Our empirical specification is grounded in the “happiness equations” literature, which models life satisfaction as a function of individual circumstances and the macro-institutional environment (Frey and Stutzer, 2002; Van Le and Tran, 2024). Accordingly, we view life satisfaction as generated by a stable country component (culture, institutions), time effects, and observed covariates capturing health, income, employment, demographics, religiosity, social trust, and expectations, along with time-varying country conditions. Into this structure, we insert an exogenous high-frequency shock (i.e., the Reported Social Unrest Index), which captures the incidence and intensity of conflict or unrest domestically (and spillovers from neighbors) and shifts perceived security and uncertainty.

The econometric model is structured as follows (Blanchflower and Oswald, 2011; Howell and Howell, 2008; Riitsalu and Murakas, 2019; Van Le and Tran, 2024):

Wellbeing_i,c,y,m = α_c + γ RSUI_c,y,m–1 + Z' X_i,c,y,m + N' C_c,y + λ_y + τ_m + ε_i,c,y,m, (1)

where Wellbeing_i,c,y,m represents the life satisfaction of resident i in country c during­ month m and year y; RSUI indicates the incidence and intensity of social unrest ­domestically and in neighboring regions. We introduce a 1-month lagged effect of RSUI on individual well-being, a novel contribution compared to studies that typically assume yearly or quarterly average effects. Two sets of control variables, X and C, include personal, community-related, and national characteristics, such as health status, income, personal faith, demographics, and job status (Blanchflower and Oswald, 2011; Dolan et al., 2008; Howell and Howell, 2008; Voukelatou et al., 2021), general societal trust (Waldinger and Schulz, 2023), government trust (Helliwell et al., 2020), neighbor relationships (Waldinger and Schulz, 2023), economic expectations, and democratic factors (Bartels, 2015; Dolan et al., 2008; Howell and Howell, 2008). α_c denotes time-invariant country-specific factors; λ_y accounts for year fixed effects; τ_m — for month fixed effects, and ε represents error terms.

Regarding government policy, we explore interaction terms between social unrest and various types of government actions, using data from the iMaPP data­base, which provides dummy indices for government tightening and loosening actions across 17 instruments monthly. The main proxy for government actions is a discrete variable recording tightening (+1), loosening (–1), or no change (0), cumulated over months and divided into (i) aggregate tightening policy and (ii) aggregate loosening policy across the 17 instruments (Alam et al., 2019). In our setting, the “harm” of state intervention is operationalized as a decline in subjective well-being conditional on RSUI, with moderation by policy stance (iMaPP). Thus, we do not claim that intervention is inherently harmful; rather, indecision or miscalibration is what raises welfare costs during unrest.

Equation (1) then specifies the model:

Wellbeing_i,c,y,m = α_c + γ₁RSUI_c,y,m–1 × activeG_c,y,m +

+ Z' X_i,c,y,m + N' C_c,y + λ_y + τ_m + ε_i,c,y,m ; (2)

Wellbeing_i,c,y,m = α_c + γ₂RSUI_c,y,m–1 × activeGc,y.m¯ +

+ Z' X_i,c,y,m + N' C_c,y + λ_y + τ_m + ε_i,c,y,m, (3)

where activeG ∈ {–1, 0, +1} denotes policy actions encompassing both tightening and loosening measures, and activeGc,y.m¯​ is the country-level average stance. The coefficient of the interaction term γ₁ in Equation (2) indicates the current effects of active or inactive government policies on subjective well-being, reflecting monthly responses. The coefficient γ₂ in Equation (3) assesses their general effect on subjective well-being using the average value of the moderating variable and examines whether, during periods of high social unrest, an inactive government — in contrast with an active one — maintains consistent policies that generally enable citizens to feel more (+) or less (–) happy. This study employs high-dimensional fixed-effect estimators, which provide advantages over traditional fixed-effect models by accommodating a larger number of individual- or entity-specific fixed effects. Thus, they effectively mitigate bias resulting from unobserved heterogeneity and enhance the precision of estimates in panel data analysis (Belloni et al., 2014).

