Presidential Election Predictor

November 4, 2008
It's now a series of individual state races. I've reduced the percentage of variance applied to a national shift from 60% down to 50%. There are still some significant unknowns that apply to all states: Which campaign will do a better job of getting out the vote? Will younger voters actually vote with a historic turnout rate? Are there national patterns in how previously undecided voters break on election day? But on election day, even these issues have a significant state-to-state variation. Of course the uncertainty in pollng data varies independently from state to state. Hence, I give equal weight to national uncertainty and state-by-state uncertainty.

The state-by-state futures at midnight (PDT) last night implied a 87.55% chance of an Obama victory with 60% of the variance applied to the national shift. Reducing that to 50% for election day increases the probability to 88.59%. Now, in the late morning, Obama's direct futures are trading at 92.8%, and the predictor based on state-by-state futures prices (use the form at the form at the bottom of this page) stands at 90.7%. IEM WTA futures for a popular-vote victory have been trading at an average price of 91.74% today, although the most recent price is 93.5%. The FiveThirtyEight predictor has been updated to 98.9%.

November 3, 2008
I'm going to reverse my decision about Ohio and say that state is still in real play, so Obama's minimum electoral count is 291, rather than 311. There really are no scenarios where McCain loses Ohio and wins the election, so I'm moving Ohio from the Possible Tipping Point category to McCain's Must Win category. All of the in-play states are in McCain's Must-Win category. If McCain loses any of those states, it's pretty much over. All of the possible tipping point states are trading over 87% for Obama, and McCain needs to win at least two of those states totaling 23 electoral votes to win the election. (Specifically, McCain needs Pennsylvania + 1 of NV/VA/CO/NM/NH or Virginia + 2 of NV/CO/NM or all four of NV/CO/NM/NH.) Other than that, I really don't have much to change since this afternoon.

Obama's direct Intrade futures finished the day at 91.1%, up almost half a point from this afternoon. The inferred predictor based on state-by-state futures (use the form at the form at the bottom of this page) is essentially unchanged at 87.55%. IEM futures retreated slightly back toward McCain, finishing the day at 91.2%, down 0.2 points since this afternoon. The FiveThirtyEight predictor hasn't been updated since this afternoon, so it still stands at 98.1% for Obama.

November 3, 2008
In the late afternoon (PDT), Obama's direct Intrade futures are trading at 90.65%, and the inferred predictor based on state-by-state futures (use the form at the form at the bottom of this page) stands at 87.59%. IEM WTA futures for a popular-vote victory are trading at 91.4% for Obama, and the poll-based FiveThirtyEight predictor gives Obama a 98.1% chance of electoral victory. All four predictors are at all-time highs.

As judged by futures trading, all of the Possible Tipping Point States are moving toward Obama, and all of them are moving decisively, except Florida (which is stalled out at 75% for Obama). None of the states in McCain's Must-Win category are moving significantly, ranging from 62.35% for North Carolina down to 19.35% for Arizona (which I just moved out of the Safe McCain category. Based on late movement, futures traders seem to have decided that Obama is strongly on track to win at least 311 electoral votes, and McCain is strongly on track to win at least 130 electoral votes. (He could possibly lose two congressional districts in Nebraska, but he will not lose the state). The remaining states (in order from likely Obama to likely McCain) are: FL, NC, MO, IN, MT, ND, GA, AZ. Of that group, Missouri is the true toss-up. The Nebraska congressional districts are in the mix with Montana and North Dakota.

Now I'll give my two cents on predicting an outcome. This includes all-of the non-quantitative information that I've picked up during the campaign. If there's a surprise outside of the Florida-to-Arizona list, I'd say that McCain could pick off Ohio. I'm not predicting that, but it's the best chance for a real surprise. That means that Obama will win with a minimum of 291 electoral votes. On the other end of the spectrum, Arizona probably isn't much in play. If Obama wins Arizona, then I'd give about a 50/50 chance of a 400-electoral-vote landslide. There are good reasons to believe that the other states could go either way in almost any combination—even McCain winning Florida while losing North Dakota or Montana. I'd agree that Missouri is the true toss-up, with no favorite in that race. If I'd pick a mild upset, then it would be Indiana for Obama. I think the enthusiasm of his suporters and his superior ground game could net him that state. Call it Obama 364, McCain 174.

