Instrumental variable methods are a standard tool for estimating price elasticities from observational data, but they can go wrong when the underlying time series are autocorrelated. In a revised arXiv paper, Hirth and colleagues argue that explicitly modeling the time structure with causal graphs can fix this problem. They introduce causal mechanism time graphs—acyclic graphs that still allow prices and quantities to be determined simultaneously, while providing an interpretable notion of intervention.
Using these graphs, the authors derive instrumental variable estimators that remain consistent under weaker assumptions than naive approaches. They also point out that when two presumably valid estimators give very different results, that discrepancy can be used to reject incorrect model assumptions. This gives researchers a practical diagnostic for improving their models of economic dynamics.
The method is demonstrated on the German electricity market, using both simulated and real-world data. The results highlight how important it is to account for structural autocorrelation when doing IV-based analysis. For economists working with high-frequency market data, this offers a more reliable route to estimating demand responsiveness.