## 2.1  The German Socio-Economic Panel (SOEP)

## Source details

**Canonical URL:** [2.1  The German Socio-Economic Panel (SOEP)](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023248-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023248-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023248-print-pdf.pdf.json)

---

### Data and sample
- SOEP: annual representative survey of persons, households and families in Germany; established in 1984.
- Coverage: residents of former West and East Germany, immigrants, and resident foreigners; individuals within households surveyed if at least 16 years of age.
- Annual interviews: approximately 41,000 households and approximately 69,000 individuals, nearly half of whom are women.
- Study period used: 2002-2018.
- Number of observations referenced: 362,488.
- Note: sample stability over time highlighted as an advantage.

### Subjective well-being (SWB) measure
- SWB operationalized as life satisfaction.
- Survey question: "How satisfied are you with your life, all things considered?"
- Response scale: 11-point scale ranging from 0 (completely dissatisfied) to 10 (completely satisfied).
- Empirical pattern: average life satisfaction score relatively stable at slightly above 7, with a decline during the 2008 financial crisis.
- Controls exploited: age, gender, marital status, employment status, educational achievement (years of schooling), total income, number of people living in the household, and health.

### Monetary policy shocks: construction and identification
- Source of surprises: Euro Area Monetary Policy Event-Study Database (EA-MPD) of Altavilla et al. (2019).
- Asset-price changes used: changes in the overnight index with maturities of 1 month, 3 months, 6 months or 1 year.
- Aggregation: first principal component of the four OIS-maturity surprises computed to obtain a single measure.
- Data frequency and coverage: daily data from January 1, 2002 to December 31, 2018; all scheduled ECB meetings during this 17-year period included.
- Event-window definition in EA-MPD: price changes computed as difference between median quote during 13:25-13:35 before the press release and median quote for 15:40-15:50 after the press conference.
- Note on informational content: central bank announcement surprises can include Delphic and Odyssean components; positive co-movement between interest rates and stock market following surprises signals “central bank information shocks”.

### Empirical approach: prima-facie and two-step aggregation
- Prima-facie: construct "clean" SWB by regressing life satisfaction on individual fixed effects and time fixed effects (month, year, month*year); 362,488 observations grouped into 20 equally sized bins based on EONIA.
- Two-step aggregation:
  - First-stage microeconometric OLS: W_{i,mdt} = α + γ X_{i,mdt} + β_d + γ_m + θ_t + ε_{i,mdt}.
  - Compute quarterly average residual ̄ε_{qt}.
  - Second-stage: ̄ε_{qt} = α + β MP_{qt} + ξ_{qt}, where MP_{qt} is quarterly average of EA-MPD monetary policy shocks.
- Key findings from Table 1 specifications:
  - Baseline indicates a negative association between life satisfaction and monetary policy surprises.
  - Dynamic construction (column 2): a monetary surprise of one standard deviation (1.42 basis points) reduces life satisfaction by approximately 2 points.
  - Inclusion of macro controls and time dummies maintains negative effect.
  - Including three lags of MP surprises preserves coefficient stability.
  - Robust regression confirms negative effect with smaller magnitude.
- Identification note: sample starts in 2002 because intraday OIS data 1999–end-2001 are very noisy.

### Empirical approach: event-study at individual level and identification strategy
- Merge each panel observation (individual * interview date) with the corresponding monetary surprise from EA-MPD.
- Event-study design: compare individuals responding within a w-day window after the announcement to those responding within the w-day window prior to the announcement; responses on the day of the announcement and outside windows are dropped.
- Treatment intensity: use direction and intensity of each surprise (variable D_{i,mdt} MP_{mdt} where D = 1 if response in w-day post-announcement window, 0 if in w-day pre-announcement window).
- Baseline window choice: w = 7 days.
- Example: LTRO May 7, 2009 produced 3-month OIS decline of -1.50 basis points; D_{i,mdt} MP_{mdt} = -1.50 for responses May 8–May 15, 2009 and zero for responses Apr 30–May 6, 2009.
- Estimation model: W_{i,mdt} = α + β D_{i,mdt} MP_{mdt} + γ X_{i,mdt} + θ_t + γ_m + ζ_d + ε_{i,mdt}.
- Fixed effects and controls: year and month fixed effects; some specs include year*month fixed effects; individual time-varying controls included.
- Panel structure: event-window construction leads to sparse repeated observations; average individual appears only 2.5 times; baseline exploits between variation.
- Robustness checks: Mundlak quasi-fixed effects, individual fixed effects on full sample, ordered logistic regressions.
- Identification assumptions and checks:
  - Interview attrition assumed random; Appendix tests show participant characteristics not significantly different before/after events.
  - Monetary surprises constructed to be exogenous to other macro news; year*month dummies used to capture correlated macro shocks.
  - Variant excluding most important monetary events estimated to guard against nonrandom attrition around major events.

### Results: baseline event-study estimates and interpretation
- Main qualitative finding: tightening monetary surprises decrease life satisfaction; easing shocks have mirror positive effects.
- Robustness: changing fixed effects or controls has marginal impact; inclusion of year*month dummies decreases coefficient slightly (e.g., from -1.05 to -0.98) but order of magnitude remains.
- Alternative MP definition (binary easing = "-1", tightening = "1") yields negative coefficient of smaller magnitude.
- Estimation methods: ordered logistic and generalized ordered logit produce similar substantive conclusions.
- Magnitude and economic interpretation:
  - Two-step aggregation: one standard deviation (1.42 basis points) reduced life satisfaction by approximately 2 points (dynamic specification).
  - Baseline event-study: a one standard deviation increase (1.92) in MP surprises reduces the life satisfaction indicator by 0.024 points.
  - Monetized equivalent: a one standard deviation MP surprise is equivalent to a drop in household income of approximately 4%.
  - Unemployment-equivalent: the drop caused by a one standard deviation MP surprise is equivalent to approximately 4% of the drop in life satisfaction caused by being unemployed.
- Ordered logistic marginal effects:
  - A one-point increase in the surprise effect reduces probability of answering '8', '9' or '10' by 0.17%, 0.11% and 0.04%, respectively.
  - Example: likelihood of answering "9" is 33.86 when monetary surprise = 0 and predictors at mean; it becomes 33.42 when monetary surprise increases by one standard deviation.
- Standard errors and clustering:
  - Standard errors clustered at individual level in all specifications.
  - Sensitivity: clustering at month*year, event window, date, and household levels produces standard errors close to baseline; main specification coefficient remains statistically significant (1% level or better).

