## Good Will Hunting: Do Disasters Make Us More Charitable? (Section 1 & Section 2)

## Source details

**Canonical URL:** [Good Will Hunting: Do Disasters Make Us More Charitable? (Section 1 & Section 2)](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023022-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023022-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023022-print-pdf.pdf.json)

---

### Abstract and central finding
- Research question: whether the COVID-19 pandemic altered prosocial tendencies and charitable donations.
- Data: novel daily dataset of debit and credit card transactions covering January 1, 2019 to October 2, 2022 for three European countries (Estonia, Latvia, Lithuania).
- Central result: the COVID-19 pandemic and government interventions have no significant effect on how much people contribute to charities as a share of total spending.
- Interpretation: a higher preference for precautionary savings during the pandemic appears to outweigh altruistic behavior, while government welfare programs crowd out private charitable donations.

### Context and motivation
- Pandemic scale reported in text: over 671.2 million COVID-19 cases and more than 6.8 million deaths worldwide.
- Prior literature: many studies find increased prosocial behavior after disasters but are often based on small-scale surveys or experiments and focus on localized events.
- Contribution: combines daily POS debit and credit card transaction data with daily COVID-19 incidence and government intervention indices to study actual charitable donations at high frequency across three Baltic countries.

### Data overview and descriptive statistics
- Transaction data source: Swedbank POS debit and credit card transactions covering about half of POS transactions in the Baltics; point-of-sale data exclude cash withdrawals and include both in-person and online transactions.
- Sample period for Swedbank POS data used in regression analysis: January 1, 2020 to October 2, 2022 (daily).
- Spending categories: 33 categories including “religious and charitable giving”.
- COVID-19 and policy data source: Oxford Covid-19 Government Response Tracker (OxCGRT).
- Policy indices used (each ranges between 0 to 100): stringency index; containment and health index; economic support index; overall government response index.
- Summary statistics (sample size = 3,018 observations unless noted):
  - Share of charitable donations in debit and credit card transactions: Mean 0.0020 percent; Std. dev. 0.0027; Minimum 0; Maximum 0.0348 percent.
  - COVID-19 infections: Mean 923; Std. dev. 1,828; Minimum 0; Maximum 15,412.
  - Deaths: Mean 6; Std. dev. 9; Minimum 0; Maximum 79.
  - Stringency index: Mean 36.6; Std. dev. 20.7; Minimum 0; Maximum 87.0.
  - Containment and health index: Mean 42.4; Std. dev. 16.4; Minimum 0; Maximum 76.7.
  - Economic support index: Mean 53.0; Std. dev. 35.7; Minimum 0; Maximum 100.0.
  - Government response index: Mean 43.7; Std. dev. 16.5; Minimum 0; Maximum 74.9.
- Aggregate changes reported:
  - Aggregate level of charitable giving in the Baltics declined from an average of 0.1164 percent in 2019 to 0.0028 percent in 2020 and 0.0029 percent in 2021.
  - Country-level changes in the share of charitable contributions in total spending:
    - Estonia: declined from 0.0019 percent to 0.0016 percent.
    - Latvia: declined from 0.3453 percent to 0.0044 percent.
    - Lithuania: increased from 0.0019 percent to 0.0023 percent.
- Card transaction volume dynamics:
  - Second quarter 2020: total card transactions in the three Baltic countries declined by an average of 2.5 percent year-on-year after growing at an annual rate of 10 percent in the first quarter.
  - 2020 country-level card transaction changes: Estonia −4 percent; Latvia +2 percent; Lithuania +14.5 percent.
  - 2021 recovery pace: Lithuania +26.2 percent; Latvia +17.9 percent; Estonia +12.3 percent.
- Time-series properties: panel unit root tests (Im-Pesaran-Shin) indicate variables are stationary in levels after logarithmic transformation.

### Empirical strategy
- Baseline panel specification:
  - log(charity_{i,t}) = β1 log(covid_{i,t}) + β2 X_{i,t} + η_i + μ_t + ε_{i,t}
  - charity_{i,t}: share of charitable contributions in total debit and credit card transactions in country i at time t.
  - covid_{i,t}: number of new COVID-19 cases or deaths.
  - X_{i,t}: vector of health and economic policy measures (stringency index, containment and health index, economic support index, overall government response index).
  - η_i: country fixed effects; μ_t: time fixed effects.
- Robust standard errors clustered at the country level to account for heteroskedasticity.
- Panel includes three countries and daily observations; number of observations reported in baseline tables ranges (e.g., 2,078 to 2,586 depending on specification).

