A History Of Bad Polling 

October 1st, 2015 | D. Stone 

polls

A strange phenomenon has defined opinion polling in Canada over the last two elections. It happened in 2011, it happened in 2008 and it's happening again in 2015. In 2011, Eric Grenier nearly perfected the technique of averaging polls to determine a more accurate outcome on election night. This is something that very few individual polling firms have ever been able to accomplish. Accuracy is the driving principle for most polling firms, but there have been a few that have seemed driven by something else all together.

Achieving near 100% accuracy in the election polling business is almost impossible in Canada. This reality seems to have given some polling firms a different priority. One of those popular polling firms is EKOS, which is piloted by Liberal Party donour and sympathizer, Frank Graves. Instead of bothering with trying to achieve accuracy, Graves' firm has seemed more interested in experimenting with something called “the bandwagon effect”. This means convincing people to jump on or off a certain political bandwagon, which can be achieved by either inflating or deflating a certain party's perceived support. We can't know for sure what happens behind the closed doors of EKOS, but we can look at the firm's history of inaccuracy.

In 2008, just one day before the Federal Election on October 14th, Angus Reid predicted that the Conservative Party would win 37% of the popular vote. Another poll, by a now defunct company called Asking Canadians, put Conservative support at 38%. EKOS put the Conservatives below 35%, among the lowest of all other polling firms. The Conservatives went on to win the 2008 election with 37%. Although the difference falls within the margin of error, this wouldn't be the last time EKOS would low-ball Conservative support. In 2011, the numbers were far more suspicious.

In 2011, just one day before the Federal Election on May 2, Forum Research pegged Conservative support at 36%. Both Nanos and Harris-Decima predicted Conservative support to be 37% and 36%. The odd poll out belonged to EKOS, which put Conservative support at only 33.9%. The Conservatives went on to win the 2011 election with 39.6%.

Leading up to election night, dating back to April of the same year, EKOS was the only polling firm that refused to give Conservatives 35%, while all the other firms were pegging the party's support between 36% and 38% as far back as March.

To make 2011 even stranger, it was EKOS that first began suggesting an NDP surge in the middle of April. It's important to note that no other polling firm was suggesting an NDP surge until EKOS did first. The first initial EKOS polls in mid-April suggested a Liberal-NDP tie, while Harris-Decima, Nanos, Innovative Research and Abacus all pegged NDP support between 18-22% and Liberal support between 26-28%.  By April 24th, the real Orange Crush had started. We can't fully attribute this to EKOS or confirm that EKOS may have single-handedly ignited the Orange Crush, but we can confirm that EKOS was the first firm to suggest a consistent NDP surge when others weren't. This only raises red flags because EKOS seems to have been completely incapable of similarly predicting Conservative support in the past.

But Frank Graves is a Liberal Party supporter, so he wouldn't want the NDP to surge, right? Wrong. As most of us know, the one thing that Liberals and New Democrats have a passion for is defeating Stephen Harper. By mid-April of 2011, it was becoming more and more obvious that Michael Ignatieff's version of the Liberal Party would never gain enough traction to defeat Harper. 

Jack Layton has always been a popular political figure in Canada, but his charisma was never a match for the left's overall fear and hatred of Stephen Harper. Over and over again, the NDP's low poll numbers failed to convince Canadians that the NDP could ever have a chance. The late NDP surge in opinion polls erased that doubt.


The Forum Anomaly


On August 24, Forum Research released one of the most bizarre and unrealistic poll results of the 2015 election campaign. Their interactive voice recorded phone survey of 1,440 people put NDP support at 40% nationally. No other polling firm had results anywhere near similar to Forum's results.

Forum's news release that day made the claim, “NDP on track for majority government – Conservatives tumble to third place.” But a big problem would begin to emerge over the following week. No other polling firm would ever end up corroborating Forum's claim that the NDP was holding 40% of the public's support. The very next day, Ispsos-Reid and Innovative Research pegged NDP support at only 33% and 32%. As the week went on, Nanos, Abacus and Leger would confirm that NDP support was no higher than 31%.