### 4.2 Robustness checks — sensitivity and alternative specifications
- Baseline event window: 7 days before/after announcement; alternative windows tested: 5 days, 3 days and 1 day.
  - 5-day window: estimate -1.061*** (0.225).
  - 3-day window: estimate -0.736** (0.324).
  - 1-day window: estimate 0.068 (0.460) — explained by day-of-week effects or need for information diffusion.
  - Placebo test (lag policy surprises by one week): no impact on life satisfaction.
  - Continuous-surprise specification (effect lasts until a new announcement): effect on life satisfaction smaller than half the size of baseline; continuous surprise estimate -0.443*** (0.090).
- Event-study decomposition (pre/post leads and lags): 7-day post-announcement coefficient estimates not significant in pre-treatment period, supporting absence of parallel trends.
- Sensitivity to outliers/major events (Table 4):
  - Winsorize shocks at 5%; exclude events with fewer than 100 or 1000 observations; exclude observations with standardized residuals > 1.96; exclude observations with large Cook’s distance; Hubert robust estimators.
  - Results similar in all cases.
- Alternative estimators for unobserved heterogeneity (Table 5):
  - Mundlak quasi-fixed effects, fixed effects, random effects (full sample): coefficients remain significant but ~half the baseline magnitude in some specs.
  - Pooled OLS with lagged satisfaction term: coefficient intact.
  - Panel random-effects ordered logit with means of time-variant controls and BUC fixed-effects ordered logit: broadly in line with baseline.
- Alternative intraday asset price measures (Table 6):
  - OIS at 1M, 3M, 6M, 12M yield coefficients similar to baseline.
  - Euro Stoxx50-based coefficient not statistically significant.

### 4.2.2 Pure monetary policy surprises
- Euro Stoxx50 surprises have no effect on life satisfaction.
- Delphic vs Odyssean shocks:
  - Delphic: convey state-of-economy news; positive comovement between short-term rates and stock markets.
  - Odyssean: pure monetary policy shocks associated with negative comovement.
- Using Jarociński and Karadi (2020) approach to isolate "pure" monetary policy shocks (negative comovement identifies pure shocks):
  - Table 7 (pure shocks) reproduces baseline specifications and finds pure monetary policy shocks significantly decrease individual life satisfaction and are strongly consistent with baseline.
  - Pure shock coefficients example: -0.885***, -0.826**, -0.681**, -1.092***, -0.845*** (standard errors: (0.220), (0.322), (0.294), (0.238), (0.230)); MP surprise (-1/1) -0.028*** (0.009).

### 4.2.3 Google Trends: measuring shock diffusion and public attention
- Motivation: intensity of high-frequency OIS reactions may relate to how much monetary policy decisions resonate with the general public via media transmission.
- Data: daily Google Trends search volume index for the term "EZB" ("Europäische Zentralbank") from Germany (January 1, 2002 to December 31, 2018).
- Findings:
  - Average Google index on ECB meeting days greater than on other days: monetary event sample mean 15.939 vs other days mean 6.926.
  - Scatter-plot of absolute monetary policy surprises vs "EZB" search volume shows positive relationship.
  - Interpretation: intensity of intraday financial market surprises relates to broader public attention; Google Trends is a plausible proxy for shock intensity.
- Constructed Google-based shock variable:
  - Use Google index and OIS surprises; sign the Google index according to OIS direction (make index negative in case of easing).
  - Estimate model using adjusted Google-based shock variable alongside MP surprise.
- Results with MP surprise diffusion (Table 9):
  - MP surprise diffusion coefficients reported: -0.002***, -0.002**, -0.002*, 0.000, -0.002* with standard errors (0.001) for each coefficient.
  - Adj. R-squared reported: 0.262, 0.264, 0.264, 0.252, 0.263.
  - Observations reported (source text): 9039590395903959039590395.
- Key interpretation: "the more a monetary event arouses public interest, the more strongly SWB will respond."

### Heterogeneity of effects (Table 10) and difference-in-difference
- Easing vs tightening:
  - Easing surprise coefficient: 0.934** (standard error: 0.398).
  - Tightening surprise coefficient: -1.031*** (standard error: 0.343).
  - Economic effect of a one standard deviation surprise stronger for tightening than easing.
- Time heterogeneity:
  - After 2009: -0.895*** (standard error: 0.226).
  - Before 2009: -1.345*** (standard error: 0.309).
  - Interpretation: coefficient decline over time consistent with downward trend in magnitude of surprises; overall impact relatively constant since 2003.
- Age heterogeneity (selected coefficients and standard errors):
  - 0.290 (0.997), -1.138*** (0.361), -1.420*** (0.319), -1.006*** (0.356), 0.020 (0.500).
  - Impact significant for middle-aged individuals.
- Income heterogeneity (selected coefficients and standard errors):
  - -0.051 (0.639), -1.391*** (0.409), -1.502*** (0.331), -0.822*** (0.274).
  - Tightening shocks have no significant effects on low-income individuals; negative effect strongest for the middle class.
- Gender:
  - Women coefficient: -1.205*** (standard error: 0.262) — decrease in life satisfaction stronger for women.
- Difference-in-difference (DID) exploiting financial literacy by age (treated: ages 25-70; control: under 25 and over 70):
  - DID linear OLS estimates (DID - MP surprise): -1.166**, -1.179**, -1.275***, -1.480***, -1.188*** (standard errors: (0.458), (0.460), (0.460), (0.459), (0.460)); Adj. R-squared: 0.262, 0.263, 0.263, 0.252, 0.261; Observations: 142839 (repeated).
  - DID ordered LOGIT: coefficients -1.207**, -1.268**, -1.397***, -1.555***, -1.207** (standard errors: (0.538), (0.541), (0.541), (0.538), (0.538)); Pseudo R-squared: 0.080, 0.081, 0.081, 0.075, 0.080; Observations: 142839.
  - DID results consistent with baseline.

