A cruise-and-crash model of the cost of ruling

Since the publication of Nannestad and Paldam ([1999] on the cost of ruling, their finding that incumbents lose votes in the order of 2 to 3 percentage points per term has gained the status of something like an “inductive law” of politics (Budge [2019]. We suggest that this generalization conjures up an inaccurate image, that of a gradual reduction of a ruling party’s vote over the course of an incumbency or “spell.” Our own analyses of national and state or provincial elections in Australia, Canada, Germany, and the United States demonstrate that this interpretation is not correct. Instead, while they remain in office the ruling party cruises along, averaging more or less the same share of the vote election after election, until they fail to win another term, at which point all or nearly all of the cost of ruling, like a balloon payment on a loan, comes due. Moreover, on average the magnitude of the loss is independent of the length of their stay in office. This means that a long incumbency confounds the expectations fostered by the “inductive law” because the overall loss – which we find to be reasonably constant across a wide spectrum of data sets – divided by a spell length exponentially distributed from one to many terms cannot be a constant. Nannestad and Paldam's assertion of 2.25 percentage points per term vote loss is at best a description of a mid-range value. Our approach suggests that the cost of ruling is the product of two country-specific constants: an exponential decay of spell functions and an invariant percentage change in votes between the ousting election and the mean winning vote that is independent of spell duration. Thus, our method comes close to matching the inductive law at the mean, but also accounts for its variability along the number of terms in a spell. Moreover, standardizing the loss relative to the mean win vote yields a more general measure that cuts across countries and types of government. We estimate that in most developed democracies the depth of the sink ranges between 15% and 25% below the mean win vote.