- My book: Bayesian Analysis for the Social Sciences (Wiley; Amazon).
- 111th U.S. Senate ideal point estimates:
- 111th U.S. House ideal point estimates:
- Next Australian election, Centrebet prob. of ALP win: 0.78 (2010-02-10, time-series)
This was an interesting design. Pick 4 marginals in each of the 5 mainland states, sample size of 800 across those 4 seats. Pollbludger posted a summary of the results from a WA paper.
Lets take two reads of these data, first as in terms swings to the ALP from 2004 (2PP percentage points), by state, and seats that would fall (consulting my graph of margins broken out by state).
NSW: 7.5; Macquarie, Wentworth, Lindsay, Eden-Monaro, Bennelong, Dobell, Page, Paterson, Cowper, Robertson. (10, including all four seats surveyed).
Vic: 4.6; no seats fall on that swing.
Qld: 5.1; Bonner and Moreton (2); Blair was not surveyed.
SA: 4.5; Kingston, Wakefield, Makin (3); Kingston was not surveyed.
WA: 0.8; no seats fall on that swing.
So thats only 15 gains, including Bennelong and Wentworth. Suppose Tasmania is a clean sweep for Labor, so add in 2 more. Maybe Solomon in the NT as well. Then we’re at 18 gains. Much tighter than you get given some of the projections from headline national 2PP numbers etc.
Now lets try this exercise with the absolute 2PP numbers:
NSW: 53 vs 48.1 in 2004, a 4.9 percentage point swing; 5 seats.
Vic: 49 vs 49 in 2004, no swing; 0 seats.
Qld: 51 vs 42.9 in 2004, 8.1 swing; 7 seats.
SA: 51 vs 45.6 in 2004, 5.4 swing, 4 seats.
WA: 50 vs 44.6 in 2004, 5.4 swing, 2 seats.
So thats 16 gains before we get to Tasmania and the NT. Again, thats a tighter picture than some of the 90+ ALP seat scenarios that are out there.
If you do a rough averaging of the marginal seat polling done by Galaxy and Newspoll of late, you’d have to attach high probability to Labor winning the 16+ seats it needs to get over the line. Newspoll had bigger marginals-by-state swings in Vic and Qld, in particular.
For what it is worth, my own seat count is the mid 80s, which I will post.
RSS feed for comments on this post.
Sorry, the comment form is closed at this time.
Powered by WordPress
Bad Behavior has blocked 397 access attempts in the last 7 days.
A great article. It’s interesting that the event most talked about in the dailies is the application of a national 2PP swing applied uniformly across Australia. In my 50 years following Australian elections it has never occurred. State variations are explainable by factors such as state based newspapers, state news on TV and radio, different demographic and economic changes occurring in different states, individual popularity of state governments, regional issues (such as Ringwood/Frankston toll freeway in Melbourne’s east etc. State variations are what can deliver a narrow win or a much larger win given the same national 2PP swing. In this election we have more state based info than ever before. If properly applied (given the margin of error) it allows psephologists to more closely predict the result (but things can change even in the last week – highly unlikely now).
On all figures published in the last three months (and especially the last 3 weeks)there is clear evidence that on election night we should get close to the following suggested 2PP numbers:
NSW: 54.8 vs 48.1 in 2004, a 6.7 percentage point swing; 8 seats.
Vic: 55.0 vs 49.0 in 2004, a 6.0 percentage point swing; 4 seats.
Qld: 51.3 vs 42.9 in 2004, a 8.4 percentage point swing; 8 seats.
SA: 52.6 vs 45.6 in 2004, a 7.0 percentage point swing, 5 seats.
WA: 47.6 vs 44.6 in 2004, a 3.0 percentage point swing, 2 seats.
Tas: 59.0 vs 54.2 in 2004, a 4.8 percentage point swing, 2 seats.
NT: 55.1 vs 52.1 in 2004, a 3.0 percentage point swing, 1 seats.
ACT: 66.6 vs 61.5 in 2004, a 4.9 percentage point swing, 0 seats.
AUS: 53.7 vs 47.3 in 2004, a 6.4 percentage point swing, 30 seats.
Even within each state some seats below the figure won’t fall and some above will fall but it is on a state basis that the compensation between above and below balances out. The state variances account for why the national 2PP required changes from one election to the next.
Comment by davo1943 — Sunday November 18, 2007 @ 2:22 pm
Totally agree with you davo. It’s the non-uniform nature of swings which is the key. And the reason that swings aren’t uniform is that each state, each electorate, each voter is different. The whole system is too complex to estimate in such a fine-grain. Really, this “marginal” polling thing is a nonsense. Just get the big samples from all the polls, combine them all together, look at the medium term trend, consider its place in the historical data-set, then factor in your own observation of the empirical evidence from non-poll sources such as politicians’ gaffes and the colour of their ties etc. My own prediction is somewhere between 52-53% and a narrow to modest win for the ALP.
Comment by new aussie — Sunday November 18, 2007 @ 6:43 pm