The forecasting model presented here is a revised version of a model developed prior to the 1996 election (Holbrook 1996) and is essentially a referendum model. The original model regressed the incumbent party percent of the two-party vote on presidential popularity, an aggregate measure of satisfaction with personal finances,¹

and a dummy variable coded “1” for years in which the incumbent party had held the White House for at least two consecutive terms and “0” for all other years. The first two variables are intended to capture the political and economic performance of the incumbent administration, while the latter variable (borrowed from Abramowitz 1988) is based on the idea that it may be easier to convince voters that it is “time for a change” if the incumbent party has held the White House for at least two consecutive terms. This model provided a fairly accurate forecast of the 1996 election and also had close out-of-sample post-dictions of elections from 1952–1992. However, the 2000 election represented a significant bump in the road, and the model over-predicted Gore's percent of the vote by approximately 10 percentage points.

The impact of the 2000 election on the performance of the model is evident in Table 1. Here we see that prior to the 2000 election the model fit the data points quite well, but that the fit takes a real hit with the inclusion of the 2000 election. Based on the values of presidential approval (58%), aggregated personal finances (138), and tenure (1), the model predicted a Gore victory with 60.3% of the two-party vote. In addition to the poorer fit, the post-2000 results leave us with the highly implausible conclusion that presidential approval has no statistically significant impact on election outcomes. Clearly, this called for some rethinking. However, to avoid falling into the trap of completely revamping the model in response to each new data point, I tried instead to preserve the essence of the model while, at the same time, accounting for the errant prediction of the 2000 election.

I have explained in considerable detail “what went wrong” in my post-2000 evaluation of the model's performance (Holbrook 2001), so I will just touch on a couple of points very briefl y before presenting the revised model. In a nutshell, Gore did not receive the kind of credit for the economy that had been given to previous incumbent party candidates. The reasons for this were two-fold. First, Gore himself appeared desperate to distance himself from President Clinton and, hence, did not run on the Clinton economic record. Given the strength of that record (unemployment was at about 4% and second quarter gross domestic product grew at an annual rate of 6.4%), Gore was putting himself at a significant disadvantage, especially considering that candidate statements about the economy hold substantial sway over the contours of public opinion during the campaign (Shah, Watts, Domke, Fan, and Fibison 1999). In addition to Gore's reluctance to run on the record of a popular president during relatively strong economic times, voters were exposed to mixed signals regarding the direction of the economy. While overall economic indicators remained relatively robust during the summer and fall of 2000, there were a few highly visible signs of economic distress that may have infl uenced the economic information to which voters were being exposed. Specifically, rising gasoline prices and a declining stock market represented sources of dissonant economic news, despite the broader, generally favorable economic picture.

We can get a sense of the tone of economic information during the campaign by utilizing data from the Surveys of Consumers, a national public opinion survey conducted on a monthly basis by the Survey Research Center at the University of Michigan. Each month, respondents are asked: “During the last few months, have you heard of any favorable or unfavorable changes in business conditions?” and “What have you heard?” Responses to these questions are used to create an index (100 + (percent favorable news - percent unfavorable news)) that summarizes the overall tone of economic news respondents reported hearing. Normally, one would expect that the tone of economic news would have been relatively positive for a year like 2000, when the general economic indicators were strong. However, the value of that indicator in the fall of 2000 was 79, which means that the number of people who reported hearing negative economic news was 21 percentage points higher than the number who reported hearing positive news. This is roughly the same as in1960 (71) despite the overall stronger economic picture in 2000.²

What is particularly important for the forecasting model is that this measure has generally tracked very closely with the personal finances variable that represents the economic component of the model. Generally, that is, until 2000. From 1956 to 1996 the correlation between May personal finances and fall economic news³

was .88; with the inclusion of data from 2000 that correlation drops to .56. The 2000 numbers defy historical expectations: one would historically expect that record high (for presidential election years) levels of satisfaction with personal finances in May would be met with reports of relatively positive economic news in the fall, but the economic news people were hearing in the fall of 2000 was decidedly negative. Whether the explanation for this lies in Gore's reluctance to trumpet the Clinton-Gore economic record, or in the infl uence of specific indicators, such as gasoline prices and the stock market, is well beyond the scope of this paper. But what is significant is that there was a clear disconnect between economic performance and the information context of the fall campaign. I have argued previously (Holbrook 2001) that this made it more difficult for voters to translate their still-positive economic attitudes into support for the incumbent party.