TheTostiMonster / YouTube NCP founder Matt Singh was previously an interest rate trader at Barclays Investment Bank.
The General Election was not just a dark day for Labour, it also was for Britain's pollsters.
The most reputable polling firms - from YouGov, to Opinium, to ComRes - all agreed that Labour and the Tories were neck and neck heading into voting day and that a hung parliament would be the outcome.
The final forecasts were so inaccurate that the British Polling Council set up an inquiry into why the Conservative vote was so dramatically understated.
Yet, while the results of last year's election left Britain's top pollsters reeling and red-faced, Barclays interest rate trader turned polling analyst Matt Singh became a political celebrity overnight.
Singh told Business Insider that his political blog, Number Cruncher Politics (NCP), was set up originally as a "bit of fun" and as a means of allowing him to combine his knowledge of the markets with his interest in
On the eve of the general election, Singh was the only analyst who was predicting a Conservative majority victory.
Now, Singh is a wanted man in the worlds of politics and finance, and all eyes are on the research of Number Cruncher Politics ahead of the EU referendum. Business Insider spoke to him on Monday morning.
Business Insider: How did it feel when you found out that NCP's General Election forecast was correct?
Matt Singh: It always feels good to be vindicated, particularly in light of how some people reacted to the prediction when I first put it out.
I remember my first reaction when the BBC exit poll was released was "let's get the tweet out" because by that point I only had a few thousand followers and I was quite keen to cash in on the success. As the hours went on and the results came in I thought to myself this is quite a result.
BI: What do you mean by how some people reacted?
MS: I write as a neutral. I don't make judgments or give subjective opinions. But anytime I make a prediction about what will happen - as opposed to what should happen - it upsets people who are on the wrong side. Last time it was mainly Labour people who were upset because it showed Labour doing badly.
I don't mind sensible debates about stuff. When people start being aggressive I just ignore them. Some people enjoy debating with trolls and taking them on but I only tend to reply to people who have sensible questions or want to have an adult debate.
Below are extracts taken from the post that Singh published on May 6, last year which predicted a majority victory for the Conservatives.
BI: So, where did the opinion polls go so wrong in 2015?
MS: What happened essentially was a problem that had been building up over time but previously hadn't caused a big effect. Polls don't have to reach everyone to produce decent results but samples need to be representative of the population. The problem at the last general election was that pollsters had samples that were not representative.
Now, a sample can be made to look representative: they can get the right proportions of men, women, ages, regions, occupation types, who they voted for in 2010, and so on. But within that sample, you still have lots of other variables.
For example, most polls got the UKIP vote share completely right. What they got wrong was where the increased UKIP vote was coming from. Pollsters thought it was coming far more from the Tories, but in reality, it was coming almost evenly from the Tories and Labour. So, as a result of not having a representative enough sample, they got the movements from voters from one election to another wrong.
BI: Do you feel greater pressure to be accurate after your success last year?
MS: There's certainly a lot more attention from markets, media, both sides of the campaign, and the public. Having been a trader I am kind of used to pressure. It's a different kind of pressure as its so public but it's just something that comes with what I'm doing. It's not something I think about day to day.
BI: What sort of work is NCP doing ahead of the EU referendum?
MS: I am doing an ongoing forecast, which is now live on the website. There are three steps to it: Firstly, aggregate the polls. Then we analyse whether the polls are actually right. Thirdly, we predict how public opinion is likely to change in the run-up to the referendum. For example, we know from the past referendums that the closer you get to voting day, the status quo option tends to do better.
But our forecast doesn't just give you a single prediction but a range of possible outcomes and how likely they all are.
Singh has been working with pollsters and advising them on how to make their samples more representative.
MS: The big difference between the EU referendum and the general election is before the general election you had pretty much all the polls saying the exact same thing. They tended to differ on things like the UKIP share, but in terms of the Labour-Conservative gap, they all had it at neck and neck and all turned out to be wrong.
This time, they're all saying different things so the question is which one is right? If this widespread trend continues, then some polls will likely get it right, but others will get it really wrong.
BI: What does the future hold for NCP?
MS: At the moment, the international interest has never been higher. Last week, more than half of my traffic was from outside the
BI: What work have you been doing outside of the blog?
MS: The work I do outside the blog is a lot more technical. It's helping pollsters look at how they can be more representative, where they can change methodology, and how they can do things differently. On the financial side, companies want to talk to me about probabilities and what my research means for their own analysis.
BI: Finally... what's gonna happen in the EU referendum?
MS: The forecast is live on the site and at the moment, the model is suggesting a 78% chance of Remain winning.
This is a bit higher than what the betting markets are saying but not by a huge amount. The caveat to that is we haven't had that many polls in the last couple of weeks so the data it's based on, is a little bit older than I would like.