### Channels linking monetary surprises to life satisfaction (Table 12)
- Two mechanisms distinguished:
  - Direct effect: individuals anticipate easing shocks will improve their own economic conditions.
  - Indirect effect: individuals expect a positive effect on the economy in general.
- Empirical proxies: "How concerned are you?" and expectations for life satisfaction in one year and five years.
- Regression results (MP surprise on channels):
  - Econ conditions: coefficient -0.648*** (standard error: 0.237); Adj. R-squared: 0.257; Observations: 137562.
  - Fin conditions: coefficient -0.022 (standard error: 0.231); Adj. R-squared: 0.250; Observations: 143860.
  - Life satisfaction 5Y: coefficient -0.614** (standard error: 0.253); Observations: 645652.
  - Life satisfaction 1Y: coefficient -2.088*** (standard error: 0.554); Observations: 3615.
- Interpretation:
  - Monetary surprises significantly deteriorate individuals’ views about the economy in general.
  - Surprises do not affect individuals’ degree of concern about their own economic conditions.
  - Tightening surprises reduce future perceptions of life satisfaction, stronger for one-year expectations.

### Robustness and complementary numeric summaries (selected exact figures preserved)
- Baseline linear OLS MP surprise coefficients: -1.054***, -0.988***, -0.722***, -0.993***, -1.004*** (standard errors: (0.182), (0.260), (0.249), (0.174), (0.192)); Observations: 142839.
- Ordered LOGIT baseline: -1.275***, -1.123***, -0.783***, -1.255***, -1.275*** (standard errors: (0.210), (0.302), (0.290), (0.200), (0.210)).
- MP surprise (-1/1) in some specs: -0.024*** (standard error: 0.006); ordered LOGIT analog: -0.026*** (standard error: 0.007).
- Window sensitivity (Table 3): 5 days -1.061*** (0.225); 3 days -0.736** (0.324); 1 day 0.068 (0.460); placebo -0.000 (0.000).
- Google Trends summary (reported means and s.d.):
  - Google Trends "EZB" index mean for monetary event sample: 15.939; No: 6.926.
  - "EZB2" mean: 6.382, -0.205.
  - Index standard-deviation: 6.772, 4.487, 6.751, 4.252.
  - Nb. Obs (as reported): 1484,6011484,601.
- Appendix — Table A2 summary statistics (mean and std. dev.):
  - Life satisfaction: 7.09 1.78.
  - Monthly income: 28772 044.75.
  - Age: 49.51 6.48.
  - Household size: 3.18 1.64.
  - Gender: 0.72 0.44.
  - Unemployed: 0.07 0.25.
  - Married: 0.61 0.48.
  - Years of education: 11.95 3.35.
  - Observations: 120,400.

### Policy-relevant conclusions
- Tightening monetary surprises reduce self-reported life satisfaction; easing surprises increase likelihood of high responses ('8', '9' or '10') on the 11-point life satisfaction scale.
- Findings are robust across OLS and ordered logit estimators, window lengths, exclusion of major events/outliers, alternative estimators, separation of monetary vs information shocks, and a DID framework.
- Heterogeneity: effects vary by shock sign, age, income, and gender — stronger negative impact for middle-aged, middle-income, and women.
- Implication: central bank communication shapes welfare and should consider broader audience impacts beyond financial market participants; media coverage is a likely transmission channel.

*Source: wpiea2023248-print-pdf - 2.1  The German Socio-Economic Panel (SOEP)*

### 2.1  The German Socio-Economic Panel (SOEP)

### 2.1  The German Socio-Economic Panel (SOEP)

### Data and sample
- SOEP: annual representative survey of persons, households and families in Germany; established in 1984.
- Coverage: residents of former West and East Germany, immigrants, and resident foreigners; individuals within households surveyed if at least 16 years of age.
- Sample stability: high degree of stability over time highlighted as an advantage.
- Annual interviews: approximately 41,000 households and approximately 69,000 individuals, nearly half of whom are women.
- Study period used: 2002-2018.
- Number of observations referenced: 362,488.

### Subjective well-being (SWB) measure
- SWB operationalized as life satisfaction.
- Survey question: "How satisfied are you with your life, all things considered?"
- Response scale: 11-point scale ranging from 0 (completely dissatisfied) to 10 (completely satisfied).
- Empirical pattern: average life satisfaction score relatively stable at slightly above 7, with a decline during the 2008 financial crisis.
- Additional controls exploited due to dataset granularity: age, gender, marital status, employment status, educational achievement (years of schooling), total income, number of people living in the household, and health.

### Monetary policy shocks: construction and identification
- Monetary policy surprises sourced from the Euro Area Monetary Policy Event-Study Database (EA-MPD) of Altavilla et al. (2019).
- Asset-price changes used: changes in the overnight index with maturities of 1 month, 3 months, 6 months or 1 year.
- Aggregation: first principal component of the four OIS-maturity surprises computed to obtain a single measure.
- Data frequency: daily data from January 1, 2002 to December 31, 2018; all scheduled ECB meetings during this 17-year period included.
- Event-window definition in EA-MPD used: the monetary-event window (price changes computed as difference between median quote during 13:25-13:35 before the press release and median quote for 15:40-15:50 after the press conference).
- Note on informational content: central bank announcement surprises can include Delphic and Odyssean components; positive co-movement between interest rates and stock market following surprises signals “central bank information shocks”.

