The State of Play One Week Out

So, with one week to go, how does the campaign look? Truth is, this remains one of the hardest elections to read of the last 11 years. Polling data has started to leak out in small dribs and drabs, but it remains nearly impossible to obtain what we’d really want: the Ficha Técnica. Without being able to scrutinize the actual methodology and periodicity being used to produce polls, it’s very hard to know what to make of things.

Worst yet, the bits and bobs that have found their way to my inbox are inconsistent, ranging from a massive government landslide to results that leave the opposition on the cusp of 83 seats. Alarmingly, Jesse Chacón’s GIS XXI is not the pollster with the most positive forecast for the government. (Apparently, GIS XXI has made an effort to get serious about its methodology after its 2008 polling catastrophe.) 

An undercurrent of panic is now noticeable in sectors of the opposition, as the government seems to be running strong in some areas that ought to have been safe for us. By some reports, the precriminations are already flowing around bits of the MUD, as some shift from campaigning to trying to figure out how to spin an embarrassing defeat to their advantage in the unending behind-the-scenes power games there. To my mind, this is insanely premature, as it’s by no means a settled matter what’s going to happen next Sunday.

Most of the bad news we’re seeing comes in the form of surveys where about a third of those polled say they’re going to vote for the government, a little less than that say they’re going to vote for the opposition, and a final third say they haven’t decided yet, or they won’t vote, or refuse to answer the question. 

Typically, pollsters take that raw data and apply a "likely voter model" to try to come up with a "polarized popular vote distribution". In effect, they end up assigning that final third either to abstention or to one of the two camps using methods that mix statistical sophistication, guessword and just plain voodoo. This is inevitable because, as one political scientist put it, 

When pre-election polls are in the field voters are a population that technically does not yet exist. While some individuals are nearly certain to vote and others are nearly certain not to (with a great many in between), the population of those who will actually vote is, prior to the deadline for voting, not yet a population, but in the process of becoming one."  [Emphasis in original]

I think this helps explain why so many have been so cagey about releasing results: more than telling us something about what poll respondents answered, polls this cycle tell us something about the pollster’s likely voter model. Or, worse, about the statistical method they use to compensate for the fact that they don’t poll small towns and villages. Luis Vicente Leon, to his credit, is careful to foreground the caveats. 

Because one thing we do know is that just asking people whether they plan to vote is not an especially useful way of finding out whether they actually will or not. Estimating likely voters is much more complex than that: part art, part science. And in this cycle, you have to add the additional wrinkles that are that screwy map and rural over-representation, adding a whole other layer of complexity to forecasting exercises. All of which drops us right back where we were last week: without any really clear polling to go on, but with a strong hunch that turnout is going to be a problem for the government this time. 

Honestly, my level of optimism about 26S has been all over the place these last few days: fluctuating depending on my mood, the weather, and the time of day. As you can probably tell, it fluctuates from one paragraph to the next in this very post! A week ago, I was sure we wouldn’t have a very bad day: now, I’m no longer sure about that. Really, the race is still wide open.