Search This Blog

18.8.08

Improper analysis misses significant displacement effect

If you’re going to analyze prospects for a particular electoral contest, you need to use the right data. Unfortunately, an otherwise well-done piece on the contest featuring incumbent Democrat Sen. Mary Landrieu and challenger Republican state Treasurer John Kennedy was marred by reliance on poorly-conceptualized analysis.

An erroneous remark came from a New Orleans-area demographer who asserted “Senator Landrieu will not be losing African American votes as a result of Katrina.” Facts simply do not back up this claim. The comparison made was between the 328,443 blacks who voted on Nov. 5, 2002 and the 310,428 who voted on Oct. 20, 2007 – which is not the correct comparison and uses the incorrect metric.

Of some concern is that the first date was the runoff involving Landrieu’s last Senate race while the second was the primary for statewide elections. A comparison of more similar things would have been looking at data for either the primary date for each election, or the runoff date for each, a minor problem. But a huge conceptual error was made because the nature of turnout for each differs because different offices were up for grabs. This causes not only different levels of turnout, but different levels of registration.

If one cares to study election cycles in Louisiana, there is a hierarchy to turnout and registrations, roughly mirroring each other. The highest turnout elections are those with presidential candidates on the ballot, followed by congressional contests, with state contests dragging the rear. Tracking monthly statistics, for example one can see about six months before a presidential election a building large increase in registrations which pays off on the first Tuesday after the first Monday in November with the largest turnout of any kind of election requiring statewide voting. Afterwards, registrations begin to tail off dramatically until the next similar kind of election (congressional) looms.

In other words, comparing number registrants in 2002, a higher-stimulus election, to those in 2007, of lower stimulus, is like comparing apples and oranges. Further, focusing on the raw numbers of a single group, rather than numbers of all groups and their turnout rates, distorts the picture even more.

A better comparison is between the figures of Nov. 5, 2004, the last presidential election, and the latest registrant statistics which will be close to the ones applicable to the presidential election this fall, and applying to them the turnout rate differential between the 2003 and 2007 state elections. This keeps the comparisons between elections of the same kind, and allows us to convert, since actual turnout is what we are interested in but we don’t have that yet for 2008, registrations to turnout.

For the 2004 election, there were 1,934,953 whites, 870,103 blacks, and 116,658 other races on the Louisiana books. For Aug. 3, 2008, there were 1,884,947 whites, 874,896 blacks, and 121,315 others. However, raw registrants turn into voters at different rates. In 2007, as I have noted elsewhere, the 4 percent decline in turnout was attributable almost entirely to sharply lower participation rates among blacks in the state that were largely a product of displacement in three parishes from Hurricane Katrina – that is, while some had their names on the books, much more than in the past they were not present to vote nor wished to vote absentee.

A year will have passed and people continue to trickle back into New Orleans but the fact is 365,015 blacks voted in 2003, a gap of almost 55,000 to 2007, and registrations respectively for those races for blacks were 812,905 and 843,674. (An academic paper I presented earlier this year predicts the effect for 2008 to be a net loss of 48,000 Democrat votes, of which most were blacks.) To say the effects of Katrina will not have a substantial negative impact on votes for a Landrieu candidacy simply ignores reality.

No comments: