Failure to sample can be dangerous to your health

Harry Shannon

Background: Municipalities in Ontario held elections last week. A couple of days before, the Hamilton Spectator published a letter complaining that an opinion poll reporting on the race for Mayor was useless. I responded and the letter was in the print edition a few days later. Below is the letter I sent; there were some minor edits in the version published.

Re: Useless poll, Letters, October 22

The letter writer claims that 679 people are less than 0.0001% of the Hamilton population. That’s wildly wrong, by a thousand-fold.

As well, if you’re doing a poll of mayoral voting intentions, you have to omit children and non-citizens. The City of Hamilton’s total population is nearly 600,000, so let’s go with 500,000 eligible voters. That means the poll included about 0.14%.

More importantly, a well-done survey can give you a good idea of overall opinions, even if it includes only a fraction of the population. Smaller surveys are less accurate and have a wider margin of error (19 times out of 20). In this case, it was 3.7%, so the proportion of people intending to vote for Andrea Horwath was 36% plus or minus 3.7%, between about 32% and 40%.

If you don’t accept that sampling can be useful, the next time your doctor wants some blood work, your logical response should be: “A sample of my blood isn’t enough. To find out what’s going on, you’d better take it all.”