### Main empirical results (Section 1 baseline)
- COVID-19 prevalence effect:
  - Coefficients on COVID-19 infections (or deaths) are consistently positive across specifications but do not reach statistical significance at conventional levels.
  - Example coefficients (Table 2, columns):
    - COVID-19 infections: 0.000153, 0.000149, 0.000011, 0.000132, 0.000012.
- Government intervention effects:
  - Government interventions generally have negative coefficients, indicating declines in charitable giving associated with stronger policy responses; these negative coefficients are not statistically significant in baseline results.
  - Example coefficients:
    - Stringency index: -0.000028, -0.000012.
    - Containment and health index: -0.000240.
    - Economic support index: -0.000209, -0.000211.
    - Government response index: -0.000538.
- Table 2 details:
  - Number of observations across columns [1]–[5]: 2,581; 2,581; 2,083; 2,586; 2,078.
  - Number of countries = 3 in all reported specifications.
  - Adjusted R-squared values: 0.56, 0.56, 0.55, 0.56, 0.55.
- Robustness and interpretation:
  - Results are consistent across alternative specifications.
  - Authors interpret findings as evidence that increased precautionary saving motives and difficulties using physical cash during NPIs may have counteracted altruistic impulses.
  - Economic support measures may have crowded out private donations by changing perceptions of need when governments provide direct income transfers.

### Empirical approach and additional estimates (Section 2)
- Estimation: models estimated in first differences with country fixed effects and time fixed effects; robust standard errors clustered at the country level.
- Sample coverage reported in tables: up to 33 countries and daily observations.
- Key first-differences regression estimates (selected reported figures):
  - Table (first set):
    - COVID-19 infections: 0.000224 [0.000]
    - COVID-19 deaths: 0.000501 [0.000]
    - Stringency index: -0.000308  -0.002249 [0.002] [0.003]
    - Economic support index: 0.003380 0.003478 [0.002] [0.001]
    - Number of observations: 2,022 1,279
    - Number of countries: 33
    - Adj R2: 0.56 0.60
  - Table (second set, five specifications):
    - COVID-19 deaths: 0.000120 0.000121 0.000263 0.000139 0.000274 [0.000] [0.000] [0.000] [0.000] [0.000]
    - Stringency index: -0.000050  -0.000121 [0.000] [0.000]
    - Containment and health index: -0.000080 [0.000]
    - Economic support index: -0.000425 -0.000448 [0.000] [0.000]
    - Government response index: -0.000424 [0.000]
    - Number of observations: 1,846 1,846 1,463 1,846 1,463
    - Number of countries: 33 (all)
    - Adj R2: 0.59 0.59 0.60 0.59 0.60
- Main findings reiterated:
  - Overall result: the COVID-19 pandemic and government interventions have no statistically significant effect on the share of charitable donations in total consumer spending, as measured by debit and credit card transactions.
  - COVID-19 variables often have positive coefficients but weak statistical support.
  - Government intervention indices often show negative coefficients but effects are not statistically significant.
- Interpretation (Section 2 emphasis):
  - Precautionary savings and uncertainty about health and economic outcomes likely reduced charitable giving.
  - Lockdown measures limited opportunities to make cash donations; economic support measures may have crowded out private donations.

### Conclusions and policy implications
- Empirical conclusion: no statistically significant evidence that the COVID-19 pandemic or associated government interventions materially changed charitable giving as a share of total debit and credit card spending in Estonia, Latvia, and Lithuania over the sample period.
- Behavioral mechanism highlighted: heightened precautionary savings and uncertainty likely dampened charitable giving despite potential altruistic responses to the disaster.
- Policy implication: public economic support measures that strengthen the social safety net may crowd out private charitable contributions.

*Source: IMF Working Paper WP/23/22 — Serhan Cevik, "Good Will Hunting: Do Disasters Make Us More Charitable?" (Section 1 and Section 2).*

### Section 1

### Good Will Hunting: Do Disasters Make Us More Charitable? (Section 1)

### Abstract and central finding
- Research question: whether the COVID-19 pandemic altered prosocial tendencies and charitable donations.
- Data: novel daily dataset of debit and credit card transactions covering January 1, 2019 to October 2, 2022 for three European countries (Estonia, Latvia, Lithuania).
- Central result: the COVID-19 pandemic and government interventions have no significant effect on how much people contribute to charities as a share of total spending.
- Interpretation offered: a higher preference for precautionary savings during the pandemic appears to outweigh altruistic behavior, while government welfare programs crowd out private charitable donations.

### Context and motivation
- The COVID-19 pandemic produced large human and economic costs: over 671.2 million COVID-19 cases and more than 6.8 million deaths worldwide (figures reported in the text).
- Prior literature: many studies find increased prosocial behavior after disasters but are often based on small-scale surveys or experiments and focus on localized events.
- Contribution: combines daily POS debit and credit card transaction data with daily COVID-19 incidence and government intervention indices to study actual charitable donations at high frequency across three Baltic countries.

