2017 French Election Macron Victory, with slight outperformance
On the eve of the second round of voting of the 2017 French Presidential Election, the SMI continues to point to Emmanuel Macron as the winner. The consensus call is that Macron will in fact win tomorrow. As the SMI forecast is in-line with consensus it might not appear of interest. However, SMI appears to have been the first to forecast a Macron win on December 12, 2016 with this post “Social Media: Emmanuel Macron France’s Next President”.
At the time, he was struggling as a candidate, with PredictIt, a prediction betting site, giving him just an 8% chance of becoming the president. He was on few radar screens and it does not appear that any forecasting method, other than SMI, gave him any real chances of winning the election until months later.
Macron, had (and still has) many strikes against him. He is not an experienced politician, he founded his supporting political party in 2016, he is extremely young at 38, and he had worked for the government of the Socialist Party (linking him to a highly unpopular government). Additionally, there were multiple other candidates with much more positive backgrounds. Even under such unusual real-world circumstances, SMI quickly identified Macron as the leading candidate, while depreciating most of the others, including the then-leading Fillon and Le Pen.
One of the most interesting insights into SMI is that it tends to provide information well ahead of other sources such as betting markets, polls and pundits. In the French election, it appears to have once again provided fairly accurate information months in advance of other sources.
In terms of second round voting, SMI is in tight agreement with polls. Averaging the last week of polls (which includes eleven individual polls) shows that Macron has 61.7% and Le Pen 38.3%. SMI puts their respective support levels at 63.4% and 36.6%. In other words, SMI is generally in-line with polls and approximates the most bullish poll over the last week for Macron. SMI is pointing to slight out-performance for Macron, but generally within the margin of error for the more Macron-positive polls.
An interesting point to note is that Social Desirability Bias does not appear to have had a significant impact on the French election. Large and consistent divergences between polls and SMI during the US election highlighted the existence of Social Desirability Bias. In the French election, however, there has been very limited evidence of a similar phenomenon. So, though some political observers have mentioned the possibility of hidden Le Pen voters coming out on election-day to upset the polls, this does not appear to be likely.