• C++ Programming for Financial Engineering
    Highly recommended by thousands of MFE students. Covers essential C++ topics with applications to financial engineering. Learn more Join!
    Python for Finance with Intro to Data Science
    Gain practical understanding of Python to read, understand, and write professional Python code for your first day on the job. Learn more Join!
    An Intuition-Based Options Primer for FE
    Ideal for entry level positions interviews and graduate studies, specializing in options trading arbitrage and options valuation models. Learn more Join!

PhD in CS VS. Master in CS + Master in Finance

Joined
3/29/08
Messages
77
Points
18
Which of the following two backgrounds is more competitive for a decent quantitative position at Wall Street (e.g., portfolio management or automatic trading)?

1) PhD in Computer Science from a good CS program, major in machine learning and statistical data analysis.


2) Master in Computer Science + Master in Finance (assume both programs are decent).

I'm now in a good CS program and will get a master degree next year. I'm considering to continue my PhD in CS or to apply another finance master program (e.g., MF @ Princeton). I don't know which choice is more promising for my future career life.

Thanks.
 
Just out of curiosity, have you worked in finance before? If not, I would look for jobs or internships before committing to a PhD program or another MS program. I'm sure you'd hate to finish 47 years of school before realizing you don't like what you prepared yourself for.
 
Just out of curiosity, have you worked in finance before? If not, I would look for jobs or internships before committing to a PhD program or another MS program. I'm sure you'd hate to finish 47 years of school before realizing you don't like what you prepared yourself for.

thanks for reply.
btw, i'm already in a phd program. my choice is 1) to finish it or 2) to leave with only a master degree and go for another MFE.
 
Choice
1) Finish your PhD. If you can't find a job, people still call you Dr.
2) Got master + MFE. Regardless of whether you find a job, people won't call you Master.
I would pick choice #1.
Based on your questions, I can see that you have little idea of what kind of career tracks you want. Please look at my signature, go to the Master reading list for MFE and read all the career guide on the first part.
It helps me a lot.
 
To add what other have said, focus on getting the skills and the profile needed to succeed in this field. A degree is like the candy wrap, it packages the goodies within and make it presentable and appetizing. But it is the candy what the employers are looking for. From that standpoint, choosing phD or MS does not really matter in the grander scheme of things (other people might disagree).

I see that you are deciding between derivatives pricing and empirical finance/algo trading careers...

So like Andy suggested, go over all othe reading materials listed on his sig file. That should help you decide.
 
Choice
1) Finish your PhD. If you can't find a job, people still call you Dr.
2) Got master + MFE. Regardless of whether you find a job, people won't call you Master.
I would pick choice #1.
Based on your questions, I can see that you have little idea of what kind of career tracks you want. Please look at my signature, go to the Master reading list for MFE and read all the career guide on the first part.
It helps me a lot.

Andy makes a good point in that the PhD does leave you with more options overall, especially if you want to go back to academia. Since you are talking about positions in either portfolio management OR algorithmic trading...well, I'm sure you know they are very, very different. I would highly suggest looking for jobs or internships to help you flush out some of these choices. I've worked with people who had top CS degrees (MIT, Cal Tech, etc) and while they were absolutely brilliant when it came to software development, they could not handle the stress of automated trading. Its a very different atmosphere than most people anticipate. And that's not to pick on CS graduates--there are a lot of MFE students who really have little idea of what they're getting into.
 
Choice
1) Finish your PhD. If you can't find a job, people still call you Dr.
2) Got master + MFE. Regardless of whether you find a job, people won't call you Master.
I would pick choice #1.
Based on your questions, I can see that you have little idea of what kind of career tracks you want. Please look at my signature, go to the Master reading list for MFE and read all the career guide on the first part.
It helps me a lot.

thanks. I will read.
 
To add what other have said, focus on getting the skills and the profile needed to succeed in this field. A degree is like the candy wrap, it packages the goodies within and make it presentable and appetizing. But it is the candy what the employers are looking for. From that standpoint, choosing phD or MS does not really matter in the grander scheme of things (other people might disagree).

