National Journal regularly conducts an “insiders poll” of elite pundits and/or elected officials, who are offered anonymity in exchange for (hopefully) honest responses. The aggregate results are typically reported by party.

The results are generally not widely reported outside the Beltway, but the most recent insiders poll has made an unusually big splash. Here’s the key question, which was asked of members of Congress:

Drudge is currently running the headline “POLL: Only 13% Of Congressional Republicans Believe Global Warming Caused By Man…” and Kevin Drum noted that the percentage of Republicans saying yes seems to have gone down from last year:

It’s not just that Republican members of Congress aren’t convinced that global warming is a man-made problem — gotta keep those campaign donations from Exxon rolling in, after all — it’s that the number who believe this has actually gone down over the past year.

Down! What could possibly have happened over the past nine months to make them less likely to believe in human-induced global warming?

However, the sample sizes on the poll are (necessarily) tiny. We have no idea if 13 percent is accurate for the GOP caucus as a whole. Only 35 Republicans voted in the 2006 poll and 31 in the current version, and those that vote may not be representative.

Also, if you consider raw votes, the number of Republicans believing global warming is a man-made problem went from 8 out of 35 to 4 out of 31 — not a big difference, especially if different members responded each time. In terms of Drum’s comment, we can’t statistically rule out the null hypothesis that the probability remained constant between polls.

Of course, I’m troubled by these numbers on a deeper level — they’re surely far too low with respect to the scientific evidence. But we should be careful about how literally we take the results of this type of poll.