Effect of Information Cascades on the Presidential Election
Less than two weeks from the election, the homepages of nearly all the major American news sources have been inundated with articles regarding the most recent poll results, and each candidate’s seeming endless source of advertising budget. The above article from NBC News specifically addresses early voting results in one of if not the single most important swing state this election: Ohio. However, the article is less focused on the actual results from early voting as it is with how each campaign has been so eager to say their side is “winning” the early poll battle. Considering that the number of votes that will be cast November 6th is vastly greater than the sum of all the early votes cast thus far, it is particularly interesting the both Mr. Obama and Mr. Romney are so eager to claim early voting momentum. Their reasoning, I believe, is directly related to our recent class discussion on Network Dynamics of Population Models, specifically regarding Information Cascades.
How, one might ask, can the election be viewed as an Information Cascade? After all, one of the ingredients in the Information-Based Imitation models we discussed in class was that people make decisions sequentially over time, and thus utilize previous players’ choices as a source of information. The election, on the other hand, occurs for the most part on a single day. However, if we consider only the ‘Undecideds’- people who are still deciding who to vote for, or whether or not to vote- we can analyze the election as occurring over an extended time period, with undecided voters making their “decision” some point between now (when they’re still undecided) and when they ultimately cast their vote. Intuitively, a large part of that decision will be made based on each voter’s opinion of the candidates, and their position on certain issues. However, we can see from the importance each campaign is placing on claiming this supposed early-voting momentum that another key component that may influence voters is other voters. For instance, after the first Presidential debate, Mr. Romney’s better-than-expected showing resulting in a steady increase in his polling numbers. Note that this increase was steady, not instantaneous. If voters were relying solely on the debates when responding to polls, Mr. Romney’s numbers should have shot up once and then leveled (considering only the effect from the debate). However, when we consider the Information Cascade that stemmed from his debate victory, it makes perfect sense that polls immediately following the debate would show only a slight change, and that the ‘momentum’ that Mr. Romney captured from his victory would result in further increase in polling percentage. In essence, this momentum that each candidate is so eager to claim for himself is the information each person in our Information-Cascade-game draws from the previous players’ choices. When one candidate starts doing better in the polls, voters see that positive increase and utilize this information-based reason (as opposed to direct-benefit effect) when deciding him or herself on who to vote for, with the result that the surging candidate does even better.
This sort of Information Cascade, just like the urn example from class, relies on less “actual” information, such as a debate performance, as it does on information gleaned from the resulting candidate’s momentum after such a performance. As such, both Mr. Romney and Mr. Obama may be increasingly focused on claiming that momentum for themselves rather than directing their efforts towards swaying voters on more substantive grounds.