### Empirical approach: prima-facie and two-step aggregation
- Prima-facie: constructed a "clean" SWB measure by regressing life satisfaction on individual fixed effects and time fixed effects (month, year, month*year); grouped 362,488 observations into 20 equally sized bins based on EONIA to trace relationship.
- Two-step aggregation approach:
  - First-stage microeconometric OLS: W_{i,mdt} = α + γ X_{i,mdt} + β_d + γ_m + θ_t + ε_{i,mdt}.
  - Compute quarterly average residual ̄ε_{qt}.
  - Second-stage: ̄ε_{qt} = α + β MP_{qt} + ξ_{qt}, where MP_{qt} is quarterly average of EA-MPD monetary policy shocks.
- Table 1 findings (various specifications of equation 2):
  - Baseline indicates a negative association between life satisfaction and monetary policy surprises.
  - Dynamic construction (column 2): a monetary surprise of one standard deviation (1.42 basis points) reduces life satisfaction by approximately 2 points.
  - Inclusion of macro controls and time dummies (columns 3 and 4) maintains negative effect.
  - Including three lags of MP surprises (column 5) preserves coefficient stability.
  - Robust regression (column 6) confirms negative effect but with smaller magnitude.
- Note on sample start: identification requires sample to start in 2002 because intraday OIS data 1999–end-2001 are very noisy.

### Empirical approach: event-study at individual level and identification strategy
- Merge each panel observation (individual * interview date) with the corresponding monetary surprise from EA-MPD.
- Event-study design: compare individuals responding within a w-day window after the announcement to those responding within the w-day window prior to the announcement; responses on the day of the announcement and outside windows are dropped.
- Treatment intensity: use direction and intensity of each surprise (not just occurrence); variable of interest D_{i,mdt} MP_{mdt} where D is 1 if response falls within w-day post-announcement window and 0 if within w-day pre-announcement window.
- Baseline window choice: w = 7 days (trade-off between sample size/timing of diffusion and contamination by other news).
- Example: LTRO announcement May 7, 2009 produced a 3-month OIS decline of -1.50 basis points; for that event D_{i,mdt} MP_{mdt} = -1.50 for responses May 8–May 15, 2009 and zero for responses Apr 30–May 6, 2009; observations on May 7 and May 16 dropped.
- Estimation model (general): W_{i,mdt} = α + β D_{i,mdt} MP_{mdt} + γ X_{i,mdt} + θ_t + γ_m + ζ_d + ε_{i,mdt}.
- Fixed effects and controls: year (θ_t) and month (γ_m) fixed effects; some specifications include year*month fixed effects; individual time-varying controls X_{i,mdt} included.
- Panel structure: because event-window construction leads to sparse repeated observations, average individual appears only 2.5 times in empirical estimates; baseline exploits between variation rather than within variation.
- Robustness checks: quasi-fixed effects (Mundlak, 1978) and individual fixed effects estimated on full sample without event windows; ordered logistic regressions also estimated given ordinal dependent variable.
- Assumptions and checks:
  - Interview attrition assumed random; tests in Appendix ensure participating individuals’ characteristics not significantly different before and after events.
  - Monetary surprises constructed to be exogenous to other macroeconomic news; specifications with year*month dummies included to capture possible correlated macro shocks.
  - To account for potential nonrandom attrition around major events, variant excluding the most important monetary events estimated.

### Results: baseline event-study estimates and interpretation
- Baseline pooled OLS and ordered logistic estimates of equation (3) reported; main finding: tightening monetary surprises decrease life satisfaction; easing shocks have mirror positive effects.
- Robustness across specifications: changing fixed effects or controls has marginal impact; inclusion of year*month dummies decreases the coefficient slightly (e.g., from -1.05 to -0.98) but order of magnitude remains.
- Alternative MP definition (binary easing = "-1", tightening = "1") yields negative coefficient of smaller magnitude.
- Robustness to estimation method: ordered logistic and generalized ordered logit produce similar substantive conclusions.
- Magnitude and economic interpretation:
  - Table summary: in earlier aggregation exercise, a monetary surprise of one standard deviation (1.42 basis points) reduced life satisfaction by approximately 2 points (Table 1, dynamic specification).
  - In baseline event-study interpretation: a one standard deviation increase (1.92) in MP surprises reduces the life satisfaction indicator by 0.024 points.
  - Monetized equivalent: a one standard deviation MP surprise is equivalent to a drop in household income of approximately 4%.
  - Unemployment-equivalent: the drop caused by a one standard deviation MP surprise is equivalent to approximately 4% of the drop in life satisfaction caused by being unemployed.
- Ordered logistic marginal effects:
  - A one-point increase in the surprise effect of monetary policy reduces the probability of answering '8', '9' or '10' on the 11-point life satisfaction scale by 0.17%, 0.11% and 0.04%, respectively.
  - Example marginal probability: likelihood of answering "9" is 33.86 when monetary surprise = 0 and predictors at mean; it becomes 33.42 when monetary surprise increases by one standard deviation.
- Standard errors and clustering:
  - Standard errors clustered at individual level in all specifications.
  - Sensitivity tested by clustering at month*year, event window, date, and household levels; standard errors remain close to baseline; coefficient in main specification remains statistically significant (1% level or better in alternatives).