November 2, 2008
The futures markets seem to have lost a little hope for John McCain today. There wasn't any bad polling news for McCain. In fact, the news was good—it just wasn't good enough. So, let's talk about the polling first. For this site, that means the predictor at FiveThirtyEight, which moved 2.5 points toward McCain. That improved his percentage from 3.8% to 6.3% (meaning that Obama's percentage now stands at 93.7%). Since McCain's probability was down to 2.8% at FiveThirtyEight just two days ago, the polls are moving in his direction. The problem is that he's at the edge of elimination. Even with typical tightening built into the model, he's improving faster than the trailing candidate typically does this late in the race.

So what's with the futures markets? Well, this is one of those situations where the difference between polling and futures markets shows up. The futures markets have been giving McCain something like a 15% chance of pulling off a victory, whereas a good mathematical analysis of the poll data suggests that's wildly optimistic for McCain. Apparently futures traders have held out some hope for a McCain comeback (or some doubt about the strength of the Obama lead), and they've felt that some type of October surprise could upset the applecart and throw McCain back into the game. Well, there's now one full day of campaigning left and the surprise hasn't materialized. Meanwhile, there are a lot of votes in the ballot box already, so the impact of a late-breaking surprise would be muted.

The direct futures at Intrade moved sharply toward Obama today, setting an all-time high at 89.05%, which was a jump of 4.3 points. The inferred predictor based on state-by-state futures (use the form at the bottom of this page) is also an all-time high for Obama at 86.15%, up 3.18 points. Finally, IEM WTA futures set an all-time high at 87.2%, up 3.4 points. I've been reporting average daily prices for IEM, because the volume is low, but it's worth noting that the closing price was 88.85%, which is another 1.6 points above the average.

November 1, 2008
Another day without much change. Obama's direct Intrade futures rose 1.25 points to 84.75%, while the inferred predictor based on state-by-state futures (use the form at the bottom of this page) rose 0.46 points to 82.97%. IEM WTA futures, however moved in the other direction, with Obama's probability for a popular vote victory down 0.55 points to 83.8%.

That gives us a pretty consistent set of numbers from the futures markets. Ironically, though, the first day that IEM futures for an Obama win in the popular vote traded below Intrade futures for an Obama win in the electoral college was also the first day that FiveThirtyEight gave a higher probability of a McCain electoral win than a McCain win in the popular vote. Go figure. In any case, neither candidate has a strong structural advantage in the electoral college. There is about a two percent chance that the electoral winner will not win the popular vote, and that's split pretty evenly between McCain and Obama. The inferred predictor based on state-by-state futures now matches the direct futures if the percent of variance attributed to national shifts is set to 45.1%. I think the real value is probably between 50 and 60% in the last few days before the election.

FiveThirtyEight now gives Obama a 96.13% chance of electoral college victory, down 1.07 points. Finally, I moved West Virginia to the Safe McCain category.

October 31, 2008
McCain's direct futures at Intrade went up today, but not for any apparent reason. There really wasn't much good news in the polling. With a 1.3 point improvement, his futures are now trading at 16.5% (i.e., 83.5% for Obama). The inferred predictor based on state-by-state futures (use the form at the bottom of this page) is essentially unchanged, up 0.01 points to 82.51% for Obama. IEM WTA futures traded at 84.35%, up 0.1 points, and the FiveThirtyEight predictor set an all-time high for Obama at 97.2%.

October 30, 2009
McCain didn't make up any more ground today, but he didn't lose ground either. I moved Arkansas to the Safe McCain category, and West Virginia could follow soon. Obama's direct Intrade futures are at 84.8%, down 0.05 points. The inferred predictor based on state-by-state futures (use the form at the bottom of this page) is at 82.5%, down 0.48 points. Again, most of the change was caused by a single trade of California shares, this time CALIFORNIA.DEM. (Note: Intrade seems to report the closing price as the last price, if a future has not traded at all during the current session. Since closing prices are not necessarily equal to the last trade price, that means that the "Last" price is not necessarily the last traded price.)

The IEM WTA futures traded at 84.25% for Obama, up 0.35 points, and the FiveThirtyEight predictor is up 0.6 points to 96.3%.

