The new and much improved Caracas Chronicles Forecasting Tool for the 26S Elections is now available. You can download it here. (Users of older versions of Excel should download this version. Descárgala en español aquí.)
This new tool, developed with invaluable help from longtime-reader and professional-number-cruncher Omar, improves on the old-Swingometer in several ways.
First, you’ll notice I’m no longer calling it a "Swingometer" because, in this version, you forecast the Opposition’s Share of the Nationwide Popular Vote, rather than the swing. I think it’s much more intuitive to do it this way.
As in the old Swingometer, you type in a single number and the software does all the rest.
New sheets show region-by-region forecasts and an analysis of Urban vs. Rural voting trends. I also include a graph that illustrates the Cochinada: the way the CNE’s rigged map systematically screws the opposition by over-representing the places where the government is most likely to do well.
The biggest change, though, is under the hood. The new forecasting tool no longer assumes a "parallel shift" in the vote in response to changes in the government’s popularity. In plain words, it no longer assumes that all circuits move in lockstep with one another.
Instead, it models each circuit’s response to a change in the government’s popularity independently, on the basis of a statistical analysis of how that circuit has responded to changes in government popularity (as measured by Consultores 21 polling) ahead of voting in 2007, 2008 and 2009.
The way we do this is by using estimates of circuit-by-circuit point elasticities – i.e., the relationship, over the last few cycles, between the change in the government’s popularity at the national level and the government’s electoral performance in each circuit.
By estimating elasticities on a per-circuit basis, we are able say that while a 1% fall in the government’s national popularity "costs the government" 0.67% of the popular vote in Maracay, that same fall in popularity costs it 1.19% of the popular vote in Cocorote.
Now, the caveats – and there are many. This circuit-by-circuit exercise is prone to a lot of measurement error: we’re forecasting on the basis of very limited data – just a handful of data points per circuit, ahead of the 2007, 2008 and 2009 elections. This is – um – not ideal. But it’s the best we can do with the data we’ve got.
We believe – which is code for "we hope" – that the error that this thin database inevitably builds into the model will turn out to be unbiased: the circuits where we overestimate the opposition’s chances will hopefully balance out the circuits where we overestimate the government’s chances. For this reason, we’re considerably more confident about the overall forecasts this tool generates than about the circuit-by-circuit forecasts it yields.
We’ll find out on September 27th whether these predictions pan out or not. For now, all I can say with confidence is that this is the best forecast we can come up with given the state of the data publicly available right now.
Those links again:
- Download the .xlsx file – for versions of Excel released after 2007 – here.
- Download the .xls file – for versions of Excel released before 2007 – here.
- Descarga la versión en español aquí.