Since December 21, 2012, gender-based premiums in insurance contracts have been prohibited in the EU [1]. If gender-based price discrimination were to be allowed, as it had been before, what would result? In car insurance, for example, the average premium for men would rise and the average premium for women would fall, since men would be more likely to be involved in car accidents. This is because, in addition to the costs accrued to insurance companies, male enrollees may have greater need for such coverage.

This example is what is known as third-degree price discrimination. In third-degree price discrimination, different prices are offered according to the buyer’s externally identifiable characteristics, as in the example above where different premiums exist across gender.

In addition to fairness, these issues are often evaluated in terms of efficiency resulting from transactions between sellers and buyers (also called “economic welfare”): whether it will increase or decrease. Economic modeling is utilized regarding preference and market competition, and economic welfare in the current situation is directly compared to economic welfare in the “counterfactual” situation to predict and evaluate a public policy or law. In the example above, the current situation is one in which price discrimination does not take place, and the counterfactual situation is one in which it is permitted. Welfare evaluation requires the task of predicting what would occur in such a hypothetical situation.

In a recently published article of the IJIO, A sufficient statistics approach for welfare analysis of oligopolistic third‐degree price discriminationTakanori Adachi of Kyoto University, proposes an alternative approach. He shows it is not always necessary to explicitly consider the counterfactual situation, but it may suffice to calculate “sufficient statistics” based on information such as current sales volumes and prices.  More specifically, if these relationships satisfy certain conditions, then a shift to the hypothetical situation will necessarily reduce economic welfare.

If price discrimination is practiced in the status quo, then in the counterfactual situation, economic welfare would rise for the group that originally had higher prices because prices would fall. Conversely, in the group that originally had lower prices, economic welfare will deteriorate because prices will rise. Therefore, if the former improvement is smaller than the latter deterioration, then it is undesirable to move to the counterfactual situation through policy or other means,

In order to confirm this, it is not necessary to predict behavioral results in the counterfactual situation, but only to measure/estimate sufficient statistics using the data observed in the current situation. In particular, for each market, we only need to calculate, among others, the profit margin (i.e., the difference between price and marginal cost), the pass-through (based on the curvature of demand), and the index measuring the degree of competition. If these indices satisfy the conditions described in the article’s analysis, it can be said that gender-specific premium setting is undesirable from the efficiency standpoint.

It should be noted, however, that even if this condition is not met, it does not mean that economic welfare will improve. In this sense, it may be compared to hypothesis testing in statistics because the null hypothesis that “economic welfare will increase” cannot be positively supported even if the condition does not hold.

This “sufficient statistics approach” (Chetty 2009; Kleven 2021) has the future potential as a “third way” that complements the “causal inference approach” that does not posit a specific behavioral model and the “structural inference approach” that describes specific behaviors and market structures in detail (Adachi and Fabinger 2022; Barnichon and Mesters 2022).

It is expected that the pursuit of such a methodology will advance our understanding based on economic reasoning. This is because the “causal inference approach” is biased toward statistical models, leaving the mechanism as a black box, while the “structural inference approach” may be biased toward a particular structure from which conclusions are drawn.

(Full paper published in IJIO Volume 86, January 2023)



Adachi, T., Fabinger, M., 2022. Pass-through, welfare, and incidence under imperfect competition. Journal of Public Economics 211, 104589.

Barnichon, R., Mesters, G., 2022. A sufficient statistics approach for macro policy evaluation. Unpublished.

Chetty, R., 2009. Sufficient statistics for welfare analysis: A bridge between structural and reduced-form methods. Annual Review of Economics 1, 451-488.

Kleven, H. J., 2021. Sufficient statistics revisited. Annual Review of Economics 13, 515-538.