Non-recurrent road congestion, accident and incident duration
Martin Adler (VU University Amsterdam)

Road congestion is one of the main externalities of car use. We focus on the effects of car accident duration, and the duration of other types of incidents (e.g. breakdown of car), on non-recurrent road congestion on the Dutch highway network. By using a panel data set and estimating a location-fixed-effects model with many control variables, we aim to determine the causal effect of incident duration on non-recurrent congestion. We demonstrate that for accidents, incident duration has a strong positive effect on non-recurrent congestion. For example, an increase of incident duration from 60 to 70 minutes increases non-recurrent congestion by 160 vehicle-loss-hours. Our results indicate that a decrease of 10 minutes in average incident duration for all highway accidents entails a 40 million € gain to society. This suggests that there are strong benefits of incident management.

*******************

Weak identification robust inference in dynamic panel data models with additional endogenous regressors
Rutger Poldermans (UvA)

We consider the linear dynamic panel data model with additional endogenous regressors. We analyze the accuracy of various weak identification robust GMM statistics and several tests for weak identification.