…BUT SHOULD WE HAVE BETTER POLLING YARDSTICKS?
[Ottawa – May 7, 2015] After an exciting month-long campaign, the 29th general election of Alberta general election has come to a close and we at EKOS are pleased that we accurately projected Rachel Notley’s majority victory. There were, however, some incongruities between our last poll and the final election results. We overestimated the Alberta NDP by 3.7 per cent and underestimated the Progressive Conservative Association by 5.3 points. Both differences fall outside the margin of error (albeit barely). Our final poll results for the Wildrose Party, the Alberta Liberal Party, and the Alberta Party were all well with the margin of error of the actual results.
We polled continuously over a two-week period and the results were remarkably stable. Polling continued up until Sunday evening, less than 48 hours before the polls opened and there was no evidence of any late shifts. We also do not believe there was some last minute cluster of voters who jumped on the PC bandwagon in a last-ditch effort to block an NDP victory.
Rather, the culprit appears to be turnout, which was not at all unexpected. As we noted in our final release, NDP support was much higher among younger age groups, while the Progressive Conservatives held a comparative advantage with seniors. In a jurisdiction where more than two-fifths of eligible voters don’t vote and where there are enormous disparities young and old people in terms of voter turnout, it is not surprising that the race was mildly tighter than expected. Ideally, we would have weighted our final results to the demographic composition of voters in the 2012 election, rather than the population as a whole. Unfortunately, Elections Alberta does not make these statistics publicly available.
Is this really a useful measure of “good” polling?
This campaign represents the third time in a row that EKOS has successfully predicted the outcome of a provincial election (after Quebec and Ontario in 2014) and as such, we feel that we have the tools and methods in place to properly monitor the trends and rhythms of future elections. However, we would like to take this opportunity to discuss why we feel that the closeness of final polling numbers to the election outcome is not the best measure of what constitutes “good” polling. Good polling, in our view, does not focus purely on predicting the exact percentages of votes each party will receive on Election Day. Such precision is extremely difficult, if not impossible, particularly at the provincial level where two-fifths of eligible voters do not vote and there are there are systematic differences between voters and non-voters.
Rather, good polling should accurately chart the rhythms of the campaign. It should tell the reader what patterns are evolving and explain the forces underlying voter response. Although the polling industry in general acquitted itself very well during the campaign, only one other firm clearly predicted a majority government. In our view, having one’s week-old or two-week-old numbers within the margin of error of the final result without having predicted the outcome is less impressive than predicting the final outcome and understanding why it occurred.
We at EKOS believe that we comported ourselves well on these fronts. We suggested that the NDP was heading toward a majority late last week and we confirmed our prediction on Monday. Furthermore, we carefully examined the underlying trends and explained how NDP fortunes were driven by a huge and unprecedented lead with university graduates and noted the influence of the labour movement. We discussed Albertans’ anxieties over middle class stagnation and how this may have been playing into the election. We noted how the NDP’s base was essentially a traffic-light coalition of progressive party supporters. Finally, we believe that we provided a unique insight between this historically important race and the federal race which is about to begin in earnest.
Closing comments
In closing, we would like to thank the 1,544 Albertans who generously took the time to participate in our surveys over the course of this campaign. We applaud the candidates who had the courage to throw their hats in the ring and we congratulate Rachel Notley’s on securing a majority government. We congratulate the voters who had the commitment to actually show up and be counted. We also thank our sponsors at iPolitics who have staked out an interest in a more ambitious program of research. We also acknowledge the skill and commitment of our colleagues and competitors who enter the arena of political polling in an era of diminished resources and rising methodological challenges.
Click here for a PDF version of this report: Full Report (May 10, 2015)
One of the most accurate polls makers that I enjoy following.
In fairness, it seems every pollster underestimated the degree of PC support without it having a major impact on the actual results.
In the recent UK general election the polls were completely wrong this time. 37 per cent of people who voted elected a majority Conservative Government whereas the polls ( How many companies I do not know) generally suggested a Labour minority government . This inaccuracy stems, in my opinion and yesterday in the opinion of Andrew Coyne of the Edm Journal and National Post, from the fact that in the First Past The Post (FPTP) election method or system of both Canada and the UK the per cent of popular vote cannot produce the per cent of seats and hence the number of seats for each party and hence the majority or minority status of a given party in the final result.Would your job be much more accurate if the electoral system was based on Proportional Representation such that the per cent of seats held will be made to closely match the actual popular vote per cent.