Who will win the next election? That is the $64, $64 thousand or $64 billion question (depending on which generation you represent), and is the question on which political pollsters like to think they provide some insights. Which election are we talking about? If you are American, it is clearly the 2020 Presidential election, but if you live in Canada, it is probably the general election on October 21 of this year. If you are British, we’d be talking about the election that Boris Johnson may call to sort out the Brexit mess, and if you are Australian you might be thinking about the most recent election where, once again, the actual results confounded the pollsters. In fact, when it comes to political polling and the “next election”, we could be talking about any country where public opinion actually counts (i.e. not China) and where reading the pulse of the electorate is part of the political process. Not that pollsters get it right most of the time these days.
I referred to last May’s Australian election where the polling firms consistently had the Labor Party winning. The electorate did not agree and returned Scott Morrison’s Liberal Coalition government which had been written off by most of the polls. In 2016 the polls in the US had Hilary Clinton consistently ahead with the New York Times national polling average putting her chances of winning at 84%. We know the result. Brexit was not supposed to pass in the referendum held in Britain, according to most polls, yet it squeaked through with 52% support. Polling is a fraught science these days with a number of notable lapses. We are all familiar with the old refrain that such and such a poll is accurate within 3% nineteen times out of twenty. That is the escape hatch and seems like quite a bit of wriggle room, but it is insufficient cover when the poll gets it totally wrong.
The pollsters of course work very hard after the fact to identify why and where they went wrong. All sorts of theories have been advanced from the inaccuracy of telephone polls, to the inability of the younger generation to make up their minds, to the natural proclivity of many to deny that they will support someone who may be perceived as a divisive or unpopular candidate. Then there is the old adage that the only poll that counts is the one held on election day. People can make unpredictable decisions when they walk into that polling booth and really consider their options.
That is why each pollster has their own tricks of the trade to try to eliminate as far as possible the uncertainties inherent in polling and sampling. Credibility is what it is all about. If you can’t predict the results of a popular election, how accurate would your polling be when it comes to determining what kind of soap people will buy, or whether they prefer Coke or Pepsi? Pollsters live by their own credibility and while sometimes political polls are commissioned, and the results closely held (i.e. internal polls), many political polls are done as part of the political process leading to an election, and are made public. This provides the pollsters with free advertising and name recognition. Getting those polls out into the public eye gives a polling firm publicity which can be used to leverage the other more specialized commercial polling and market research that is their bread and butter. However inaccurate political polls could undermine their brand.
The fact that each pollster has their own “secret sauce” to try and improve the accuracy of results is an important differentiator. The innovations and techniques that they use to filter out discrepancies and reduce the margin of error becomes their own intellectual property (IP), and that IP can be protected by copyright. This brings me to the nub of the issue being played out in Canada in a dispute between the public broadcaster, CBC (Canadian Broadcasting Corporation) and some political pollsters with regard to the use of their polls to predict the forthcoming October election.
According to coverage in Blacklock’s Reporter, a specialized online policy publication in Ottawa, a number of Canadian pollsters have demanded the CBC stop copying their research without permission or payment, and aggregating it into a new product. Blacklock’s reported that the CBC’s pollster, Eric Grénier, claimed that lifting poll results from public websites without payment is performing a “public service”. (“We have the right to decide how we’ll use public information”). Nik Nanos, Chair of Nanos Research Group, a large polling firm, disagrees. Blacklock’s quotes him as saying, “
The issue is further complicated by the fact that Nanos does commissioned polling for CBC’s rival, private network CTV.
Another pollster, Frank Graves of EKOS Research, is quoted as saying;
“I would really prefer they weren’t using our data for purposes we don’t agree with…They are making a living off this. The CBC is making revenues from our work…They were aware we didn’t want them using it…There should at least be consent to use it. We never even got to the stage of permission, let alone payment or discussion on use of the data. No one even asked…There is an art and a science to doing this…It’s not something that magically comes out of a computer. It costs a lot of money. It takes experience, effort and skill in assembling sound polling data. That’s our intellectual property, and it should be treated like any other copyright material that cannot be reproduced without written permission.”
The CBC of course could commission its own research, but generally does not do so. Rather it aggregates the work of others to produce its own “poll tracker”, a compilation of polls. The result is a prediction of numbers of seats likely to be won by each party. While simplistic, it is easy for the public to understand and relate to, and as a result is arguably more popular than carefully curated data from individual pollsters. And, for the CBC, the costs of collection are basically free, other than Grénier’s time.