October 29, 2008
Today was nicer to McCain—at least in the futures markets. Obama's direct Intrade futures dropped 2.45 points to 84.85%. That might seem like a modest change, but that means that McCain's futures improved from 12.7% to 15.15%, which is almost 20% higher. The change was precipitated by some good national polling numbers. The national numbers were not reflected at the state level, however, and as a result, the inferred predictor based on state-by-state futures (use the form at the bottom of this page) dropped only 0.61 points to 82.98%. There's a little sleight of hand here, because Intrade's trading data is inconsistent for California. The last trade for CALIFORNIA.REP was at 2.5%, and that was just before I posted last night. There are no reported trades between then and now, but Intrade's summary pages report the last trade was at 6%, even though there has been no trading today. California is such a large state, that nudging Obama's certainty of victory in California can be seen easily in the inferred predictor. I used 2.5% as the last trade for CALIFORNIA.REP resulting in an agreggate probability of 97.15% for Obama to win California. If I were to use the 6% value, then Obama's aggregate probability of winning California would drop to 95.4%, and the inferred predictor would drop to 82.36%.

Obama's IEM WTA futures traded at 83.9% today, down 2.15 points, and the FiveThirtyEight predictor dropped half a point to 95.7%.

October 28, 2008
Obama's direct Intrade futures are at 87.3%, down 0.55 points. The inferred predictor based on state-by-state futures (use the form at the bottom of this page) is now 83.59, up 0.05 points. The IEM WTA futures traded at 86.05 today, down 0.70 points, and Obama's FiveThirtyEight predictor is down 0.5 points today to 96.2%

October 27, 2008
Things are not much changed from yesterday—mostly just some ongoing creep in Obama's favor as election day draws nearer. Direct Intrade futures are at 87.85% for Obama, while the inferred predictor based on state-by-state futures (use the form at the bottom of this page) is now 83.54%. The direct futures are up from yesterday to a new high and the inferred predictor is down slightly, with the result that the gap between the direct futures and the inferred probability based on state-by-state futures is at an all-time high of 4.31 points. The two values can be brought into line if the percentage of variance due to a possible national shift is set to 31.9%. As I've said before, I believe that value is too low, and I recommend a value of 60% during the period close to the election.

The IEM WTA futures traded at 86.75%, up more than a point from yesterday, but still below the all-time high. Obama's FiveThirtyEight predictor was unchanged at 96.7%.

October 26, 2008
Most of the predictors show continued creep in Obama's direction today. Direct futures at Intrade were at 87.5% for Obama tonight. The inferred predictor based on state-by-state futures (use the form at the bottom of this page) is at 83.78%. Both are all-time highs. IEM WTA futures are at 85.4%, down 0.2 points, and off 1.7 points from the high a week ago. The FiveThirtyEight predictor is 96.7%, up one point, and also at an all-time high.

October 25, 2008
Obama's direct Intrade futures are at 87.05%, up 0.35 points. The inferred predictor based on state-by-state futures (use the form at the bottom of this page) is now 83.4%, up 0.27 points. The IEM WTA futures traded at 85.6% today, up 0.65 points, and Obama's FiveThirtyEight predictor is up 0.8 points to 95.7%.

October 24, 2008
Obama's direct Intrade futures are at 86.7%, up slightly from yesterday. The inferred predictor based on state-by-state futures (use the form at the bottom of this page) is now 83.13%, down slightly from yesterday. The IEM WTA futures are are down 0.65 points to 84.85%. The FiveThirtyEight predictor is down to 1.4 points to 94.9%.

October 23, 2008
Obama's predictors are mostly at or near all-time highs. His direct Intrade futures are at 86.65%, up about half a percentage point. The inferred predictor based on state-by-state futures (use the form at the bottom of this page) comes in at 83.27%. The FiveThirtyEight predictor took a jump of almost three percentage points in Obama's favor and now stands at 96.3%. Only the IEM WTA futures are off their all-time highs, trading today at 85.6%.

It's probably good to keep in mind that the dollar volumes at IEM are small compared to Intrade. For example, today's trading represented something like $1300 for WTA futures at IEM vs. $85,000 for direct futures at Intrade. The trading in state-by-state futures at Intrade may be small for individual states, but all together represents a volume similar to the trading of direct futures. Another thing to note is that Intrade direct futures should trade higher than IEM WTA futures, because Obama still has a better chance of winning the necessary electoral votes than he does of winning the popular vote.

October 22, 2008
The gap between Obama's direct Intrade futures (86.2%) and the inferred predictor based on state-by-state futures (81.99%, use the form at the bottom of this page) is the highest that it has been this election season at 4.21 percentage points. Because McCain's chances are rather remote either way, this is a big fractional difference in McCain's futures (13.2% vs. 18.01%). Part of the difference may have to do with the futures for the state of Pennsylvania. Those have moved strongly for McCain. As far as I can tell, there is no polling or other data behind the price change, but the McCain campaign has made it clear that they will fight for Pennsylvania as a top priority—perhaps as the top priority.

