Voodoo analysis: Goldman Sachs in trouble
Again, the title is not a fact, but a premonition. Since my last premonition (sadly) came true, maybe there is something to this voodoo analysis after all.
One aspect of the subprime CDO and subsequent contagion that hasn’t gotten much attention (yet) is the influence and impact of risk management systems. In one of my first articles on the blog, I wrote about CDOs and risk management. If you’re anything like me, you’re sick of bloggers saying things like “As I said way back when…” implying that they saw everything coming long before everyone else, so I’ll try to tread lightly.
About a year and a half ago, Bill MacMahon, then head of risk management at GS (not sure if that is still true today), came to our (I should say my former) office to give a presentation on risk management at GS. I’ve never seen such a hard sell in all my life. I felt like I was watching an infomercial. He was completely confident that they had their risk under air tight control. If I hadn’t spent time working in risk management myself, I might have believed him, but I have worked in risk management and I’ve seen what goes on. I know how things are measured and there is a lot more subjectivity than those risk reports containing 4 digit risk metrics might imply. Since that time, I’ve told other analysts that I thought GS would be the first major broker to bite the dust if markets crashed.
I’m not saying I think we are in the midst of a market “crash”, but certainly things are ugly and getting uglier. Especially for structured products like CDOs. Risk management systems across the board are now throwing yellow flags and risk managers are sweating because they know that their risk systems have not adequately quantified the risk associated with CDOs.
I think it is possible GS bites the dust. That would be really bad for markets (apocalyptically bad). Regardless, I think they are getting hit really hard now due to repricing of risk, particularly in CDOs.
I have a love/hate relationship with GS. I would love to work there, but their interview process rubs me the wrong way so the chances of that ever happening are nil. One of my first job interviews was with GS in their credit derivatives group back in 2002. I had no idea what I was walking into. I had only slept about 1.5 hours the night before [Back in grad school, I was a machine. One of my mottos was "sleep is for undergrads!"]. They flew me in from Chicago the day of the interview, so I woke up around 2:30am, left the apartment around 3:30am, flew to NYC, arrived at their office around 9am, just in time for the interview to begin. I thought the interview MIGHT last 3-4 hours. First 4 hours I was kicking @$$, but was starting to feel the lack of sleep (plus travel fatigue) creeping in. At the 6 hour mark, I was still sharp but noticeably fading. By hour 9, I was barely functional. Who would have thought an interview could last 9 hours?!?! The last interview was with a guy who had a PhD in stochastic PDEs and he totally mutilated me. I didn’t get the job *surprise*.
Anyway, that was my first dose of those irritating “brain teasers” that quant interviews are so famous for. One Russian dude asked me a question about an optimal path from point A to point B given that you have to first touch some point on a separate line. I simply noted the analogy to light rays and quoted Fermat’s principle and invoked the law of reflection. Done in seconds. His response? “I don’t like that solution, give me another one.” I thought for a second, and noted a more traditional approach which might involve writing the distance function and minimizing it, which obviously yields the same answer. Again, “I don’t like that answer either, give me another one.” At this point, I complained, “I just gave you two perfectly good answers. What is wrong with them?” He replied, “You American’s are so stupid. Every Russian high school student can answer this question in their sleep!” I challenged him to show me his answer. He then proceeded to “reflect” point B across the line and drew a straight line from point A to the reflected point B’. Sure, that is a perfectly valid solution, but I didn’t think it was particularly more insightful than invoking Fermat’s
theorem principle. ARGH!
Anyway, that experience left me scarred. A few years later, I was making a switch and decided to give GS another shot. We set up a phone interview and the person on the line started asking me about Fibonacci numbers. I had flashbacks. I told the guy, “If these are the kinds of questions you’re going to ask me, I think we’re done.” I think I am probably the only person to ever hang up the phone on GS during a phone interview. Probably left me blacklisted
I love GS. I think their great, but I’m just not willing (or able) to jump through the hoops required to get through the interview process. C’est la vie.