ELECTION ANALYSES IN THE BRAVE NEW WORLD OF CONVERGENT MEDIA?
by Inderpal Bhandari, executive editor at large

Nation-wide elections were held on Nov. 3rd for gubernatorial, senate and house seats. Extensive media coverage followed the closing of polls. All the major broadcast and cable channels dedicated substantial air time to covering the races as votes were counted. While watching the media coverage of the elections, I found myself musing about the possibility of getting the election analyses directly from a data mining program in the near future, as opposed to the situation today, where we try to soak in the wisdom of the talking heads on television.

The day after, I tried to summarize what I had learned about the elections from the extensive media coverage over and above what I already knew. First, the strong showing of the Democratic party was clearly unexpected. The talking heads had informed me that during the off-year elections of a second term president, the party in the White House traditionally loses votes. In contrast, it looked like the Democrats were actually picking up seats on Nov. 3rd. Second, I learned that the average turnout had been high, 38% versus a number in the low 30's as is traditionally the case. This was especially true of the minority vote. Women, African-Americans, and Hispanics had turned out in large numbers. Furthermore, they voted for the Democrats, explaining the strong performance of the Democratic party. Third, I learned that the percentage of minority votes that made up the votes cast has been steadily increasing over the past few elections. In the case of African-Americans, this was true even though they had declined as a percentage of the overall American population. The clear implication was that the minority vote had become a critical factor and would continue to become even more important in future elections.

If you review the interesting facts in the preceding paragraph, they all have a distinct form, namely, a statistic that is markedly different from a traditional expectation. But, that is precisely what data mining programs can do, leading one to think about the possibility of getting interesting information about election results directly from a data mining program, instead of scanning the airwaves listening to the talking heads.

For example, data of the election results could be made available on the Internet and users could interact with that data using a decision support program. Or, if one preferred the more passive mode of TV watching, Pointcast or some other such push facility could simply pump interesting nuggets of information as they were discovered by a data mining program to one's computer screen. Or, better still, one's TV screen, perhaps by including the nuggets in the vertical interval of the broadcast. We could get the big picture in a few such snapshots. No need to sit glued to the TV. No need to listen to the talking heads.

Or, is there? Some talking heads are smart. One such person speculated that the high turnout could be attributed to Hillary Clinton's efforts in getting out the vote, clearly, an interesting hypothesis. If there was a facility by which we could explore election data, I would have been interested in following up on this hypothesis by checking the turnout in places where Hillary Clinton had campaigned. I expect that these smart talking heads will co-exist with the automatic extraction of nuggets.

I also think that much of what can be learned from interviews with voters does not always come through as clearly from an automated analysis of aggregate trends. For example, take Judy Malone-Stein, a voter in Houston who was quoted in the New York Times. "I voted against the Republicans for being such pinheads, for not getting down to issues. It's a dangerous time in this country when people start prying into people's private lives and not taking care of business". To which one may add, Q.E.D. Clearly, there is no substitute for a precise, succinctly-worded opinion as above. It is much more than mere sound bite.

In summary, I think that when people can directly explore data about events through sophisticated computer programs, it will impact the coverage of those events on TV for the better. It may even exert a much-needed, evolutionary pressure on the talking heads to improve their act.

---

Inderpal Bhandari can be reached via http://www.virtualgold.com