Dead Beattie?
Glenn Milne has found some research that suggests Beattie may be in a worse position than many thought. The research was from John Black and John Lockwood of Australian Development Strategies (ADS).
If I understand it correctly, ADS undertakes ecological modelling of voting intentions by polling booth. They have developed a multiple regression model for how booths vote based on the demographic characteristics of the surrounding locale. ADS then compares the model’s predictions with the actual vote. Incumbents typically do better than the model predicts, because they have a personal following in the electorate.
In the case of the recent Gaven by-election, the sitting member, Bob Poole, under performed against the model at the last state election. He was not a popular member, and he scored two per cent less than the ADS model predicted. As a consequence, according to Milne, Black and Lockwood argue that the swing in the recent Gaven by-election should be reassessed from 8.4 per cent to 10.4 per cent. This, Milne asserts, is bad news for Beattie.
The best way to measure whether Beattie is on the way back, post-Gaven, or in deeper do-do is to try to adjust the swing against the ALP, taking onto account this personal-vote quotient. Which is what Black and Lockwood have done. They have been benchmarking sitting MPs for 30 years and, using computer modelling, have been able to estimate the personal votes of the sitting MPs in all three seats, Chatsworth, Redcliffe and Gaven. In Chatsworth the personal vote for the retiring MP was 5.7 per cent. In Redcliffe it was 3.9 per cent. And in Gaven the figure for the unpopular Poole was -2 per cent.
In Chatsworth and Redcliffe, the standing of the local candidates was pulling the Labor vote up. In contrast, Poole was suppressing the natural ALP vote in Gaven.
Without taking into account this personal-vote factor, the two-party preferred swing against the ALP in Chatsworth was 13.9 per cent. In Redcliffe it was 8.3 per cent. And in Gaven it was 8.4 per cent. Given the extraordinary circumstances surrounding Poole’s departure, this led to the dominant theory that Beattie was clawing back support after the so-called Dr Death hospitals crisis.
But what happens when you take into account the personal-vote factor? Suddenly the swing against Labor in Chatsworth is reversed to 8.2 per cent, in Redcliffe it’s down to 4.4 per cent, but in Gaven it jumps to a whopping 10.4 per cent. So much for Beattie clawing back.
I have a couple of difficulties with Milne’s argument. While I agree that the personal vote of an incumbent is lost at a by-election, one must also factor the protest vote into any analysis of by-election results. I would express this as follows.
By-election swing = loss of incumbent personal vote + protest against government + other factors.
The critical factor issue with protest votes at by-elections is that they are not repeated at the subsequent general elections. By way of example, the Canberra Liberals’ Brendan Smyth won the 1995 Canberra by-election, but lost the seat in the 1996 general election that saw Keating decisively deposed by Howard.
If the typical by-election swing were five per cent, I would typically allocate two per cent (in must electorates) to the incumbent and three per cent to a protest vote against the government. In a seat with an unpopular local member, retiring in unusual circumstances, the protest vote may well be higher. Without examining the scale of the protest vote, Milne’s argument is incomplete.
My second issue with the analysis is that it does not appear to have comprehended the changing demography of the seat. In the 2004 election there were 29,040 electors enrolled. In the 2006 by-election that number had increased to 32,223. I suspect the demographic modelling was based on data from the 2001 Census, when the enrolment in the seat was 24,357. Since 2001, it is highly likely that there have been significant changes in the population characteristics of the Gaven electorate that were not factored into the most recent analysis. Those changes may have explained the deviation between the model and the vote at the 2004 election.
In summary, while I can see the logic of Milne’s argument, I am not convinced about the reliability of the underlying analysis. The cited analysis has not independently assessed the by-election protest vote effect, nor has it considered the changing demography in Gaven.