“Overfitting” is a phenomenon in statistics where models are needlessly complex. Overfit models are very elaborate yet inaccurate. Such models look precise and sophisticated, but the precision is illusory: they reproduce the past while obscuring the future. The same dynamics can emerge in regulatory frameworks and models when detail accumulates faster than insight.

Economic regulation in the UK has increased in complexity over recent decades with the introduction of cost-based regulation, more stringent performance incentives, and the growing capital program across sectors. As a result, parts of the regulatory model have become overfit, such as granular adjustments to the WACC, or the Totex modelling in water and wastewater. Other elements are underfit, such as CAPM-based Cost of Equity or Totex modelling in aviation, where there are great gains from making better use of data, even if this implies more complex models. Overfitting is arguably the bigger problem.

View pdf document