On Wednesday this week, a series of by-elections will be held in parts of the country. The results will be examined for what clues they might give to political trends. But history suggests that one should be a little sceptical about how useful this can be as a tool. This is mainly because while many wards will see around half of their residents voting in local elections, by-elections can be decided by as few as 4% of voters. This reduces their use in trying to predict future political shifts.

Listen to this article 7 min Listen to this article 7 min In the months before national and provincial elections, it is entirely human to try to predict the result. In the weeks before this year's elections much brain energy was spent trying to divine the accuracy of various polls.

In the end, despite all of that thinking by the commentariat, very few people foresaw the ANC losing so many votes and MK gaining so strongly.

The point is often made that by-elections are a useful predictor of political trends. This can be for important reasons.

First, they are actually elections. In other words, they are the results of people actually casting ballots, rather than telling a pollster what they will do.

As the recent elections in the US have shown, for the thousandth time, polls can be misleading for the simple fact that voters will not always say who they are voting for.

A measure of sentiment

But actual voting is different. One presumes that voters are being entirely honest about their choices (That said, voters may make different decisions in by-elections because their choices might be limited).

In short, by-elections provide the best information available between elections to...