I suppose we are stuck with incessant opinion polls, especially before an election. For a long time, I've wondered how much the polling affects the way people vote when they get to the ballot box. I have suspected that oftentimes they can produce "self-fulfilling prophecies" of the outcome. In fact, one political commentator was quoted in The New Zealand Herald as saying that one reason for the low turnout this election was the fact that the polls had persuaded many that the result that eventuated on the day was a "foregone conclusion". Sometimes, of course, the polls can get it very wrong. This election they seem to have got in pretty right.
In recent days, I've been wondering whether anyone has done any analysis of the effect that polls have on voting decisions and outcomes. As it turns out, someone has. On Radio New Zealand's Sunday morning "Media Watch" today (5th December), the topic of the influence of polls on voting was discussed (Google "Radio New Zealand", look for the link to Chris Laidlaw's "Sunday" programme, and you find there a link to the programme).
A PhD student at the University of Otago, her name is Michelle Nichol (if I caught it correctly), has studied this question, and the answer is that, yes, polls definitely affect the outcome. But just to back up a minute, I should say that the discussion began with the presenter saying that there had been a record number of polls this election. And they had come "thicker and faster", partly because of the shortened period of the campaign preceding polling day. These polls, he opined, had been "partly to inform the public, but mostly to give the media something to report on in their own news outlets".
Michelle's study has shown, as I said above, that polls definitely affect the outcome of elections. She offered the opinion that Parliament would have looked different without them. There are three ways in which polls have an influence. These have to do with "the bandwagon effect", "the underdog effect", and the effect polls have on "strategic voting".
The "bandwagon effect" refers to the way in which a popular party (as shown in the polls) attracts votes to that party. Her research has shown that this effect dominates over "the underdog effect". There can be a six percent to ten percent movement towards a party in the ballot on account of the "bandwagon effect". This is especially the case if one party is well out in front, as was the case this time. The more popular the party the more strongly the "bandwagon effect" operates. So in this case, the polls do have a very strong tendency to act as "self-fulfilling" prophecies.
The "underdog effect", as the name suggests, has the result of boosting votes for a party that is perceived to be the "underdog", to be receiving a poor showing (perhaps in the public mind, an unfair showing?) in the polls. But in this case, the effect is likely to produce only a two to three percent gain in a party's share of the vote. You can listen to the broadcast for interesting commentary on how the "underdog" effect (and to a degree the "bandwagon effect") influenced the fortunes of both Act and NZ First. It seems that polling is likely to be more inaccurate in the case of minor parties, and to that extent perhaps more damaging to their fortunes.
Should polls be banned for a period (say a week or two) before an election? Michelle was inclined to think not. For one thing, it is likely to drive polling "underground", with the result that misleading polls may circulate, or that (even if the polls are a reasonably accurate reflection of public opinion) certain groups will learn of the results and others will not. Besides that, she said, democratic societies consider that one cannot ban the taking of polls, though some may wish to ban the publishing of them. But in this internet age that is pretty well impossible. Some countries, however, do ban the publishing of polls before an election: many Eastern European countries do, she said, but also Greece, Switzerland (interesting this), and Spain.
One effect of the media's delight in running polls, research has shown, is that often the media gives more space, or attention, to polls rather than the issues. Someone has done a study, and has discovered that in this election (or it may have been an earlier one) ten percent of coverage relating to the election was given to the poll results, while a further 23% was devoted to discussion of coalition possibilities based on the polling.
So, are polls a good thing or a bad thing? These are my reflections, now: as I said at the outset, they are probably inevitable, and perhaps they are helpful to a degree in helping the electorate make up its mind. I tend to agree with Michelle that it is not wise to ban polls. But what is required is to educate voters to be "critically aware" of the influence of polls; and also to be thoughtful about making up their minds, and try to arrive at educated decisions. Perhaps, cumulatively and over time, polls give a good indication of the way in which voters are likely to vote. But on individual polls, people should consider this: the sample in relation to the overall number of voters is extremely small, miniscule, almost to the point of vanishing in percentage terms. In this latest pre-election round of polling, I noticed that a number of the polls were based on a sample of 750 voters (and the margin of error posted was usually +/- 3.5% or thereabouts). Against the total number of people on the electoral roll, a sample of 750 voters works out at about 0.00002289% of the electorate. Or, if you prefer, of the 2,254,581 people who actually cast their vote on November 26th, the sample represents 0.00003326% of those who voted. Over against the other percentages that are thrown around in an election e.g. the 5% of total votes required for a party to enter Parliament (short of winning an electorate, some of which were won in marginal seats in this election on about 12,000 to 15,000 votes), those are pretty small percentages.
