The result of the electoral reform referendum in BC was a disappointment to advocates for proportional representation (PR). With 61% of voters voting for first past the post, the referendum produced a clear mandate to stick with the current electoral system. This support was not only large, but it was spread across the province. In my previous post, I note that first past the post had broad support everywhere except for the centre/east-side of Vancouver and Victoria. In this post I look at some of the factors that led ridings to vote for PR. Predictably, ridings that supported the NDP or Greens were more supportive, but there were also a substantial number of NDP ridings that voted for first past the post. Ridings with a greater population density were also more likely to vote for PR, though not by very much. Interestingly more ethnically diverse ridings were more likely to vote for first past the post and there was no strong relationship between the support for second and third place candidates and support for PR.
To look at some of the factors that influenced the outcome of the referendum I compared the success of PR to both 2017 election results and demographic data for different ridings.* This is not ideal as a measure of whether PR did better with some voters than others. Survey data looking at individual vote choice would be better, but to my knowledge, survey data for the referendum is limited. In this post I walk through a number of individual variables that could predict referendum outcomes and then look at them together to compare the impacts of the different variables against each other. I look at support for the NDP and Greens in the previous election, the percentage of voters in a riding that speak a non-official language (unfortunately BC stats does not have immigration or visible minority population data for different ridings), population density, and support for second and third place candidates.
NDP Vote Share
Because the referendum campaign was so partisan (with the NDP and Greens supporting PR and the Liberals approaching it) one would expect support for the NDP and Greens within ridings to match support for PR. This turns out to be the case. The upward sloping line in the graph below shows that the more likely a riding was to support the NDP in 2017, the more support in the riding there tended to be for PR. The “x’s show each riding’s support for PR and for the NDP. The fact that all but two of the ridings that voted in favour of PR were also ridings where the NDP won at least 40% of the vote highlights the extent to which support for the NDP tended to match support for PR.
At the same time, the graph shows a large number of outlier ridings where NDP support did not translate in to support for PR. These ridings are shown by the “x”s below the trend line on the right side of the graph. Further, the correlation between NDP support and support for PR was not strong enough to lead to majority support for proportional representation in most ridings in which the NDP did well. While the trend line is upward sloping, it also shows a trend in which ridings need at least 60% support for the NDP for them to be expected to see majority support (marked by the red line in the graph) for PR. Indeed, of the ridings in which the NDP won at least 40% of the vote in the last election, only 14 (32% of all ridings in which the NDP won more than 40%) saw majority support for PR. The NDP were able to translate some of their support in 2017 into support for electoral reform, but not nearly enough to give it a chance of passing.
Green Vote Share
There is a stronger relationship between support for the Green party and support for PE. This is expected. The Green party would benefit the most from electoral reform and therefore had the strongest incentive to mobilize their supporters in favour of it. Green voters also had a strong interest in supporting PR so that they could increase their party’s representation in the legislature. The graph below shows this relationship, and that any riding where the Greens saw at least 30% of the vote saw either majority support for PR or close to majority support.
Unfortunately for advocates of electoral reform, there simply are not enough ridings in which the Greens got a large share of the vote for the correlation between Green support and support for PR to make much of a difference in the referendum. There were only 7 ridings in 2017 where the Greens got close to or more than 30% of the vote. Winning majority support (or close to majority support) in these ridings was not enough to substantially affect the referendum result.
Advocates for PR made the argument that electoral reform would benefit immigrant and ethnic minorities. I am skeptical of this argument. Geographic concentration under first past the post should make it easier for immigrants and minorities to get elected in diverse ridings and force parties to be responsive to their concerns. Nonetheless, it is still worth looking at whether advocates’ argument gained traction with minorities. Ideally, I would be able to measure this by looking at whether ridings with larger numbers of immigrants or visible minorities voted for proportional representation, but unfortunately BC Stats does not have data for the number of immigrants or visible minorities in particular ridings. As a result, I use the percentage of individuals who report a mother tongue other than English, French, or an indigenous language as a proxy for the number of immigrants and minorities in a given riding.
The graph below shows no relationship between how ethnically diverse a riding is and support for PR. The trend line is slightly downward sloping. This indicates, if anything, a negative relationship (though the shallow slope of the line suggests one should not read too much into this negative relationship). Later in this piece, I will show that when one controls for party support, density, and a number of other factors, this negative relationship gets stronger. There is no evidence here that advocates for electoral reform were able to convince immigrants and ethnic minorities to support PR. If anything, ridings with more immigrants and ethnic minorities were more likely to support first past the post.
There was a substantial amount of concern expressed during the referendum about what would happen to local representation under PR. The different systems in the referendum were designed to provide some degree of local representation, but this was easier to do in more densely populated parts of the provinces. Any proportional system would require combining some ridings together. This is less of a problem in densely populated urban areas where ridings are quite small, but is a problem in rural areas with ridings already cover a large area. Thus, one would expect that smaller, more densely populated ridings, would be more likely to support PR.