In any event, traders are more optimistic about McCain's overall chances than they are about McCain's chances on a state-by-state basis. Alternatively, traders may think that the state polling numbers have more room for surprises this year, but that McCain has relatively little chance of pulling off anything that looks like a national swing. It is possible to bring the state-by-state inference into agreement with the overall futures if the percent of variance due to national shift is only 33%. From what I've been able to tell, this is just too low. On the other hand, McCain does not have the luxury of moving a few similar states. He has to move very different states. Even if we assume that McCain can win Pennsylvania, he has to do it in a way that doesn't have him losing Ohio or Florida, or even a combination of states like New Hampshire, Virginia, Colorado and Nevada. So, McCain needs to win Pennsylvania and Ohio and Florida and at least one of New Hampshire, Virginia, Colorado and Nevada—and that's even assuming that he wins both Missouri and North Carolina. He's losing in every one of those states, most by significant margins. Whatever trick he finds to swing Pennsylvania has to also bring those other very different demographics along for the ride.

The other predictors are IEM WTA futures at 86.65% and the FiveThirtyEight predictor at 93.5%.

October 21, 2008
The gap widened between Obama's direct Intrade futures at 85.05% and the inferred predictor based on state-by-state futures (use the form at the bottom of this page) at 81.28%. The change in the inferred predictor would have been greater by about a third of a percentage point, but I reduced the percentage of variance from nationwide shifts from 65% to 60%. Over the last four days, the gap has gradually widened by more than 2 percentage points. The biggest jump was today, and half of that shift was due to a drop in Obama's Pennsylvania futures from 89.4% to 86% on limited trading. I'll see what the market price is tomorrow and decide whether to take PA out of the "Obama Safe" category.

The other predictors today were IEM WTA futures at 86.75% and the FiveThirtyEight predictor at 93.4%.

October 20, 2008
Again, not much movement in the four predictors today—which is good news of a sort for McCain, since the clock it ticking on his campaign. If his probability of victory isn't fading away, then he's making some progress. Obama's direct Intrade futures (84.65) were up slightly again today, but the inferred probability based on state-by-state futures (use the form at the bottom of this page) was unchanged at 81.8%. IEM WTA futures (86.8%) moved slightly for McCain, and the FiveThirtyEight predictor (92.5%) moved a little over half a point for McCain.

October 19, 2008
Not much movement in the four predictors today. Obama's direct Intrade futures (84.35%) moved less than half a point toward Obama. Obama's inferred probability based on state-by-state futures (81.8%, use the form at the bottom of this page) moved less than half a point toward Obama. IEM WTA futures (87.1%) moved just over half a point toward Obama. The FiveThirtyEight prediction (93.1%) moved half a point toward McCain.

October 18, 2008
The futures markets have been settling after the debate bump. Obama's direct Intrade futures (84.0%) are still up a couple of percentage points relative to pre-debate levels, while Obama's inferred probability based on state-by-state futures (81.55%) is up about one point. I've also been tracking IEM WTA futures (86.4%), which are also up about one point since pre-debate levels, and the FiveThirtyEight prediction (93.6%), which has shown a shift over two percentage points toward McCain.

This is probably a good moment to talk about the difference between the FiveThirtyEight prediction, which is based directly on polling data and historical variance between polling data and electoral outcomes, as opposed to the futures markets, which are based on the sentiments of traders informed by polling data, historical outcomes, debate performances, news events, etc. The futures markets, including my inferred probability based on state-by-state futures, all behaved similarly after the debate, showing a slightly improved situation for Obama. In general terms, I'd attribute that to a pre-debate assessment that John McCain had a small, but non-negligible, chance of pulling off a game changer on Tuesday night. When that didn't happen, there was a little bump in Obama's direction as measured by futures prices. At the same time, actual polling data showed a slight improvement in McCain's direction. This is entirely consistent, since the improvement in McCain's polling numbers wasn't as big as pre-debate futures traders thought it might have been. On the other hand, FiveThirtyEight's model doesn't have event-based expectations built into it.

So, McCain did make up more ground than typical in a few day period three weeks before the election, but he didn't make up as much ground as pre-debate futures traders thought that he might have. As a result, the poll-based model developed by Nate Silver at FiveThirtyEight and the futures prices got a few points closer this week.

October 17, 2008
I've moved Minnesota to the "Safe Obama" category. McCain claimed back some of Obama's gains from the last few days. The direct futures value for Obama are 83.75% and the inferred probability based on state-by-state futures is 81.87%.

