As Forrest Maltzman beautifully noted in his eulogy, Lee didn’t like to leave things undone. One exception is the only paper he and I wrote back in 2004, which means quite rightly that the blame must fall on me.
Our paper analyzed the extent to which daily perceptions of campaign negativity (measured in the Annenberg survey) tracked actual negative attacks by the Bush and Gore campaigns in the 2000 elections. Our answer: not at all. Instead we found that perceptions stayed remarkably consistent throughout the campaign with one major exception: on October 11 (the day of the second debate) a sudden but lasting change occurred (see the graph below). Up until then, the Bush campaign was thought to be more negative than the Gore campaign by about ten percentage points. Yet, Bush more than neutralized that difference in one day.
What happened? It wasn’t an unusually negative attack by Gore nor was it Gore’s performance in the second debate. Almost all interviews were held before the debate actually took place and the result holds when we remove the other survey respondents from the analysis. Instead, we suggest that it was a media effect:
Tens of millions of Americans watched the first debate, but the lasting effects of that evening’s developments – especially Gore’s exasperated sighs and overbearing demeanor – were not registered until more than a week later, when, on the eve of the second debate, virtually every discussion focused disparagingly on his first-debate performance.
A nice twist in the paper is that we applied a Bayesian change point model to identify the precise day in which a shift in perceptions occurred. We submitted the paper as a research note and received an invitation to revise and resubmit from a respectable journal. Nevertheless, we thought it would make more sense to wait until the 2004 Annenberg survey came out to see if similar sudden shifts in perceptions of campaign negativity occurred. Unfortunately, we never got around to this (and I promised Lee’s spouse not to remind him of this). Although I know it won’t help my Sigelman number, I have now made the paper available on SSRN (the link to it from my web-site had been broken for years).