Image Source: Vancouver Observer
Andy Yan with Mark Heeney
In the aftermath of the results of the Metro Vancouver transit plebiscite, BTAworks wanted to see what kinds of demographic characteristics were the most correlated to “No” and “Yes” votes for a municipality. We looked at a number of variables such as percent of workers who were reliant on transit, median household income, type of housing as a percentage of city housing stock, residential tax rates, percentages of renters and owners, education levels, and the number of registered cars per 1,000 residents and how each might or might not be correlated with the percentage of “No” or “Yes” votes for each City in Metro Vancouver.
Using citywide data from the 2011 Census and National Household Survey from Statistics Canada and the results of the transit referendum from Elections BC, we performed a basic linear correlation table for the percentage of “Yes” and “No” votes with each demographic characteristic. A special credit to Metro Vancouver for its very rich data site on the region and its member municipalities.
Correlation coefficients of above .7 are considered strong correlations, .5 to .7 are considered moderate correlations, .3 to .5 are a low correlation, and 0 to .3 being negligible (essentially irrelevant) correlations while correlations with a minus sign indicate a negative relationship between variables (eg. the more of x, the less of y). A reference guide on correlation coefficients can be found here. Moreover, the classic statistician disclaimer of “correlation is not causation” still stands. This is also tempered with the caveat that given the small size of dataset (22 municipalities) that is currently available, observations may change such as if the data points and voter turnout increases from citywide characteristics to census tract levels of observations.
Education was the variable that was strongly correlated in our study to the “Yes” or “No” vote in the referendum. The higher a city had in terms of a percentage of population with only a high school education (0.746), the higher the percentage in most cases of a No vote. Conversely, the cities that had a greater population (in percentage terms) with university education (-0.732), the higher the percentage “Yes” votes.
The moderate correlations with a “No” vote were percentages of residential ownership and residential tax rates. Municipalities with high percentages of home ownership (0.608) and high residential property taxation (0.507) had a correlation to a high percentage of No votes.
For low correlations for a “Yes” vote, municipalities with high percentages of apartments (5 stories or higher) (-0.420) as part of their housing stock and a workforce reliant on transit (-0.307), they saw higher rates of a “Yes” vote.
Interestingly, median household income (0.0518), density (-0.106), percent of population renting (-0.148), registered cars per 1000 residents (0.192), voter turnout (0.093) were unimportant, with a negligible correlation for either a “Yes” or a “No” vote.