Modelling negative campaign advertising: evidence from Taiwan

Abstract

Why do election candidates go negative? This question about one of the most distinct elements of campaign behaviour has vexed political scientists for many years. Many explanatory variables have been proffered, in isolation and in ad hoc combinations. But extreme variation in the predictive power of these variables across elections has confounded attempts to develop a general model for explaining campaign attack ads. Taking advantage of the increasing number of polities in which negative advertising has been observed, this article applies a series of explanatory models to explain negative advertising in national and subnational elections in Taiwan. The analysis replicates prior research findings in the Taiwan context and improves the predictive power of the basic model. It also demonstrates that the aggregated measure of negativity commonly used in the literature is based on non-reliable scaling methods. Modelling an alternative, disaggregated measure that accounts for the various types of content contained in negative messages produces significantly different results. Negativity in advertising is a heterogeneous phenomenon and its components work in different ways or, more accurately, candidates do not decide just to ‘go negative,’ but make decisions to attack on policy, ideology, personality or strategy, based on a range of contextual and agent-based conditions.