Lies, damn lies, and polls

HOW meaningful are opinion polls? Dramatic headlines last week after the November 2005 EMRS poll showed, apparently, that Tasmanian Labor is in trouble.

It is, in fact, arguable that opinion polls say anything much useful at all.

Here I will attempt to explain just what opinion polls can and can’t tell us, and in light of that see, in a considered way, where public opinion might be as 2005 comes to a close. To conclude I will offer some insight into what might influence public opinion between now and the Tasmanian election, due sometime before September 2006.

Polling methodologies and hazards

There are two organisations that regularly poll Tasmanian political views: local firm EMRS and the Melbourne based Roy Morgan International. Considering they are sampling the same population shouldn’t the results be similar? Well, not always. Consider the table below for August this year:

Morgan had a differential of 13 percentage points between the two major parties, while EMRS had 21 percentage points. How can that be from two random, presumably representative samples from the same population?

The question. Just a subtle difference in the wording of the question, or how it is asked, can cause variations in outcomes. EMRS asks, “If a state election was held tomorrow which political party would you be likely to vote for?” Morgan asks, “If a state election were being held today which Party would receive your first preference?”

“Likely to vote for” is not quite the same as “receive your first preference”. I am likely to visit my aunty for Christmas, but there again I may not. See my point?

The questioner. Another variable is how the questions are asked. Morgan conduct their surveys face to face, while EMRS use the telephone. There is a mountain of literature I won’t go into here arguing one way or the other over which method gives the most accurate result (Interpreting opinion polls: some essential details); what is important for this discussion is that the methods are different, and therefore have the potential to show different outcomes.

And both face to face and telephone are different ways of giving a response to actually casting a ballot. Do people always tell the truth when asked a question by a pollster? The voting booth is a very private place: you can do whatever you like and no one will ever know.

The sample. How representative is the sample polling organisations use? Telephone polls are going to exclude all people without a telephone, and the growing number who only use mobiles. They all vote. And we don’t know how representative of the wider population face to face interviews are either. In addition, EMRS and Morgan both have different ways of dealing with undecides and may use different calculations to weight their samples to reflect the population, or indeed they may not weight the sample at all.

Margin of error

There are also mathematical factors to consider in any sampling activity. Every survey has a margin of error that is dependent on the sample size: the larger the sample the lower the margin of error. Both Morgan and EMRS use a sample size of about 1,000.

It works like this: a survey of 1,000 respondents has, statistically, a 3.1 per cent error margin with a 95 per cent confidence level. So when a pollster finds 50 per cent Labor support, for example, it means it is between 46.9 per cent and 53.1 per cent. Possibly — there is one chance in 20 of something completely different.

The graph below charts the EMRS poll results during 2005. It illustrates how the margin of error can ascribe an entirely different result to that reported:

The bars on the data points are set on 3.1 per cent to indicate the range where the voters might be. It can be seen, for example, that the Labor and Liberal results for November overlap so it is possible, on this survey result, that the Liberals are in front of Labor. Or, if you look at the extreme points of the error margin bars for the months of August and November, it is just as possible that public opinion has not moved much at all.

The graph shows the differential between Labor and Liberal shrunk from 21 percentage points to 4 in the three months from August to November – from almost certain Labor majority government to almost certain hung parliament. Do voters really go on such a roller coaster ride of opinion? It is possible of course, but it equally feasible that the actual voting population is somewhere else within the 6.2 per cent spread around the reported data points.

Then again there is a five per cent chance it is all a load of rubbish. If you want to be 99 per cent confident, the margin of error stretches to 4.1 per cent, or an 8.2 spread around the data points.

Where are we now?

Notwithstanding, taking all the above qualifications into account, opinion polls can still be useful. Since the 2002 election it is obvious that support for Labor has softened while the Liberals’ position has strengthened. The Greens appear to be bobbing along at about a quota. The graph below shows the Morgan Poll for the three years since the 2002 election:

I added the trend lines to help illustrate the point. I think the differences between Labor and Liberal can be best explained by what I have described previously as the “Jim Bacon factor”. Between the 2002 election and August 2004 (Jim Bacon left politics in mid 2004) both Labor and Liberal tracked more or less around their respective election results: Labor in the low 50s and Liberal in the mid-high 20’s. At the three polls since we see Labor drop into the mid 40’s and the Liberals climb into the low 30s.

