IT IS hard to make predictions, the old saying goes, especially about the future. When future involves Brazil’s presidential race, the first round of which takes place on October 5th, the task is harder still. That has not stopped number-crunchers trying. Neale El-Dash of PollingData.com.br, a website, has made a valiant attempt at “tropicalising” Nate Silver, a statistician and blogger who rose to stardom during the 2012 US election. Mr Silver took polls released each week, then aggregated and weighted them to come up with a prediction, framed in terms of probability of victory for the main contenders. Our chart shows how Brazilian hopefuls’ chances, calculated in a similar fashion by Mr El-Dash, have shaped up since the campaign was upended by the tragic death in a plane crash in mid-August of Eduardo Campos, a centrist candidate.
President Dilma Rousseff’s probability of re-election went from 55% in mid-August to 25% a few weeks later. At that point Marina Silva, a popular former green activist and running-mate to Mr Campos who had vaulted to the top of the ticket, began to look like a shoo-in. But the past weeks have seen Ms Rousseff resurgent, as ample media exposure let her play up own successes and bash her main rival, who has also been on the receiving end of attacks by the third-placed Aécio Neves from the main centre-right opposition.
Why have Mr El-Dash’s numbers swung so wildly? There are several reasons, ranging from the nature of Brazilian democracy to the niceties of statistical analysis.
Start with the democratic process. In the months before the first-round vote the race in the country is akin to American primaries, notes Mr El-Dash: a riot of hopefuls with varying name-recognition vying for undecided voters. Voter allegiances shift more readily in such conditions, affecting polls, and therefore poll aggregates, too.
Full Article: Gauging Brazil’s election: Crunch time | The Economist.