Off the Kuff Rotating Header Image

Crosstab versus 538

You are familiar with the Congressional race projections from G. Elliott Morris at The Crosstab, which I’ve noted here and here. He uses a probabilistic model for each district. Which as it happens is also the approach taken by FiveThirtyEight in their model. You can see all those projections here. You know what that means: Let’s compare the two!


Dist   XTab    538
==================
CD02  12.9%   7.8%
CD03   5.5%   0.8%
CD06  18.2%   6.6%
CD07  52.0%  49.3%
CD10  18.1%   2.6%
CD14   5.2%   1.8%
CD17   3.5%   0.5%
CD21  18.1%  17.8%
CD22  17.4%  14.2%
CD23  84.6%  72.4%
CD24  25.3%   4.4%
CD25   9.1%   7.9%
CD27   5.6%   0.5%
CD31   9.3%  20.1%
CD32  41.4%  11.7%

Overall, 538 is a bit more pessimistic about the individual Texas races. Where the Crosstab sees CDs 02, 06, 10, and especially 24 as lower-tier possibilities, 538 basically views them as nearly impossible. The one place 538 is more bullish than the Crosstab is in CD31; I’d love to understand the particulars behind that difference. But by far the most striking difference is in CD32, which the Crosstab has as slightly leaning red, while 538 sees it has almost a lock for the Republicans, behind not only CD31 but also CDs 21 and 22 on the Dem pickup list. That’s well out of line with the view of other national prognosticators, and at least one poll. Doesn’t mean they’re wrong, just that any time there’s this big a difference of opinion it’s notable. I’ll check back on this later to see if anything changes.

Related Posts:

One Comment

  1. Joel says:

    The “beauty” of probabilistic models (for the publishers) is that you can NEVER say they were wrong.

    Which is also why they are useless for anyone seeking insights.