adoption as mostly exogenous and to examine voluntary responses to rising infections.

### Regression discontinuity (RD) analysis and main results
- RD setup
  - Running variable: time; treatment date: introduction of national stay-at-home orders (event-study style RD).
  - Narrow bandwidth: time window of 20 days around adoption of stay-at-home orders for baseline estimates.
  - Mobility indicator: mob_{i,g,a,t} = percentage of people moving in province i, gender g, age group a ∈ {[18,24]; [25,44]; [45,64]; [65+]}, at time t.
  - Estimation equation (local linear regression) includes province fixed effects (α_p), day-of-week fixed effects (τ_dow), treatment indicator stay_{j,t}, women dummy women_{i,a}, date_{j,t} (days since stay-at-home order), interactions including stay × date and women × stay, and clustering of standard errors at the province level.
  - Identification assumption: stay_{j,t} × date_{j,t} absorbs smooth relationship between date and the error term around introduction of lockdown.
- Key RD findings (baseline, 20-day window)
  - Introduction of stay-at-home orders led to a sharp drop in mobility for both men and women: the percentage of people leaving their homes in a day declined by about 15 percentage points.
  - Additional gender gap:
    - Age 25–44: women’s mobility declined by 3 percentage points more than men’s when stay-at-home orders were imposed (column (2) result).
    - Age 18–24: gap of 2.3 percentage points (column (1)).
    - Age 45–64: gap of 1.7 percentage points (column (3)).
    - Age 65+: no disproportionate effect of lockdowns on women’s mobility (column (4)).
  - Evidence from Northern Italy with two discontinuities (February 23 local school closures; March 9 national lockdown):
    - Divergence in mobility between men and women started at school closures.
    - Men’s mobility declined marginally when schools closed; women’s mobility showed a clear discontinuity.
    - Heat map across Northern Italian provinces shows mobility of men and women similar before February 23; darker colors after school closures indicate disproportionate reduction in women’s mobility across provinces; national stay-at-home orders led to further widening of the mobility gap.
    - Lack of a clear vertical pattern in province ordering by COVID-19 cases per population on March 9 indicates effects of school closures and lockdowns on gender mobility difference were not correlated with local intensity of the COVID-19 crisis.
- Interpretation
  - The largest mobility gender gap for people aged 25–44 suggests childcare needs largely explain the disproportionate impact of lockdowns on women’s mobility: school closures increased childcare responsibilities and reduced women’s mobility more than men’s.

*Italic: Source — wpiea2020282-print-pdf*

### 4.2    Local projections

### 4.2    Local projections

### Methodology and identification
- Estimation approach: local projection regressions (Jordà, 2005) using province-level Vodafone mobility data disaggregated at NUTS-3 level; sample of 163 provinces in Italy, Spain, and Portugal between January 1 and June 29, 2020.
- Endogeneity mitigation:
  - Control for the log of daily COVID-19 cases (ln∆cases) with weekly lag length to track pandemic stage.
  - Include lags of the dependent variable to capture pre-existing trends.
  - Exclude provinces that (i) registered the first 100 (cumulative) cases, (ii) had the highest number of COVID-19 cases by the end of June 2020, and (iii) had more than five percent of the country’s total confirmed cases when the lockdown stringency index reached its maximum (exclusions listed in notes).
  - Province, gender (or age-group), and time fixed effects; standard errors clustered at the province level.
- Lockdown stringency index: reconstructed from the University of Oxford’s Coronavirus Government Response Tracker as a simple average of nine sub-indicators, excluding public information campaigns and considering only measures adopted at the national level.
- Key regression interaction: lockdown stringency × gender dummy (womeni,a) to measure differential effects on women; lockdown stringency × age-group dummies to measure heterogeneous age responses.

