Which districts voted against the HST?
by Chris Auld
Stephen Gordon laments British Columbians’ failure to ratify the HST, which will reduce our standard of living in B.C. for many years.The case in favour of the HST was overwhelming and no expert opposed the HST in public. Which people voted against their own best interests?
The Globe and Mail’s Chris Hannay presents some graphs showing proportion voting for the HST against certain demographic characteristics at the electoral district level. Districts with higher incomes tended to vote to keep the HST:
and a similar pattern holds for education: unconditionally, districts with more educated people were more likely to vote to keep the HST. Hannay implies that these effects are really the same effect: more educated people also tend to earn more, so the education—vote correlation is just another way of seeing that higher income people voted in their class interest.
An alternate explanation draws on my post from a couple of days ago on education and beliefs that economists understand the effects of taxation on the economy: uneducated people tend to place low weight on expert analysis, certainly including economic analysis. The direct impact of the HST on prices was easy for everyone to observe—they’re printed on cash register receipts. But the indirect effects, the effects on changes in embedded prices, were hard to observe. The HST was not in place long enough to observe effects on investment and growth. One had to yield to expert opinion to draw the correct conclusion that the HST is good policy.
We can disentangle the effects of income and education with some simple regressions. The data that the Globe’s Hannay used are readily available, 2006 Census data from here, and HST vote data by district from Elections B.C. I also wanted to control for party in power, which I found listed here. I ran simple OLS regressions. Some results:
Specifications (1) and (2) confirm that the relationships between the proportion voting for the HST and income or education are substantively and statistically significant. Unconditionally, a 10% increase in median income is associated with 7.7% greater pro-HST votes (t=7.5), and a 10% increase in the proportion of people with no greater than high school education is associated with 3.4% (t=3.9) lower support for the HST (note these are percent changes, not percentage point changes).
Specification (3) controls for both education and income. The effect of income remains substantial and highly statistically significant, but the effect of education falls to less than a third of its unconditional level and loses statistical significance. However, column (4) adds dummies for NDP and Liberal MLAs (relative to independents), for Vancouver ridings, and for the proportion of the population who are immigrants. Holding these additional variables constant dramatically reduces the effect of income, which is quite highly correlated with votes in the 2009 election results, and the effect of education returns to roughly its unconditional magnitude and statistical significance. Other things equal, which party won the last election has little explanatory power; NDP-held ridings were 8% less likely to vote for the HST, but the effect is not statistically significant.
What can we conclude from these results? The idea that income drives everything has some superficial support: when we statistically hold income constant and look across districts with different education levels (as in column (3)) income and not education appears to be important, but once we consider more demographic characteristics (as in column (4)) income loses much of its predictive power. Both income and education appear to have independently influenced voting patterns. Finally, the proportion of immigrants appears to have had a large effect even after holding income and education constant, although it’s not at all obvious why, so perhaps other omitted variables explain that result.
An important caveat to keep in mind is that correlations between aggregate outcomes such as these tell us nothing at all about the underlying individual–level relationships. For example, we cannot conclude from these results that higher income or more educated people tended to vote for the HST—that may very well be true, but we cannot draw that conclusion from this sort of data. We can only conclude that regions in which there are a higher proportion of high income or relatively highly educated people were more likely to have a relatively high proportion of votes to keep the HST.