After the brawl, the duel. The French Presidential election has been in full swing for the best part of a year for many voters, but we’ve now entered the final fortnight where the top two candidates face off. Polls show Macron consolidating his position and his lead returning to roughly 55:45, but I feel markets are too confident that the polls are spot on.

Learning from our mistakes

The French polls are good, but they aren’t infallible.

Back in 2017, Macron defeated Le Pen very soundly (66:34). What is sometimes forgotten is that he defeated her even more convincingly than the polls predicted. The final week saw him average just over 60% in the polls, an overperformance (or polling error) of about 5%. Since the result (Macron winning comfortably) was what people expected this didn’t raise too many eyebrows, but it shouldn’t be ignored*.

The same was true last weekend. Macron v Le Pen was forecast and delivered: But Melenchon beat his polls by a solid 4-5%, getting 22% of the vote when he was polling 17-18% in the final days of the campaign.

A surprise is underpriced

It is likely that the polls will be roughly right, and a narrow Macron win has got to be a best guess for what the future holds. But betting markets are acting as if anything else is almost unthinkable, and it isn’t.

Back in 2017 Le Pen was considered to have crashed and burned in the sole head to head debate between the pair – what if she trounces Macron this time (or he trounces her again)? A shift of a few points in the polls between now and next weekend is still plausible.

It is for this reason, and the historical margin of error discussed above, that The Economist’s statistical model for the election has Macron as ‘only’ 83% likely to win (at time of writing). That’s a strong favourite, but it isn’t as strong as betting markets think.

Le Pen’s odds, down to barely 3/1 or 7/2 before the 1^st Round, have swung out to longer than 8/1 on the exchanges and not far off that at various bookies. Meanwhile, markets offering odds on the margin of victory have become fixated on a result within a very narrow range around the current polling.

William Hill, for example, will give you less than evens that Macron polls 50-55% on election day – despite him polling very near the top of that range already. 7/4 is available on him getting 55-60%, and a rather massive 16/1 on him polling 60-65%. That figure, which William Hill gives an implied probability of roughly 6%, would be more likely than not if the campaign pans out as it did last time: With Macron growing his lead and then beating the polls like he did in 2017 and Melenchon did just last week.

I’m not saying that any of these surprises will happen, they probably won’t. But betting is about the odds you get compared to the probabilities you forecast: And these ‘unexceptional upsets’ have become significantly overpriced. I cashed out my Le Pen position after the first round when her odds were 4/1, but have bought back in a little as the odds continue to shift. I’ve also put a small sum on Macron getting 60-65%, which seems ‘unlikely but entirely plausible’ and far too long at double-digits.

*(Incidentally, the same thing happened in the UK in 1997, one of the biggest polling misses for decades. Labour were polling almost 50% in the final week or two, but got ‘only’ 43% on election day.)

Pip Moss posts on Political Betting as Quincel. He has bets on Le Pen winning at 8/1 (but will profit from either candidate) and Macron winning with 60-65% of the vote at 16/1. You can follow him on Twitter at @PipsFunFacts