### Differential impact of lockdowns by gender (local projections and RD corroboration)
- Local projection point estimates (age group 25–44):
  - A full lockdown leads to a decline in mobility for both sexes, with the trough reached after seven days.
  - Women’s mobility falls by 26 percentage points seven days after the introduction of lockdowns; men’s mobility declines by about 21 percentage points seven days after the introduction of lockdowns.
  - The women-minus-men mobility differential is statistically significant until the lockdown impulse weakens 15 days after initial tightening.
- RD estimates by age group (Table 1 coefficients preserved):
  - Stay-at-home coefficients:
    - 18–24: -19.60*** (0.59)
    - 25–44: -12.95*** (0.46)
    - 45–64: -12.52*** (0.41)
    - 65+: -11.76*** (0.46)
  - Women × stay-at-home coefficients:
    - 18–24: 2.31*** (0.50)
    - 25–44: -3.24*** (0.35)
    - 45–64: -1.74*** (0.34)
    - 65+: 1.55*** (0.57)
  - Observations and R-squared:
    - Observations: 13,909; 14,102; 14,151; 13,102 (columns 1–4)
    - R-squared: 0.87; 0.86; 0.87; 0.82
- Local projection trough estimates (Table 2: lockdown stringency and women interaction at trough):
  - Lockdown stringency (trough coefficients):
    - 18–24: -27.07*** (1.80)
    - 25–44: -20.60*** (1.60)
    - 45–64: -19.75*** (1.57)
    - 65+: -15.19*** (1.77)
  - Women × lockdown stringency (trough coefficients):
    - 18–24: -3.58*** (0.95)
    - 25–44: -4.73*** (0.67)
    - 45–64: -3.69*** (0.73)
    - 65+: 2.33 (1.46)
  - Days after the shock at trough: 7 for all columns.
  - Observations and sample:
    - Observations: 18,798; 18,798; 18,830; 17,872
    - Provinces: 163; 163; 163; 157
    - R-squared: 0.94; 0.94; 0.94; 0.91
- Summary gender findings:
  - Local projections corroborate RD: lockdowns disproportionately reduce women’s mobility.
  - Largest gender differential observed for ages 25–44.

### Voluntary social distancing and gender
- For a given lockdown stringency, mobility also responds to rising COVID-19 infections (ln∆cases).
- Age group 25–44 local projection:
  - A doubling of daily COVID-19 cases leads to a contraction in mobility by about 0.3 percentage points 20 days after the introduction of the lockdowns.
  - The effect of rising infections on mobility is similar across men and women: no statistically significant gender difference in voluntary social distancing responses.

### Heterogeneous effects by age (local projections and RD)
- RD evidence (section 5.1):
  - Stay-at-home orders drastically reduced mobility across all age groups but the drop was stronger for younger cohorts.
  - Percentage point drops at stay-at-home introduction (RD graphic and estimates): over 20 percentage points for people below 25; 11 percentage points for people aged 65 and above.
- Local projection evidence (equation (3) and Figure 8):
  - Mobility declines sharply across all age categories during the first ten days after a lockdown.
  - Troughs and timing:
    - People aged 18–24: trough of 30 percentage points nine days after the introduction of lockdowns.
  - Differential relative to 65+:
    - The mobility differential of 18–24 vs 65+ is the largest and statistically significant from the beginning of the projection horizon.
    - The drops for 25–44 and 45–64 become statistically larger relative to 65+ a few days after the lockdown impulse.
- Local projection trough gender-gap summary across age groups (20-day horizon; largest mobility gap between women and men):
  - Age 25–44: gender differential of 4.7 percentage points (largest).
  - Age 18–24: gender differential of 3.6 percentage points.
  - Age 45–64: gender differential of 3.7 percentage points.
  - Age 65+: fall in women’s mobility is statistically indistinguishable from men’s.
- Voluntary social distancing by age:
  - Youngest cohorts react most forcefully to rising infections.
  - A doubling of COVID-19 cases leads to a fall in mobility by about one percentage point after 20 days for people aged 18–24.

*Source: 4.2 Local projections (wpiea2020282-print-pdf).*

### 24.  A rise in infections leads also to reduction in mobility for people aged 25–44 and 45–64, even though the

### 24.  A rise in infections leads also to reduction in mobility for people aged 25–44 and 45–64, even though the

### Impact of infections and lockdowns by age
- Mobility declines for people aged 25–44 and 45–64 in response to a rise in infections; the effect is more modest than for younger cohorts.
- Mobility for people aged 65+ remains broadly unchanged and actually increases marginally towards the end of the projection horizon.
- Panel 9b (described in the source) confirms that mobility declines more for all working-age groups relative to people 65+.
- Two possible explanations:
  - A rise in infections reduces business activity in contact-intensive businesses (e.g., bars and restaurants), reducing employment in sectors where many young people work.
  - Younger cohorts may be more sensitive to fear of the virus and use more media/social media emphasizing danger; consistent with Bordalo et al. (2020) survey evidence that perceptions about health risks decline sharply with age.

### Quantitative RD estimates by age (selected coefficients from Table 5 and Table 6)
- Table 5 (robustness of RD results by age, several specifications):
  - Stay-at-home: 11.515*** (baseline) with (0.441)
  - 18–24 × stay-at-home: -9.238*** with (0.505)
  - 25–44 × stay-at-home: -2.526*** with (0.445)
  - 45–64 × stay-at-home: -1.482*** with (0.489)
  - Observations: 55,264 (baseline column)
  - R-squared: 0.85 (baseline column shows 0.85; table entries include 0.852)
- Table 6 (RD results by age across provinces, split samples):
  - Stay-at-home ranges: -9.866*** to -13.328*** (columns (1) and (2) shown)
  - 18-24 × stay-at-home: -8.307*** to -10.188*** across splits
  - 25-44 × stay-at-home: -3.463*** to -1.666*** across splits
  - 45-64 × stay-at-home: -2.970*** to -0.007 across splits
  - Observations across columns: 27,914; 27,350; 28,115; 27,149; 27,089; 28,175; 14,659; 14,659
  - R-squared across columns: 0.854; 0.870; 0.810; 0.908; 0.865; 0.840; 0.876; 0.876

