Erweiterung der Propensity Score Gleichung zur Vermeidung von Fehlspezifikationen? Eine Monte Carlo Simulation

Abstract

Propensity score matching is a semi-parametric method of balancing covariates that estimates the causal effect of a treatment, intervention, or action on a specific outcome. Propensity scores are typically estimated using parametric models for binary outcomes, such as logistic regression. Therefore, model specification may still play an important role, even if the causal effect is estimated nonparametrically in the matched sample. Methodological research indicates that incorrect specification of the propensity score equation can lead to biased estimates. Augmenting the propensity score equation with terms that represent potential nonlinearity and nonadditivity, as proposed by Dehejia and Wahba and more recently by Imbens and Rubin, represents a means of avoiding such bias. Here, we conduct a Monte Carlo simulation and show that the misspecification bias is rather small in many situations. However, when the propensity score equation and/or the outcome equation are characterized by strong nonlinearity and nonadditivity, the misspecification bias can be severe. Augmentation is shown to reduce this bias, often substantially. The Dehejia-Wahba (2002)

Agricultural Credit and Extension Services: Does Their Synergy Augment Farmers’ Economic Outcomes?

“…The most reliable way to address such an issue is to employ an experimental study, such as a randomized control trial [31]. In randomized control studies, treatment and control groups are randomly assigned, and the potential confounding variables are equally distributed throughout the groups [47]. However, a randomized controlled trial requires interventions, which are costly and not always possible.…”

“…In nonexperimental studies, the treatment is nonrandom [33,47]. In such cases, the potential observed and unobserved confounding variables may influence both the response and the treatment variables, which causes a selectivity bias dilemma.…”

“…Many studies apply the matching method as a useful tool to relieve potential selection bias and endogeneity [33,49]. PSM imitates randomization while assigning the treatment [47] by matching the treated observations with the untreated observations. In this way, it imputes the unobserved counterfactual response variable by matching the treated observations with the closest untreated observations based on pretreatment features.…”

“…There may be a large gap between the two closest individuals' propensity scores available for the match, which causes poor matches. Therefore, in addition to using nearest neighbor matching and kernel matching, this study also applies the caliper restriction to mitigate this potential problem [47] by establishing a threshold for the maximum difference between the matched observations. If the difference is beyond the threshold, then the treated unit is omitted to avoid a biased estimation.…”

Participation of Brinjal Farmers in Large and Small Wholesale Markets: Factors Influencing Farmers’ Decisions and Impact on Producers’ Prices

“…To address the selection bias, this study applies quasi experimental identification propensity score matching (PSM) because it imitates randomized experiments. PSM define randomization [37] while assigning the treatment by matching the treated observations with the untreated observations. Many studies even apply the matching method as a useful tool to relieve potential selection bias issues [36,38].…”