26 November 2009

How the Tories get 28% more votes thanks to CRA

Support for the ruling Conservatives went down in the last quarter compared to three months earlier. But you’d never know that if you read the news release from Corporate Research Associates.

That’s because CRA torques its news releases. Here’s how CRA presents its information in a misleading way.

1. Release early, for no apparent reason. CRA normally polls in one month and releases results early the next month. For the November polls, CRA has sometimes released results as late as December 12.

For some unknown reason, CRA released the November 2009 report a week earlier than usual.

Coincidentally it was right before a crucial by-election.

2. Release out of sequence. CRA usually releases the Newfoundland and Labrador results last, cycling through its other provincial quarterly results before getting to Newfoundland and Labrador. For some unknown reason, CRA released the NL results first AND posted them online before the end of business the day they were released.

3. Reporting as share of decideds boosts apparent results for Tories by 28%.

cra november 09The chart at right shows the CRA number in red and the correct number in blue.

There’s a huge difference between the two. It shows the Tory support as being 17 and 18 percentage points higher than it actual is in CRA’s polling.

Put another way, CRA’s way of showing the numbers inflates Tory support by 28%. You get that number by taking 17 percentage points as a share of 60 percentage points. In the last result, the Tory number is artificially inflated by 24% because of the dubious reporting method.

And not everyone does it. In fact, researchers shy away from this sort of reporting because it distorts results.

Just check the other pollsters and see what they do. You’ll be surprised.

4. Hide the trends. Reporting results as a share of decideds masks the real trends, or, as in the past three quarter gives the wrong trends. Tory support isn’t up and stable, as suggested by the CRA torqued version. CRA’s own numbers - presented more accurately - show support for the Tories going down.

And what’s more it has been declining steadily since November 2007.

So what would the ordinary Newfoundlander or Labradorian think if they heard that from news media instead of the torqued version? The partisans won’t care: they’ll be leaping forward to note the Tories are still miles ahead of the opposition. Anyone using that line is likely a Tory partisan or one being spun by them.

But if ordinary people had heard the whole story presented accurately over time, would their opinion change over that same time?

Bet on it.

Now there’s also a suspicious pattern of results through 2009 – varying over nine months by less than one half of one percent - but that’s a whole other issue.

5. Don’t tell what you know and can tell.

As we know from polls released through access to information in September, CRA knows a lot more about public opinion in the province than they tell.

Opinion results vary by region of the province. Opinions sometimes run differently in one region compared to the overall picture. They also vary by age, sex, education and income.

If ordinary people knew all that, perception of continued high satisfaction across the province or increasing voter support would change and odds are it would change radically.

But people can’t know since CRA hides information from the public.

6. Don’t tell all you know. The people at CRA know they ask questions on behalf of the provincial government - yes, they pay for questions every quarter - but ethically it can’t report those results. However, the people at CRA also know that information they can’t say tells a very different story from what they do say.

Did you know last August that people were actually dissatisfied with government performance on something like health care?

Well, that story in the Telegram didn’t get as wide coverage as the original torqued news release which was carried by most media, including VOCM.

CRA could find a way to tell all they know, ethically, if they wanted to.

7. Report questions you didn’t ask. CRA routinely tells you that people in the province are completely satisfied or mostly satisfied with the ruling Conservatives.

They only problem is that is an answer they never got.

CRA regularly asks about satisfaction but they use a standard break-down that gives respondents a moderate option - “somewhat” - and a high option: “mostly”.

They report two high options that CRA never asked. You can see this in Table 3b in the link above. The question is described one way at the top and another way at the bottom.

And before you try it, remember that it is very unusual for people to respond outside the range they are given.

But if they got ones outside the range, ethically CRA would have to report the full range of responses including the information on the scale as they set it up themselves. If there were no “somewhat” they’d have to say that.

But since they don’t report that way, you can be pretty much guaranteed, CRA is torquing the meaning but changing respondent answers. The moderate category “somewhat” becomes the high end category “mostly” and “mostly” becomes “completely”.

Couple that with the data they withhold – variation by region, age sex and so on - and you can get a very different picture of the province’s people and their opinions than the one offered up by CRA every quarter.

No matter what way you slice it though, CRA results are presented in a way that is misleading and in some cases it is grossly misleading.

And when will conventional news media start questioning what they are getting when the evidence of torquing is overwhelming?

Good question.

But as you can see, there are lots of ways to goose a poll.