• My book, Bayesian Analysis for the Social Sciences (Wiley; Amazon).
• 111th U.S. Congress ideal point estimates (updated daily).
  Senate: PDF, CSV.
  House: PDF, CSV, scatterplot vs Obama vote share.
• Next Australian election, Centrebet prob. of ALP win: 0.77 (2009-11-24, time-series)

A blog post based on this graph (thumbnail below) headed to the Bullring.

This came in today’s Crikey (an anonymous tip):

I am a big punter and I have tried to have several bets of the order of $500 to $1,500 on a range of seats at odds ranging from $3.50 to around $1.80. The betting agencies have not taken them and even after long conversations with the betting agencies and discussion with the chief bookmakers. When the bets have been agreed, the odds have been adjusted just to get a bet of that size on. So stories of punter having $12,000 on Howard to win Bennelong are crap. I challenge any punter looking to have a winning stake above about $1,000 (thats $500 at 2 to 1 or $1,000 at evens).

Thats how thin these markets are. I’m guessing that the bookies must be worried that they are going to get creamed by staffers etc with inside information (polling, news about a scandal about to break, etc), and so a red flag goes up when they are exposed to a four figure loss. Hmmm…. The way Centrebet sat out for a day or two at the height of the Howard leadership speculation is certainly consistent with this.

EconoMan’s tale re his proposed $2,000 wager at Centrebet getting shot down to $300 is just the tip of the iceberg? Update: see EconoMan’s comment, below

I’ve been prattling on for a month now about the split between the national betting market (and the national polls) and the seat-by-seat betting markets (e.g., see here).

Barry Cassidy highlighted the disjuncture between the national betting markets (and the polls) and the seat-by-seat betting markets on Insiders yesterday:

The Coalition took fresh heart this week from a new opinion poll that closed a previously yawning gap by eight points. Newspoll now has Labor on 55 per cent two-party preferred to the Coalition’s 45.

And as a result, punters snapped up those tempting odds that were available for just a day or two.

The Coalition has been backed in from $3.70 to $2.80, the biggest movement with the bookies all year.

Still, Labor, at $1.43, remains the red-hot favourite.

But dig deeper, and there is considerable encouragement for Coalition supporters.

That’s because there is a real disconnect going on here.

Labor needs 16 seats to win government in its own right, and here they are, the 16 most marginal.

Now, the punters have Labor favourites to win 14 of those seats. The exceptions are Malcolm Turnbull’s seat of Wentworth and John Howard’s seat of Bennelong.

So, if Labor is such a red hot favourite, then they’ll be favourites to win seats beyond the 16 most marginal, right?

Well they’re not.

They’re favoured to win Blair in Queensland – but that’s it. 15 seats in total. Herbert in Queensland is line ball.

Indeed, the Herbert picture is interesting. Here are the Herbert prices, as of lunchtime Sunday, and you’ll see they’re split around 50-50 across the 3 agencies I am following:

+-------------+------------+----------+------+------+-------+---------+
| agency      | date       | division | ALP  | Coal | Other | ALPprob |
+-------------+------------+----------+------+------+-------+---------+
| portlandbet | 2007-09-23 | Herbert  | 1.62 |  2.1 |   101 |  0.5595 |
| sportingbet | 2007-09-23 | Herbert  | 1.85 | 1.85 |    26 |  0.4828 |
| centrebet   | 2007-09-23 | Herbert  |  1.9 |  1.8 |    41 |  0.4758 |
+-------------+------------+----------+------+------+-------+---------+

Averaging across those 3 agencies, we get a ALP win probability of .51, and so I count Labor tipped to win in 76 seats at this point, but only barely.

On the other hand, if you sum the average seat-specific ALP win probabilities, you get 77.86 seats, which is the expected number of ALP seats given the prices in the seat-by-seat markets at the moment. That particular market summary popped up above 75 seats on the back of that 59-41 Newspoll from 3 weeks ago and has stayed there since.