October 16, 2008
I've moved Iowa and Pennsylvania to the "Safe Obama" category. Unless there is an electoral shift that puts those states back into play, Obama only needs to secure 24 more electoral votes from the remaining possible Tipping Point States. Minnesota is on the cusp of moving another 10 electoral votes to the Safe category. The direct futures value for Obama (85.25%) is still higher than the inferred probability based on state-by-state futures (83.31%).

October 15, 2008
Today was another surge for Obama as judged by futures traders. I've moved Wisconsin out of the "Possible Tipping States" category to the "Safe Obama States" category. In order to keep some balance to the number of states in the Tipping-States category, I've dropped Iowa down—but I'll be moving states to the Safe category when their futures prices go above 90%. I've also moved Arkansas out of the Safe McCain category.

This is the first time that I'm doing an update while the direct futures prices are more favorable toward Obama (84.5%) than the inferred predictor based on the state-by-state futures (82.68%), although I've seen the pedictors cross during the day.

October 14, 2008
I'm going to take a moment here to discuss the various predictors of the presidential election that are available on the web. Please keep in mind that a prediction is a statement about the probability of an outcome, not a statement that an outcome will or will not occur. There are five sites that either give a prediction in terms of probability, or give enough information to infer a probability.

The first is the Princeton Election Consortium. This site does not give a direct predictor, but it does give an electoral vote estimator with a 95% confidence interval—very nice and very meaningful. The confidence interval is based on the margin of error for aggregated poll data. The polls are combined to determine a probability for each state for each candidate. Those probabilities are combined into a histogram of possible electoral outcomes. The median outcome with 95% confidence interval is determined from that.

This is not, however, a prediction of the electoral outcome. The probabilities in that meta-analysis are only for recent polling. There is no attempt to determine or estimate the amount that the overall national or individual state electorate may change their opinions or intentions between now and the time that the election is held.

The second site is FiveThirtyEight. This site does provide a direct predictor. The underlying analyis is very similar to the Princeton Election Consortium, but FiveThirtyEight goes a step further and uses demographic data, historical data about national and state-by-state shifts in polling and actual election returns. This is very important, and the model that they use is sophisticated and their methodology is fairly transparent. I feel they are the best site for aggregating and interpreting poll data.

Their model of electoral shifts includes state-by-state shifts, national shifts, and shifts in demographically similar states. This is fairly similar to my analysis, but we use very different underlying data…more about that later.

The third site is Intrade. This is a futures market with four different futures for the outcome of the presidential election. Each of these futures is a separate predictor of the overall electoral outcome (if you believe that the possibility of a third-party candidate winning the election is negligible). In short, the futures are "Obama", "McCain", "Democrat", and "Republican". Because the futures are actively traded, the four futures never indicate a single value, but it's straightforward to aggregate them. (I use the bid and ask prices, as well as the sales prices, to determine the most consistent market value.)

The prices of the futures track the polls pretty directly, since that's the main source of information available to futures traders. But they do not reflect the polls in a well-defined mathematical way. The trading market is making an implicit assessment of the factors that the Princeton Election Consortium does not address. The futures prices are completely aggregated. They may or may not be biased, but it is arguable that they are better at removing biases than pollsters are, and they are certainly better at incorporating the intangible aspects of the electoral outcome, such as how large will the black or youth turnout be?

The fourth site is Iowa Electronic Markets. This is also a futures market, but their futures are different. They have two futures based on the U.S. Presidential Election. The easiest to understand is the "Winner-Take-All" market, based on the popular vote This is a direct predictor, but it does not include an assessment of the probability that the popular vote winner fails to win in the Electoral College

The fifth site is this one. Although less slick than the others, it also provides a direct predictor of the electoral outcome. By visiting the site and clicking on the button on the bottom of the page, you get a direct prediction of the electoral outcome, based on Intrade prices for each state. The Intrade prices in the web form are updated each night, and the cgi includes a historical tracker of the daily predictor.

The model behind the form is similar to that used by FiveThirtyEight, in that it allows for both state-by-state shifts and also for national shifts. The default 60% split between national shift and state-by-state shift is based on the typical ratio of polling uncertainty in a state to the state-by-state Intrade futures prices.

