I interviewed with 22 people to get a job at Goldman Sachs. But a computer could have done a better job
“I was interviewed by 22 people before I was hired by Goldman Sachs,” says Dr Ewan Kirk, chief investment officer at quant hedge fund Cantab Capital Partners. Algorithms might have done a better job though, he believes.
Getting into any front office finance job is a tough slog, and only a tiny proportion of people make the cut. But banks, private equity firms and hedge funds could make better hiring decisions if they take people out of the process, says Kirk.
Kirk ran a team of 120 people working in Goldman Sachs’ Strategies Group in London before co-founding Cantab Capital Partners in 2006. Cantab has said previously that it receives around 700 CVs a year, but only hires a handful of people.
Cantab, he says, once tried to base its hiring purely on the CVs it received. “We decided that we’d just hire them, rather than go through the interview process. Then set them loose in a quantitative systematic mathematics environment,” he said.
Did it work? No. “At least half the people who apply are probably nuts, so you have to actually meet them to find out whether they’ll fit in,” he said. “Nuts is probably the wrong word. They’re just too clever.”
Generally, though, the recruitment process should be more automated, he believes: “There’s a whole load of research that shows people in interviews pick people like them, which is a bad a thing. They pick the wrong people, and could make better decisions if they based it on anonymised CVs,” he said at the London School of Economics Alternative Investments Conference.
To some extent this is already happening. Investment banks have started employing the services of the likes of HireVue during the recruitment process at a graduate level. It uses "predictive analytics" to sift through 15,000 traits of existing or past top performers at the firm and applies this to an automated interview process for potential new recruits.
Kirk delivered a talk about bias against algorithms in finance - or what he called "systematic squeamishness" - and suggests that most people in going into the industry have a skewed view of computer-driven strategies. They view them as more prone to error, and are much less forgiving of machines making mistakes than humans, he said. Algos are better than humans in almost every facet of life, he added, but they still make mistakes and this is what is seized upon.
"This is my thing. I'm right at the geeky end of finance - as massively geeky as you can get," he said. "You probably aren’t as geeky as me, which is why you haven't thought about statistics and find it very difficult to believe that a computer can do a better job than you can."
Algorithms are only correct around 52% of the time, he says, but it's human nature to focus on the 48% time they are wrong.
"If there’s an algorithm which would give you a set of rules that would make you better at your job, why would you not use it?" he said. "It's very hard for people to use these rules as they feel like they’re deskilling their job, or deskilling the edge that they have. But these rules can ultimately make the outcome better."
Photo: Getty Images