Although Republican candidate Jim Tedisco is leaving no stone unturned in the special election in New York's 20th Congressional District -- including objecting to the ballot of Kirsten Gillibrand, the woman whom he hopes to replace in the Congress -- the results tallied so far suggest that he is bound for defeat.
So far, according to results released (.pdf) by the New York Secretary of State, Democrat Scott Murphy has received 53.0% percent of the 1,870 the absentee ballots thus far counted by election officials. This translates to a net gain of 112 votes, enough to put him 47 votes ahead of Tedisco once Election Day results -- adjusted to reflect the state's re-canvassing -- are added back into the total.
Notably missing from the absentee count so far is Saratoga County, where Tedisco racked up his largest margin over Murphy on Election Night. Will Tedisco gain enough votes from Saratoga to overtake Murphy?
Well, he might - if Saratoga were the only county with absentee ballots outstanding. But there are plenty of votes still to be counted in Murphy's stronger areas too. Only about 16 percent of absentee ballots have been counted in Coumbia County, only 12 percent in Warren County, and none at all in Washington County. Murphy won between 55.8 and 56.6 percent of the vote in each of these counties on Election Day.
In fact, the counties in which absentee ballots have been counted so far appear on balance to be highly representative of the district as a whole. If one takes a weighted average of the Election Day margin based on the number of absentee ballots processed in each county so far, one finds that 49.8% of the vote in these counties went to Murphy -- incrementally less than the 50.0% (actually, 49.98%) that he actually received. If anything, the ballots counted so far have come from (very, very slightly) red-leaning areas.
What this means is that Murphy is in fact overperforming among absentee ballots, as we and other observers anticipated that he might. In fact, the pattern has been rather predictable. Take a look at the following table:
Four counties -- Greene, Rensselaer, Delaware, and Essex -- have completed processing their absentee ballots. In each of those counties, Murphy received between 2.7 percent and 3.4 percent more of the vote among absentees than he did on Election Day. Although the pattern is a bit more irregular in counties that are only partially done with their count, we feel safe in saying that Murphy has some systematic advantage here.
Let's project out the outstanding ballots based on the following assumptions:
-- Murphy will overperform his election day result by 3.2 percent in each county,
as he has thus far.
-- The counting will cease when 87 percent of the absentee ballots returned in each county have been counted. Why 87 percent and not 100 percent? Because some absentee ballots will be illegal (or will be frivolously objected to as being so). In the four counties that have completed their count so far, an average of 87 percent of the absentee ballots returned were in fact agreed to be legal and counted.
We project Murphy to pick up roughly 500 additional votes once all absentee ballots have been counted. That would give him an overall win by somewhere between 500 and 600 votes.
This projection is fairly crude and relies on nothing other than simple algebra. So let's try a slightly more advanced version. In this case, I've employed regression analysis to estimate the fraction of absentee ballots that each candidate will receive in each county based on two factors: the election day totals and the partisan distribution (.pdf) of the absentee ballots returned. The model uses these factors to attempt to explain the actual distribution of absentee ballots in counties which have counted some or all of them thus far.
This version of the model posits a larger gain for Tedisco in Saratoga County -- a net of 132 votes rather than 33 as projected by the algebraic analysis. On the other hand, it projects a much larger gain for Murphy in Columbia County, where the absentee returns were disproportionately Democratic and where he has received more than two-thirds of the absentee ballots processed thus far. Overall, the "advanced" model projects almost exactly the same result -- about a 500-vote gain, and 550-vote win, for Murphy.
I am candidly a bit unsure about how to calculate the standard error for a forecast such as this one, but it appears to be, using conservative assumptions, roughly +/- 190 net votes. That would imply a 95 percent confidence interval of between Murphy +173 and Murphy +913, and that Tedisco has only about a 1-in-500 chance of prevailing.
EDIT: One caveat: Tedisco, somewhat mirroring the strategy of the Norm Coleman campaign, has started to challenge a ridiculously high fraction of ballots. These challenges may already be depressing Murphy's numbers, and therefore our estimate of his eventual margin of victory may be low. On the other hand, if Tedisco continues to escalate the frequency of his challenges in the interim, he may be able to hedge against Murphy's reported gains or possibly even to take a paper lead.
The challenge to Gillibrand's ballot, by the way, was incredibly dumb. If Tedisco and his staff are looking at the numbers objectively, he should know that his odds of winning are extremely long -- much longer than Norm Coleman's were at a comparable point in the counting process. Tedisco, however, only has to wait 19 months to challenge Murphy again rather than six years. Re-matches for House seats, moreover, have a much better track record than re-matches for Senate seats. So there are far more tangible consequences to him of losing the spin war and impairing his credibility going-forward, something which he may now have done irretrievably. If Tedisco has any semblance of political aptitude, he will withdraw the challenge and apologize on behalf of the overzealous staffer/lawyer who made it.