How to hire a junior quant – part III

I have nothing to add to the waterfall of comments on the current events which put the industry on the verge of collapse. I just can’t help but notice the striking similarities to the 2001 IT bubble. The technicalities are vastly different of course but here is the repeating scenario. The Wall Street arrives at some false and comletely unjustified risk-return assumptions about something – computer technology in 2001 and securitization of bad loans at present – and starts pumping obsene amounts of money into the idea. The law of supply and demand kicks in and everybody starts playing along. IT startups sprung up left and right back then and mortgage brokers set out to lure financially illiterate to unbelievably cheap loans now. Which brings up to the main subject of the recent posts in this blog.

I am talking about people going to professions for which they have neither aptitute nor skills. I already mentioned proliferation of java “developers” in 2001 and financial “mathematicians” in 2007 in this blog. Could it be that it was those people who caused the bubbles to burst? See, the programmer who has no clue about computer science can not deliver a software product of reasonable compexity. Just can’t. What happens? Eventually the startup begins to show signs of weakness, its stock plunges from $300 to $3, and the investors lose (huge) money. What about now? Could it be that the models developed by armies of clueless analysts underestimated the risks in the subprime loans? And the executives were too happy with the short-term bottom line in both cases?

I have recently had a questionable pleasure to observe analysts at two big financial services (the word “MBS” is somehow associated with both of them) companies in very different parts of the country (one of the firms is almost dead by now by the way). Those were supposed to be seasoned professionals – not some fresh graduates of 1-year “Master”’s program. The sight was sad to say the least. The mediocrity was flowrishing.

They say that the time of complex derivatives is over so the demand for finmath programs would inevitably fall. I would say it’s for the better. I wouldn’t admit to those programs anybody without graduate degree in applied math, statistics or physics in the first place. I would even say that the programs themselves are utterly wrong approach to the problem because what is really needed is degree in economics on top of degree in math/statistics/etc. That first degree would take care of the bigger part of the finmath programs automatically. Regrettably, this is not going to happen. Just like the standards for software developers were lowered during the IT bubble and never recovered, it looks like there is not much to expect in the area of financial math either.

But I digress. You know, I have been actually planning to write about the ultimate way to hire a junior analyst all along. At first I planned to have a post on technical questions. Of course you can have them calculate some tricky limit testing their knowledge of undergraduate analysis to the limit so to speak or you can give them an elegant problem to ponder but I don’t thik that is enough to assess sombody’s abilities for practical work.

In my opinion the best way to deal with candidates for a quantitative job is to give them homework (perhaps at the end of some short interview aimed to weed out grossly inadequate). The work may involve obtaining a data set, cleaning it, coming up with a model, implementing a model and analysing results. Or it can be analysis of several technical papers and implementing some ideas from them. Or coming up with ideas on some recent project that your team was working on. A few days later have them do a presentation of the results and go through detailed discusion on each and every point. That’s it. See, you don’t even have to come up with tricky brainteasers. If (s)he even shows up with the results – that alone tells a lot. The discussion would naturally highlight all the strengths and weaknesses of the person. Treat them as colleagues not enemies you have to defeat intelectually. You do not want to have enemies in the current market…

How to hire a junior quant – part II

Well it certainly looks and feels like after the recent subprime circus the whole industry is actively downsizing. The supply of junior (and not so junior) quantitative analysts is more plentiful than ever before. That brings memories of infinite amount of available java “developers” after the IT bubble of Y2K finally bursted. Consequently, it has become even harder to select a person with ample knowledge and skills which brings us to the second part of the series -

Brainteasers
One piece of advice – don’t! The brainteasers are said to assess “how a person thinks under stress”. Well, actually they turn out to be assessing completely different things of which the interviewers usually are/(choose to be) unaware. Admit to yourself that none of the brainteasers you are going to ask have been invented by you personally. You know that almost all of them come from folklore. Have you ever wondered how old they actually are? I am going to reveal this secret right now: the majority of the brainteasers I have encountered (in real life, in quant web forums, interveiw preparation guides etc) can be found in one ultimate source …drumroll… – the wonderful Martin Gardner’s books on recreational mathematics! You can find all the usual suspects there, the simplest ones being St.Petersburg paradox, “probabilty of second child being a girl”, the Birthday paradox and many more. Given the fact that lots of puzzles in those books are actually reprinted from his old column in Scientific American, it is a safe bet that a particuarly nasty brainteaser you are going to give to a shivering candidate is actually no less than fifty years old.