What is interesting is that the four sites that offer an electoral predictor give different results. The inferred predictor here at MathRec closely tracks the direct futures prices at Intrade, but there is a consistent offset, as I've discussed previously. Both of these are significantly lower than the futures price of the IEM Winner-Take-All market for Obama, even though it is quite clear that Obama has a significantly greater chance of winning the Electoral College than the popular vote. The predictor at FiveThirtyEight is significantly higher than the futures price at IEM, even after adjusting for the possibility that Obama wins the Electoral College without winning the popular vote.

I'll discuss possible reasons for the discrepancies in my next post.

October 12, 2008
Apparently South Carolina's futures price last night was an aberration. I've moved it back to the Safe McCain category.

The inferred probability and direct futures prices have been stable for four days now at about 78.8% and 78.0% respectively (for Obama). Ohio, Florida and Nevada have stayed in the 67–70% range after being priced over 70% (very briefly); Colorado and Virginia have drifted further to Obama in the 75–80% range. McCain must win all of those states to win the election, unless he can win New Mexico and New Hampshire (both currently priced at 80+%), in which case he could afford to lose Colorado or Nevada.

October 11, 2008
South Carolina and North Dakota have moved out of the "Safe McCain" category. It certainly doesn't look like either state could tip the election, but futures traders are no longer too sure that McCain will win those states. Iowa is priced on Intrade as safer for Obama than either Arkansas or Louisiana are for McCain, but I'm leaving those three states as they are for now.

October 9, 2008
A lot of little things have changed, but they are all ways of saying that Obama is way ahead. Colorado is no longer the pivotal state. Nevada has passed Colorado's futures price, and stands to give Obama his 269th electoral vote. But lots of states have moved into the 70–75% probability range for Obama. In addition to Nevada and Colorado, Ohio, Virgina and Florida have all made it to the 70% mark. Any of those five states would get Obama to 269. That also means that every "Possible Tipping Point State" is now rated as at least 70% probability for Obama. Michigan has moved out of contention—it's hard to say whether McCain pulled out because it was out of reach or whether it's out of reach because McCain pulled out. North Carolina and Missouri have both moved over the 50% line. North Carolina has moved decisively, rating a 60% probability for Obama. Furthermore, West Virginia seems to be more seriously in play, with Obama futures there running at 39%. As a result, the overal inference from state-by-state futures is now at 78.56% for Obama. The direct futures are in line with that at almost 77%.

A major sea change is the only way for McCain to get back into the race.

October 8, 2008
For a little while this afternoon the direct futures and the inferred probability based on state-by-state futures were switched, with Obama futures valued higher than the inferred probability. That wouldn't be unusual, except that it had not happened in the last month, with the difference ranging between about two points and about six points.

Even more unusual was the coordinated buy of McCain shares with a sale of Obama shares (between 8:44 and 9:05 PM EDT), involving about $139,000. That was essentially one day of trading in a 20-minute span. It increased the value of McCain's futures by over 50%...for about half an hour. After 30 minutes, McCain's futures had sunk back by eight of the twelve points they had been boosted during the coordinated trading. There's still a residual change of about three percentage points in both futures. There was no coordinated trading on the PRESIDENT.REP2008 and PRESIDENT.DEM2008 futures, but those futures still moved a few points in the same time frame, presumably because investors saw the better value in the equivalent Obama and McCain futures. The REP and DEM futures are now within one point of their afternoon values.

So, is someone spending a lot of money to move the market value, or is someone doing a very poor job of managing their investment in the futures market, since a gradual buy and sell would have netted a far better value? The $139,000 figure that I calculated was based on the assumption that someone was acquiring a new position. If someone was actually liquidating an existing position, then they were cashing in a lot more than $139,000—almost $300,000. I estimate that trading more gradually would have made a difference at least $25,000 in either case. Mischief or stupidity? Or just one whopper of a margin call (which I suppose is a subset of stupidity)?

October 7, 2008
Today is four weeks before the election. I've lowered the default percentage of the variance applied to the national shift from 70% to 65%. This change added 0.14 percentage points to Obama's inferred probability of overall victory—reflecting the reality that McCain needs a collective shift of many states if he is to win the election on November 4.

I also moved Georgia from a Safe McCain State to Must-Win State for McCain. If I was to have any consistency at all, I needed to put Georgia on the list of in-play states or remove Michigan and Iowa.

October 6, 2008
Yesterday's drop in Obama's direct futures seems to have been a result of the unusual trading. The predictor inferred from the state-by-state futures held steady (moving from 71.54 to 71.84), while the direct futures rebounded by more than four percentage points to 69.35. (I calculate the direct futures using all four of the Intrade offerings: 2008.PRES.OBAMA, 2008.PRES.McCAIN, PRESIDENT.DEM2008, and PRESIDENT.REP2008.)