In brief, the coefficient γ in Equation (1) captures the effect of the (log) RSUI on life satisfaction; the a priori expectation is a negative and statistically significant association. The coefficient γ₁ in Equation (2) identifies the short-run moderating effect of government interventions when policy actions are observed at the monthly frequency. The coefficient γ₂​ in Equation (3) captures longer-run moderation by replacing monthly actions with the country-specific average policy stance over the study period.

Equations (1)–(3) assume a linear, additive relationship between RSUI (and its interactions) and well-being conditional on controls and fixed effects. Identification relies on the exogeneity of the incidence and intensity of social unrest domestically and in neighboring regions, mathematically:

E[ε_i,c,y,m | α_c, λ_y, τ_m, X_i,c,y,m, C_c,y, RSUI_c,y,m–1, ∙] = 0.

Although the incidence and intensity of social unrest—domestically and in neighboring regions—are plausibly exogenous to individual respondents, we mitigate residual simultaneity by entering RSUI with a 1-month lag, absorbing country, year, and month fixed effects, and conditioning on rich covariates at the personal, community, and national levels. Because key regressors vary at the country—month level while the outcome is individual-level, residuals may cluster within countries and within months; accordingly, we report two-way clustered standard errors (country and year—month). To address multicollinearity and leverage, we provide variance inflation factors (VIFs) for time-varying covariates and influence diagnostics.

The study utilizes data from three primary sources. First, to assess well-being, we collect responses from the World Values Survey (WVS) to the questions: “All things considered, how satisfied are you with your life as a whole these days?” and “Taking all things together, would you say you are very happy/quite happy/not very happy/not at all happy?” The WVS is a pivotal dataset aimed at examining global cultural and social trends through standardized questionnaires administered across diverse populations. Since 1981, it has collected data from over 100 countries in multiple cycles designed to investigate changes in values and beliefs over time. The WVS employs rigorous methodological approaches, such as face-to-face interviews or telephone surveys conducted with representative samples from each participating country. Studies frequently utilize WVS data to examine the correlates of either happiness or subjective well-being and other factors, such as cultural values (Diener and Oishi, 2000; Inglehart et al., 2008) and democratic governance (Oishi and Diener, 2003). Furthermore, the WVS provides information on socio-demographic characteristics, personal life, and community aspects, including levels of trust in neighbors and government.

The second data source is the Reported Social Unrest Index (RSUI), which measures levels of social unrest based on media reports analyzed through big data techniques (Barrett et al., 2022). Covering 130 countries from January 1985 to May 2020, the RSUI offers a consistent, high-frequency, real-time monitoring indicator of social unrest. Spikes in the RSUI align closely with major social disturbances domestically and in neighboring countries, validated by external sources. The Armed Conflict Location & Event Data (ACLED) dataset2 is ­another proxy for assessing social unrest, such as violent events, and draws on near real-time resources for recording political violence and protest data globally. This dataset is also used for robustness checks.

The Integrated Macroprudential Policy (iMaPP) Database3 is the third data source, which accounts for tightening and loosening government policy (Alam et al., 2019). The iMaPP database focuses not only on the disparate types of policy actions but also on the direction in which they are implemented. Notably, dummy indicators characterize the use of macroprudential policy instruments, which are useful in estimating the effects of these instruments per policy action, as demonstrated in previous studies.

It should be noted that these dummy-type indices do not provide information concerning the magnitude of policy actions, as Alam et al. (2019, p. 6) noted: “It is possible, for instance, that some instruments are adjusted frequently but with small changes, while other instruments are adjusted less frequently but with large changes.” The current study assumes that an active government is more likely to use instruments in response to situations of unrest, whether the changes are small or large. In addition, from various sources we collect other national-level control variables, such as GDP growth from the World Development Indicators (WDI) and democracy level from the V-Dem dataset.⁴

Table 1 presents all variables, units, sources, and descriptive statistics. Appendix 1 lists the countries included in the final merged dataset, Appendix 2 outlines the main components of iMaPP.

