Last year, a group of researchers from 'Universitat de Valencia', a Spanish University, decided to join the quest for improving election outcomes prediction. The lab was excellent, since Chancellor election was going to start, and this year there were 4 candidates, a number unusually high. Even further, the fact that out subjects of study were university professors gave us a controlled environment… or this is what we thought.
Taking advantage of the options that current technology offers, for the first time in Spain, it was performed a prospective study of an election process in an academic institution. This work tried to test three forecasting procedures (a Delphi panel of experts, an on-line survey to professors and an information market) in order to reduce the uncertainty of the process, to bring information to the people of the university and to compare the goodness-of-the-fit of the forecasting approaches. We wanted to gather as much experience and knowledge on all three methodologies, in order to assess the different results and warnings about the difficulties and problems that a project of this kind could find. In fact, a paper relating our little adventure is being refereed at the moment.
Regarding the obtained outcomes, the survey forecasts were considerably more accurate than the opinions of the experts, although, as it happens in almost all the electoral surveys, ours also suffered from non-response bias; moreover, later on we managed to know that 'the dark side of the election machinery' moved the strings in order to add noisy signals through the survey respondents. And for the information market, a tool that is already being successfully used to predict electoral results in the US, and the technique we most wanted to test and that we managed to make operative through a website and text messaging, was restrained and couldn't be researched and compared with the other two. Regretfully, we were accused of organising a raffle with the election outcomes and trying to enrich ourselves with this last experiment. Holy cow!
Hopefully, no researcher was injured during the experiment, neither legal proceedings were taken against.