*Source: wpiea2023248-print-pdf - 2.1  The German Socio-Economic Panel (SOEP)*

### 4.2  Robustness checks

### 4.2  Robustness checks

### 4.2.1  Sensitivity
- Baseline window: 7 days before and after the focal announcement.
- Alternative windows tested: 5 days, 3 days and 1 day.
  - Columns (1), (2), and (3) in Table 3 report estimates for 5-day, 3-day and 1-day windows using the full data sample in which monetary policy shocks last until a new shock occurs.
  - Results: findings largely unafffected when using shorter windows, except for the one-day window.
    - Suggested explanations for the 1-day window result:
      - Model specification does not consider proper daily effects and people are systematically happier on Fridays than on Wednesdays.
      - People do not react directly to an announcement but need time to absorb the new information; "most people do not follow central bank news closely" (Blinder and Svensson, 2015).
  - Column (4) in Table 3: placebo test that lags policy surprises by one week — no impact on life satisfaction.
  - Columns (5) and (6): both consider the 7-day impact after each announcement; (5) includes all observations, (6) is limited to observations within the 7-day impact period — coefficients consistent with baseline.
  - Last column of Table 3: transforms the monetary policy surprise into a continuous variable (effect lasts until a new announcement) — effect on life satisfaction is smaller than half the size of the baseline estimate.

- Event-study design (equation 3): decompose the 7-day period around each announcement into event lags (pre-event) and leads (post-event).
  - Event lag and lead coefficients and confidence intervals plotted in Figure 2.
  - Coefficient estimates for the 7-day post-announcement period (lead effects) are not significant, confirming absence of parallel trends in the pretreatment period and reinforcing the validity of monetary surprises.

- Sensitivity to major events and outliers (Table 4):
  - Column (1): monetary policy shocks winsorized at the5% level.
  - Columns (2) and (3): exclude policy events with fewer than 100 or 1000 observations, respectively.
  - Column (4): exclude observations with standardized residuals greater than 1.96.
  - Column (5): exclude observations with Cook’s distance greater than a common rule-of-thumb threshold.
  - Column (6): estimates using the robust estimators of Hubert (1964).
  - Results: very similar in all cases, indicating conclusions are not driven by outliers or major events.

- Alternative estimators modeling individual unobserved heterogeneity (Table 5):
  - Column (1): Mundlak’s quasi-fixed effects estimator (explores within and between variation).
  - Column (2): fixed effects.
  - Column (3): random effects (full sample).
    - Coefficients of interest remain significant but are approximately twice as small as baseline.
  - Column (4): pooled OLS with lagged term of satisfaction — coefficient estimate remains intact.
  - Columns (5) and (6): account for ordinal dependent variable and individual heterogeneity:
    - Panel random-effects ordered logit including means of time-variant controls (quasi-fixed effects spirit).
    - "blow and cluster" (BUC) fixed-effects ordered logit model à la Baetschmann et al. (2015).
    - Results broadly in line with baseline findings.

- Alternative intraday asset price measures (Table 6):
  - Assets tested: OIS 1M, OIS 3M, OIS 6M, OIS 12M and Euro Stoxx50.
  - Coefficients using OIS at different maturities are similar to baseline.
  - Euro Stoxx50-based coefficient is not statistically significant.

### 4.2.2  Pure monetary policy surprises
- Euro Stoxx50 surprises have no effect on life satisfaction, although a positive impact might be expected.
  - Reasoning: stock markets do not systematically react to ECB communication in the opposite direction of short-term rates; an interest rate cut can signal either bad macroeconomic news or good news via more accommodating policy.
  - ECB announcements convey information about both policy stance and ECB assessments of the economic outlook.
- Delphic vs Odyssean shocks (Andrade and Ferroni, 2021):
  - Delphic shocks: convey news about the state of the economy and lead to positive comovement between short-term interest rates and stock markets.
  - Odyssean shocks: pure monetary policy shocks associated with negative comovement.
- Robustness check following Jarociński and Karadi (2020): use "pure" monetary policy shocks purged from central bank information shocks (identified by negative comovement between interest rates and stock prices, whereas contemporaneous information shocks show positive comovement).
  - Table 7 reproduces baseline specifications from Table 2 using pure shocks.
  - Results: pure monetary policy shocks significantly decrease individual life satisfaction and show strong consistency with baseline results.

### 4.2.3  Google Trends: A new way to measure shock impact
- Motivation:
  - Baseline assumes intensity of high-frequency OIS reactions relates to how much monetary policy decisions resonate with the general public.
  - Channel: media transmits central bank information; assumption that the (absolute) strength of a surprise affects dissemination.
  - High-frequency surprises can impact short-term welfare (Jarociński and Karadi (2020) find contemporaneous decline in real GDP in a structural monthly VAR).
- Use of Google Trends to test whether larger high-frequency surprises induce higher public attention to the ECB:
  - Data: daily Google Trends search volume index for the term "EZB" ("Europäische Zentralbank") from Germany.
  - Comparison (Table 8): average index on the day of an ECB meeting vs all other days from January 1st, 2002 to December 31, 2018.
    - Also compare using a "clean" index purged of day-of-week effects.
    - Observation: significant positive difference between means of the two samples — ECB meetings increase search volume.
  - Figure 4: scatter-plot of absolute monetary policy surprises vs "EZB" search query volume shows a positive relationship.
    - Interpretation: intensity of intraday financial market surprises is related to broader public attention; Google Trends is a plausible proxy for shock intensity.
- Advantages and drawbacks of Google Trends index:
  - Advantage: each day of the event week has a different value, allowing shock intensity to fade or be nonlinear over the week.
  - Drawback: Google Trends data are potentially endogenous.
- Construction of a new shock variable:
  - Rely on Google index and OIS surprises.
  - OIS surprises used to distinguish tightening vs easing shocks and to set sign of Google index accordingly.
  - In practice: change sign of the Google index to be negative in the case of easing shocks.
  - The model is then estimated using this adjusted Google-based shock variable alongside MP surprise variable.