[Update: a less technical version of this post appears at the Bulletin's Bullring.] After a couple of frenized days in national politics last week, we’re essentially at the same place we’ve been for months and months: Labor well ahead and remaining on track for a victory of historic proportions.

The Galaxy poll was conducted “at the weekend” for News Ltd tabloids, and so of course we’re stuck with the maddening lack of detail one associates with a Courier Mail/Daily Tele poll report: no sample size, so we can’t compute confidence intervals, or assess whether differences/trends are statistically significant, or whatever. The Galaxy web site isn’t much help either; PollBludger has a table summarizing the last six months of Galaxy numbers. Mumble has the odd link to background docs from Galaxy that sometimes makes their way on to the Courier’s site; their August poll was [tex]n=1,004[/tex], so I think its safe to assume that this poll is also with a sample size around about 1,000.

Hence, ignoring departures from simple random sampling that are additional sources of uncertainty (i.e., quota sampling, post-stratification weighting, the assumption that preferences will flow as in the 2004 election, integer rounding), the standard error is [tex]\sqrt{.56(1-.56)/1000} = .0157[/tex] and so a conventional 95% confidence interval around the published result is [tex]56 \pm 1.96 \times 1.57 = [52.9, 59.1][/tex].

Note that almost all of the confidence interval lies above Labor’s best 2PP result in the post-1949 era, the 1983 result of 53.2%. Indeed, given this poll result and the assumptions stated above, the probability that Labor 2PP support lies above 53.2% is .963.

Under the implausible assumption of uniform swing, Labor needs 52.11% 2PP to win a bare majority (76 seats) in the House of Representatives (the resulting 4.85 percentage point swing is just enough to bring Dobell over the line, i.e., after the 2005 redistribution, the AEC lists Dobell at a notional 45.16 ALP 2PP). Given the Galaxy poll result and the assumptions stated above, the probability that Labor 2PP support lies above this 52.11% threshold is .993.

LOL site. Yeah, ok, you can over do anything. But in the sleek sans serif world we’re in (like this blog layout!), a little Copperplate pizzazz might go a long way (but only a little). Just be careful how you mix and match (i.e., the Seabiscuit poster at underconsideration is kinda nice, the high contrast against the dark background nuking the serifs, at least at this resolution).

Following on from the comments here, check this out:

+-------------+------------+----------+------+------+---------+---------+
| agency      | date       | division | ALP  | Coal | Other   | ALPprob |
+-------------+------------+----------+------+------+---------+---------+
| portlandbet | 2007-09-20 | Ryan     | 3.15 |  1.3 |     101 |  0.2895 |
| sportingbet | 2007-09-20 | Ryan     | 2.75 |  1.4 |      26 |  0.3257 |
| centrebet   | 2007-09-20 | Ryan     | 5.25 | 1.12 | 17.2208 |  0.1669 |
| portlandbet | 2007-09-21 | Ryan     | 3.15 |  1.3 |     101 |  0.2895 |
| sportingbet | 2007-09-21 | Ryan     | 3.05 | 1.33 |      26 |  0.2932 |
| centrebet   | 2007-09-21 | Ryan     |  3.5 | 1.25 | 17.2208 |  0.2498 |
+-------------+------------+----------+------+------+---------+---------+

Nice work EconoMan.

And an elaboration at the BullRing.

Ross Gittins (SMH) on the futility of a lot of proposed remedies to housing issues:

Because the fundamental cause of hard-to-afford prices is demand exceeding supply, the only genuine solutions involve either increasing supply or reducing demand.

What don’t work are efforts to make high prices more affordable by increasing the first-home owners’ grant, cutting stamp duty, introducing shared ownership schemes or a subsidised saving scheme.

All those measures would work if you were the only person who benefited from them. That is why they sound like they would help. But because all the other would-be home buyers you are competing against also benefit, the attempt to make prices more affordable ends up pushing them higher.

Gittins goes on to talk about how Australia’s tax code is written to encourage investment in homes etc, how removing some of these advantages would do a lot to ease the affordability crisis, and that any such proposal is politically DOA.

