This is not just a chance for me to brag about being right – it is about me learning what I failed to account for to ensure more accurate predictions in the future; it is about explaining my mistakes to you, the hypothetical reader, rather than just shrugging them off; it is about transparency, and validity, and all those things; and it is also a chance for me to brag about being right.
So, let's start at the top:
Okay, we all know Obama won. Now I called it while official news sources were still saying it was 50-50, but I also said it would be close, and basically, it wasn't. Not even close to close. So I was righter than the mainstream media but wronger than I would like. Why?
Well, I - like many other sources - was looking at nation-wide polling. And on the nation-wide vote I would have been right, with Obama beating Romney by a margin of less than 3%. However I, and many others, failed to take the electoral college system into account. If you didn't hear about the EC during the coverage, it basically works like this: the 50 states get a certain number of votes on the EC. This number is their senate seats (2 per state) plus their house seats (between 1 and 53 depending on population). D.C. Is also counted as a state, getting 3 votes (2 for it's imaginary senators and 1 for its illusory house member determined by its real, but puny, population). This makes 539 votes all up, so a candidate needs 270 to win a majority and take the throne. In most states the winner takes all the votes, thus California gave 55 votes to Obama. The exceptions are Maine (4 votes) and Nebraska (5 votes) who divide their votes; if Barak Obama got 40% of Nebraskan votes and Mitt Romney got 60%, for example, Nebraska would give 2 votes to 'The Big O' and 3 to 'Mittens'. (Side note: I love that our federal elections are organised, you know, FEDERALLY! None of this state-by-state interpretation of electoral codes that mean you cannot vote outside your hometown.)
So while the popular vote was pretty evenly split, Obama was a fraction ahead in the key 'swing states', and most of them fell to him giving a landslide of votes on a tiny margin as preferred POTUS.
I predicted no net change, giving the final result as Democrat 51, Republican 47, Independent 2. In actual fact there was a 2-seat swing to Dem (53, Rep 45, Ind 2).
What went wrong here is pretty straight forward: all of the safe and predicted seats fell as I expected, including the independents in Maine and Vermont (Go Bernie Sanders!). The flaw was my ego in playing guesswork with the seats I considered tossups. In the mid-terms my senate predictions were equally accurate, and the tossups fell 50-50. I noted prior to this election that tossups could just as easily go all one way as split down the middle, but without thinking I called the 7 tossups 4-3 to the Democrats. There was no real logic behind this, and it didn't influence the predictions for the balance of power. In the real world, the tossups went 5-2, and that is the 2 seat swing (from a 1 seat gap in the tossup predictions to a 3 seat gap).
In my defence I did back the tossups favouring the Democrats. I was also only out by one seat, if that gives you any confidence in my predictions. But this is where transparency is important. Tossups are tossups, and if I even call their split in the future I certainly should not be factoring it into the final predictions.
Two years on, I haven't reviewed my midterm predictions for the house. Don't hold your breath either. It will probably never happen. I am reviewing the 2012 predictions now:
Puerto Rico R.C. – I predicted a Democrat other than Pedro R. Pierluisi Urrutia, favouring Rafael Cox Alomar. Alomar came second, but Pedro R. Pierluisi Urrutia held the seat. Right party, wrong member. This was just a lack of research and a reliance on trends which I said at the time were inaccurate. Basically, I was just plain wrong. I don't plan to correct this in the future because mistakes happen and not even I have time to fully research and cover Puerto Rico every two years.
American Samoa Delegate – I called this a safe Democrat seat. It is a safe Democrat seat. Eni Faleomavaega returns, as he has for over two decades.
D.C. Delegate – As with American Samoa. Always was Democrat, still no change in 2012. Eleanor Holmes Norton is returned.
Guam Delegate – Although one of the closer non-voting seats, it was pretty clearly Democrat. I was correct in my predictions that Madeleine Bordallo would return.
Mariana Islands Delegate – Another close seat, another Democrat win, another correct prediction. Gregorio Sablan returned.
U.S. Virgin Islands – Democrat seat is Democrat. Donna Christian-Christensen is returned as predicted.
So, 5 out of 6 right on a candidate level, and 6 out of 6 on a party level. That sound's alright, and I'm happy with that, but that was the easy part. Now, the voting seats:
Firstly, NC7 has not been called. I don't know why, but it hasn't. Secondly LA3 has decided it's Republican, as predicted. It just hasn't decided which Republican. Charles Boustany and Jeff Landry will be fighting it out on December 8.
As far as settled seats are concerned, there were some errors made. Five seats called for the Democrats turned Republican (NE1, NE2, NE3, KY6, MI1), while nine called Republican were in fact Democrat (CA36, FL14, FL18, GA12, MA6, MI11, NH1, NY18, TX23). Of some concern, I had some of these down as safe.
Reassessing the data I used for NE2 – the only data I can find in the post-election deluge of actual results – I should have ruled NE2 Republican, not Democrat. I suspect NE1, NE2 and NE3 (that is, the entire state of Nebraska) should have been called Republican, and I pasted it in the wrong column. I will spend some time tracking down the other data and working out if this is all the result of my clerical errors, or if I made some bad calls. I expect a mix, but hopefully none of my predicted “Safe” seats changed.
On the up said, I got 411 of 425 right (plus ten tossups which I called a 7 to 3 split in favour of the Republicans, but wisely didn't factor into the final results. In fact it was a 6 to 4 split to the Republicans) so that makes 96.7% accuracy. Many of those were predictable, but I'm not unhappy with that result.
I could list each governorship individually, but there is only one I want to pay detailed attention to; Montana was the only one I got wrong. I had Montana as a predicted Republican win, but it went all Democrat on me.
A predicted win, as opposed to a safe win, means there was some room for doubt when I called it, but I was confident enough to label it and move on. A predicted win means a closer election than a safe win, but this election was very close. The designation Tossup would have been more appropriate, and looking back on my meandering dialogue I wonder if I even considered Tossup as a possibility of gubernatorial elections. I will definitely remedy this next election.
So, how close was the election? 8,674 votes close. The relatively minor Libertarian Party got 17,364 votes, but as these parties rarely (if ever) win governorships I ignored them. I don't intend to change this, as minor parties just bog down the number crunching with no predictable value in the outcome. However, in a two-horse Dem vs Rep race, these 17,364 votes could have put the Republicans ahead.
Did people voting for the Libertarian Party throw my results? Possibly. In terms of it's political position it is pretty well in the centre of the major parties, so it's difficult to guess where the votes would have landed otherwise (or, indeed, whether those voters would have turned up). On the one hand the Libertarian Party is in favour of small government and less financial regulation – strong Republican platforms – while also supporting same-sex marriage and open immigration which are far closer to Democrat positions.
I guess, all in all, getting one predicted position wrong is par for the course. After all, if we could predict results with 100% certainty, we wouldn't need to vote.
Presidential prediction was accurate in terms of outcome, but well
off as far as vote numbers. Next election, closer attention should be
placed on electoral college divisions of votes. 100%.
Senate predictions were spot on. The tossups did not fall evenly,
and this should not be assumed in future predictions. 100%, excluding
House numbers were generally accurate. How much of the inaccuracy
stems from clerical error rather than poor prediction is under
investigation. 96.7%, excluding tossups.
Gubernatorial elections largely as predicted. One error in a very
close seat. Tossups should be considered next election. 90.9%