### Data overview and descriptive statistics
- Transaction data source: Swedbank POS debit and credit card transactions covering about half of POS transactions in the Baltics; point-of-sale data exclude cash withdrawals and include both in-person and online transactions.
- Sample period for Swedbank POS data used in regression analysis: January 1, 2020 to October 2, 2022 (daily).
- Spending categories: 33 categories including “religious and charitable giving”.
- COVID-19 and policy data source: Oxford Covid-19 Government Response Tracker (OxCGRT).
- Policy indices used (each ranges between 0 to 100): stringency index; containment and health index; economic support index; overall government response index.
- Summary statistics (sample size = 3,018 observations unless noted):
  - Share of charitable donations in debit and credit card transactions: Mean 0.0020 percent; Std. dev. 0.0027; Minimum 0; Maximum 0.0348 percent.
  - COVID-19 infections: Mean 923; Std. dev. 1,828; Minimum 0; Maximum 15,412.
  - Deaths: Mean 6; Std. dev. 9; Minimum 0; Maximum 79.
  - Stringency index: Mean 36.6; Std. dev. 20.7; Minimum 0; Maximum 87.0.
  - Containment and health index: Mean 42.4; Std. dev. 16.4; Minimum 0; Maximum 76.7.
  - Economic support index: Mean 53.0; Std. dev. 35.7; Minimum 0; Maximum 100.0.
  - Government response index: Mean 43.7; Std. dev. 16.5; Minimum 0; Maximum 74.9.
- Aggregate changes reported:
  - Aggregate level of charitable giving in the Baltics declined from an average of 0.1164 percent in 2019 to 0.0028 percent in 2020 and 0.0029 percent in 2021.
  - Country-level changes in the share of charitable contributions in total spending:
    - Estonia: declined from 0.0019 percent to 0.0016 percent.
    - Latvia: declined from 0.3453 percent to 0.0044 percent.
    - Lithuania: increased from 0.0019 percent to 0.0023 percent.
- Card transaction volume dynamics:
  - Second quarter 2020: total card transactions in the three Baltic countries declined by an average of 2.5 percent year-on-year after growing at an annual rate of 10 percent in the first quarter.
  - 2020 country-level card transaction changes: Estonia −4 percent; Latvia +2 percent; Lithuania +14.5 percent.
  - 2021 recovery pace: Lithuania +26.2 percent; Latvia +17.9 percent; Estonia +12.3 percent.
- Time-series properties: panel unit root tests (Im-Pesaran-Shin) indicate variables are stationary in levels after logarithmic transformation.

### Empirical strategy
- Baseline panel specification:
  - log(charity_{i,t}) = β1 log(covid_{i,t}) + β2 X_{i,t} + η_i + μ_t + ε_{i,t}
  - charity_{i,t}: share of charitable contributions in total debit and credit card transactions in country i at time t.
  - covid_{i,t}: number of new COVID-19 cases or deaths.
  - X_{i,t}: vector of health and economic policy measures (stringency index, containment and health index, economic support index, overall government response index).
  - η_i: country fixed effects; μ_t: time fixed effects.
- Robust standard errors clustered at the country level to account for heteroskedasticity.
- Panel includes three countries and daily observations; number of observations reported in baseline tables ranges (e.g., 2,078 to 2,586 depending on specification).

### Main empirical results
- COVID-19 prevalence effect:
  - Across specifications, coefficients on COVID-19 infections (or deaths) are consistently positive, suggesting that higher infections/deaths could be associated with increased charitable donations.
  - These coefficients do not reach statistical significance at conventional levels and indicate only a weak association.
- Government intervention effects:
  - Government interventions (stringency, containment and health, economic support, overall government response) generally have negative coefficients, indicating declines in charitable giving associated with stronger policy responses.
  - These negative coefficients are not statistically significant in the baseline results.
- Baseline table notes:
  - Example coefficients (Table 2, column specifications):
    - COVID-19 infections reported coefficients include 0.000153, 0.000149, 0.000011, 0.000132, 0.000012 (format preserved from table).
    - Stringency index reported coefficients include -0.000028, -0.000012.
    - Containment and health index reported coefficient -0.000240.
    - Economic support index reported coefficients -0.000209, -0.000211.
    - Government response index reported coefficient -0.000538.
  - Number of observations in Table 2: 2,581; 2,581; 2,083; 2,586; 2,078 across columns [1]–[5].
  - Number of countries = 3 in all reported specifications.
  - Adjusted R-squared values shown as 0.56, 0.56, 0.55, 0.56, 0.55 across columns.
- Robustness and interpretation:
  - Results are consistent across alternative specifications.
  - Interpretation emphasizes that increased precautionary saving motives and difficulties using physical cash during NPIs may have counteracted any increase in altruistic impulses.
  - Economic support measures may have crowded out private donations by changing perceptions of need when governments provide direct income transfers.