3. Results

3.1. Main results

Table 2 displays the impact of social unrest on life satisfaction using a high-dimensional fixed-effects estimator. All coefficients (columns 1–4) are statistically significant and negative, indicating that, as expected, social disturbances reduce life satisfaction. Column 2 controls for personal variables (e.g., health status, personal faith, demographics, income, and job status), while columns 3 and 4 additionally account for community and national factors. As more variables are included, the coefficients exhibit larger magnitudes, suggesting that social unrest has a greater detrimental effect on well-being when external factors are taken into account. We reestimate the model, replacing the 1-month lag of RSUI with two-, three-, and four-month lags (Appendix 3). In specifications without controls, the two- and three-month lags are negative and statistically significant. However, once we include covariates together with country and time fixed effects, these coefficients lose statistical significance. The four-month lag is insignificant in all specifications. These patterns suggest a short-lived well-being response to unrest and align with our interpretation that the public’s emotions adjust rapidly, while government actions typically operate with non-trivial implementation lags. Accordingly, we retain the 1-month lag as the primary specification.

The coefficients of the control variables align with findings from prior research. Better health status is associated with a significant increase of approximately 2.8 points in life satisfaction from ‘very poor’ to ‘very good’ status. Personal faith and demographic factors such as gender and age also corroborate existing studies­ (Bartels, 2015; Dolan et al., 2008; Howell and Howell, 2008). Furthermore, higher income correlates positively with life satisfaction, with individuals in the wealthiest group reporting approximately 1.3 points greater satisfaction than those in the poorest group. Unemployment is linked to a decrease of about 0.35 points in satisfaction.

Regarding community-related features, trust in both community members and government positively influences satisfaction. Accordingly, individuals who trust their community and government report higher satisfaction levels (approximately 0.17 points per category). Neighbor characteristics, including ethnicity, and behaviors such as substance use, also affect life satisfaction (Waldinger and Schulz, 2023). Notably, when controlling for both trust in people and government in columns 3 and 4, the coefficients of RSUI increase compared to column 2, suggesting that social disturbances may decrease social trust among respondents (Delhey and Newton, 2003). In other words, reducing trust may be one mechanism of the nexus under consideration.

By employing a large sample covering over 200,000 individuals worldwide, the study supports Inglehart et al. (2008) finding that individuals in democratic countries are happier than those in autocratic ones. Moreover, the study confirms that economic expectations, proxied by GDP growth, predict life satisfaction better compared to using GDP per capita (Blanchflower and Oswald, 2011), as seen in previous studies (Voukelatou et al., 2021).

Table 3 presents the estimated results for Equations (2) and (3), examining the moderating effects of government actions on the relationship. Column 1 reveals a statistically significant and negative impact, suggesting that an active government is generally associated with lower satisfaction levels among its citizens during periods of unrest. Columns 2–4 explore the effects of the current government response to these events, specifically focusing on tightening actions (column 3) and loosening actions (column 4). The findings indicate either that governments are more likely to introduce new policies in more vulnerable situations or that policy changes in these areas exacerbate the negative effects of social disturbances on satisfaction (Inglehart et al., 2008; Oishi and Diener, 2003).

As previously indicated, the iMaPP database encompasses 17 government instruments. Accordingly, we delve into a detailed examination of how each action moderates the relationship. Based on our final dataset, it is noteworthy­ that only 4 of these 17 instruments showed changes during the study period. These were reserve requirements (RR), systemic liquidity and funding (liquidity­), loan loss provision (LLP), and capital requirements (capital). Definitions of these instruments are provided in Appendix 2. Figure. 2 displays the coefficients of all interaction terms, while the regression table format is presented in Appendix 4.

