Ipsos Responds

Last week I noted that Ipsos polling out of the Kentucky Senate race appeared to be based on an unlikely projection of voter turnout. Ipsos sent along a response over the weekend, which I'll repost here:

Now to the blog piece, there was mention that our turnout estimates were incongruent will past Kentucky board of election turn out figures—that they were too high. On your specific point, we agree that the official Kentucky electoral turnout results for 2006 are lower than our estimates (approximately 50% versus 70%).

We, however, shy away from using electoral board results as such data sources are notorious for double counting people that have moved and not culling those who have died. In our estimate, such data sources over-estimate by 20 to 30% the actual registered vote count.

Instead, we rely on the Census Bureau's CPS (Current Population Survey) for our estimates. In November of every electoral year, the CPS includes a registered voter and turnout module on the survey and produces official estimates of turnout. See http://www.census.gov/hhes/www/socdemo/voting/publications/p20/2006/tables.html for an example.

For full disclosure, in the below table, I include our estimates of registered voter turnout for Kentucky based on CPS data.

Percent Turnout in Kentucky by Registered Voters and Citizen Adult Population

Source: Current Population Survey, US Census

	1998	2002	2006
Percent Registered Voters who Voted	68%	68%	67%
Percent Total Population of Citizens who Voted	46%	45%	49%

Source: Current Population Survey, US Census


I think it worthwhile to point out that the Kentucky election results that you cite indicate that in 2006 there were a total of 2,766,288 registered voters, while the CPS estimates that there were 2,240,000 registered voters—an approximate 500,000 voter difference. If the Kentucky registered voter numbers were right, that would mean that about 91% of the adult citizen population is registered to vote--which is too high by any standard.

For these very reasons, we use the CPS estimates—with all the caveats in using sample surveys—as our turnout estimates.

Hope this clarifies our rationale. I hope this was not too esoteric. Keep up the good work.

Best,
Cliff Young
Managing Director, Public Sector
Ipsos Public Affairs

The substance of the Ipsos email is well taken. I might still note, however, that focus on voter enthusiasm -- and likely voter screens more broadly -- may obscure more than inform. This isn't to say that voter enthusiasm does not matter, because it clearly does. But one need look back only to the 1998 midterms (h/t @tkbva) when Republican voters were significantly more enthusiastic than Democratic voters but the Democrats ended up posting unprecedented gains for the party of a President in his sixth year in office. As has been said elsewhere before, an enthusiastic vote is worth as much as an unenthusiastic one -- and assumptions about who is going to vote are just that: assumptions, which may be right but also may be wrong.

[UPDATE at 4:30]: A good point from desmoinesdem in the comments:

I would add that traditionally the RNC has funded a lot of the GOTV in the states, and this year the RNC is very cash-poor. If the Democrats can manage strong GOTV despite the weak economy and the Republicans don't have the money to implement their traditional 72-hour plan, the GOP will leave seats on the table.

The Republican National Committee has all of $3 million in the bank, so it's yet unclear where money for getting the GOP vote out will come from.