The graph below shows that this is indeed the case. While there is a substantial degree of variation and a large number of outliers, the general trend suggests that the more densely populated a riding is, the more voters in the riding backed electoral reform. Like with support for the NDP though, support in even the more densely populated ridings tended to be below 50%. This suggests that advocates of proportional representation needed to do a better job of winning over the easier voters to convince, those for whom enlarged ridings and less fine-grained local representation would have been less of a concern.
Votes for Losing Candidates
One of the major concerns often noted with first past the post is the potential for voters to end up “wasting” votes. Votes that do not go to one of the top two candidates in a riding have no impact on who wins the riding. Further, voters that vote for a candidate that finishes second may feel unrepresented in the legislature because their vote did not go towards the election of an MLA. The number of people who feel this way will not be equal across ridings. In ridings where a candidate wins with a large vote share most voters may feel satisfied with their representative and therefore with the electoral system that led that representative to be elected. In ridings where a large number of voters voted for a losing candidate a large number of voters may be unsatisfied with the electoral system that led to election of a candidate they did not like.
The graph below shows that proportional representation did not tend to receive more support in ridings with more voters that supported a candidate that finished second or third. There is a positive trend line, but it is largely driven by three ridings in which few voters voted for a losing candidate and few voters supported proportional representation. If one looks at ridings where 40%-60% of voters voted for a second or third place candidate one sees a range of levels of support for proportional representation. This, coupled with a lack of statistical significance in the tests that combine different variables shown later in this post, suggest that PR did not do particularly well in ridings where large numbers of voters supported losing candidates. Advocates of PR often argue that such systems are better than first past the post systems because they ensure every vote goes towards electing a representative. The results from the referendum suggests that these arguments failed to gain traction in the ridings where they would have been most relevant,
It is worth looking at the impacts of these different variables in relation to each other. To do this I run four multivariate regression analyses that include a couple of variations of the variables discussed above. To get at whether combined NDP/Green support affected support for proportional representation I ran two models that look at combined support instead of support for each party. To look at whether support for a third-place candidate had a stronger impact than support for a losing candidate I ran a test that looked only at support for the third-place candidate. The results for all of these models are shown in the table below. Ultimately, I found that the model using the variables discussed above, coupled with a measure for the margin of victory of the winning candidate, best explained variation in support for PR across ridings and so used that model as a basis for comparison. This is listed in the table as model 4.
The table shows that support for the Green party had the strongest impact on whether a riding voted for electoral reform. For every percentage point of support the Green party won in a riding the riding’s support for proportional representation increased by an average of 0.97 percentage points. Support for the NDP had the next strongest impact on support for PR, with a percentage point increase in support for the NDP increasing support for PR by 0.73 percentage points. Finally, density had a small, but statistically significant, impact on support for PR. For every 100 people per km2 increase in density in a riding support for PR increased by 0.16 percentage points. Ethnic diversity decreased support for proportional representation. For every one percentage point increase in the number of people in the riding that speak a non-official language as their mother tongue support for PR dropped by 0.23 percentage points. When support for the NDP and Greens are controlled for, votes for losing parties actually has a negative impact on support for PR, with ridings with a larger number of voters for second and third place candidates seeing lower support for proportional representation. This result is only statistically significant at the 90% confidence level in the final model, so I do not have much confidence in it.
The graph below shows the predicted impacts of the four most important variables for different values of each variable. The dotted lines show the margins of error for the impact of each variable for a 95% confidence level. The strongest impacts are for the two partisan variables with the Green party having a stronger impact on support than the NDP (though the differences between the two parties is within the margin of error for a 95% confidence level). In contrast to the support for the NDP and Greens, the impact of population density is quite small. Finally, the impact of minority language speakers on a riding is substantial but also smaller than that for support for parties.
An analysis of where PR won a substantial number of votes suggests that partisanship played an important role in the referendum. Ridings with more Green and NDP supporters generally saw higher levels of support for proportional representation. At the same time, the NDP did not convert their supporters into PR supporters at a high enough rate to lead to a pro-proportional representation vote in the referendum. The analysis also shows that, while population density did play some role in determining support for PR, that advocates could have done more to win over voters in high density ridings where local representation would have been less of an impediment to support for electoral reform. These results further show evidence that two arguments that advocates for electoral reform made failed to get traction. Most notably the less well supported claim that immigrants and ethnic minorities benefit from PR failed to lead more diverse ridings to vote for PR. Second, the argument that PR ensures more people elect a representative they prefer, while true, failed to get lead ridings with larger numbers of people that voted for a losing candidate to vote for PR. When examining what worked and what did not in their campaign, advocates for electoral reform should pay careful attention to a number of these areas where they could have made more gains or where their arguments failed to gain traction.
*Demographic data comes from BC stats.