jackman.stanford.edu/blog
• 114th U.S. Senate
• ideal point estimates svg pdf csv 12/29/15
• scatterplot against 2012 Obama vote share svg pdf
• roll call object: RData
• 114th U.S. House
• ideal point estimates svg pdf csv 4/30/16
• scatterplot vs Obama vote share svg pdf
• roll call object: RData
• Bayesian Analysis for the Social Sciences Wiley; Amazon; errata as of 3/6/15

## Sunday August 29, 2010

Filed under: Australian Politics,statistics — jackman @ 6:48 pm

Nice to see this work by Christine Neill and Andrew Leigh getting plenty of media this morning.

Andrew’s off to parliament (holding the swing against Labor to -1.1% 2PP in Fraser). His election is fascinating for those of us with geeky inclinations; I’m intrigued and quietly optimistic as how Andrew’s tenacious commitment to evidence-based, highly-relevant policy research will carry over into his new life.

A straw in the wind, perhaps? The Herald’s economic correspondent, Peter Martin, actually went into some of the inferential details underlying the research on the gun buyback effects, with a reference to “diffs-in-diffs” appearing in his article:

They [Neill and Leigh] used what is known as a difference-in-differences approach, exploiting the fact that some states withdrew guns more quickly than others and examining whether their firearm suicide rates fell faster.

A previous study had found no nationwide effect, noting that firearm suicides began falling before the buyback. However, Dr Neill and Dr Leigh found that states such as Tasmania that withdrew guns quickly had a much bigger decline in firearm suicides than states such as NSW that withdrew more slowly. Whereas the earlier study had found an increase in suicides by other methods, suggesting substitution, Dr Neill’s study found no evidence of substitution within any state.

## Friday August 27, 2010

Filed under: Australian Politics — jackman @ 5:04 pm

Labor has blown out in the Centrebet post-election market. The current price puts them at their lowest level to date, back to the 2:1 odds we saw on the day after the election.

Comments Off on Coalition favored in post-election betting

## Thursday August 26, 2010

Filed under: Australian Politics,R — jackman @ 1:32 am

We’re starting to get more divisions reporting two-candidate preferred numbers by vote type. The emerging picture (literally) is one in which Labor’s performance on the pre-polls and postals is lagging its performance in ordinary votes. On the other hand, Labor seems to be doing well among absentee voters (the regression line sitting above the 45 degree line).

These numbers are in flux, and the graph below will change a little.

Details: each panel shows a scatterplot between Labor’s 2PP percentage of the vote in ordinary votes (horizontal axis) and other forms of voting (absentee, pre-poll, postal). Â Each plotted point is an electoral division (in which the Labor candidate is one of the two-candidate-preferred candidates). Â The gray line is a 45 degree line; if Labor’s 2PP vote shares were the same irrespective of the type of ballot, we’d have the data falling on this line. Â If the points tend to lie below that line, we have evidence of Labor doing better in the “ordinary” voting. The dark line is a linear regression.

Continuously updated PDF; preview JPG below.

## Wednesday August 25, 2010

Filed under: Australian Politics — jackman @ 6:29 pm

The volatility in the post-election betting market is really something to behold. I’m plotting the Centrbet odds hourly. Labor just made it back to 50-50.
Update: 5pm AEST. And Labor went back below 50-50. Arb opportunities?

Comments Off on volatility in post-election betting market

### pre-poll numbers? (and Hasluck, Brisbane)

Filed under: Australian Politics — jackman @ 6:12 pm

Will the AEC give us a break-out of pre-poll voting?

I see we have pre-poll numbers in Hasluck, but not in Brisbane, two seat counts we’re all watching pretty closely.

Wyatt (Lib) beat Jackson (ALP) in Hasluck in ordinary 2PP votes 50.22 to 49.78, is behind on the absentee count (48.54 to 51.46), but is cleaning up in pre-polls 52.22/47.78 and postals 54.08/45.92. That one must be close to being over, I would think. The margin is going in the wrong way for Jackson, now out to 50.47/49.53 overall, and a margin of 350 votes.

