Martin Huber

1 Causal mediation analysis aims at understanding the mechanisms through which a treatment affects an outcome of interest. It disentangles the treatment effect into an indirect effect, which operat...

Interference or spillover effects arise when an individual's outcome (e.g., health) is influenced not only by their own treatment (e.g., vaccination) but also by the treatment of others, creating challenges for evaluating treatment effects. Exposure mappings provide a framework to study such interference by explicitly modeling how the treatment statuses of contacts within an individual's network affect their outcome. Most existing research relies on a priori exposure mappings of limited complexity, which may fail to capture the full range of interference effects. In contrast, this study applies a graph convolutional autoencoder to learn exposure mappings in a data-driven way, which exploit dependencies and relations within a network to more accurately capture interference effects. As our main contribution, we introduce a machine learning-based test for the validity of exposure mappings and thus test the identification of the direct effect. In this testing approach, the learned exposure mapping is used as an instrument to test the validity of a simple, user-defined exposure mapping. The test leverages the fact that, if the user-defined exposure mapping is valid (so that all interference operates through it), then the learned exposure mapping is statistically independent of any individual's outcome, conditional on the user-defined exposure mapping. We assess the finite-sample performance of this proposed validity test through a simulation study.

We propose a novel application of graph attention networks (GATs), a type of graph neural network enhanced with attention mechanisms, to develop a deep learning algorithm for detecting collusive behavior, leveraging predictive features suggested in prior research. We test our approach on a large dataset covering 13 markets across seven countries. Our results show that predictive models based on GATs, trained on a subset of the markets, can be effectively transferred to other markets, achieving accuracy rates between 80% and 90%, depending on the hyperparameter settings. The best-performing configuration, applied to eight markets from Switzerland and the Japanese region of Okinawa, yields an average accuracy of 91% for cross-market prediction. When extended to 12 markets, the method maintains a strong performance with an average accuracy of 84%, surpassing traditional ensemble approaches in machine learning. These results suggest that GAT-based detection methods offer a promising tool for competition authorities to screen markets for potential cartel activity.

The majority of Swiss children attend mandatory and cost-free kindergarten at age four. We examine the effect of this policy on maternal labour market outcomes. Using administrative data from Switzerland, we exploit the birthday cut-off for kindergarten entry in the same or in the following year and apply a non-parametric regression discontinuity design (RDD). We find that mandatory kindergarten has a statistically significant positive effect on the labour market attachment of previously non-employed mothers, increasing their employment probability by 4 percentage points. In contrast, there are no significant effects on other groups or in the total sample of mothers.

In social sciences and economics, causal inference traditionally focuses on assessing the impact of predefined treatments (or interventions) on predefined outcomes, such as the effect of education programs on earnings. Causal discovery, in contrast, aims to uncover causal relationships among multiple variables in a data-driven manner, by investigating statistical associations rather than relying on predefined causal structures. This approach, more common in computer science, seeks to understand causality in an entire system of variables, which can be visualized by causal graphs. This survey provides an introduction to key concepts, algorithms, and applications of causal discovery from the perspectives of economics and social sciences. It covers fundamental concepts like d-separation, causal faithfulness, and Markov equivalence, sketches various algorithms for causal discovery and discusses the back-door and front-door criteria for identifying causal effects. The survey concludes with more specific examples of causal discovery, e.g., for learning all variables that directly affect an outcome of interest and/or testing identification of causal effects in observational data.

In social sciences and economics, causal inference traditionally focuses on assessing the impact of predefined treatments (or interventions) on predefined outcomes, such as the effect of education programs on earnings. Causal discovery, in contrast, aims to uncover causal relationships among multiple variables in a data-driven manner, by investigating statistical associations rather than relying on predefined causal structures. This approach, more common in computer science, seeks to understand causality in an entire system of variables, which can be visualized by causal graphs. This survey provides an introduction to key concepts, algorithms, and applications of causal discovery from the perspectives of economics and social sciences. It covers fundamental concepts like d-separation, causal faithfulness, and Markov equivalence, sketches various algorithms for causal discovery, and discusses the back-door and front-door criteria for identifying causal effects. The survey concludes with more specific examples of causal discovery, e.g. for learning all variables that directly affect an outcome of interest and/or testing identification of causal effects in observational data.

This paper investigates the finite sample performance of a range of parametric, semi-parametric, and non-parametric instrumental variable estimators when controlling for a fixed set of covariates to evaluate the local average treatment effect. Our simulation designs are based on empirical labor market data from the US and vary in several dimensions, including effect heterogeneity, instrument selectivity, instrument strength, outcome distribution, and sample size. Among the estimators and simulations considered, non-parametric estimation based on the random forest (a machine learner controlling for covariates in a data-driven way) performs competitive in terms of the average coverage rates of the (bootstrap-based) 95% confidence intervals, while also being relatively precise. Non-parametric kernel regression as well as certain versions of semi-parametric radius matching on the propensity score, pair matching on the covariates, and inverse probability weighting also have a decent coverage, but are less precise than the random forest-based method. In terms of the average root mean squared error of LATE estimation, kernel regression performs best, closely followed by the random forest method, which has the lowest average absolute bias.