The point I am trying to make is that the chances that the interviewee has merely memorized all your clever brainteasers are very high. So by observing the correct answer you actually achieved nothing – and you have no reasonable way to figure that out. On the other hand this whole “thinking under pressure” concept has nothing in common with real life stressful “thinking” sutuatiations where one has a plethora of external resources to his/her disposal. Once again I reiterate the value of a skill of being able to locate an answer quickly by any means possible which includes finding it in books, research papers, web-published lecture notes, forums etc. Brainteasers provide no help here.

Finally here is one more thing that I have come to believe. By giving a brainteaser the interviewer acts exactly as a child giving a riddle to another child – deep down there always is unconscious hope to get a feeling of satisfaction from witnessing the counterparty’s failure. The reason why I consider that to be true is that often solving the brainteaser brings the candidate much fewer positive “points” than it brings negative ones when the question is not answered correctly.

The next installment will be devoted to the remaing part of a typical interview – technical questions.

How to hire a junior quant

I started writing this post about two months ago when graduation ceremonies were in full steam and cohorts of newly minted financial mathematics gurus were getting braced to join an army of candidates to junior quant positions. Little did they (and I) know of the future fruits of the summer: credit crisis, astronomical losses and massive downsizing across the financial industry and, most importantly (for them) universal failures of mathematical modes to, well, model. Anyway, if somebody is still hiring, this short note, I hope, would be of some use. Here we go…

Nowadays there are dozens of guides (paper and online) telling you what to expect in the interview, how to prepare, what to wear etc. In other words they are all written from an interviewee’s standpoint. Reading almost any of them will result in two sure results – 1) the jobseeker will get nervous to the point of breakdown of thinking (s)he does not know anything and 2) the quant interviewers are jerks drawing sadistic pleasure from a candidate’s sweating at unsolvable brainteasers and calculus tricks which are only remembered by undergraduate students the night before an exam. I was surprised to find out that usually there is nothing available to those who need to interview somebody other than a frayed printout of abovementioned brainteasers and some forum postings made by (un)lucky candidates who failed a particularly mean question in a recent interview.

So here is something for the other side of the game.

There are generally two kinds of people applying for junior quant jobs. First kind is fresh graduates of PhD or full-time financial
mathematics MS (full time is crucial here) programs. The second type is people with real life job experience who decided to change careers and/or have been unsatisfied in their current jobs. Most of them come from the same financial math programs which they did part-time. Here are their relative strengths and weaknesses.

Fresh graduates
This folk is supposed to have very sharp knowledge of the theory. Analysis, PDEs, stochastic calculus, they are walking encyclopedias of graduate mathematics. Don’t expect them to be able to match a theoretical model to practical problem though. They can derive anything from Black Scholes to BGM, but often unable to see what the actual problem requires. They also tend to propose the most complicated model known to them from textbooks. Don’t expect them to be decent programmers as well. The academic environment doesn’t teach robust software development and usually puts emphasis on correctness of the results given very limited ranges of artificial (or real but carefully prepared by instructor) data.

People with experience
Those are a very diverse crowd as they come from radically different jobs from pure finance to pure IT. All things equal, they are usually much much better at coming up with time saving trade-offs and shortcuts in modeling but can be somewhat rusty on the math. The latter should not outweigh the experience though as long as they can quickly locate the answer. Depending on background, you can find exceptional programmers among them.

Naturally the two groups need to be dealt with quite differently and you might want to decide upfront whether the position’s requirements eliminate one of the groups altogether. E.g. admit to yourself if you just want an assistant for boring stuff you don’t want to do yourself and to share your wisdom with or you need a more or less independent member of the team.

Just like in hypothesis testing there are two kinds of errors that can be made: you hire somebody who is not suitable for the job or you decide to reject a person who would turn out to be perfect candidate. Let us see how easy to make either of those errors asking usual interview questions.

Resume questions
First questions tend to be related to the resume – those are the easiest to come up with. People with experience should be able to tell you about their previous jobs but fresh graduates have not much to talk about in that department. They are normally asked about their classes and what they have learned. The real goal here is to determine whether there are any embellishments in the resume. Now to the errors. Normally it is very easy to determine if there is much more stuff in a resume than the person in question is actually skilled at so making this type of error has relatively low probability. On the other hand it is also easy to write off as unqualified somebody who did not include some of his valuable (for you) skills in the resume. Either they are too modest and do not think that the skills are that valuable and/or rare or they are from those perfectionists who does not consider themselves entitled to state in the resume anything that they are not 100% confident at. They are direct opposites of the people whose resumes are direct copies of their financial math curriculums. My preferred way to conduct a quant interview (see below) allows to deal with both.