We have a postal vote and absentee vote breakout for Brisbane, but no separate pre-poll numbers. Gambaro (Lib) is ahead with about a 190 vote margin (2PP), 50.3 to 49.7. Absentees and postals are going for Arch Bevis 56.5/43.5 so far; he’ll need a few more postals to maybe scrape over the line, but he could run out of postal voters to count.

Brisbane is currently the tightest seat count going and lagging a little in the counting, relative to say Hasluck or Corangamite.

Update (3.30pm August 26, AEST): Brisbane postals are now running 52.2 vs 47.8 against the Labor incumbent; a margin of about 100 votes there. Absentees about the same margin, but running the other way. Overall, the Coalition candidate Gambaro is ahead by 650 votes, roughly the lead she recorded in ordinary votes, the seat currently showing 50.5 to 49.5 2PP. That one could be sliding away from Labor. The AEC will stop calling it a “close seat” once the margin gets wider then 0.5 of a percentage point.

Update 5.26pm AEST: Pre-polls being counted separately in 18 divisions now, absentees in 37 and postals in 87. Code below the fold. Brisbane and Hasluck no longer considered “close”.
(more…)

Comments Off on pre-poll numbers? (and Hasluck, Brisbane)

## Tuesday August 24, 2010

Filed under: Australian Politics,R,statistics — jackman @ 9:48 pm

For the last two cycles I’ve done some simple regression analysis of the informal vote.Â  I saw Possum have his go at it, using a specification that is virtually the same as what I’ve run in the past (2007, 2004).

The 2010 edition follows.Â  As usual, electorate-level informality in House of Reps voting increases with (a) the number of candidates on the ballot; (b) the percentage of the electorate residing in non-English-speaking households (NESH); (c) does the state have optional preferential voting in their state legislative elections (NSW & QLD); but decreases with (d) percentage of the electorate with tertiary qualifications.

The basic linear spec gets you:

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept)  4.196338   0.304983  13.759  < 2e-16 ***
OPTRUE       1.618110   0.146803  11.022  < 2e-16 ***
neshp        0.104462   0.005328  19.606  < 2e-16 ***
Nominations  0.126415   0.045005   2.809  0.00566 **
unip        -0.139205   0.010577 -13.162  < 2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 0.8898 on 145 degrees of freedom
Multiple R-squared: 0.7977,	Adjusted R-squared: 0.7921


Not too shabby for 4 linear, additive predictors.

You can do a little better with semi-parametric terms (thin-plate smoothing splines, via the mgcv package in R) in the NESH and tertiary predictors, and an interaction with OP/non-OP:

Formula:
InformalPercent ~ OPf + s(neshp, by = OPf) + s(unip, by = OPf) +
Nominations

Parametric coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept)  4.11470    0.19677   20.91  < 2e-16 ***
OPfTRUE      1.62839    0.11146   14.61  < 2e-16 ***
Nominations  0.11768    0.03362    3.50 0.000636 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Approximate significance of smooth terms:
edf Ref.df     F  p-value
s(neshp):OPfFALSE 2.554  3.200 23.78 7.17e-13 ***
s(neshp):OPfTRUE  7.515  8.424 95.63  < 2e-16 ***
s(unip):OPfFALSE  3.057  3.769 26.34 1.81e-15 ***
s(unip):OPfTRUE   2.558  3.170 50.91  < 2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

R-sq.(adj) =  0.895   Deviance explained = 90.8%
GCV score = 0.45629  Scale est. = 0.39945   n = 150


Update: by request, the four smooth terms from the GAM.
PDF

### banning election betting

Filed under: Australian Politics — jackman @ 12:56 pm

I’m quoted in a story in today’s The Australian, under the headline “Call for ban on election bets”.