*Source: wpiea2023248-print-pdf — 4.2 Robustness checks.*

### section is the same as our baseline model. Table 9 reports the results with our new MP surprise

### wpiea2023248-print-pdf - section is the same as our baseline model. Table 9 reports the results with our new MP surprise

### Results: diffusion of monetary policy surprise and baseline corroboration
- All specifications show a negative relationship between monetary surprises and life satisfaction, corroborating the results of the baseline approach.
- Finding: "the more a monetary event arouses public interest, the more strongly SWB will respond."
- Table 9 estimates (MP surprise diffusion):
  - MP surprise diffusion coefficients: -0.002***, -0.002**, -0.002*, 0.000, -0.002*
  - Standard errors reported: (0.001) for each coefficient
  - Adj. R-squared: 0.262, 0.264, 0.264, 0.252, 0.263
  - Observations: 9039590395903959039590395

### Heterogeneity of effects (Table 10)
- Easing vs tightening:
  - Easing surprise coefficient: 0.934** (standard error: 0.398)
  - Tightening surprise coefficient: -1.031*** (standard error: 0.343)
  - Economic effect of a one standard deviation surprise is stronger for tightening than easing.
- Time heterogeneity (before vs after 2009):
  - After 2009: -0.895*** (standard error: 0.226)
  - Before 2009: -1.345*** (standard error: 0.309)
  - Interpretation: decrease in coefficient estimates over time consistent with downward trend in magnitude of surprises; overall impact relatively constant since 2003.
- Age heterogeneity (columns reported across quantiles):
  - Age-category coefficients include: 0.290, -1.138***, -1.420***, -1.006***, 0.020 (standard errors: (0.997), (0.361), (0.319), (0.356), (0.500))
  - Impact significant for middle-aged individuals.
- Income heterogeneity:
  - Coefficients reported: -0.051, -1.391***, -1.502***, -0.822*** (standard errors: (0.639), (0.409), (0.331), (0.274))
  - Tightening shocks have no significant effects on low-income individuals; negative effect strongest for the middle class.
- Gender:
  - Women coefficient: -1.205*** (standard error: 0.262)
  - Decrease in life satisfaction stronger for women.
- Notes: Adj. R-squared and observations vary across columns (examples include 0.262, 0.260, 0.159, 0.216, 0.293; and observations such as 142839, 61373, 44335, 65063, 45382, etc.).

### Difference-in-difference design (Table 11)
- Identification strategy: exploit variation in financial literacy by age (treated: ages 25-70; control: under 25 and over 70).
- DID linear OLS estimates (DID - MP surprise):
  - Coefficients: -1.166**, -1.179**, -1.275***, -1.480***, -1.188*** (standard errors: (0.458), (0.460), (0.460), (0.459), (0.460))
  - Adj. R-squared: 0.262, 0.263, 0.263, 0.252, 0.261
  - Observations: 142839 (repeated)
- DID ordered LOGIT estimates:
  - Coefficients: -1.207**, -1.268**, -1.397***, -1.555***, -1.207** (standard errors: (0.538), (0.541), (0.541), (0.538), (0.538))
  - Pseudo R-squared: 0.080, 0.081, 0.081, 0.075, 0.080
  - Observations: 142839 (repeated)
- Robustness: DID results continue to show strong consistency with baseline findings.

### Channels linking monetary surprises to life satisfaction (Table 12)
- Two potential mechanisms distinguished:
  - Direct effect: individuals anticipate easing shocks will improve their own economic conditions.
  - Indirect effect: individuals expect a positive effect on the economy in general.
- Empirical proxies: response to "How concerned are you?" and expectations for life satisfaction in one year and five years.
- Regression results (MP surprise on channels):
  - Econ conditions: coefficient -0.648*** (standard error: 0.237)
  - Fin conditions: coefficient -0.022 (standard error: 0.231)
  - Life satisfaction 5Y: coefficient -0.614** (standard error: 0.253)
  - Life satisfaction 1Y: coefficient -2.088*** (standard error: 0.554)
  - Adj. R-squared reported for econ/fin conditions: 0.257, 0.250
  - Observations: 137562, 143860, 645652, 3615 (as reported per column)
- Interpretation:
  - Monetary surprises significantly deteriorate individuals’ views about the economy in general.
  - Surprises do not affect individuals’ degree of concern about their own economic conditions.
  - Tightening surprises reduce future perceptions of life satisfaction, stronger for one-year expectations, highlighting role of optimism and expectations in transmission.

### Robustness and complementary findings (summarized from main tables)
- Baseline linear OLS and ordered LOGIT consistently find negative MP surprise effects:
  - Example baseline MP surprise coefficients: -1.054***, -0.988***, -0.722***, -0.993***, -1.004*** (standard errors: (0.182), (0.260), (0.249), (0.174), (0.192))
  - Ordered LOGIT baseline: -1.275***, -1.123***, -0.783***, -1.255***, -1.275*** (standard errors: (0.210), (0.302), (0.290), (0.200), (0.210))
  - MP surprise (-1/1) in some specifications: -0.024*** (standard error: 0.006); ordered LOGIT analog: -0.026*** (standard error: 0.007)
  - Observations in baseline: 142839
- Window sensitivity (Table 3):
  - 5 days window: -1.061*** (0.225)
  - 3 days window: -0.736** (0.324)
  - 1 day window: 0.068 (0.460)
  - Placebo test (lag MP surprise by one week): -0.000 (0.000)
  - Continuous surprise: -0.443*** (0.090)
- Robustness to major events/outliers, alternative estimators, different shock definitions, and "pure monetary surprise" specification all continue to show negative impacts (see respective table coefficients preserved in source).
- Google Trends and diffusion:
  - Google Trends "EZB" index mean for monetary event sample: 15.939; No: 6.926; "EZB2" mean: 6.382, -0.205
  - Index standard-deviation: 6.772, 4.487, 6.751, 4.252
  - Nb. Obs: 1484,6011484,601 (as reported)
- Pure monetary surprise (Table 7) coefficients: -0.885***, -0.826**, -0.681**, -1.092***, -0.845*** (standard errors: (0.220), (0.322), (0.294), (0.238), (0.230)); MP surprise (-1/1) -0.028*** (0.009)

### Policy-relevant conclusions
- Tightening monetary surprises reduce self-reported life satisfaction; easing surprises increase likelihood of high responses ('8', '9' or '10') on the 11-point life satisfaction scale.
- Findings are robust across OLS and ordered logit estimators, window lengths, exclusion of major events/outliers, alternative estimators, separation of monetary vs information shocks, and a DID framework based on age categorization.
- Heterogeneity suggests policy communication effects vary by shock sign, age, income, and gender (stronger negative impact for middle-aged, middle-income, and women).
- Implication: central bank communication shapes welfare and should consider broader audience impacts beyond financial market participants; media coverage is a likely transmission channel.

