The merits of adding complexity: non-separable preferences in spatial models of European Union politics

This paper challenges the assumption of separable preferences that has been applied throughout the existing empirical research on European Union (EU) legislative politics. Yet our analysis reveals that non-separable preferences are in fact a widespread phenomenon in EU politics. In many cases actors’ spending preferences are conditional upon the expected policy outcome, but not vice versa. In order to appropriately model such non-reciprocity, we propose a simple modification to the standard Euclidean utility function. Applying simulation techniques, we demonstrate that overlooking non-separable preferences may have caused a substantial bias in the empirical evaluation of competing models of EU legislative politics. Specifically, models that constrain the set of feasible outcomes to either the Winset and/or the core must rely on a correct specification of actors’ utility functions. Therefore, a false assumption of core separable preferences significantly disadvantages these models vis-à-vis unconstrained models. Moreover, our findings underline the agenda-setting power of the European Commission for proposals that involve either a spending decision or the delegation of power from the national to the European level.