Nonparametric Instrumental Variable Methods for Dynamic Treatment Evaluation

Abstract

We develop a nonparametric instrumental variable approach for the estimation of average treatment effects on hazard rates and conditional survival probabilities, without model structure. We derive constructive identification proofs for average treatment effects under noncompliance and dynamic selection, exploiting instrumental variation taking place during ongoing spells. We derive asymptotic distributions of the corresponding estimators. This includes a detailed examination of noncompliance in a dynamic context. In an empirical application, we evaluate the French labor market policy reform PARE which abolished the dependence of unemployment insurance benefits on the elapsed unemployment duration and simultaneously introduced additional active labor market policy measures. The estimated effect of the reform on the survival function of the duration of unemployment duration is positive and significant. Neglecting selectivity leads to an underestimation of the effects in absolute terms.

Instrumental variable estimation of dynamic treatment effects on a duration outcome

“…where I(•) is the indicator function. This is the so-called no anticipation assumption (see Abbring and Van den Berg (2003); Van den Berg et al (2020a)), which is standard in the literature on dynamic treatment effects. It means that the hazard rate is not affected by the treatment before the latter happens.…”

“…This restriction is in contrast with our nonseparable nonparametric identification approach. Finally, Van den Berg et al (2020a) study a very specific one-sided noncompliance context, where the time to treatment is either equal to the instrument or to ∞ (untreated). Instead, the present work does not assume one-sided noncompliance.…”

Instrumental Variable Estimation of Dynamic Treatment Effects on a Duration Outcome

“…Another paper is Van den Berg et al (2020a), which studies a one-sided noncompliance setting, where the time to treatment is either equal to the instrument or ∞ (untreated).…”

“…In contrast, the present work does not assume one-sided noncompliance, and our method allows us to estimate effects over the whole population. Van den Berg et al (2020a) also applies their estimation strategy to evaluate the causal effect of a reform of the unemployment insurance system in France. Our approach could also be applied to their dataset.…”

Bounds on Average and Quantile Treatment Effects on Duration Outcomes Under Censoring, Selection, and Noncompliance

“…Another relevant related literature is that of duration models (e.g., Heckman and Singer, 1984;Ham and LaLonde, 1996;Eberwein et al, 1997Eberwein et al, , 2002Abbring and Van den Berg, 2003;Van den Berg et al, 2016, Ba et al, 2017. This literature usually addresses censoring, selection, and sometimes noncompliance/endogeneity, by imposing parametric, separability, or proportionality assumptions.…”