*Italic: Source — wpiea2023248-print-pdf (section is the same as our baseline model. Table 9 reports the results with our new MP surprise).*

### References

### wpiea2023248-print-pdf - References

### References (selected bibliography)
- Akay, A., Bargain, O., and Elsayed, A. (2020). Global terror, well-being and political attitudes. European Economic Review, 123:103394.
- Altavilla, C., Brugnolini, L., Gürkaynak, R. S., Motto, R., and Ragusa, G. (2019). Measuring euro area monetary policy. Journal of Monetary Economics, 108:162–179.
- Andrade, P. and Ferroni, F. (2021). Delphic and odyssean monetary policy shocks: Evidence from the euro area. Journal of Monetary Economics, 117:816–832.
- Atkinson, A. and Messy, F.-A. (2012). Measuring financial literacy: Results of the oecd/international network on financial education (infe) pilot study.
- Baetschmann, G., Staub, K. E., and Winkelmann, R. (2015). Consistent estimation of the fixed effects ordered logit model. Journal of the Royal Statistical Society: Series A (Statistics in Society), 178(3):685–703.
- Bholat, D., Broughton, N., Ter Meer, J., and Walczak, E. (2019). Enhancing central bank communications using simple and relatable information. Journal of Monetary Economics, 108:1–15.
- Binder, C. (2017). Fed speak on main street: Central bank communication and household expectations. Journal of Macroeconomics, 52:238–251.
- Blanchflower, D. G., Bell, D. N., Montagnoli, A., and Moro, M. (2014). The happiness trade-off between unemployment and inflation. Journal of Money, Credit and Banking, 46(S2):117–141.
- Blinder, A. and Svensson, L. E. O. (2015). Comments and discussion. Brookings papers on economic activity, pages 209–225.
- Blinder, A. S., Ehrmann, M., De Haan, J., and Jansen, D.-J. (2022). Central bank communication with the general public: Promise or false hope? De Nederlandsche Bank Working Paper.
- Blinder, A. S., Ehrmann, M., Fratzscher, M., De Haan, J., and Jansen, D.-J. (2008). Central bank communication and monetary policy: A survey of theory and evidence. Journal of economic literature, 46(4):910–45.
- Carvalho, C. and Nechio, F. (2014). Do people understand monetary policy? Journal of Monetary Economics, 66:108–123.
- Cavallo, A., Cruces, G., and Perez-Truglia, R. (2017). Inflation expectations, learning, and supermarket prices: Evidence from survey experiments. American Economic Journal: Macroeconomics, 9(3):1–35.
- Cieslak, A. and Schrimpf, A. (2019). Non-monetary news in central bank communication. Journal of International Economics, 118(C):293–315.
- Clark, A. E. (2018). Four decades of the economics of happiness: Where next? Review of Income and Wealth, 64(2):245–269.
- Coibion, O., Gorodnichenko, Y., Knotek II, E. S., and Schoenle, R. (2020a). Average inflation targeting and household expectations. Technical report, National Bureau of Economic Research.
- Coibion, O., Gorodnichenko, Y., Kumar, S., and Pedemonte, M. (2020b). Inflation expectations as a policy tool? Journal of International Economics, 124:103297.
- Coibion, O., Gorodnichenko, Y., and Weber, M. (2019). Monetary policy communications and their effects on household inflation expectations. Technical report, National Bureau of Economic Research.
- Colciago, A., de Haan, J., and Samarina, A. (2019). Central bank policies and income and wealth inequality: A survey. Journal of Economic Surveys, 33(4):1199–1231.
- de Haan, J. and Sturm, J.-E. (2019). Central bank communication. In The Oxford handbook of the economics of central banking, page 231. Oxford University Press.
- Di Tella, R. and MacCulloch, R. (2008). Gross national happiness as an answer to the easterlin paradox? Journal of Development Economics, 86(1):22–42.
- Di Tella, R., MacCulloch, R. J., and Oswald, A. J. (2001). Preferences over inflation and unemployment: Evidence from surveys of happiness. American economic review, 91(1):335–341.
- Diener, E. and Biswas-Diener, R. (2002). Will money increase subjective well-being? Social indicators research, 57(2):119–169.
- Dincer, N. and Eichengreen, B. (2007). Central bank transparency: Where, why, and with what effects? NBER Working Papers 13003.
- Dincer, N. and Eichengreen, B. (2009). Central bank transparency: Causes, consequences and updates. NBER Working Papers 14791.
- Dräger, L., Lamla, M. J., and Pfajfar, D. (2016). Are survey expectations theory-consistent? the role of central bank communication and news. European Economic Review, 85:84–111.
- Ehrmann, M., Georgarakos, D., and Kenny, G. (2022). Credibility gains from communicating with the public: Evidence from the ecb’s new monetary policy strategy. In The European Central Bank Blog.
- Eklou, K. M. and Fall, M. (2020). The (subjective) well-being cost of fiscal policy shocks. International Monetary Fund.
- Frank, R. H. (2003). Happiness and economics: How the economy and institutions affect human well-being. Const. Pol. Econ., 14:253.
- Frey, B. S. (2017). Research on well-being: Determinants, effects, and its relevance for management. Technical report, CREMA Working Paper.
- Gardner, J. and Oswald, A. J. (2007). Money and mental wellbeing: A longitudinal study of medium-sized lottery wins. Journal of health economics, 26(1):49–60.
- Gürkaynak, R., Sack, B., and Swanson, E. (2005). Do actions speak louder than words? the response of asset prices to monetary policy actions and statements. International Journal of Central Banking, 1(1):55–93.
- Gürkaynak, R. S. (2005). Using federal funds futures contracts for monetary policy analysis.
- Gürkaynak, R. S., Sack, B. P., and Swanson, E. T. (2007). Market-based measures of monetary policy expectations. Journal of Business & Economic Statistics, 25(2):201–212.
- Haldane, A., Macaulay, A., and McMahon, M. (2020). The 3 e’s of central bank communication with the public. Available at SSRN 3518605.
- Haldane, A. and McMahon, M. (2018). Central bank communications and the general public. In AEA papers and proceedings, volume 108, pages 578–83.
- Hofstetter, M. and Rosas, J. N. (2021). The poor and the rich: Preferences over inflation and unemployment. Journal of Money, Credit and Banking, 53(4):875–895.
- Issing, O. (2019). The long journey of central bank communication. Mit Press.
- Jarociński, M. and Karadi, P. (2020). Deconstructing monetary policy surprises—the role of information shocks. American Economic Journal: Macroeconomics, 12(2):1–43.
- Kahneman, D. and Krueger, A. B. (2006). Developments in the measurement of subjective well-being. Journal of Economic Perspectives, 20(1):3–24.
- Kryvtsov, O. and Petersen, L. (2021). Central bank communication that works: Lessons from lab experiments. Journal of Monetary Economics, 117:760–780.
- Kumar, S., Afrouzi, H., Coibion, O., and Gorodnichenko, Y. (2015). Inflation targeting does not anchor inflation expectations: Evidence from firms in new zealand. Technical report, National Bureau of Economic Research.
- Kuttner, K. N. (2001). Monetary policy surprises and interest rates: Evidence from the fed funds futures market. Journal of monetary economics, 47(3):523–544.
- Layard, R. et al. (2015). A new priority for mental health. Paper EA035, Centre for Economic Performance, London School of Economics and Political Science, May.
- Lewis, D. J., Makridis, C., and Mertens, K. (2019). Do monetary policy announcements shift household expectations? FRB of New York Staff Report, (897).
- Lusardi, A., Mitchell, O. S., and Curto, V. (2010). Financial literacy among the young. Journal of consumer affairs, 44(2):358–380.
- Lusardi, A., Mitchell, O. S., and Curto, V. (2014). Financial literacy and financial sophistication in the older population. Journal of pension economics & finance, 13(4):347–366.
- Mundlak, Y. (1978). On the pooling of time series and cross section data. Econometrica: journal of the Econometric Society, pages 69–85.
- Ram, R. (2009). Government spending and happiness of the population: additional evidence from large cross-country samples. Public Choice, 138(3):483–490.
- Romer, C. D. and Romer, D. H. (1998). Monetary policy and the well-being of the poor.
- Ter Ellen, S., Larsen, V. H., and Thorsrud, L. A. (2021). Narrative monetary policy surprises and the media. Journal of Money, Credit and Banking.