October 5, 2008
I changed the make-up of "Obama's Must-Win States" moving NM, MN and WI into "Possible Tipping Point States" in exchange for MI. This is not an indication that NM, MN and WI have become more vulnerable—just that they are more likely to be part of a McCain victory than MI. At the same time, I moved WA, NJ and OR to "Safe Obama States". That is an indication that they are less vulnerable.

The direct futures on Intrade shifted about 2.7 points toward McCain in the last day. The inferred predictor based on state-by-state futures did not move. The difference tonight stands at 6.5 percentage points, which is fairly large. Generally I'd expect the state-by-state futures to follow about a day later, but the trading patterns on Obama and McCain futures were a bit odd today, with large sell-offs of Obama futures and simultaneous buys of McCain futures at about 9:30 AM and again at about 3:00 PM EDT. It will be interesting to see whether the direct futures rebound, the state-by-state inference drops, or the unusually large separation holds up for a little while. I would bet against the latter. This is another good time for arbitrage, if you're so inclined...

October 1, 2008
I've had to edit my headings for the sections below. I decided not to change the state groupings, because these particular groups really highlight where the candidates stand. When the race was very close, I identified the tipping point states that could swing either way in a close race. That is, either candidate could reasonably lose any of these states and still win the overall election. The other states that could be in play were just grouped as "Leaning Obama" or "Leaning McCain". I've replaced those names with "Obama's Must-Win States" and "McCain's Must-Win States". The problem is that we now have a "Leaning McCain State" leaning Obama (North Carolina). That futures price may or may not swing back to McCain, but I prefer to keep the state groupings than to keep the names of the groups.

September 29, 2008
As of this writing, Obama now leads in every likely tipping-point state. He needs to win one of CO, VA, OH or FL, or he can win both NH and NV. Colorado remains at the pivot point, and Obama captures 269 electoral votes if the states maintain their current ordering, and he wins Colorado. That is sufficient for Obama to win the election, because Democrats will certainly control a majority of state delegations to the House of Representatives. Note, however, that McCain could conceivably win Maine's second congressional district, along with FL, OH, NV, VA and NH to reach 270 electoral votes. There's no separate polling available in Maine's second district, but Bush only did two percentage points better in that district than he did in Maine as a whole, so both districts are likely safe for Obama.

Obama's direct futures rose about four points today, while the predictor inferred from the state-by-state futures rose only about one point. The separation now stands at 3.3 percentage points.

I had a dry run at executing an arbitrage between state futures and direct futures (without actually putting my money where my mouth is). I set hypothetical limit orders to sell five shares each of FL, CO and NV futures for Obama, while buying 10 shares for Obama to win overall. My orders would have been filled for FL at 51.7 and CO at 68 and withdrawn for NV. My order for Obama would have been filled at 57.3. This would have given my an initial exposure of $97.45 ($24.15 for FL, $16.00 for CO, and $57.30 for direct futures).

I then placed hypothetical limit ordres to buy five shares each of Obama futures for FL at 52.0 and CO at 69.0. When each order was filled, I sold an equal number of Obama direct futures at an average market price of 62.3 (paying a total commission of $0.50 on the ten shares). The liquidation of my existing positions would have returned $101.30 ($24 for FL, $15.50 for CO, and $62.30 less $0.50 for direct futures). That would have been a one-day net of $3.85 on an initial exposure of $97.45. It would be difficult to generalize that to larger trades, because the market volume in the state-by-state futures isn't that great, but it does suggest that it's possible to identify times when the state futures and the national futures are out of step.

September 28, 2008
The separation between the direct futures at Intrade and the overall predictor based on state-by-state futures has widened farther. It now stands at 6.43 percentage points. The predictor based on state-by-state futures has been climbing steadily over the last week, increasing by almost 10 percentage points. At the same time, the direct futures have climbed more slowly, increasing by less than 7 percentage points. Over the last four days, the difference has been even starker. The state-based predictor has risen by 7 percentage points, while the direct futures have risen by only 2 percentage points. I may have to try a little arbitrage by buying direct futures for Obama (or selling them for McCain) and selling selected state futures for Obama (or buying them for McCain). Later when the gap narrows, it should be possible to liquidate the position for a net gain.