After looking at what I have written so far I decided I should put the remaining material to the follow-up. In the next posting(s) I will tackle two other common types of interview questions – brainteasers and “technical questions”, and propose what I think is a very productive way to interview for junior (or not so junior) quantitative jobs.

Collection of ignorance

Continuing my rant on incompetency, the problem of poor general mathematical culture seems to be pervasive among financial mathematics (financial engineering, mathematical finance, computational finance, whatever catchy name universities come up with) students. I have a few favorite embarassments that I witnessed personally. The first happened in the middle of lecture on risk management when a student (who during introductory lecture boasted her advanced degree in statistics) asked why there was a division by n-1 (and not just n) in the formula for sample standard deviation. The lecturer probably thought it was hopeless to say words like “unbiased” and just resorted to “degrees of freedom” handwaving. I found the whole situation extremely ironical given the environment of graduate math program.

My second example shows the level of knowledge that students choose (not) to maintain after a class is over. A student is asked what Girsanov theory is about. This particular student has just finished rather technical class on fixed income derivatives and “sat through” two quarters of stochastic calculus a year ago. The answer he manages to come up with is literally “Girsanov theory helps us to switch to risk-neutral world where all math works out easily”. That is it. The word “measure” is not even mentioned.

Finally, the last case comes not from university but from, disturbingly, real life. A person calling herself quantitative analyst explains PCA to other “quants”. At some point she casually explains that eigenvalues are equal to lengths of corresponding eigenvectors. All the “quants” seemed quite comfortable with the explanation.

All those (and many other) observations draw some joyless conclusions, notably that financial mathematics jobs attract herds of people with abysmal mathematical capabilities. It also may just be the consequence of general decline in mathematical education (brilliantly illustrated in this paper by Vladimir Arnold). What is worse, majority of those people manage to get the jobs and that is not surprising at all given the fact that majority of hiring managers in not a top tier firms are MBA types with vague understanding of math. It would be hard to imagine those “mathematicians” thriving in a demanding environment which would put their skills to real test.

More examples for my collection are welcome.

Fooled by intuition

I have heard too many times in various financial math classes that one of the goals were to “develop intuition” about some concept. More often than not it turned out to be just a poor excuse for not going into technical details. I hope the reason for that was the low expectation of the audience’s intellectual capabilities rather than the lecturer’s sloppy preparation. The proper course of events would be, naturally, to get the students to firmly grasp the subject in the finest detail being as technical as possible and achieve lucid understanding. Tons of homework and discussions would help here. Then the intuition would grow by itself just by virtue of the concept’s being now a natural component of a student’s “cloud of knowledge”. All that is not practical of course if there is no foundation to build upon. In such a case the lame “we need to develop intuition first” is the only way to mark a topic as “covered”.

The recent article by Dan Goldstein and Nassim Taleb “We don’t quite know what we are talking about when we talk about volatility.” presents an example of glaring incompetence among portfolio managers, quantitative analysts and students of graduate programs in financial engineering i.e. people who are supposed to know a thing or two about volatility. It turned out that when asked for quick answer they did not really see a difference betwean absolute mean deviation and sample standard deviation. So much for the advanced degrees and important looks. Not surprisingly all of them could write the obvious formulas correctly, it was the intuition that fooled them.

“Those who cannot remember the past…

…are condemned to repeat it.” (George Santayana.) I doubt there exist many people capable of rediscovering stochastic calculus but it is surprising how many “financial mathematicians” claim to know it while demonstrating utter ignorance of the foundations. The 2004 paper “A short history of stochastic integration and mathematical finance the early years, 1880–1970″ by Protter and Jarrow documents the history of mathematical finance in great detail. I am convinced that knowing the events leading to discoveries of certain difficult to understand mathematical facts makes the facts themselves much more natural and accessible. If you have ever been puzzled by, say, Itō isometry or Doob-Meyer decomposition, the paper will show all the gradual refinements the ideas underwent. There are also also all kinds of less mathematical stories in the text – like how American and European options got their names.

In conclusion, here is another, less famous, part of the quote above: “…when experience is not retained, …, infancy is perpetual.”

Hello world!

WordPress.com supports !