I’m quite certain that insiders are betting, and that there is a certain amount of “friends and family” betting in the thinner seat-by-seat markets.

This said, I think banning “insiders” from election betting is impractical and counter-productive. I told the reporter that, but that didn’t fit with the narrative, nor the space the story got.

If “strategic” betting by insiders moves the market, then it is always open to people possessing beliefs to the contrary to take advantage of what they see as an “out-of-equilibrium” price.

Casual inspection reveals the big national polls to be the big market movers, so I doubt that these markets are capable of being subject to extensive and/or prolonged manipulation by insiders (betting sincerely or strategically or whatever).

In addition, this isn’t like sports betting. A football match is decided by the players on the ground and the refs, maybe the coaches and selectors too. Banning them from betting (somehow) might make sense in that context. But elections aren’t like that (“teeth are harder than chalk…”); while polls and betting markets might shape expectations about who will win, Australia has compulsory turnout (you can’t legally opt out if you think your team is losing, or winning) and then some 13M+ voters decide the outcome.

John Wanna was quoted as saying that “senior public servants, MPs, candidates and anyone who is directly involved in the campaign should sign something to say they won’t bet”.

Good luck with that. Who is this set of people “directly involved” with the campaign? What about journos? What about people working at polling companies? Ad agencies? Their family members? Their friends? And we would have these people would “sign something?”

Update: and I’m sure John Wanna’s take on this is more nuanced than the short story in The Oz could convey.

## Monday August 23, 2010

Filed under: flight nerdery — jackman @ 2:57 pm

A guy at the VAT refunds office at Heathrow was trying to get a refund on 22 iPhone 4s. The HM Customs officer was apoplectic. “I have made a determination that this is a commercial export”. The guy persisted. Turns out he isn’t the one seeking to leave with the iPhones, but it is his buddy, who doesn’t speak English. Then it turns out that the buddy doesn’t have a passport. HM Customs officer blows up again. Not a bad effort, since I don’t know how you can get through security and into the VAT refunds office without a passport.

They got a VAT refund approved for 2 iPhones and pissed off.

My own story was quite uneventful. LHR-SFO, on a UAL 777 (Flightaware). I was in Heathrow a good 4 hours before departure. I watched ABC election coverage from Australia in the Star Alliance Lounge (wickedly decent bandwidth). It was a long flight back (11 hours) with unfavorable headwinds and a kink to the north once we got to Nevada, coming into the Bay Area from the north on the Golden Gate 5 STAR. This gave me a lovely view out my side of the aircraft, seat 15A, looking down on Sonoma and Marin counties, San Francisco and the Bay.

### Centrebet market down

Filed under: Australian Politics — jackman @ 9:31 am

2am Sydney: Centrebet’s Australian election market went down sometime between 9pm and 10pm Monday night and hasn’t come back up yet.

Sportingbet.com.au seems closed too, but I’m not even sure if they had a post-election book going.

Recent Betfair quotes included the ALP at 2.02 and the Coalition at 1.80.

This despite a recent ABC News online story that the market is still “powering away”.

Hmmm…

Comments Off on Centrebet market down

## Sunday August 22, 2010

Filed under: Australian Politics — jackman @ 2:19 pm

The post-election betting market kicked off yesterday with the Coalition firm favorites. Yesterday afternoon Centrebet had prices of 2.85 (Labor) to 1.40 (Coalition), corresponding to a .33 probability of Labor forming the next government.

Ar 3pm Sydney time this afternoon that is into 1.90 to 1.80 (.49 probability of a Labor win). I’m updating the prices at the top of my blog every hour; the situation is very fluid as the disposition of in-doubt seats slowly gets resolved, the correct 2PP counts get computed in places like Denison etc, and the intentions of the Independent MHRs might become clearer.

I’ll produce a time-series graph of these hourly movements too.

Update: 4pm — Labor now favored. 1.85 to 1.90.

Next Page »