Tag Archives: RealClearPolitics

Election 2020: Keep Your Eye on Rasmussen Reports

Posted on by

In my previous post I contrasted the results of polling by Rasmussen Reports with two indicators published by RealClearPolitics: its the “poll of polls” and its summary of election betting markets. Although Rasmussen’s numbers (as of September 30) look bad for Trump, they’re not as bad as the numbers produced by most polls and betting markets.

Why is that?  Rasmussen’s polls yield better — more accurate — results than most other polls because Rasmussen’s polls are unbiased. It’s not surprising, therefore, that Rasmussen has an excellent track record. Many pollsters and pundits try to dismiss Rasmussen as pro-Republican, or to denigrate Rasmussen’s methods. This is a classic example of psychological projection because most polls are systematically biased toward Democrats.

There are two reasons for that. Pro-Democrat pollsters (and their media allies) don’t like to publish bad news about Democrats. By the same token, underestimating the electoral prospects of Republicans is a devious form of election-rigging: It helps to demoralize Republican voters and therefore reduce pro-Republican turnout at election time.

How biased are the other polls? On average, extremely biased. The following graph shows the relationship between Rasmussen’s polling on the 2020 presidential election and the average of the dozen-or-so polls tracked by RealClearPolitics:

If I had removed Rasmussen’s poll from RCP’s average, the result would have been more stark, but it’s stark enough as it is. Rasmussen’s (presumably) accurate poll (White House Watch) would have to show Trump leading Biden with 70 percent of likely voters before the RCP average would show Trump tied with Biden.

The moral of the story: I won’t cite the RCP “poll of polls” again.

I will however cite RCP’s summary of betting markets. They don’t estimate the split of the popular vote, but they do measure the degree of confidence that one or another candidate will win. Unfortunately, there is growing confidence on the part of bettors that Biden will win.

I will close with a reminder of what’s at stake in this election: liberty.

RealClearPolitics’ Misleading “Poll of Polls”

Posted on by

For my latest analysis of trends in RCP’s polls and several others, see “Election 2016.”

REVISED AND UPDATED 10/23/16

A lot of commentators cite the “poll of polls” at RealClearPolitics.com. You know the one I mean; it looks like this:

rcp-poll_1
rcp-poll_2

The graph is followed by a long list of historical polling results, on which the graph is based. It all looks authoritative. But it’s misleading.

Take the values for September 22, 2016, which show a spread of 2.1 points in favor of Clinton.  However, the values for September 22 represent polling that was done between September 8 and September 21. That’s quite a lag and it badly distorts what’s really happening in the Trump-Clinton race.

So I reconstructed the “poll of polls,” as follows:

  • Assigned a date to each poll that coincides with the central date of the polling period it represents.
  • Computed, for each poll, the spread in favor of (or against) Clinton.
  • Arranged the polls in chronological order, according to central date.
  • Averaged the spreads for polls having the same central date.

Because only one or two polls are assigned to many dates, I added a trendline to emphasize the pattern that emerges from the many polls included in RCP’s “poll of polls.” Here’s a graph of the result, for polls conducted since August 1:

clintons-lead-deficit-in-rcp-polls

That’s a much more realistic depiction.