Of course it is possible to assign any number of reasons for this seven percentage point transfer from Labor to Liberal in late 2004. The usual bad news stories on health, housing, the environment and ministerial blunders — such is the lot of government — have occurred almost without pause for the whole time since the election. But Jim Bacon’s departure is the one seismic event in Tasmanian politics during the period. That I think is a reasonable explanation for what is reflected in the polls.

Recently I made some assessment how various primary vote results might relate to seats won. For a hypothetical election outcome of 47 per cent Labor/34 per cent Liberal there is a 12.5 per cent probability of a hung parliament and an 87.5 per cent probability of a Labor majority. That discussion can be found on Online Opinion here The Tasmanian Liberals must climb a mountain.

Where to?

So here we are, somewhere between three and nine months adrift from an election. Following are my observations on some factors that might influence public opinion up until that election day:

Governments tend to win campaigns. In Australia and elsewhere in the western world there is a tendency for incumbents to win the election campaign; that is, their position improves during the 4-5 weeks leading to election day.

Below is a table comparing the Morgan poll one month prior to the most recent Australian elections, and the eventual election results (data for Territories not available):

I am not trying to declare an inviolate rule here as obviously incumbents do not always improve during the campaign, but governments should improve during the campaign period, all other things being equal:

• Incumbency provides for more resources, authority and critical mass for conducting the campaign. This should give the edge in at least quantity in media coverage and exposure.

• Opinion polls, it is argued, understate support for the government as voters are “free” to express concern for the issues of the day. Come election day, when the vote “matters”, cautious voters return to the incumbent.

It is hard for governments to keep winning. Countering the above point is the fact that it gets harder for incumbents to keep winning government. This is sometimes referred to as the “cost of ruling”. Simply by being in office, government parties lose popularity because they risk being held responsible for the wide range of social and service delivery problems all governments face.

The cost of ruling may be one explanation for to the apparent slide in support for Tasmanian Labor. It will be interesting to see if voters think three Labor terms is too many and are prepared to punish them on ballot day.

It’s the economy stupid. Some commentators have suggested that the premier may go to an election early if the economy starts to slide. Maybe he will, but it is my view that, paradoxically, rocky economic times ahead may work in the favour of the incumbent. The reason is that the economy, as an issue, is now seen by voters to be “owned” by state Labor. The worse the economy looks, the more worried the voters become, and the more likely they are to be afraid of risking an untried opposition. Perhaps for the Tasmanian Liberals, it is better if the voters to take the economy for granted so that the election can be fought on issues more favourable to them.

You can’t win a scare campaign from opposition

When the economy is strong whoever is in government benefits. Tasmanians have gained from the Australian property boom, the result of which has been a dramatic rise in the wealth and aspirations of many “ordinary” people. The natural tendency of this risk-averse, self-interested, middle class demographic is to support the party in situ.

You can’t win a scare campaign from opposition. There is little evidence that oppositions win elections with muscular tactics. I can’t think of an election in Australia over the past 30 years which can be explained primarily by voters being swept along by opposition momentum, by a leader on a charger showing the way to salvation. Even the Whitlam and Hawke landslides can be explained just as easily by voters being tired of dysfunctional, creaking, long-term governments well past their use-by date, rather than any “It’s Time” factor.

At state level, Bacon, Bracks, Beattie and Rann beat the incumbent by being quiet and efficient, unobtrusive but creditable and, most importantly, keeping their election platforms free of wild promises or unrealistic schemes. No leap of faith required by the voter there.

On this score, the Tasmanian Liberal opposition will have the devils job selling their “majority government or bust” election tactic now that they have embarked on a scare campaign – if their press release is any guide: Labor is preparing for a Labor Green Accord unless it signs majority government pledge. The logic appears to be to frighten voters into believing that supporting Labor will lead with certainty to the “dark days”, “unmitigated disaster” and “terrifying future” of a Labor/Green accord.

Voters do not respond well to such “sky will fall in” campaigns from opposition. The danger is that a voter subsequently spooked will look for the party that can best deliver the outcome – and on majority government that is Labor. Every time the Liberals mention “majority government” they only push voters towards Labor.

The gravitas and authority that comes with incumbency means that governments can much more readily mount and win a scare campaign. Labor, I am sure, will be delighted with the opportunity to out-scare Tasmanians on the question of majority government.

Earlier, THE PSEPHOLOGIST:
Poll: the crucial Green vote