The findings confirm that government actions, particularly those directed at tightening, negatively moderate the relationship between RSUI and subjective well-being. This effect is observed predominantly in actions related to RR, LLP, and capital instruments. Conversely, loosening government actions using the RR instrument enhances public satisfaction. These results elucidate why governments often implement such measures as reducing reserve requirements during periods of unrest, thereby stimulating economic activity by enabling increased lending capacity among banks. This lowering of constraints on foreign currency deposits also facilitates enhanced support for businesses and consumers with diverse funding needs, potentially leading to increased consumer spending, investment, and overall economic stability during periods of unrest (Cerutti et al., 2017; Galati and Moessner, 2018).

Regarding the liquidity instrument, tightening actions such as raising liquidity coverage ratios, liquid asset ratios, net stable funding ratios, and core funding ratios ensure that banks maintain stable funding by holding more high-quality liquid­ assets and imposing restrictions on external debt. At the same time, loosening these ratios and easing external debt restrictions enable banks to increase lending (Arena et al., 2020; Cerutti et al., 2017; Galati and Moessner, 2018).

Notably, the most common government practice poses the threat of reducing subjective well-being given an increase in unstable situations. Accordingly, tightening capital requirements compels banks to allocate a larger portion of their capital as reserves, restricting their lending capacity and potentially reducing credit availability for businesses and consumers. Similarly, increasing LLP requires banks to set aside higher provisions for potential loan losses, further constraining their lending capabilities. These measures can lead to decreased investment, higher borrowing costs, and lower economic growth, thereby contributing to growing public discontent during economic downturns (Akinci and Olmstead-Rumsey, 2018; Alam et al., 2019; Basten, 2020; Galati and Moessner, 2018).

The study proposes several strategies to enhance the robustness of current findings. First, the effects of RSUI are analyzed based on thresholds where dummy variables are created, taking the value of 1 if RSUI values exceed the 90th, 75th, and 25^th percentiles, respectively. These results are presented in columns 1–4 of Table 4 and align consistently with previous findings. As the percentile thresholds increase, the coefficients of the social unrest index on life satisfaction are larger. This finding indicates that the higher the severity and scale of unrest are the more significantly it damages the subjective well-being of the population. Specifically, the top 10% of regions with the highest instability experience a decrease of more than 0.184 points in satisfaction compared to the remaining 90%.

Second, an alternative proxy for measuring subjective well-being is consider­ed, specifically the perception of happiness. Results in columns 5–8 mirror those in columns 1–4 but the dependent variable used is the individual perception of happiness. Once again, the findings are consistent. It is noteworthy that if the study seeks to assess immediate emotional states or short-term fluctuations in well-being, feeling of happiness may be more appropriate. Conversely, for evaluating overall life satisfaction and the influence of stable factors on well‑being — the main purpose of this study — satisfaction with your life would be preferable (Diener et al., 1985, 1999; Lucas et al., 1996).

Third, an alternative proxy for RSUI is employed with two requirement criteria­: (i) it must cover data on the monthly level, and (ii) it must reflect the ­severity of disturbances in relation to armed conflict. The first criterion ensures the study model remains consistent with previous research, while the second criterion aims to consider recent social unrest associated with wars (deaths in armed conflicts), such as those in Gaza; thus, focusing on this unrest category ensures policy relevance. Data on deaths in armed conflicts, sourced from the ACLED dataset, serve as a substitute proxy for RSUI, and the results are presented in Table 5. The regression coefficients are negative and statistically significant, as anticipated. However, due to data limitations, conducting regressions with control variables is impractical, limiting our study to presenting results with simple regressions.

Regional heterogeneity is also examined by estimating the effects separately for five World Bank regions (Africa, the Americas, Asia, Europe, and Oceania). The results, presented in Appendix 5, show negative and statistically significant coefficients for the Americas, Asia, and Europe, aligning with the expectation that higher unrest reduces subjective well-being by increasing uncertainty, disrupting economic activity, and weakening perceptions of safety. The largest adverse ­effect is observed in Europe, while Africa displays a positive coefficient, indicating a distinct regional pattern.