Here’s my cliff notes version of “standard deviation” and why this is as important as the actual poll number: suppose several polls have tracked various candidates over time, and, from these polls, two leaders emerge. One leader’s poll numbers when plotted as a line describe peaks and valley, going up and down, and up and down again. The other leader has a similar average but this candidate’s numbers are fairly consistent, such that, when plotted on a graph, the numbers describe a smooth line. While both leaders in our hypothetical have similar averages, there is a way to gain some insight as to who is actually doing better: for each candidate, we can calculate a “standard deviation,” which is the mathematical expression of how widely values are dispersed around the average.

Using Excel, you can compute standard deviations for both candidates and, in the case of the candidate with a high average based on wildly fluctuating polls, more than likely the computed standard deviation will be “high” relative to that of the candidate with consistent poll numbers. There is no agreed upon threshold for determining when a standard deviation is “low” or “high”, but I’ve always used anything below 2.0 as the threshold for what constitutes a “low” standard deviation. In tracking elections in 2004 and 2006, I noticed that winning candidates have standard deviations at or below 1.5 *that were also decreasing over time*, even if they were consistently trailing in many polls.

This was the case of G.W. Bush in New Mexico in 2004. In that race, Bush had a strong support base as reflected in low and decreasing standard deviations even as Kerry consistently led in polls. In contrast, Kerry had a high standard deviation indicating a number of his supporters supported him only tepidly, a situation that Bush and Rove exploited successfully.

In summary, analysts and pundits typically track presidential politics like a horse race, comparing poll numbers *at a given point in time* to say “so and so” is leading “such and such.” In applying a rudimentary tool like “standard deviation,” you can obtain other equally important insights with respect to a *series of polls*, particularly on whether support for a candidate is hardening or softening.

### Like this:

Like Loading...

*Related*

This entry was posted on August 15, 2007 at 10:17 am and is filed under Uncategorized. You can follow any responses to this entry through the RSS 2.0 feed.
You can leave a response, or trackback from your own site.

## Leave a Reply