Economics: An econometric analysis of the effects of Austrian marriage subsidies

 

Note: The following article was written by Lucky Bheemreddy L6 (20bheemreddyl@students.watfordboys.org)

1- Introduction

This academic year I undertook a research project examining whether Austrian marriage subsidies (grants given to those that get married) stimulated greater investment levels in the economy. Having completed and published a literature review, I extended this project and independently implemented econometric analysis to empirically assess whether Austria's 1986 marriage subsidy reform generated a measurable effect on national investment behaviour, rather than simply altering the timing and incidence of marriage. The policy is designed under the assumption that marriage promotes greater financial stability, higher household saving, and ultimately stronger investment outcomes at the macroeconomic level.

This implies a causal chain running from marital status to household financial integration, and from household behaviour to aggregate capital formation. However, as pointed out in my literature review, marriage does not necessarily lead to unified financial decision-making or increased long-term saving. This raises an empirical question whether marriage subsidies translate into detectable changes in national investment, or primarily affect social outcomes without altering underlying economic behaviour?

2- Research Design and identification strategy

To identify causal effects, the analysis employs a Synthetic Difference-in-Differences (Synthetic DiD) framework, which combines elements of synthetic control methods and traditional difference-in-differences estimation.

Rather than relying on a single comparison country, this approach constructs a counterfactual “Synthetic Austria” using a weighted combination of comparable European economies including Greece, Italy, Czechia, Germany, Spain, France, and the EU average. The weights are chosen to ensure that pre-treatment trends in investment closely match Austria’s observed path prior to the policy change.

The treatment year is defined as 1986, when the Austrian government announced and implemented the subsidy reform that triggered a sharp increase in marriage rates. The identifying assumption is that, in the absence of the policy, Austria’s investment trajectory would have continued to follow the path of its synthetic counterpart. Any systematic divergence after 1986 can therefore be interpreted as the potential macroeconomic effect of the marriage subsidy.

This strategy allows for stronger causal inference than simple before-and-after comparisons, as it explicitly controls for contemporaneous macroeconomic trends shared across European economies.

3- Synthetic DiD

Figure 1: “Synthetic DiD: Austria vs Synthetic Austria”

Figure 1 plots Austria’s observed gross fixed capital formation (GFCF) as a share of GDP against a synthetic control, obtained from the World Bank World Development Indicators (WDI) database. In the pre-treatment period, the two series move closely together, indicating a strong pre-reform fit. This alignment suggests that the synthetic control provides a credible approximation of Austria’s counterfactual investment trajectory- the path investment would likely have followed in the absence of the 1986 marriage subsidy reform.

Following the introduction of the subsidy, a divergence emerges between Austria and its synthetic counterpart. In the late 1980s, Austria’s investment-to-GDP ratio rose above the synthetic series, producing a positive gap. This pattern coincides temporally with the “marriage boom” identified by Frimmel, Halla, and Winter-Ebmer (2014), during which the announcement of the subsidy’s suspension induced a sharp increase in marginal marriages. One possible interpretation is that the policy temporarily altered household formation patterns in ways that could have affected short-term economic behaviour, such as increased housing demand, durable goods purchases, or household-level capital accumulation, which are components of GFCF.

However, this divergence is not sustained. By the early 1990s, the gap between Austria and the synthetic control narrows substantially, and the two series begin to reconverge. This re-alignment suggests that any investment response associated with the subsidy was transitory rather than structural. Overall, between 1980-2020 the absence of a persistent post-treatment gap implies that the reform did not generate a lasting shift in Austria’s aggregate investment trajectory.

This supports the broader interpretation developed in the literature marginal marriages and selection effects (Frimmel, Halla, and Winter-Ebmer, 2014; Kan and Laurie, 2014; Ludwig and Brüderl, 2018; Killewald and Lundberg, 2017): policy-induced changes in marital status, particularly marginal marriages, may influence short-term household behaviour, but these micro-level adjustments do not necessarily scale into sustained changes in national investment outcomes.

4- Statistical Significance and Confidence Interval Evidence

To formally assess statistical significance, yearly treatment effects are computed as the gap between Austria’s observed GFCF share and its synthetic control.

These gap estimates are paired with placebo-based confidence intervals, constructed by assigning “treatment” status to each donor country and calculating the resulting distribution of gap values. This non-parametric approach captures the range of effects that arise purely from random cross-country variation rather than policy intervention.

Across the post-treatment period, the confidence intervals consistently include zero, indicating that the estimated effects are not statistically distinguishable from no effect at conventional significance levels. For example the 1995 gap is 3.21 percentage points while the 97.5% confidence interval= -2.12 to 7.72
This pattern is repeated across multiple post-reform years, reinforcing the conclusion that the observed divergence in Figure 1 lacks statistical robustness.

To complement the confidence interval analysis, a formal permutation test is conducted using the ratio of post-treatment to pre-treatment RMSPE as a test statistic. This metric evaluates whether Austria’s post-reform divergence is unusually large relative to the distribution of placebo effects generated by treating each donor country as if it had received the subsidy. The results are: Austria RMSPE ratio = 2.13 and Permutation p-value = 0.556

A p-value of 0.556 implies that more than half of the placebo-treated countries exhibit effects of equal or greater magnitude than Austria’s estimated gap.

Formally, this prevents rejection of the null hypothesis that the observed post-1986 divergence occurred by chance. In econometric terms, the marriage subsidy does not generate an investment response that is statistically distinguishable from normal cross-country macroeconomic variation.

5- Implementation

All estimation and inference were implemented in R, using specialised econometric and data-processing libraries: synth for synthetic control estimation, dplyr for data manipulation, readxl for importing World Bank data, ggplot2 for visualisation.

The World Bank GFCF series was formatted into a balanced panel, with countries indexed by numeric unit identifiers and years treated as integer time variables. The synthetic control was constructed using GFCF as both the predictor and dependent variable, optimising pre-treatment fit over the 1980–1986 window while the full observation windom extends to 2020. Placebo tests were generated by iteratively reassigning treatment to donor countries and recalculating RMSPE ratios and gap distributions. Confidence intervals were derived from the standard deviation of placebo gaps for each year.

6- Econometric Conclusion

The preceding econometric evidence does not support the conclusion that Austria’s marriage subsidy reform generated a measurable or lasting effect on national investment behaviour. Synthetic control analysis reveals a post-treatment divergence that dissipates by the early 1990s, while the permutation test yields a p-value of 0.556 and placebo-based confidence intervals that consistently contain zero, indicating the observed gap is statistically indistinguishable from chance. One key takeaway is that the short-term divergence in GFCF cannot be straightforwardly attributed to the subsidy itself. A limitation being the problem of cross-correlation: investment and marriage rates may move together not because one drives the other, but because both respond to the same underlying macroeconomic conditions such as GDP growth, interest rates, or consumer confidence. The 1986 subsidy reform appears to have accelerated household formation decisions that were already underway given broader demographic trends, rather than fundamentally altering economic behaviour. Under this interpretation, any observed co-movement between the marriage boom and the short-term GFCF divergence may be driven by shared cyclical reasons rather than a causal mechanism. Ultimately, this suggests that marriage subsidies of this kind operate primarily as social policy instruments, with no robust transmission mechanism to aggregate capital formation. This positions the policy as a sub-optimal use of government expenditure when taking into account its opportunity cost.