In recent days, I've been wondering whether anyone has done any analysis of the effect that polls have on voting decisions and outcomes. As it turns out, someone has. On Radio New Zealand's Sunday morning "Media Watch" today (5th December), the topic of the influence of polls on voting was discussed (Google "Radio New Zealand", look for the link to Chris Laidlaw's "Sunday" programme, and you find there a link to the programme).
A PhD student at the University of Otago, her name is Michelle Nichol (if I caught it correctly), has studied this question, and the answer is that, yes, polls definitely affect the outcome. But just to back up a minute, I should say that the discussion began with the presenter saying that there had been a record number of polls this election. And they had come "thicker and faster", partly because of the shortened period of the campaign preceding polling day. These polls, he opined, had been "partly to inform the public, but mostly to give the media something to report on in their own news outlets".
Michelle's study has shown, as I said above, that polls definitely affect the outcome of elections. She offered the opinion that Parliament would have looked different without them. There are three ways in which polls have an influence. These have to do with "the bandwagon effect", "the underdog effect", and the effect polls have on "strategic voting".
The "bandwagon effect" refers to the way in which a popular party (as shown in the polls) attracts votes to that party. Her research has shown that this effect dominates over "the underdog effect". There can be a six percent to ten percent movement towards a party in the ballot on account of the "bandwagon effect". This is especially the case if one party is well out in front, as was the case this time. The more popular the party the more strongly the "bandwagon effect" operates. So in this case, the polls do have a very strong tendency to act as "self-fulfilling" prophecies.
The "underdog effect", as the name suggests, has the result of boosting votes for a party that is perceived to be the "underdog", to be receiving a poor showing (perhaps in the public mind, an unfair showing?) in the polls. But in this case, the effect is likely to produce only a two to three percent gain in a party's share of the vote. You can listen to the broadcast for interesting commentary on how the "underdog" effect (and to a degree the "bandwagon effect") influenced the fortunes of both Act and NZ First. It seems that polling is likely to be more inaccurate in the case of minor parties, and to that extent perhaps more damaging to their fortunes.
Should polls be banned for a period (say a week or two) before an election? Michelle was inclined to think not. For one thing, it is likely to drive polling "underground", with the result that misleading polls may circulate, or that (even if the polls are a reasonably accurate reflection of public opinion) certain groups will learn of the results and others will not. Besides that, she said, democratic societies consider that one cannot ban the taking of polls, though some may wish to ban the publishing of them. But in this internet age that is pretty well impossible. Some countries, however, do ban the publishing of polls before an election: many Eastern European countries do, she said, but also Greece, Switzerland (interesting this), and Spain.
One effect of the media's delight in running polls, research has shown, is that often the media gives more space, or attention, to polls rather than the issues. Someone has done a study, and has discovered that in this election (or it may have been an earlier one) ten percent of coverage relating to the election was given to the poll results, while a further 23% was devoted to discussion of coalition possibilities based on the polling.
So, are polls a good thing or a bad thing? These are my reflections, now: as I said at the outset, they are probably inevitable, and perhaps they are helpful to a degree in helping the electorate make up its mind. I tend to agree with Michelle that it is not wise to ban polls. But what is required is to educate voters to be "critically aware" of the influence of polls; and also to be thoughtful about making up their minds, and try to arrive at educated decisions. Perhaps, cumulatively and over time, polls give a good indication of the way in which voters are likely to vote. But on individual polls, people should consider this: the sample in relation to the overall number of voters is extremely small, miniscule, almost to the point of vanishing in percentage terms. In this latest pre-election round of polling, I noticed that a number of the polls were based on a sample of 750 voters (and the margin of error posted was usually +/- 3.5% or thereabouts). Against the total number of people on the electoral roll, a sample of 750 voters works out at about 0.00002289% of the electorate. Or, if you prefer, of the 2,254,581 people who actually cast their vote on November 26th, the sample represents 0.00003326% of those who voted. Over against the other percentages that are thrown around in an election e.g. the 5% of total votes required for a party to enter Parliament (short of winning an electorate, some of which were won in marginal seats in this election on about 12,000 to 15,000 votes), those are pretty small percentages.
No comments:
Post a Comment