Alexander Busch

We document the expanding geographic scope of labor unions in the US and show how the resulting organizational networks propagate shocks across distant labor markets. The number of states the average union local (union branch) operated in doubled from 1.5 in 2001 to 3 in 2011. As a consequence, shocks to unionized workers in one state are likely to have spillovers if unions respond by reallocating resources across states. We show this mechanism at work in the context of state-level Right-to-Work (RTW) laws, which make it optional to pay union dues. A model of a multistate union predicts spillovers with a sign that depends on the relative strength of substitution and scale effects. We test for spillovers using variation in union exposure to five RTW laws adopted between 2012 and 2017. RTW laws have negative direct effects in adopting states and positive spillovers; the average union represents 1.2 fewer bargaining units in adopting states and makes up for 35 percent of this loss in worker-level coverage by substituting towards non-adopting states. Notably, more than half of the spillover occurs in non-bordering states. We find no effect on the price or quality of union services. New organizing explains 34 percent of the spillover; the rest is explained by successful contract renegotiation for existing members. Despite the substitution, unions see a persistent decline in membership, revenues, and assets.

We use a newly digitized panel of collective bargaining agreements linked to quarterly labor market data to study how sectoral agreements shaped aggregate wage inequality in Germany between 1950 and 2025.

We study the effect of explicitly and implicitly gendered pay scales in West German collective bargaining agreements on gender gaps in wages and employment.

LPDID performs the Local Projections Difference-in-Differences estimator (LP-DiD) proposed by Dube, Girardi, Jordà and Taylor (2023). LP-DiD is a convenient and flexible regression-based framework for implementing Difference-in-Differences with multiple time periods. It can provide both dynamic event-study estimates and pooled estimates of the overall average effect in a post-treatment window. Treatment can be absorbing or non-absorbing. The estimation sample is restricted to units entering treatment and “clean” controls, thus avoiding the negative-weights bias of two-way fixed-effects estimators. The command allows reweighting to estimate an equally-weighted average effect, flexible choice of the pre-treatment base period, and inclusion of control variables.