On September 1, Forum began to slowly back-track on their NDP claims, not by immediately releasing accurate results, but by slowly knocking down their NDP numbers. Their next news release on September 1 made the claim, “NDP lead slips, gap with Liberals closes – Minority government with Liberal opposition seen.” Their new results put NDP support down to 36%, still far higher than any other polling firm.

It wasn't until September 15 that Forum Research would bring its NDP numbers down to earth and up to par with other polling firms, placing the party at 30%. Their news release for that day said, “Federal NDP falters, Conservatives take lead.”

There's no indication as to why or how Forum came to its unusual numbers in August, just as there is no indication about the exact political leanings of the company's CEO, Lorne Bozinoff. The only certainty is that Forum was dead wrong on its NDP numbers in August.


The CROP Hoax


In September, an unusual poll by CROP suggested that Justin Trudeau was in danger of losing his Papineau seat to the NDP. The results of the poll were leaked to media outlets by the NDP. To make the poll numbers even more suspicious, it was confirmed that the poll was indeed commissioned by the NDP. The poll was immediately put under scrutiny by the CBC and the Canadian Press.

Why would the NDP do this?


The immediate objective was clear: to rattle the Liberal leader and put him off his game. But it also had a longer range goal: to break the three-way logjam in voting intentions by persuading Canadians not to waste their votes on a party whose leader can't even win his own seat.” – The Canadian Press, September 19, 2015


This blatant use of skewed polling data to sway public opinion was on full display and confirmed by media. This immediately brings to question the entire polling industry. The CROP incident also pushes us to demand further scrutiny of polling results that have been commissioned and paid for by third parties with specific goals.

It was determined by other pollsters who were hired by media to analyze the CROP data that the firm had over-sampled NDP voters in the Papineau riding. This begs the question: how many other firms have purposely over-sampled a certain demographic or voting block to skew their own results?


This Election


EKOS seems to have changed its course following two failed attempts at predicting and influencing voting intentions in Canada. Their most recent poll puts Harper's Conservatives in the lead at 35%. Whether this is an attempt to scare the Harper hating hordes into choosing a side, or whether the results are truly accurate, has yet to be known. If the results are accurate, they would make EKOS the first honest poll to predict a strong Conservative lead in this election.

Looking at historic election results leading up to election day by polling firms, it appears that each polling firm has had its own objective to influence public opinion. Or they've been accurately gauging the indecisive moods of Canadians.

In 2006, 2008, 2011 and 2015, one common trend suggests that pollsters might tighten their techniques and results in order to get more accurate numbers closer to election day. It also suggests that pollsters try to release more accurate numbers closer to an election in order to preserve their reputations, after spending months playing with the numbers and experimenting with the bandwagon effect.

The most accurate pollster throughout the 2011 election campaign was Nanos. The firm reflected a steady lead for Harper's Conservatives (37% on average) from April through to election night on May 2. In 2015,the numbers are much tighter and suggest that Harper's Tories could, in fact, be in a rut. This doesn't necessarily mean that Nanos is being honest this time.

EKOS's results were off by 6% in 2011, well beyond the margin of error. Perhaps this time EKOS is trying to improve its reputation – or its current Conservative lead has a different intention. Considering a left-wing campaign to strategically oust Harper, this poll could cause anxiety among the unions and organizations looking to unseat the Prime Minister, creating an effective surge in anti-Harper votes leading up to October 19.

What we do know is that the polling numbers of each pollster will tighten and become more accurate as we approach October 19. The negative press and media coverage of pollsters has been a dampener on public faith. This could be encouraging pollsters to salvage whatever might be left of their reputations. The importance of polling accuracy plays an important role in keeping the companies in business, as most polling firms spend time between elections contracting their services to private corporations that want to develop better public relations strategies. Corporations across Canada use Nanos, Abacus, Angus-Reid and other firms to gauge public opinion on several subjects, so poor performance could affect each company's bottom line.

We can only hope that the public's waning faith in pollsters forces a change in their methods. On the night of October 19, all eyes will be on Nanos, EKOS and Ipsos-Reid. Election night could be the night that makes or breaks either one of them.