Hillary Clinton will win the Iowa caucus this Thursday by a comfortable margin over opponents Senator Barak Obama and former Democratic vice presidential nominee John Edwards. On Thursday, Clinton will garner 29.3 percent of the Iowa vote to 26.9 and 24.6 for Obama and Edwards respectively.

Since the Iowa Democratic Party’s Jefferson-Jackson Dinner in early November, Clinton has averaged 28.2 to Obama and Edwards’ 27.2 to 23.4. Clinton’s number since that dinner is an improvement over her long-term average, which since January 2007 has been 26.9 percent versus 23.7 percent and 23.8 percent for Obama and Edwards respectively.

More importantly, Clinton has managed to maintain a relatively low standard deviation of 2.7 through 34 polls conducted since the Jefferson-Jackson Dinner. Obama and Edwards’ standard deviation figures are 4.1 and 3.0 respectively. Standard deviation is a rudimentary statistical tool that measures volatility around an average. The higher the standard deviation, the greater the volatility. The lower the standard deviation, the stronger the base of support for a candidate.

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

In the three weeks leading up to the Thursday vote, Clinton has exhibited a **relatively** low standard deviation of 2.7 (18 polls), all the while improving her polling average to 29.3 from the 28.2 recorded since the Jefferson-Jackson Dinner. In short, over the last 18 polls, Clinton has exhibited a higher polling average and lower standard deviation compared to her leading opponents. It is worth noting that Clinton’s standard deviation is higher than the 1.5-to-2.0 target we have employed in the past when accurately predicting winners.

In predicting races, we construct a 2-by-2 matrix that, on the “x” axis, compares long-term versus short-term polling averages and, on the “y” axis, long-term standard deviations versus short-term standard deviations. We have found that winning candidate’s combine better short-term polling average when compared against the long-term **and** short-term standard deviation that is lower than long-term standard deviation.

In other words, a winning candidate is one who garners a greater share of votes over time and, over the same period, exhibits less voter volatility as expressed by a low standard deviation. We have also found that a second place candidate will pull out a “come from behind” victory if she or he exhibits a significantly declining standard deviation while her or his opponent exhibits high and increasing standard deviation.

If the prediction is accurate, Clinton’s victory would represent a stunning turnaround since the beginning of July when former Bill Clinton began campaigning actively in the hawkeye state. As discussed earlier, prior to July, Hillary Clinton experienced wildly fluctuating poll figures resulting in correspondingly high standard deviations. There was talk of dropping out of Iowa. She trailed Edwards badly, and even worse, Edwards throughout the campaign recorded relatively low standard deviations that demonstrated a strong support base for the former Senator from North Carolina.

Data set on which the analysis is based is available here.

January 2, 2008 at 4:29 pm |

[…] Here’s another interesting post I read today by tony_daysog […]