jackman.stanford.edu/blog
• Bayesian Analysis for the Social Sciences Wiley; Amazon; errata as of 5/23/13

• 113th U.S. Senate
• ideal point estimates pdf csv 4/15/14
• scatterplot against 2012 Obama vote share pdf
• roll call object: RData
• 113th U.S. House
• ideal point estimates pdf csv 4/16/14
• scatterplot vs Obama vote share pdf svg
• roll call object: RData

## Saturday July 31, 2010

Filed under: Australian Politics,statistics — jackman @ 5:44 pm

1.54 to 2.40 (Centrebet), implied probability of an ALP win now down to .61:

I conjecture that if Newspoll gives a similar result to the Nielsen poll, then we’ll see the Coalition become the market favorite.

In fact, we’re already at a point where one could insure a “long-odds” position one might have taken on the Coalition not so long ago, say, when they were trading around $4.00 or higher. One could have bought the Coalition at$4.25 at Centrebet as recently as July 25.  Today Labor is at $1.54 (with more easing to come, I suspect). The under-round there is substantial: 1/4.25 + 1/1.54 = .88. In turn, this means you can insure a$4.25 bet on the Coalition — assuming you had the gall or foresight to make it — with a bet on Labor, locking in a minimum profit.  A $100 bet on the Coalition at$4.25 gets you a $425 payout ($325 profit) if they win.  With Labor at 1.54, you could place $200 on Labor and profit (a)$425 − $100 –$200 = $125 if the Coalition wins, or (b) 1.54*$200 – $300 =$8 if Labor wins.  Either way you are not losing money, provided you got on the Coalition at odds as long as those.

These arbitrage scenarios are hardly breath-taking, but might get more lucrative if Labor’s price continues to ease.

And note that the particular situation I’m describing is only available to punters who took the Coalition at long odds.  It is quite possible that these people have the courage of their convictions and don’t see any need or value in laying off like this; they’d rather (a) the lottery over $325 profit and a$100 loss versus (b) tieing up an extra $200 for the lottery between$125 profit and $8 profit. Indeed, if you thought the true probability of a Coalition win was .5, then the expected profit of your$100 bet at 4.25 is $325/2 –$100/2 = $112.5, while the expected profit flowing from buying$200 “insurance” is $125/2 +$8/2 = $66.50. Then again, if you really thought the Coalition was even-money, then you know something the betting market doesn’t, which at the (current)$2.40 to $1.54 is equivalent to a .39 probability of winning. In this case perhaps the best thing you could do is to get (some more?) on the Coalition. Last thought on this for now: on Dec 2nd 2009, at the height of the Liberal leadership kerfuffle, the Coalition reached$5.15.  If you took that position you have even more headroom for locking in risk-free profits.  Congrats.

## Friday July 30, 2010

Filed under: Australian Politics,statistics — jackman @ 11:36 am

It is 3.45am in Sydney, but the Nielsen poll has been out on web sites etc for a good couple of hours. 48-52 2PP with n=1,356; Labor’s primary vote is down to 36%.

The usual caveats apply: the nominal sample size isn’t the effective sample size, due to weighting etc; rounding; computing the 2PP assuming preferences flow as in the last election. So there is more uncertainty on the 2PP number than the usual V(p) = p(1-p)/n.

But if we take the poll at face-value, we have reasonable evidence that the Coalition is above 50% 2PP (probability = .93). That falls to a probability of .86 if you assume the poll has a design effect of 1.8.

The betting market has moved to 1.40 to 2.85 (Centrebet), for an implied probability of an ALP win of .67. That will almost surely move through the day, as news of the poll gets around. At this point, that is the lowest ALP win probability at Centrebet since Gillard took office. It could well be lower by the time of my regular 9am update.

## Thursday July 29, 2010

Filed under: general,politics — jackman @ 9:26 am

Naturalization! The ceremony was surprisingly moving. Over one thousand of us were sworn in at the Paramount Theater in Oakland. The call of the 109 countries we come from was wild. Australia is early in the alphabet, and I was on my feet early, but China and India must have been well over half the pool.

I’m going to do the voter registration with the kids watching tonight at home.

## Wednesday July 28, 2010

Filed under: Australian Politics — jackman @ 9:04 pm

I’ll get to something more serious on the upcoming Australian election pretty soon.

In the meantime:

(1) I was just told about the Fake Penny Wong blog, Forced Conversations and Shit Lunches. Who is writing this? Fantastic.

(2) The betting markets are displaying a little more day-to-day volatility, as more punters get interested in the election and the increased tempo in polling etc. My daily time-series graph gets updated at 9am each morning, Sydney time, with the ALP trading in the .74 to .78 range (implied probability of winning) over the campaign thus far.

Labor was at .74 (1.29 to 3.65) on Centrebet on Election morning, 2007. Almost the same story at Centrebet this morning (.74; 1.27 to 3.65).

(3) I haven’t seen much marginal seat polling this cycle, thus far. I can’t help but wonder about QLD & NSW marginals (where — I hypothesize — that the switch to Gillard hasn’t helped as much as the national polls might suggest). And in turn, might we be headed for a result where the winner of the two-party preferred vote doesn’t win the election? This happened last in 1998, and before that 1990, where in both cases the opposition won the 2PP vote but not enough Reps seats to win government.

The hypothetical scenario I’m musing over here sees a government win the 2PP nationally, but lose enough marginals to lose government. Hmmm…

## Tuesday July 20, 2010

Filed under: general — jackman @ 5:57 pm

Coarse black sand beach. Late summer afternoon. Tide turning. Ridiculously sucking beach break, typical of steep Northern California beaches. Two larger than average waves and the dune I’m sitting on shudders.

Cool.

## Monday July 12, 2010

Filed under: computing,general,statistics — jackman @ 1:30 pm

Patrick Perry (Harvard) on dynamic social network data in the Stat Dept seminars series (Tuesday, Sequoia Hall).

Thursday: Jeff Lewis (UCLA) teaches a day long class on scraping the web. I taught this last year, but since Jeff taught me just about everything I know about this topic I thought we’d all be better off if we cut out the middle man. This is listed as SOC 384 at Stanford, Meyer Library Forum, 10-2.30pm.

## Friday July 9, 2010

Filed under: Australian Politics — jackman @ 1:17 pm

Amanda Bishop does a very good Gillard. This interview has with a lovely take on Gillard’s accent, with digressions on regional/class variations in Australian accents.

“werking fehmlees”?

## Wednesday July 7, 2010

Filed under: politics,statistics — jackman @ 4:16 pm

The NYTimes reports:

The Huffington Post is venturing into the wonky but increasingly popular territory of opinion poll analysis, purchasing Pollster.com, a widely respected aggregator of poll data that has been a major draw for the Web site of The National Journal.

More than one or two polisci colleagues and friends are or have been associated with Pollster; congratulations to all.

And NYTimes has 538.com.

Should be a busy Fall.

## Tuesday July 6, 2010

Filed under: Australian Politics — jackman @ 5:54 pm

An update on Labor’s standing in the betting markets post-Rudd. Using the Centrebet prices, ALP’s win probability is back up to > . 75 (1.24:3.85) since getting down to .62 (1.51:2.50) on June 16-17.

Labor was at 1.29:3.65 on Centrebet (prob = .74) on Election morning, 2007.

Snippet from the time-series:

## Friday July 2, 2010

Filed under: flight nerdery — jackman @ 2:38 pm

1. Ryan Air takes a step closer to flying standing up. Many think they are not serious (again).