Does Living with Wealthy Peers in Mandatory Military Service Boost Future Earnings?

This study asks whether long-term economic outcomes are influenced by exposure to peers from wealthier families. Using data from over 50,000 Finnish military conscripts assigned to dorms through an alphabetical process, the researchers exploit quasi-random variation in peer socioeconomic background. They find that conscripts from high-income families significantly benefit from dorming with peers from similarly wealthy families: a one-standard-deviation increase in dormmates’ parental income leads to a 5.7% earnings increase. In contrast, conscripts from low-income families experience only small, statistically insignificant gains. The results suggest that social sorting and persistent labor market networks contribute to reinforcing economic inequality.

Full Citation and Link to Article

Einiö, E. (2024). The Long‑Term Impacts of Mixing the Rich and Poor: Evidence from Conscript Dorms. American Economic Journal: Applied Economics, Forthcoming. https://www.aeaweb.org/articles?from=f&id=10.1257/app.20240173 

Extended Summary

Central Research Question

The study investigates a fundamental question in economics and social policy: To what extent does exposure to peers from different socioeconomic backgrounds influence an individual’s long-term economic and educational outcomes? Specifically, the research explores whether conscripts in the Finnish military—assigned to dorms with others from diverse income backgrounds through an alphabet-based system—experience lasting effects on earnings and educational attainment. The question has strong implications for debates on social integration, inequality, and the role of networks in perpetuating or mitigating disparities in life outcomes.

Previous Literature

This work contributes to a broad literature on peer effects, economic mobility, and social networks. Prior studies have shown that peers influence short-term outcomes in school and military settings (e.g., Sacerdote, 2001; Carrell et al., 2009), but less is known about the persistence and mechanisms of these effects into adulthood. There is also an important connection to research on labor market networks and referrals (e.g., Granovetter, 1995; Topa, 2011), which suggest that social ties can lead to better employment opportunities and wages. Additionally, the study relates to work on the long-term effects of social mixing policies, such as housing mobility experiments (e.g., Chetty et al., 2016). Finally, the methodological approach closely mirrors Rao (2019), who exploited alphabetical assignment in Indian schools, but this study uniquely focuses on young adults in a military context with long-run outcomes.

Data

The analysis is based on an administrative dataset of over 50,000 male Finnish conscripts who served between 1996 and 2006. Military service is compulsory in Finland for men, and around 80% of male cohorts serve, ensuring a representative sample. Dorm assignments were made alphabetically within military squadrons, providing near-random variation in peer composition. Dormmates share intense daily interactions during training and downtime, creating meaningful peer exposure.

The researchers merge the conscription registry with rich administrative data from Statistics Finland, including income, employment, education, and parental background information. These data are linked through social security numbers, allowing for longitudinal tracking of individuals from conscription into middle adulthood (ages 28–42). Key variables include parental income (used to determine socioeconomic status), conscripts’ earnings, hourly wages, educational attainment, and employer identifiers.

Methods

The research design exploits the alphabetization of dorm assignments within squadrons to create quasi-random peer groups. The core methodological approach is an instrumental variables (IV) strategy, where the average parental income of the two alphabetically nearest conscripts in a squadron is used as an instrument for the average parental income of the two nearest dormmates. This setup allows for credible causal inference about peer effects while avoiding the common reflection and selection problems.

The model controls for conscripts’ own parental income, squadron fixed effects, and a wide set of predetermined characteristics (e.g., demographics, pre-service earnings, education). The authors further validate their design by showing that the instrument is unrelated to observable background characteristics and that there is no systematic alphabetical clustering by unobserved traits. They also conduct several robustness checks, including subsetting by language, age, and dorm size, and estimating models with alternative peer definitions (e.g., all dormmates rather than just the two closest).

To explore mechanisms, the authors use employer-employee linked data to test whether dormmates end up working at the same firms. They also examine educational pathways by linking conscripts to national education registers.

Findings/Size Effects

The study finds clear and heterogeneous long-term effects of peer exposure based on socioeconomic background:

1. Earnings Gains for High-Income Individuals:

   For conscripts from families above the median parental income, exposure to wealthier dormmates leads to significantly higher long-run earnings. A one-standard-deviation increase in the average parental income of the two nearest dormmates (around €30,000) is associated with a 5.7% increase in annual earnings, equivalent to roughly €2,010 per year. Over a working life, this translates to approximately €45,000 in present discounted value. This effect is statistically significant and robust across specifications.

2. Limited Effects for Low-Income Individuals:

   In contrast, conscripts from families below the median income show only small, statistically insignificant changes in long-term earnings from exposure to wealthier peers. The point estimate is near zero, and the confidence intervals rule out large effects.

3. Effects on Hourly Wages but Not Work Hours or Employment:

   The earnings effect for high-income conscripts operates almost entirely through hourly wages, not through increased work hours or employment probabilities. A one-standard-deviation increase in peer parental income raises hourly wages by 3.6% overall and by 5.2% for high-income individuals. Effects on total days worked or employment status are small and insignificant.

4. Mechanisms: Education vs. Labor Market Networks:

   Educational improvements appear to be a limited mechanism. While low-income individuals exposed to wealthier peers choose educational programs with slightly higher expected returns, the magnitudes are modest and do not translate into significant earnings gains. For high-income individuals, educational pathways are unaffected by peer composition.

   The main mechanism behind the observed earnings effects appears to be labor market networks. Dormmates from high-income families are more likely to work in the same firms years later—suggesting that peer connections facilitate access to lucrative job opportunities. In contrast, there is no evidence of long-term labor market connections for low-income conscripts, even when paired with high-income peers.

5. Wage Gap Illustration:

   The authors provide a stylized example showing that mixing conscripts from high and low SES backgrounds could reduce the wage gap between the groups by up to 14% under ideal conditions. However, because only high-SES individuals benefit significantly, current peer integration policies may have limited effects on inequality reduction.

Conclusion

This paper provides rigorous causal evidence that social exposure during young adulthood can have long-lasting economic consequences—but that the benefits of such exposure are not evenly distributed. Specifically, only individuals from already-advantaged backgrounds reap significant rewards from being assigned to live with high-income peers. These rewards manifest mainly through higher hourly wages and are driven by persistent labor market connections rather than educational improvements.

The findings have important implications for policies aimed at reducing inequality through social integration. While mixing individuals from different socioeconomic backgrounds may increase social cohesion and improve outcomes for some, it does not necessarily close gaps for those from disadvantaged backgrounds. The mechanisms that translate exposure into advantage—particularly labor market networks—appear to be segmented along class lines. Without policies that foster cross-class connections in meaningful, persistent ways, social integration alone may not be sufficient to break the cycle of intergenerational inequality.

Ultimately, the study underscores the importance of understanding both who benefits from social policies and why. It also suggests that addressing economic inequality may require more than just bringing people from different backgrounds into shared spaces—it may require deeper structural changes that ensure all individuals can access and leverage valuable social capital.