Could economics have predicted the outcome of the 2001 federal election?

One of the greatest, if under-rated, economists of the twentieth century was Harold Hotelling. He had a phenomenal insight into the nature of competition. In essence, his view was that real world competitors would compete next to each other - in the same space as it were. Further the allocation of resources that we see will deviate, often substantially, from that which an omnipotent social planner would have made. So, for example, Qantas and Air New Zealand compete on the trans-Tasman routes by flying "wing tip to wing tip". Aeroplanes take off within minutes of each other and fly to the same destinations and land within minutes of each other. Hotelling would view this a competitive outcome. Similarly, in politics the various political parties would compete in the same ideological space. In political terms competition would manifest itself by a number of largish political parties vying to occupy the centre ground. In a political system such as ours we anticipate two parties occupying the centre. Parties and voters arrange themselves in "ideological space". If the Hotelling view of competition is correct then we would expect that voters would vote for the party "closest" to themselves. The rules of the game for political parties is to locate in ideological space in such a way as to snare marginal voters from the other party while not losing your own heartland as it were. Based on this view (which Hotelling first articulated in 1929) the Australian political environment is very competitive. The "Tweedle Dee - Tweedle Dum" complaint that we so often hear is simply a manifestation of Hotelling type competition.

Hotelling competition can be very boring. Every now and again, however, there is an exciting election. The Federal election of 10 November 2001 was such an election. At the time there was heightened concern about global terrorism following the destruction of the World Trade Centre in New York and the attack on the US Pentagon. Here in Australia there seemed to be a perception that an increasing number of illegal immigrants and asylum seekers were being smuggled into the country. In August of 2001 the Federal government had used military special forces to prevent a ship carrying Afghan refugees from entering Australian Territorial waters. Immediately prior to the election a group of boat people were said to have deliberately thrown their children into the ocean as part of an attempt to gain access to Australia. In addition to issues of national security there was a high level of economic uncertainty. It was becoming clear that the world economy had been adversely affected by the events in the US. Ansett - Australia's second largest airline - went into bankruptcy in the days after September 11. This at a time when there had been a number of high profile corporate failures. The insurance industry was in turmoil and the tax system was in some flux. The coalition government had campaigned for a reformed tax system at the 1998 election and had won a majority of seats, but lost the popular vote. The introduction of the new tax system had not proceeded smoothly - the less polite might suggest it had been bungled - and the Labor Party had campaigned to "roll back" certain parts of the new tax system. All up there were plenty of exciting headlines, debates and issues to enliven what would otherwise have been a typically boring election.

The election saw the federal government re-elected with an increase in votes and seats. One of Australia's leading political scientists, Ian McAllister, has argued that the election was about national security and conservative governments are likely to do well under those circumstances. Lindy Edwards in her book "How to argue with an economist", describes a conversation between two "labor hacks" where one indicates a similar view to McAllister while another argues the Labor vote to be "soft". When the 2001 election is considered through the prism of Hotelling competition it turns out that a soft vote had more to do with the outcome than national security. In a paper presented at the 2003 Australasian Political Studies Association Conference, held in Hobart, we use Australian Election Study (AES) data, Hotelling competition and some sophisticated statistical techniques to evaluate the 2001 election.

The AES asks voters to rank themselves on a "Left - Right" spectrum and also to rank the various political parties on the same "Left - Right" spectrum. The ranking ranges from 0 (very left) to 10 (very right). We use this data to calculate a simple voter distance variable. Most voters (41.8 per cent) rank themselves as being in the centre. Unsurprisingly, the Liberal Party is seen as being centre - right and the Labor Party as being centre - left. The Democrats occupy exactly the same ideological space as the Labor Party while the National Party occupies the same space as the Liberals. To the extent that these parties occupy the same space we would anticipate long term decline for the two smaller parties - at least for the House of Representatives. The Greens are Australia's only "true" left-wing party. The data for One Nation is particularly interesting. About half the respondents placed them on the extreme right while the other half placed them on the extreme left. Perhaps voters confuse One Nation's economic policies (left) and social policies (right). We have not explored this result in any detail but we would anticipate very few votes on the extremes of the left - right spectrum. While One Nation may have nuisance value (especially given preferential voting) they do not pose any threat of forming government or the official opposition.

We combined the voter distance variable, demographic variables and issues variables from the AES in a statistical analysis of actual votes to investigate the 2001 election outcome. Some of the results we find are unsurprising while others are particularly interesting. For example, as voters move from left to right in ideological space so they are more likely to vote for coalition parties than the Labor Party or the Greens. Proving the obvious always increases the confidence one has in empirical models. On issues of national security, those voters who nominated Refugee, Defence or Immigration as being the most important issue of the election were very likely to vote for the Liberal Party as opposed to the Labor Party or the Greens. This may seem to support the McAllister view of the election. Unfortunately, however, a ranking of the importance of these issues for voters places them as being unimportant overall. For example, only 5.8 per cent of the electorate thought Defence and National Security were the most important issues at the election. The highest ranking issue was Education at 17 per cent followed by Taxation at 16.3 per cent. The third most important issue was Health and Medicare at 16.1 per cent. All up two of the three most important issues at the election can be described as being "traditional" Labor Party issues. While the voters may have strongly supported the Coalition government on refugees and national security, these issues were of secondary importance when it came time to cast ballots.

Of primary importance was that 55.2 per cent of voters thought the Coalition government had either done a good job or a very good job over the previous three years. So an incumbency effect was at play. More importantly, those voters who took the view that they would be better off in twelve months time and that the country would be better off in twelve months time were more likely to vote for the government. This we interpret as being a "hip pocket" effect. More voters trusted the government on economic issues than the Labor Party.

Finally, we looked at people who changed their vote in 2001 from the 1998 election. About 72 per cent of voters voted for the same party at each election. The Liberal and Labor Parties swapped (approximately) an equal number of voters. The Liberal Party picked up some votes from One Nation, but mostly picked up votes from people who did not vote in 1998. Unfortunately we are unable to explore in any detail why they did not vote in 1998 - perhaps they were first time voters. Labor lost votes. Our model predicts that individuals who changed their vote were most likely to change to Labor. In fact more voters changed to the Liberal Party. The big winner was the Greens. Our model predicts that 3 per cent of voters would switch to them when 14 per cent switched to them. We predicted that 8.7 per cent would switch to the Democrats whereas 13.7 per cent switched to them. The Labor vote was soft and while they lost some voters to the left, they also lost voters in their own ideological space.

There are a lot of numbers and statistics here and, no doubt, many readers' eyes have glazed over and some would have skimmed their way to this point. Our model is complex and we are in the process of tweaking it and making it more realistic. At this time we have made some unrealistic simplifying assumptions. For example, we have implicitly assumed proportional representation in the model. So the party with the most votes forms government, however, we know the party with the most seats forms government. Similarly, not all parties stand in all electorates. The final version may well be more complex. We are able, however, to make some preliminary yet pertinent conclusions. As much as the government won the last election on incumbency and hip pocket issues, Labor lost. They were unable to hold their own ground while picking up enough new ground to form government. To the extent that three semi-viable parties occupy centre-left to left positions in the ideological spectrum two things will happen. First, all will be denied office. Second, at least one of the parties will eventually disappear. Australia's preferential voting system will slow the agony but not prevent the shake out on the left of politics.

The data used in this paper comes from the Australian Election Study, 2001. We wish to state that those who carried out the original analysis and collection of the data bear no responsibility for our analysis or interpretation of the data.