### Impact of lockdowns and infections by gender
- Lockdown measures reduce the mobility of women more than men.
- The disproportionate decline in women’s mobility is concentrated among those aged 25–44 (likely linked to child care when schools close).
- Survey and region-specific evidence from Northern Italy: mobility gender gap opened at time of school closures in some regions.
- Differential estimates (selected coefficients from Table 3 and Table 4):
  - Table 3 (robustness of RD results by gender):
    - Stay-at-home (baseline): -12.95*** with (0.459)
    - Women × stay-at-home (baseline): -3.24*** with (0.353)
    - Observations: 14,102 (baseline column)
    - R-squared: 0.86 (baseline column shows 0.860)
  - Table 4 (by province characteristics):
    - Stay-at-home ranges: -12.16*** to -15.93*** across splits
    - Women × stay-at-home ranges: -2.655*** to -3.859*** across splits
    - Observations across columns: 7,203; 6,899; 7,293; 6,809; 6,992; 7,110; 3,695; 10,407
    - R-squared across columns: 0.861; 0.880; 0.817; 0.915; 0.876; 0.838; 0.881; 0.852

### Robustness checks
- RD and local projection findings are robust across:
  - Regression windows (20-day and 10-day).
  - Excluding Northern Italy regions that introduced early measures.
  - Excluding Portugal and/or Spain (where mobility series are interpolated); results closely mimic baseline.
  - Controlling for the moving average of daily COVID-19 cases at the province level (column (6) in tables).
  - Applying a logistic transformation to the mobility indicator (column (7) in tables).
  - Splitting samples by province characteristics: GDP per capita, population density, population size, rural vs not rural.
  - Replacing time fixed effects with day-of-the-week fixed effects (panels 11a–11d): core findings persist; women-men mobility differential in response to lockdowns appears more persistent without time fixed effects.
- Local projections excluding provinces with early and/or large outbreaks still confirm:
  - Lockdowns disproportionately reduce women’s mobility.
  - Lockdowns and voluntary social distancing took a larger toll on younger cohorts.
  - Estimated effects are virtually identical in magnitude to baseline.

### Interpretation and economic implications (from Conclusions)
- Gender:
  - Lockdowns reduce women’s mobility more than men’s, largely because women bear uneven child care burdens when schools are closed.
  - Stay-at-home orders have a larger impact on women aged 25–44 who are more likely to have young children.
  - Women tend to be employed in contact-intensive sectors (hospitality, personal care, retail) that have been more severely impacted.
  - Policy implication: targeted interventions to support women during the pandemic (example given: offering parental leave to both men and women to encourage equal burden sharing).
- Age / Generational effects:
  - Lockdowns benefit people above 65 by containing virus spread (they face much greater health risks).
  - Economic costs fall disproportionately on working-age people: lockdowns lead to a stronger reduction in the mobility of younger people (preventing work commutes and school-related activities).
  - Younger people also respond more to rising infections for a given lockdown stringency, possibly due to fear reactions or employment in contact-intensive sectors.
  - Younger workers depend on labor income, have less savings, and more precarious contracts—raising fairness and political economy concerns.
  - Policy implication: need for a social pact across generations to at least partially compensate younger workers for economic losses attributable to lockdowns to ensure public support for necessary measures.
- Long-term concerns:
  - Greater mobility reduction among younger generations during lockdowns suggests the potential for long-lasting scarring effects on cohorts entering the labor force during the crisis.
  - This preliminary insight warrants further investigation.