### Appendix — Table A1: SWB and monetary policy surprises (marginal effects)
- The table reports marginal effects of the ordered logit model.
- Coefficient estimates with cluster-robust standard errors computed at the individual level are reported; standard errors are shown in parentheses below coefficient estimates.
- Statistical significance indicators: *, ** and *** indicate significance at the 10%, 5% and 1% levels, respectively.
- Reported marginal effects and standard errors (as presented in the source):
  - 0: 00.003*** 0.023** 0.002*** (0.000)(0.009)(0.000)
  - 1: 0.003*** 0.025*** 0.002*** (0.001)(0.009)(0.000)
  - 2: 0.009*** 0.068*** 0.007*** (0.001)(0.023)(0.000)
  - 3: 0.020*** 0.127*** 0.016*** (0.003)(0.033)(0.000)
  - 4: 0.028*** 0.138*** 0.024*** (0.005)(0.021)(0.000)
  - 5: 0.101*** 0.286*** 0.099*** (0.017)(0.006)(0.001)
  - 6: 0.085*** 0.136*** 0.114*** (0.014)(0.026)(0.001)
  - 7: 0.070*** 0.123*** 0.260*** (0.012)(0.036)(0.001)
  - 8-: -0.169*** 0.061*** 0.340*** (0.028)(0.022)(0.001)
  - 9-: -0.108*** 0.011** 0.102*** (0.018)(0.004)(0.001)
  - 10: -0.042*** 0.003** 0.034*** (0.007)(0.001)(0.000)

### Appendix — Figure A1 and Table A2: Life satisfaction and sample summary
- Figure A1: Life satisfaction - Sample average
  - Source: SOEP data, own calculations.
- Table A2: Summary statistics - Mean values (variables, mean, std. dev.)
  - Life satisfaction: 7.09 1.78
  - Monthly income: 28772 044.75
  - Age: 49.51 6.48
  - Household size: 3.18 1.64
  - Gender: 0.72 0.44
  - Unemployed: 0.07 0.25
  - Married: 0.61 0.48
  - Years of education: 11.95 3.35
  - Observations: 120,400
  - Notes: SOEP data, own calculations

### Appendix — Figure A2: Policy news shocks, 2002–2020 (EA-MPD)
- Figure panels described:
  - (a) 1 Week OIS rate change
  - (b) 1 Month OIS rate change
  - (c) 3 Months OIS rate change
  - (d) 6 Months OIS rate change
  - (e) 1 Year OIS rate change
- Note: The Euro Area Monetary Policy Event-Study Database (EA-MPD)

*Italic: Source: wpiea2023248-print-pdf - References*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023248-print-pdf.pdf_