FiveThirtyEight reports that the progressive blog Daily Kos is suing its former pollster, Research 2000, for fraud. Analyses of the Research 2000 data have uncovered patterns that are statistically extremely unlikely to occur.
It doesn't look like any of the major Canadian pollsters are up to such shenanigans: while some of them have recurring biases (e.g. Nanos often favouring the Liberals and Ipsos often favouring the Tories, though both less so recently), it doesn't look like any of them is doctoring results in order to fake consistency. Biases are expected to arise from honest differences in methodology. You may notice that EKOS poll results are remarkably consistent from week to week, but that is due to their big sample sizes: their day-to-day results are actually very erratic, as one would expect.
In completely unrelated news: EKOS has announced that it will report polls only every 2 weeks during the summer. There will therefore not be an update this Thursday. But I'm guessing that if significant movement unexpectedly occurs, they may go back on this announcement :)
No comments:
Post a Comment