[This is an article I wrote a year ago for The Business Bulletin.]

A few weeks ago I went into a small town branch of the Royal Bank of Canada with a cheque received for some translation work. The cheque was in US dollars and I asked the teller to convert it to Canadian dollars, give me $80.00 in Canadian cash, $20.00 in US cash (to include in a card I was sending to someone in the US) and deposit the rest to my account. I received friendly and efficient service and left without thinking any complicated thoughts about what had transpired.

A survey firm called a few days later, asking me to rate different facets of that banking experience on a scale of one to ten. I told him I couldn’t do it. I do not look at other people as machines and mentally rate their performance on a scale of one to ten. That doesn’t make sense to me.

I suspect the bank intended to use the survey results for publicity purposes, informing the public of the great satisfaction rating of the Royal Bank of Canada. How can anyone trust such a poll when the respondents most likely just pick numbers out of the air so the questioner will let them get back to their work?

Sometimes it is important to consider what the statistics are actually measuring. Do statistics of traffic violations by province measure the driving habits of the population or the enforcement habits of the police? Statistics on charitable donations per capita show Saskatchewan near the top of the list and Quebec near the bottom. These stats come from the receipted donations claimed on income tax returns. Is it possible that Quebeckers are more generous in giving spontaneously without needing a receipt? Do statistics such as these shape our opinions of the people of each province?

I believe it was Mark Twain who stated that there were three kinds of lies: lies; d****d lies; and statistics. That being said, I am a strong believer in the usefulness of statistics — when dealing with inanimate objects that can be measured or counted. I took numerous courses in statistics in preparation for writing the Certified Quality Engineer exam. I worked for many years with the practical application of statistical analysis in a manufacturing setting and I am convinced that this is the most effective way of determining what is going on in an industrial process.

We first need to understand some basic principles. The sample to be measured must be chosen completely at random, there is a margin of error to be taken into account in each sample, and an average of one time out of twenty the sample will not be representative of the actual process. In a manufacturing setting, samples are taken at regular intervals. If one sample does not fall within the range established by preceding samples it may mean that the process has changed, or it may be the one time out of twenty when the sample was not truly representative of the process. The way to find out is to immediately take another sample. If this one falls within the limits established by earlier samples, it means the former sample was not representative. If measurement of the resample gives results close to the former sample it is time to sound the alarm, shut down the process and find out what has changed. Statistical methods have done wonders in tightening tolerances and reducing waste in industrial processes.

Statistical sampling of opinion is fraught with much more complexity. First off, you are dealing with opinions, which are subjective and not amenable to precise measurement. Secondly, it is hard to obtain a truly representative sample, many people might be unavailable or unwilling to participate. Thirdly, there is no way of telling if a one time poll falls on the side of the 19 times out of 20, or the 1 time out of 20. Fourthly, many polls are conducted with leading questions designed to elicit a certain type of response. Another complicating factor arises when a newspaper eliminates the no responses and no opinions and calculates a percentage using only the remaining responses. That can raise the margin of error into the stratosphere.

I could phone a few hundred people at random with the following question: “The beautiful flowers of Purple Loosestrife are no longer seen in Saskatchewan’s wetlands. Do you think this is due to: a) global warming; b) excessive use of pesticides; c) lack of pollination due to honey bee die back; or d) a dramatic increase in the number of Canada Geese?” Many people will have forgotten, or perhaps never knew, that Purple Loosestrife was deliberately eradicated ten years ago as an invasive species and will pick one of the answers supplied. I might come up with a statistic saying that 50% of Canadians believe that Canada Geese are destroying Purple Loosestrife in Saskatchewan, but such a result would be rubbish.

Too many surveys are conducted along similar lines, giving a choice of preselected answers on sensitive subjects such as abortion and gay marriage. Then the results are fed back to us as proof of what the majority of Canadians think on this particular topic. The newspapers report the results of these surveys with a slant that indicates that those of us who think otherwise are quite out of step with the times, perhaps even hinting that we are dangerous to the public good.

Such carefully manipulated polls are voices of the zeitgeist, pressuring us to think in the approved manner of our time. We should take a step back and look at what is really behind these polls, so we can think soberly and realistically. May we never be ashamed to express those sober and realistic thoughts, they may be a breath of fresh air for someone trapped in the stifling atmosphere of the zeitgeist.

We are men and women. It should not be possible for a propaganda machine to adjust and fine tune our attitudes as if we were machines.