Arnold raises an important challenge to my last post:
In measuring the importance of personality traits in predicting
political ideology, I would ask how much of political ideology can be
explained by personality. The answer, if you look at the R2 in table 6 is, “not much.” For economic conservatism, in fact, the R2 ranges from .07 to .15…In statistical research, I believe strongly that you should focus on the practical significance of results. Statistical significance is not practical significance.
The low R2 means that ideology remains fairly random. But to understand the practical significance of personality for ideology, you shouldn’t be looking at the R2; you should be looking at the coefficients. Turn to Table 6B, column 1, in the Gerber et al piece. The R2 is a mere .12; 88% of the variation remains unexplained.
Consider, though, what the coefficients mean. The paper standardizes the measure of ideology so that it has a mean of 0 and a standard deviation of 1. The personality measures range from 0-1. So the .458 coefficient on Stability means that going from the minimum level of emotional stability to the maximum level raises economic conservatism by .458 standard deviations. Doesn’t that seem practically significant to you, Arnold?
Now let’s take a more extreme example. Suppose someone had the personality least favorable to economic conservatism: a 0 on Extraversion, 1 on Agreeableness, 0 on Conscientiousness, 0 on Stability, and 1 on Openness. According to the same regression referenced earlier, this person is expected to be 1.747 SDs less economically conservative than someone with the opposite profile.
I agree that statistical and practical significance are different, but this paper’s results seem to be both.
READER COMMENTS
Dr. T
Jul 6 2009 at 6:45pm
“So the .458 coefficient on Stability means that going from the minimum level of emotional stability to the maximum level raises economic conservatism by .458 standard deviations. Doesn’t that seem practically significant to you…”
For something as “fuzzy” as personality and ideology each divided into a few arbitrary categories, my answer is “no.” (As an aside, I love the use of three significant digits when the actual total error limits probably support only one significant digit.)
To look at something roughly equivalent, a .46 standard deviation change in IQ equals 6-7 points (which is within the error limits of IQ tests). IQ testing is more reliable than personality and ideology categorization. Even so, I would not be impressed about a factor that correlates with only a 6-7 point IQ difference between groups.
Comments are closed.