In previous posts, I described how I constructed a measure of "economic thinking" and a measure of political ignorance. I also discussed some of the general contours of the debate about whether ignorance matters in the real world.
The purpose of my study isn't so much to explain the things that correlate with ignorance as it is to be sure that I've controlled for all of the demographic characteristics that explain ignorance when I look to see whether ignorance has independent explanatory power. To do that second part, I have to find all of the things that correlate with ignorance.
Again, I'm using the 2005 New Zealand Election Study, which includes a rich set of demographic variables. Some variables I want to include because theory suggests they ought to be in there. Lots of studies have found that men have more political knowledge than women: I then expect women to score higher on my ignorance measure than men. We might reasonably expect that higher levels of education correlate with less political ignorance, simply because there's a greater chance that the respondent will have been exposed to classroom discussions of how the political system works. Those currently students might be in a similar position. Age may matter as respondents accumulate political information over time. Occupation may matter as different occupations may be differently sensitive to changes in public policy and so the benefits of having political information may vary by occupation. The same holds for income. Membership in various social organizations may matter by reducing the costs and increasing the benefits of acquiring political information. People whose parents expressed a preference among political parties might have learned a bit about the political system around the dinner table while young. Being a beneficiary might increase the benefits of political information but probably not by as much as the costs of processing political information are increased by being in that set.
So, I throw these variables and a kitchen sink of other variables into a regression where ignorance appears as the dependent variable. The regression lets us say what the independent effect of each of the independent variables is, holding all the others constant. In the table below, the numbers tell you the amount by which the ignorance score is expected to change with a unit move in each of the listed variables. Since the dependent variable is mean zero, standard deviation one, the coefficients tell us the proportion of a standard deviation move expected from a unit change in the dependent variable.
The table above (sorry, you'll have to hit it to zoom in; I don't know how to make tables in Blogger) lists the interesting variables that were associated with increases or decreases in ignorance. As the dependent variable is ignorance, a positive coefficient tells us that the variable is associated with an expected increase in ignorance (and vice versa for negative coefficients).
What does it all mean? All else equal, males are about a fifth of a standard deviation less politically ignorant than females; those from farming backgrounds are about 0.38 standard deviations less ignorant than those employed in manual labour; those with a university degree are 0.37 standard deviations less ignorant than those with less than a secondary degree. Having the highest level of income means you're about 0.3 standard deviations less ignorant than folks with the lowest household income. Ethnicity matters too: in this sample, Maori know less about how the political system works even after correcting for income, education, and all the other things in the table (as well as a bunch of other things not listed in the table but listed in the paper). Not being able to place yourself on a left-right spectrum correlates strongly with ignorance as well. A reasonable argument could be made that it should have been a component of ignorance in the first place, but I didn't want to rule out off-dimensional self-placement as something explaining support for parties that don't line up easily on the traditional spectrum.
All of those are about what I was expecting, at least in terms of direction of the effect and relative magnitudes. What did I find most interesting? Left-wingers are far less ignorant than centrists or right-wingers. The effect is almost as big as switching from the lowest level of education to the highest. So, props to our friends from the left. On the other hand, folks who reckoned the prior Labour government had put in a good inning's performance were about a tenth of a standard deviation more ignorant than others.
In my next post, I'll tell you the effects of ignorance on "economic thinking". Sneak preview: it doesn't help your economic thinking, even after controlling for all of the demographic covariates.