Graves and other pollsters object to the use of their data to produce the poll tracker results, (the predicted seat count), because of the high risk of inaccuracy. Last year Ipsos, another leading polling firm, objected to similar use of its data and prominently posted the following notice on its website;
“© 2018, Ipsos Limited Partnership
This polling release and the data contained in it are the sole and exclusive property of Ipsos. They are NOT designed to support any election outcome or prediction model and no license to use the polling release or the data is either granted or implied by their publication. Ipsos does not endorse, and has no responsibility for the accuracy or result of any predictive model that incorporates this polling data. Furthermore, any use of this information to produce polling aggregations or election models without Ipsos’ written permission will be considered a violation of our intellectual property, and Ipsos reserves the right to take appropriate legal action.”
Nanos for his part has now put his polling data behind a paywall, at the symbolic sum of $4 a month. The CBC could, of course, easily afford to pay for the access but has chosen not to do so, presumably on the basis that data should be free. As a result Nanos results have been dropped from the Poll Tracker. Grénier has been reported as saying, quite rightly, that if the material is behind a paywall it is not accessible. (Contrast this with the position of anti-copyright advocate law professor Michael Geist who continually talks about a “fair dealing gap” in which he argues that it should be legal to circumvent technological protection measures, such as paywalls, in order to exercise fair dealing exceptions like news reporting or research.) I have written elsewhere (here) about the fallacy of this argument, which would allow in the digital world that which is not allowed in the offline world. There are access controls in the analog world (subscriptions, a library membership, a purchase) so why should content in the digital world be free for the hacking just because the ultimate use is a legal fair dealing?
But let’s return to the key point; is polling data protected by copyright? Is it a “literary work” that would qualify for copyright protection? According to commentary posted by the Canadian “IP boutique law firm” of Wilson Lue;
Copyright can protect original data and original aggregations of data. Originality is more likely when people — not machines — curate, annotate, select and arrange the data. Mechanically aggregating large amounts of data does not necessarily provide copyright protection.
US copyright law is clear that facts are not copyrightable. This includes basic data such as weather data (average temperatures, annual precipitation etc.), geographic data (elevations, distances), population data and so on, but compilations of data (in other words, databases) can be protected. Surely data collected and interpreted using proprietorial techniques and trade secrets, and compiled and presented in a unique format, will qualify for protection. The fact that it is publicly available is irrelevant, despite the claims by the CBC’s Grénier that because the polling data is “public information” it is free to use. Just because something is publicly available does not mean that the copyright owner has given up their rights. Think of a copyrighted photograph that appears on the internet. While it may be on the internet and not behind a paywall, this does not mean that it can be freely used. Notwithstanding this, however, copyrighted materials can be used for certain specified “fair dealing” purposes without permission of the rights-holder.
The CBC could argue that it is using the polling data for news reporting, an allowed fair dealing exception. However, it is doing more than merely reporting on the results of the polls; it is aggregating and manipulating them and in the process providing a new interpretation of what the data mean. Is this going beyond fair dealing? Indeed is it a transformational use of the original data, to use a test that has often been applied to determine fair use in the US? These are all good questions to which we are unlikely to find the answers. Despite the accusations of copyright violation, I doubt if any of the unhappy pollsters are going to take the CBC to court, or seek an injunction to stop the use of their data. Rather they will try to bring pressure on the CBC through the media and the court of public opinion.
When it comes to public opinion, what interests the public most is, “who is likely to win?” Polling gives the electoral process a horse-race like allure, although some have argued that it can also influence voters to stay home (if they think there is no chance of their candidate winning). Equally however, a candidate who is considered a slam-dunk may also suffer from low turnout from their supporters. And a tight race is an incentive for all parties to work to get out their vote. So polls have upsides and downsides. One thing they don’t have, however, is any guarantee of accuracy despite decades of fine-tuning and polling research.
It is that criticism that is making the pollsters particularly unhappy with the unauthorized use of their work by the CBC. They are concerned that it sets them up for yet another failure. Nuanced questions and polling don’t easily lend themselves to the “who is going to win X number of seats” scenario. That scenario plays well publicly, which is why the CBC and others promote stories based on polls, but in terms of predictions and outcomes, it is still true that the final poll is the one that ends up in the ballot box.
Copyright will neither prevent nor enable the pollsters to predict with 100% accuracy who is going to win. Nonetheless, the pollsters who have sunk so much money and effort into producing results that they hope will be as close to accurate as possible are surely owed the same respect with regard to their copyrights that the CBC exercises on behalf of its own programming.
© Hugh Stephens, 2019. All Rights Reserved.