*Source: IMF working paper chapter/section (wpiea2020282-print-pdf).*

### References

### References

### Key cited studies
- Acemoglu, Daron, Victor Chernozhukov, Iván Werning, and Michael D. Whinston (2020). “Optimal Targeted Lockdowns in a Multi-Group SIR Model”. NBER Working Paper No. 27102.
- Adams, Abigail, Teodora Boneva, Marta Golin, and Christopher Rauh (2020). “Inequality in the Impact of the Coronavirus Shock: Evidence from Real Time Surveys”. CEPR Discussion Paper 14665.
- Adda, Jérôme (2016). “Economic Activity and the Spread of Viral Diseases: Evidence from High Frequency Data”. Quarterly Journal of Economics 131(2), pp. 891–941.
- Alon, Titan, Matthias Doepke, Jane Olmstead-Rumsey, and Michèle Tertilt (2020). “This Time It’s Different: The Role of Women’s Employment in a Pandemic Recession”. CEPR Discussion Paper 15149.
- Anderson, Michael L. (2014). “Subways, Strikes, and Slowdowns: The Impacts of Public Transit on Traffic Congestion”. American Economic Review 104(9), pp. 2763–96.
- Aum, Sangmin, Sang Yoon Tim Lee, and Yongseok Shin (2020). “COVID-19 Doesn’t Need Lockdowns to Destroy Jobs: The Effect of Local Outbreaks in Korea”. NBER Working Paper No. 27264.
- Belot, Michèle, Syngjoo Choi, Egon Tripodi, Eline van den Broek-Altenburg, Julian C. Jamison, and Nicholas W. Papageorge (2020). “Unequal Consequences of COVID-19 across Age and Income: Representative Evidence from Six Countries”. IZA Discussion Paper No. 13366.
- Berthelon, Matias, Diana Kruger, and Melanie Oyarzún (2015). “The Effects of Longer School Days on Mothers’ Labor Force Participation”. IZA Discussion Paper No. 9212.
- Bordalo, Pedro, Katherine B. Coffman, Nicola Gennaioli, and Andrei Shleifer (2020). “Older People are Less Pessimistic about the Health Risks of Covid-19”. NBER Working Paper No. 27494.
- Chetty, Raj, John N. Friedman, Nathaniel Hendren, and Michael Stepner (2020). “How Did COVID-19 and Stabilization Policies Affect Spending and Employment? A New Real-Time Economic Tracker Based on Private Sector Data”. NBER Working Paper No. 27431.
- Davis, Lucas W. (2008). “The Effect of Driving Restrictions on Air Quality in Mexico City”. Journal of Political Economy 116(1), pp. 38–81.
- Glaeser, Edward L, Caitlin S Gorback, and Stephen J Redding (2020). “How Much Does COVID-19 Increase with Mobility? Evidence from New York and Four Other us Cities”. NBER, Working Paper No. 27519.
- Glover, Andrew, Jonathan Heathcote, Dirk Krueger, and Jose-Victor Rios-Rull (2020a). “Health versus Wealth: On the Distributional Effects of Controlling a Pandemic”. CEPR DP 14606.
- Glover, Andrew, Jonathan Heathcote, Dirk Krueger, and José-Víctor Ríos-Rull (2020b). “Health versus Wealth: On the Distributional Effects of Controlling a Pandemic”. NBER Working Paper No. 27046.
- Goolsbee, Austan and Chad Syverson (2020). “Fear, Lockdown, and Diversion: Comparing Drivers of Pandemic Economic Decline”. NBER Working Paper No. 27432.
- Hupkau, Claudia and Barbara Petrangolo (2020). “Work, Care and Gender during the COVID-19 Crisis”. CEPR Discussion Paper 15358.
- Jordà, Òscar (2005). “Estimation and Inference of Impulse Responses by Local Projections”. American Economic Review 95(1), pp. 161–182.
- Lourenco, Pedro Rente, Gurjeet Kaur, Matthew Allison, and Terry Evetts (2020). “Data sharing and collaborations with Telco data during hte COVID-19 pandemic: A Vodafone case study”. Data & Policy.
- Maloney, William and Temel Taskin (2020). “Determinants of Social Distancing and Economic Activity During COVID-19: A Global View”. The World Bank, Policy Research Working Paper 9242.
- Montenovo, Laura, Xuan Jiang, Felipe Lozano Rojas, Ian M Schmutte, Kosali I Simon, Bruce A Weinberg, and Coady Wing (2020). “Determinants of Disparities in Covid-19 Job Losses”. NBER Working Paper 27132.
- Oreopoulos, Philip, Till Von Wachter, and Andrew Heisz (2012). “The short-and long-term career effects of graduating in a recession”. American Economic Journal: Applied Economics 4(1), pp. 1–29.
- Sevilla, Almudena and Sarah Smith (2020). “Baby Steps: The Gender Division of Childcare during the COVID19 Pandemic”. CEPR Covid Economics Issue 23.

### Appendix A. Lockdown stringency dynamics
- Finding: “A lockdown tightening tends to gradually decline and dissipate after about two weeks.”
- Empirical context: “These estimated dynamics reflect the way in which Italy, Portugal, and Spain have adjusted their lockdown stringency during the sample of analysis.”
- Dynamics observation: “As illustrated in panel A.1b, countries have adjusted the stringency of lockdowns rather frequently.”
- Figure notes:
  - Panel A.1a shows the impact of a full lockdown on lockdown stringency (Percentage of people moving).
  - Panel A.1b shows Lockdown stringency (Index).
  - In panel A.1a, the x-axis denotes the number of days, the line denotes the point estimates, and the shaded area corresponds to the 90 percent confidence interval computed with standard errors clustered at the province level.

*Source: wpiea2020282-print-pdf - References*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020282-print-pdf.pdf_