News & Events

Researchers Studying Market Formula That Predicts Election Results

Brazilian researchers are studying a mechanism that uses characteristics of values and betting sites purse to try to predict the outcome of elections in the country. The system is already used in the United States and Europe and, according to studies, have been as or more accurate than traditional surveys of voting intentions.

The method is known as "predictive market." In general, is to create different types of "actions" of each candidate—as if they were companies—and make them available for trading on a virtual market through an internet site.

The price achieved for each type of paper enables researchers to make predictions about the winner of the election and even on the percentage of votes he can get.

But to work on a large scale in Brazil, experts say, the method will have to be supervised by authorities—to avoid being manipulated by interested parties.

In other countries, negotiations are made with real money—which has already attracted criticism for making the stalls like betting system—but will work in Brazil, at least during the research phase, based on a fictional currency.

The basic difference compared to traditional surveys of voting intentions is that the participant "bet" more resources for the candidate deemed to have greater chances of winning—instead of pointing the candidate who intends to vote.

Also, instead of having a proportional sample of the population, the technique aims to gather a group of motivated people who, collectively, have good information on the subject.

This "bag" attract "investors" who make decisions after analyzing voting intention surveys, market analyzes, and reports of sectors of commerce and industry.


The "predictive market" has the advantages of being cheaper than the polls on voting intentions and provide forecasts in near real time, according to researcher Ivan Roberto Ferraz, FEA (School of Economics and Business Administration), USP (University of São Paulo).

"A representative sample of the population, which reduces the cost of the method, and a predictive market can add a lot of information quickly, so you always have an updated thermometer is not necessary," he said.

The method has been used since the late 1980s abroad. The University of Iowa in the United States examined the application of the method in 15 elections held between 1988 and 2000 in the United States, Germany, Austria, Denmark, France, and Turkey.

She showed that in nine of these claims, predictive markets were more accurate than opinion polls. While traditional studies obtained margins of error of 1.9%, reached the market of the predictive indices from 1.49 to 1.58% (corresponding to two different predictive criteria used in the market).


Researchers from at least two universities are studying ways to implement the method in Brazil. One of the experiments is the site of BPrev (Stock Forecaster), USP.

According to Ferraz, responsible for the project, a group of participants being selected to act as "investors" and "sell" actions of candidates, with fictitious currency, during the election period. Stakeholders agree to participate in the survey filling a form on the BPrev ( site.

But because it is academic research, forecasts will not be made public before the official announcement of the result of the election. They will be compared with the polls and the electoral results to evaluate the effectiveness of the method.

According to the researcher, the site will also make predictions about how financial markets will react to the claim. After the elections, the tool have been tested for their effectiveness predictions about economics and even football championships.

UFMG (Federal University of Minas Gerais), political scientist Deivison Souza Cruz developed mathematical models based on the predictive method market that claim to predict not only the outcome of statewide elections but also legislative elections.

"There is no opinion poll on candidates for the legislature to be very precise, because the amount of competitors is immense. Percentage of votes of each is often less than the margin of error in the survey," he said.

The formula he created tries to anticipate the results of elections for councilors, MPs, and senators using the bag model, but taking into account other variables such as the amount of legislative seats that can be obtained by each coalition.

In the current phase of the study, Cross and his team are seeking funding to create a structure to process the data. They also want to submit an application to the Securities Commission, Ministry of Finance the authority to operate with real stakes in the future (with a maximum ceiling of about £ 1,000 per participant).

According to Cruz, although the project has the technical ability to operate already in this election (without using real money), will possibly be put into practice in the 2016 election.


According to the sociologist Valeriano Costa, professor at the Institute of Philosophy and Human Sciences at UNICAMP (University of Campinas), although it has existed for years abroad, the model's predictive power market face difficulties in Brazil.

"The degree of control over these bags must be clearly stated. 'Gambling' in the candidate have to be spontaneous and not directed, so that the market does not suffer from some sort of bias," he said.

"It must be an open and competitive market," he said. According to him, it is necessary that the participants do not communicate with each other and do not make their investments following a joint strategy.

The Superior Electoral Court told BBC Brazil there are no laws or cases ever tried in the country dealing with predictive markets.

According to the American sociologist David Fleisher, professor at UNB (University of Brasília), Brazilian researchers should find it difficult to make predictions about the elections in the country using the method due to lots of parties and candidates involved in the election—unlike the United States, where there are only two major parties.

According to Ferraz, BPrev the project, despite the peculiarities of Brazilian reality, the "bag" will be able to make predictions and, if successful, become an important tool in future elections. However, in the current phase of the research, the priority is not hitting the results, but learning more about the behavior of the predictive tool markets in Brazil.

(Editor's Note: This was originally published in Portuguese. This is a Google translation.)

Return to top of page