Ever since fiscal consolidation first took centre-stage on the political agenda of many Western democracies in the 80s, the “era of permanent austerity” has served as a reference point for welfare state schools. According to the New Politics perspective, welfare policy will follow a different pattern from that of welfare expansion by allowing for only incremental cuts in welfare programmes and by blurring partisan differences leading to policy convergence across parties. By contrast, adherents to the Power Resource Approach argue that proper measurement of welfare efforts still offers testimony to persistent differences between the Centre-left on the one hand, and Centre-Right and Conservative Parties on the other.
Neither of the major approaches in contemporary welfare scholarship, however, has properly addressed one of the most puzzling finding of the Fiscal Consolidation literature: despite the presumed electoral verdict inflicted on retrenching incumbents, re-election prospects, if anything, have improved after successful fiscal consolidations, particularly if they tackled the most politically salient items in public budgets.
In an attempt to resolve this underlying puzzle, I propose a median voter framework where certain incumbents are confronted with a credibility gap in their welfare policies. I argue that this credibility gap will compel those parties to veer most sharply towards a pro-retrenchment position that one would least expect to.
To empirically test this argument, I depart from conventional measurements of partisanship and directly turn to the underlying support base of incumbent parties, relying on cross-national electoral survey data from Eurobarometer and ISSP . I set up a Time-Series Cross-Section model to predict programme-specific welfare efforts as a function of the electoral support constellation of incumbents. My main results point indeed towards a Nixon-China logic guiding welfare retrenchment: incumbents often inflict the most main on those voting blocs that they owe their electoral mandate to.
