The final seats have yet to be allocated, so I’ll be reserving my comments on this until after the last count, but at this stage we know the FPVs of each party, and I thought it might be timely to review how the different organisations (and my good self!) did in trying to gauge how many votes each party would get, that being the idea behind such polling in the first place.
Most of you will be well aware that MRBI and Millward/Lansdowne have been rubbished for some time by some FG posters on politics.ie in particular, despite having as good a record in General Elections (I’ve outlined elsewhere why I believe you can’t compare effectiveness in predicting GE results with other elections). The posters generally pointed to RedC as the most reliable, despite having no better (and arguably a worse) record in predicting GE results. This time, despite (yet again) being held closer to polling day than the other companies final pre-polling day efforts, RedC were actually the furthest out of all the respected polling companies, as I suspected they might be.
The last MRBI poll, which I flagged as the one to watch (having been extremely close in ’07), was the first of the “final” polls to be published, leaked on Sunday evening and published in Monday’s Irish Times. They were out by 0.9% for FG, 0.4% for LP, 1.4% for FF, 1.1% for SF, 0.2% for GP, and 0.4% OTH, making their total deviation a very impressive 4.4%. Millward Brown/Lansdown came out the following day and were 1.9% out for FG, 0.6% for LP, 3.4% FF, 1.1% SF, 0.8% GP and 0.6% OTH, making their deviation a less impressive (but still solid) 8.4%. The last official poll of the campaign came from RedC (as in 2007) and as in the previous GE, they failed to make this advantage work for them, with the least accurate prediction of the final vote. They were 3.9% out for FG, 1.4% for LP, 2.4% for FF, 0.1% for SF, 1.2% for GP, and 1.4% for OTH, a whopping 10.4% overall.
As regular readers will be aware, I went out on a limb and posted my predictions on polling day and my figures were out by 1.8% for FG, 0.1% for LP, 1.1% for FF, 0.1% for SF, 0.2% for GP and 0.9% for OTH, a total deviation of 4.2%. And finally, another MB/L poll, this time an exit poll for RTE, gave figures that deviated by 0% for FG, 1.1% for LP, 2.3% FF, 0.2% for SF, 0.9% for GP and 0.1% for OTH, a total deviation of 4.6%.
In summary, the total number of percentage points the last projections were out by was as follows;
Now, that’s not to rubbish RedC, polling is a complicated discipline, but rather I think it does shoot the credibility of those who ridiculed the other company results, particularly people who poured scorn on MRBI for recording figures that they didn’t agree with, and who took particular exception to my basing projections on their data. MRBI by contrast were the closest of all the polls in the final week (even better than the Exit Poll) and the only prediction I’m aware of being closer was mine, which was based in large part on their poll anyway.
So next time you hear someone scorning a polling company, check who they are, who they support, and look at the evidence.