TL;DR
This paper quantifies the long-term financial cost of compulsory military service in the Netherlands by analyzing its impact on future earnings. Using a clever "natural experiment" approach, the authors found that conscription leads to a significant, persistent 5% reduction in earnings, equivalent to losing a year of potential work experience.
What Problem Does This Paper Address?
Estimating the true economic cost of military conscription is tricky. It's not as simple as comparing the pay received during service to what one could have earned. The Netherlands, like many countries, had a system of compulsory military service. However, the decision of who actually serves is influenced by several factors beyond just being drafted. These include passing medical and psychological exams, and crucially, a variety of exemptions (for students, breadwinners, etc.) that could be strategically used. This "selection bias" means that simply comparing veterans to non-veterans would likely underestimate the true cost. The paper aims to disentangle the causal effect of military service on future earnings, accounting for these complex selection mechanisms.
Methods
The core challenge is dealing with selection bias. The authors tackle this by employing an Instrumental Variables (IV) strategy, a common technique in econometrics for estimating causal effects when direct comparison is confounded. Instead of relying on individual-level data (which was unavailable for earnings and military service history), they leverage grouped data estimation and policy-induced variation in aggregate military enrollment rates.
Specifically, they exploit two key policy changes in the Netherlands:
- A reduction in the age for medical and psychological examinations from 19 to 18.
- The introduction of a "special exemption" category, which allowed the government to adjust the number of conscripts annually by designating certain groups as exempt.
These policy shifts created exogenous variation in the proportion of men who served in the military across different birth cohorts. The authors use this variation as an instrument. They essentially argue that while individual decisions to serve are endogenous, the aggregate enrollment rates, driven by these policy changes, are not directly influenced by individual earnings potential. They then estimate a log-earnings regression model using this instrumental variable approach.
Datasets
The study utilizes two primary data sources:
- Dutch Defense Department Data: Provided information on the proportion of each birth cohort that actually served, the proportion designated "special exemption," and the size of each birth cohort.
- Central Bureau for Statistics (Netherlands) Data: Specifically, the "Inkomen Panel Onderzoek," which provided mean and variance of earnings in 1989 and 1990, broken down by month of birth. This data represents a random sample of about 1% of Social Security data.
Key Insights and Ideas
- Conscription is a "Tax" on Future Earnings: The paper frames military service as a form of taxation, not just in terms of foregone earnings during service, but more importantly, through its lasting impact on future earning potential.
- Selection Bias is a Major Hurdle: The authors emphasize that without accounting for exemptions and selection criteria, any estimate of military service costs would be flawed.
- Natural Experiments are Powerful Tools: The study highlights how policy changes can create valuable "natural experiments" that allow researchers to isolate causal effects in situations where controlled experiments are impossible.
- Aggregate Data Can Inform Individual Effects: By cleverly combining aggregate policy variation with grouped earnings data, the authors were able to estimate individual-level causal effects.
- The Cost is Persistent: The 5% earnings reduction is not a short-term blip but appears to be a lasting effect, suggesting a significant long-term economic burden.
Key Results
The central finding is that military service in the Netherlands leads to an average 5% reduction in future earnings, approximately 10 years after service. This translates to an estimated annual cost of 2,000 Dutch guilders (fl.) (around $1,000 USD in 1989 dollars) for the affected cohorts. This cost is comparable to losing a year of potential work experience. The results were found to be robust across various specifications and sensitivity analyses.
Limitations and Future Work
- Age Profile Approximation: The authors acknowledge that estimating the age profile of earnings is crucial, especially since the affected individuals are on a steep part of their lifetime earnings curve. While they use a cubic function, they note that more complex age-dependence structures might be explored.
- Women's Data: While they attempted to include women's data as a control group, they note limitations due to data availability and potential differences in earnings patterns.
- "Active" Category: The definition and treatment of the "active" category (individuals still enrolled in education) were considered, and alternative measures were explored, showing robustness.
- Imputed Data: For the 1959 cohort, which was entirely exempt, the "special exemption" percentage had to be imputed, and the sensitivity of the results to this imputation was tested.
- Future Work: The paper suggests that further research could explore the indirect effects of military service on education and career choices, as well as the potential for a non-zero return from military service itself (e.g., through skills gained, though their findings suggest this is not the case). The study also emphasizes the importance of investigating the sensitivity of estimates to different specifications.
Summary updated