Economic voting is simple. When the economy does well, voters tend to vote for the incumbent party. When it does not, they tend to vote for another party instead. This process of reward and punishment is vital for democratic accountability. Without it, voters have no prospect of âkicking the rascals outâ and, thus, governments face no consequences for wrecking the economy and their own votersâ material well-being. Economic voting scholars used to test this theory with macro-level data. But problems of ecological inference made this task more difficult. Simply put, they could not tell how the economy affected individual voters. To address this problem, they started to use individual-level data in the early-1980s. Unlike macro-level data, these micro-level data were cross-sectional. As such, the national economy was constant for all respondents. In response, scholars stopped using economic indicators, which did not vary, and started to use votersâ self-reported economic perceptions, which did, instead. In this thesis, I examine how voters answer common economic perception items. In doing so, I focus on how their party identification affects their responses. To this end, I make three main contributions. First, I show that some people answer these items in different ways in different surveys. This is because political surveys politicise the perceptions that some partisans report. Second, I show that party identification moderates how voters respond to economic change. When things are bad, partisansâ view converge. When things are good, they diverge instead. And this runs contrary to assumptions that all voters update their views in parallel. Third, I show that this partisan bias affects economic voting models. In particular, it artificially-inflates economic voting effects. Once we hold constant, economic voting effects are no longer statistically-distinguishable from zero. Thus, I conclude that economic voting scholars should develop new theories and new items to explain how economics affects politics.