There is a broader point here, and something that I think partially underlies the government’s problems at the moment. The fact that my house has appreciated by an astronomical amount does me no good if everyone else’s house has done the same, unless perhaps I am contemplating exit/retirement to a cheaper housing market. Of course, it could be worse: you could be a renter.

Sturt has been offering a very tidy arbitrage opportunity for the last four or five days. Sportingbet has Chris Pyne at 1.80, but Centrebet has Labor at 2.95. I sound the usual caveat about Centrbet paying on the candidate, not the winning party, and the fact that all this arb stuff ignores the possibility of a minor party or independent candidate getting up. Thats over 11% profit…

You can probably also ignore the fact that Centrbet has the Liberal candidate listed as Christipher Pyne (thumbnail below).

Centrebet is also offering prices in Victorian seats; I’ve picked those up in my summaries as of today.

Headed for the Bullring, but they are having some technical issues, so I’ll post here…

A statistically significant change recorded by Newspoll, from 59/41 two weeks ago, to 55/45 this past weekend (the four percentage point change is right on the cusp of what counts as a statistically significant change in the national 2PP support levels, given the sample sizes used, and a conventional 95% level of statistical significance). And the market has moved a little in response (see the table below); averaged across the 3 agencies I am following, Labor’s position has eased to roughly where we were ahead of last week’s meeting of the Federal Parliamentary Liberal Party.

The other bit of news from the betting markets is that Labor remains favoured to win in 76 seats, but with Deakin (VIC) replaced by Herbert (QLD) as the 76th seat over the line. Again, the disjuncture between the thinner seat-by-seat markets and the national market remains quite striking, despite the slight easing in Labor’s heavily-favoured position in the national market.

National Prices as at Noon, Tuesday, Sept 18, 2007:

National Prices as at Noon, Monday, Sept 17, 2007:

Paul Sheehan maintains that last December’s departure of Arthur Sinodinos, John Howard’s Chief of Staff for most of Howard’s tenure as Prime Minister, underlies why the government is in such dire straits at the moment (from the SMH):

I’m betting this election will be on November 17 or November 24.

Why it will be held later, rather than sooner, and why Downer bungled [taking "soundings" on Howard's continued tenure after *that* 59-41 Newspoll], and why someone in cabinet was treacherous [leaking to Sky News], and why Peter Costello will not be leading the Government into the next election, are all downstream by-products of a miscalculation made last December.

On December 8 the Prime Minister announced the amicable departure of his trusted, long-time chief-of-staff, Arthur Sinodinos. Howard knew what he was losing. “Arthur has been a wonderful adviser in every way. His policy advice, his political nous and his great interpersonal skills have all contributed in a substantial way … I will miss Arthur greatly.”

He has missed him greatly. Sinodinos ran Howard’s office for nine years, and neither Howard nor his Government has recovered from his departure. The Prime Minister’s office became a bunker.

Similiar sentiments are in Malcolm Colless’ piece in the Oz blog pages (with even less detail, if thats possible).

Two reactions:

1. I tend to think the weight attached to these “trusted advisers” is overstated. At the very least, I think the ways in which they add value to an administration, a campaign, etc, is usually not sufficiently elaborated. Sheehan’s trying to say a lot in a short piece, but nowhere does he tell us what it was that Sinodinos actually did that that makes him so sorely missed, say contra his successor, Tony Nutt. Details, please…? I’m not disputing that Sinodinos’ departure hasn’t “hurt” the PM in some way (indeed, this is something you pick up talking to people in Canberra etc); just that I’d like a little more convincing before I’d accept the proposition that today’s problems for the government are the “downstream by-products” of Sinodinos’ departure.
2. Strategic exit? Have things gotten bad because Sinodinos left, or vice-versa? Indeed, maybe advisors don’t make that much difference o the electoral success of a PM or a government. But, because “good advisers” do possess at least some political nous and access to a steady stream of polling data, they are certainly well placed to make it look as if they are valuable, if for no other reason than they know when it is time to move on. Again, I’m not claiming that advisors don’t/can’t make an important contribution to a government’s success, just that I’d like more detail please.