Today we are going to take a look at the KAP performance at the Queensland state election from another point of view, namely how the established parties’ vote in the previous election in 2009 correlated with the KAP vote in 2012. The implication is that if the KAP vote is positively correlated with another parties’ vote in the 2009 election, then the KAP is drawing voters away from that other party.
To do this we are going to run a multiple variable regression of the ALP, LNP, Greens and “Others” 2009 electorate vote shares against the KAP 2012 electorate vote shares. The former will be our variables in this regression. Now we won’t get bogged down in the technical details of multiple regression, however if the results of the regression are not fully conclusive (i.e poor Adjusted R squared, F value, coefficient not statistically significant at 0.05), we will note that here.
We should note that for the 2012 KAP vote shares, we are using ECQ data from the close of counting today. The primary vote count is almost over bar the few postal votes received today.
By running a multiple variable regression for each party we are going to obtain a co-efficient, which means that for every 1% of a party’s vote in 2009, the KAP vote would increase by the size of the co-efficient. If the coefficient for the ALP was 0.25, then on average, for every 1% larger the ALP vote in an electorate in 2009, the KAP would score 0.25% more in 2012.
Firstly we ran a state wide regression using all electorates where the KAP ran a candidate, except Gladstone where the LNP did not run a candidate in 2009. This amounted to 75 seats. The coefficients are below. In the statewide regression, all parties’ coefficients were statistically significant at 95% probability.
The coefficient are as expected, the largest coefficient was for “Others” which included Family First, One Nation’s Rosa Lee Long, Pauline Hanson and various country independent candidates, such as Wally Gleeson in Warrego, Stuart Copeland in Condamine, Dolly Pratt in Nanango, Chris Foley in Maryborough and others. This vote was concentrated in Regional Queensland (12.1% vs 5.3% in SEQ) in 2009, so its correlation to the similar regionally focused KAP is unsurprising.
The next largest positive coefficient correlation was for the LNP, reflecting the regional strength of its 2009 vote. Labor also had a positive coefficient but this was in spite of its greater vote strength in SE Queensland in 2009.
The only negative correlation was with the Greens. This is also unsurprising, the Greens highest support is in inner Brisbane, which is poor territory for the KAPs’ rural focus and social conservatism. The relatively large figure for the coefficient should not be given too much weight as its flows from a primary vote which is around 10%. The take-home message is that where the Greens did well in 2009, the KAP did poorly in 2012 and vice versa.
Taking a broader look at the state-wide regression seems to confirm one well known point, the KAP did best outside of SE Queensland. But we already knew this! So let’s drill a bit lower.
We will take our 75 seat sample and break it down to a 31 seat Regional Queensland sample (all seats except Gladstone) and a 44 seat SE Queensland sample (all SEQ seats contested by the KAP).
Again, the largest coefficient is “Others”. This reflects rural seats where voters have already broken away from the major parties in significant numbers and do not have any affinity with the left-wing Greens. Therefore for these voters the KAP is an easy fit. In other words, Bob Katter has built these voters a home.
The next largest coefficient is for Labor, its coefficient is larger than the LNP’s. This would imply that seats with higher 2009 Labor primary votes had higher shares of KAP votes on average compared to seats with higher LNP primary votes.
The LNP has a smaller coefficient than Labor, but it is still positive, this implies that 2009 LNP voters did defect to the KAP in 2012. However these LNP coefficient must be treated with caution because it is not significant at 95% probability, it is close with a p value of 0.08, but not quite there. We will keep the LNP coefficient in our analysis for completeness, however it is included with this caveat.
The Greens also scored a positive co-efficient. However the Greens coefficient is not significant at 95% probability with a p value of 0.84. Thus we have to discount the Greens coefficient in this analysis.
Putting everything together in regional Queensland, it is more probable than not, that in the pool of Regional Queensland KAP 2012 voters, there are more 2009 Labor voters than there are 2009 LNP voters. This is the case even taking into account the larger LNP vote in regional Queensland in 2009 (44.1 % vs 37.8%). However these former major party voters are joined by a large amount of 2009 “Others” voters who would be presumably conservative by not voting LNP in 2009 but avoiding Labor or the Greens.
South East Queensland
For the SE Queensland regression, we only obtain statistically significant coefficients at 95% probability for the ALP and the LNP, with the ALP scoring the larger co-efficient. Both coefficients are positive but not strongly so at 0.12 and 0.08 respectively.
As in Regional Queensland, the in the pool of SE Queensland KAP 2012 voters, it is likely that there are more 2009 Labor voters than there are 2009 LNP voters.
Based on this regression analysis, it seems likely but not certain, that Labor lost a larger share of its 2009 vote to the KAP than the LNP, once the SE/Regional Queensland divide is taken into consideration. Why is this so?
Firstly, conservatives, particularly rural conservatives which voted anybody but ALP, LNP & Greens in 2009 had already defected from the LNP. These conservative rural voters flocked to the KAP. It is worth noting that the six of the seats with the seven highest KAP vote shares had an Others vote of above 10% in 2009. The conservative defectors from the LNP to the KAP had already left before 2012.
Secondly, an old unpopular government was tossed out of office at the election. Therefore, Labor supporters were more likely to leave their party at the ballot box than the LNP supporters at this election.
Thirdly, with a federal Labor government, the LNP had no incumbent government that could alienate their supporters. While conversely, if the federal polls are taken at face value, this is what the current federal Labor government has done to its supporters. Once again there was more motivation for Labor supporters to leave their party.
So what do these root causes of the composition of the KAP vote portend for the party in 2015? We will address this in our next post.