September 26, 2008
The separation between the direct futures at Intrade and the overall predictor based on state-by-state futures has opened back up. It now stands at 4.66 percentage points. Volatility after the debate is to be expected, but I don't think we're seeing that yet. I think that we saw movement toward Obama on Wednesday, and it just took a few days for the state-by-state markets to catch up. Obama's advantage of 3–4 percentage points in the state-by-state projection has held up, except when there is a sharp movement—then the direct futures tend to move first, with the state-by-state futures lagging by about one to three days.

September 24, 2008
I made two statements last night that were reflected in today's activity. The ROI table at FiveThirtyEight.com now puts New Hampshire on top by a wide margin, and the direct futures for the overall outcome moved sharply for Obama (4.2 percentage points) to close the gap with the inferred probability based on state-by-state futures (which moved only 0.7 percentage points). This brings the discrepancy between the direct futures and the state-by-state futures down to a single percentage point—closer than I have seen it in the two weeks that I've been tracking.

September 23, 2008
I've switched over to a 51-state form, which ignores the reality that Maine is two separate elections plus a statewide total, and Nebraska is three separate elections plus a statewide total. Otherwise, the model is complete. I'll only make modifications for allocation of electors by congressional district if it looks like one or more districts in Maine or Nebraska are in play this year. All 51 states are used in the calculation, because I've switched to building a probability matrix of electoral vote totals, instead of building a list of permutations for state outcomes. The states are grouped alphabetically as possible tipping point states, leaners and safe states. If you ignore the safe states, you can use the 51-state form the same way as the old 19-state form. This structural change to my calculation removed about 0.66 percentage points from that difference.

The inferred value from state-by-state futures comes in consistently about 3–4% better for Obama than the direct futures. My interpretation is that the overall futures market is underestimating Obama's advantage in the Electoral College. Without a significant national vote swing, McCain needs to pick up essentially all of the states that are near the pivot point.

I've also switched the default allocation of variance to 70% national instead of 75%. I'll reduce it twice more before the election to a final value of 60%. It made an insignificant difference of 0.07 percentage points in the overall projection.

Finally, I looked systematically at the impact that various states make on the outcome. I calculate a figure of merit that is based on the change in overall probability that results from a small shift in the z-value for each state. That result it divided by the population in the state (2007 estimate) to give a per-capita value for the impact. The top 13 states are:

It's interesting to compare this table with the similar table from FiveThirtyEight.com, run by Nate Silver. Their ROI index is currently led by Colorado, narrowly over New Hampshire. Their top 13 are: CO, NH, NV, NM, VA, PA, OH, WV, MN, MI, IN, WI, and FL. Our analyses should be fairly similar, but we use different base data—they use direct polling data with a relatively sophisticated model of voter shifts, including shifts in groups of demographically similar states, whereas I use the prices of state-by-state futures. I also normalized by population, whereas he normalized by eligible voters. (My table does not change significantly using the number of registered voters in 2004, except that NH is barely first.) Plugging state-by-state probabilities from FiveThirtyEight into my form matches their overall projection quite closely. I have not run my impact analysis using his state-by-state projections. Perhaps another day, after my impact analysis is automated...

Overview

A quick disclaimer. Although I think that this site is useful and serves a small niche, there are better sites available. In particular, I've drawn heavily from Sam Wang's Electoral College Meta Analysis. I've also paid close attention to FiveThirtyEight. My impression is that Mr. Silver doesn't adjust sufficiently for the effects of overall national shifts when he determines his state-by-state probabilities. In any case, FiveThirtyEight's overall Win Percentage does not line up with the Win Percentage from futures markets.

I believe the principal usefulness of my site is to accurately assess the effect on the overall election resulting from shifts in one or a few states. There may also be some usefulness in picking out arbitrage opportunities between state-by-state futures and the futures for the overall electoral result.

The form below takes state-by-state inputs as percentages for Obama, such as that provided by Intrade. (Look for Politics, then US Election by State.) The state-by-state odds will incorporate the possibility of a national swing in the popular vote, plus state-by-state variation. Of course there is uncertainty associated with both of these. You choose how to allocate that uncertainty, by specifying the percentage that is national. My best guesstimate of the value is 75% for a couple months out, but four years ago I shifted that to 60% as the election drew closer. (It cannot be 100%, but it can be 0%. The overall prediction doesn't work consistently with values outside of 5% to 95%.) Then just plunk in the state-by-state odds. The cgi will do the permutations for various values of the national vote swing and predict the probability of an Obama win. I've prepared another page with more details about the methodology (and the math).

The form below is populated with default values from Intrade prices as of Tuesday mornning, 11/4/08.

Steve Schaefer
Last modified: Date: 2008/11/04