### Conclusions and implications
- Empirical conclusion: no statistically significant evidence that the COVID-19 pandemic or associated government interventions materially changed charitable giving as a share of total debit and credit card spending in Estonia, Latvia, and Lithuania over the sample period.
- Behavioral mechanism highlighted: heightened precautionary savings and uncertainty about health and economic outcomes likely dampened charitable giving despite potential altruistic responses to the disaster.
- Policy implication noted: public economic support measures that strengthen the social safety net may crowd out private charitable contributions.

*Source: IMF Working Paper WP/23/22 — Serhan Cevik, "Good Will Hunting: Do Disasters Make Us More Charitable?" (Section 1).*

### Section 2

### wpiea2023022-print-pdf - Section 2

### Empirical approach and data
- Dependent variable: the share of charitable contributions in total debit and credit card transactions.
- Estimation: models estimated in first differences with country fixed effects and time fixed effects.
- Standard errors: robust, clustered at the country level.
- Sample coverage reported in tables: up to 33 countries and daily observations.

### Key regression estimates (first-differences results)
- Table (first set):
  - COVID-19 infections: 0.000224 [0.000]
  - COVID-19 deaths: 0.000501 [0.000]
  - Stringency index: -0.000308  -0.002249 [0.002] [0.003]
  - Economic support index: 0.003380 0.003478 [0.002] [0.001]
  - Number of observations: 2,022 1,279
  - Number of countries: 33
  - Country FE: Yes
  - Time FE: Yes
  - Adj R2: 0.56 0.60
- Table (second set, five specifications):
  - COVID-19 deaths: 0.000120 0.000121 0.000263 0.000139 0.000274 [0.000] [0.000] [0.000] [0.000] [0.000]
  - Stringency index: -0.000050  -0.000121 [0.000] [0.000]
  - Containment and health index: -0.000080 [0.000]
  - Economic support index: -0.000425 -0.000448 [0.000] [0.000]
  - Government response index: -0.000424 [0.000]
  - Number of observations: 1,846 1,846 1,463 1,846 1,463
  - Number of countries: 33 33 33 33 33
  - Country FE: Yes (all)
  - Time FE: Yes (all)
  - Adj R2: 0.59 0.59 0.60 0.59 0.60

### Main findings
- Overall result: the COVID-19 pandemic and government interventions have no statistically significant effect on the share of charitable donations in total consumer spending, as measured by debit and credit card transactions.
- Coefficient patterns:
  - COVID-19 variables (infections or deaths) consistently have positive coefficients in some specifications (e.g., Baltics) but these coefficients do not reach statistical significance at conventional levels and thus indicate only a weak degree of association.
  - Government intervention indices (stringency, containment and health, economic support, government response) show negative coefficients in multiple specifications, but these effects are also not statistically significant.
- Interpretation:
  - A higher preference for precautionary savings amid a global shock appears to outweigh altruistic behavior.
  - Pandemic severity entails economic damages and adverse psychological effects that increase uncertainty about health and economic outcomes, manifesting as larger precautionary savings and lower charitable contributions.
  - Government interventions may magnify the effect: lockdown measures limit opportunities to make cash donations; economic support measures designed to strengthen the social safety net reduce the tendency to make charitable donations (interpreted as crowding out).

### Contextual and supporting evidence
- Pandemic scale reported in the conclusion: over 671.2 million infections and more than 6.8 million deaths worldwide.
- Prior literature and mechanisms cited:
  - Evidence that exposure to disasters can increase philanthropy, but most prior studies rely on small-scale surveys and experiments.
  - Government welfare programs may crowd out private charitable donations (consistent with Abrams and Schmitz, 1978, 1984; Jones, 1983; Brooks, 2000; Andreoni and Payne, 2003; Garrett and Rhine, 2007).
  - Changes in payment habits during pandemics (reduced use of cash) may also affect measured charitable giving (Cevik, 2020).

### Conclusion and implications
- Contribution of the paper:
  - Combines a daily dataset of debit and credit card transactions with daily data on COVID-19 prevalence and government interventions.
  - Conducts a real-time analysis of actual charitable donations in three European countries during the COVID-19 pandemic in a regression framework.
- Policy-relevant implication:
  - In crisis contexts like the COVID-19 pandemic, precautionary saving motives and changes in payment methods can offset altruistic responses; government support programs may reduce private charitable contributions through a crowding-out mechanism.

*Source: Author's estimations and Section 2 of wpiea2023022-print-pdf*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023022-print-pdf.pdf_