I see that you are deciding between derivatives pricing and empirical finance/algo trading careers...

So like Andy suggested, go over all othe reading materials listed on his sig file. That should help you decide.



Thanks for suggestoins. I am mainly interested the quantitative roles in finance who mainly use statistical data analysis and machine learning. So perhaps asset/portfolio management or high-frequency automatic trading correspond to such roles.

To choose different programs/degrees, I am also choosing different education and training. To finish my CS PhD, I will get a deep understanding of statistical machine learning, but relatively lack of finance knowledge (merely self-study a few courses). By choosing a CS master + a Finance master, I will get much more finance knowledge, but not that deep expertise in statistics.
 
Andy makes a good point in that the PhD does leave you with more options overall, especially if you want to go back to academia. Since you are talking about positions in either portfolio management OR algorithmic trading...well, I'm sure you know they are very, very different. I would highly suggest looking for jobs or internships to help you flush out some of these choices. I've worked with people who had top CS degrees (MIT, Cal Tech, etc) and while they were absolutely brilliant when it came to software development, they could not handle the stress of automated trading. Its a very different atmosphere than most people anticipate. And that's not to pick on CS graduates--there are a lot of MFE students who really have little idea of what they're getting into.



Thanks for your information. I mentioned portfolio management OR algorithmic trading because my quantitative background (statistics, machine learning, AI, etc) is useful in these tasks. On the other hand, derivative pricing is not suitable for me since I'm not expert in PDE and Stochastic Calculus.

But what you mentioned about automated trading is very interesting for me. In fact, I also consider that quantitative portfolio management is the most feasible track for me.
 
PhD is something that not everyone is willing or able to do. If you think you want it and can do it, why not try getting PhD first. You can do MF @ Princeton later or at the same time while working on your thesis.
 
I suppose I would go for the CS PhD, on the grounds that you can learn the finance as you go along.
I can't say it brings light to my cold flabby heart as a headhunter though....

I've been talking to CS Machine learning PhDs recently and one way they go wrong is that it may
be the Machine that is learning, not the student. Too many CS ML PhDs seem to involve using some scabby package, resulting in a "cloud" thesis, ie 30,000 words that basically say "I got data to a form the cloud found acceptable and the cloud shared its wisdom with me. I now am wise in the ways of the cloud, blessed be the cloud.
Some cloud PhDs consist wholly of humping data until the cloud finds it acceptable.

But...
If you get to work on real data, and write your own code (preferably in C++) then you may learn some valuable stuff.
 
I suppose I would go for the CS PhD, on the grounds that you can learn the finance as you go along.
I can't say it brings light to my cold flabby heart as a headhunter though....

I've been talking to CS Machine learning PhDs recently and one way they go wrong is that it may
be the Machine that is learning, not the student. Too many CS ML PhDs seem to involve using some scabby package, resulting in a "cloud" thesis, ie 30,000 words that basically say "I got data to a form the cloud found acceptable and the cloud shared its wisdom with me. I now am wise in the ways of the cloud, blessed be the cloud.
Some cloud PhDs consist wholly of humping data until the cloud finds it acceptable.

But...
If you get to work on real data, and write your own code (preferably in C++) then you may learn some valuable stuff.


thanks.

You are right. a lot of CS researchers are not working towards a good algorithm that works well on real-world data. What they do is to design an algorithm, and then to FIND SOME DATASETS ON WHICH THE ALGORITHM WORKS ...... as a result, most knowledge I can get from the PhD might be something that does NOT work at all in real-world finance problems.

This is quite disappointed, and this is also why I want to pay more attention to the finance world --- absolutely a real-world area that needs some really working algorithm. But to design algorithms for finance, I need enough finance knowledge --- so I'm considering a finance master program. Such knowledge might be more useful than the knowledge I can get from a CS PhD program.
 
I agree about the selectivity of data sets, and it's a hard one to call.
For instance Fourier analysis is a perfectly respectable math technique, but is insanely useless to use on interpreting a yield curve.

However, that does give me an idea for a valid bit of research, based loosely upon the the Hearsay/Blackboard AI artchitecture.

A system that actually does the searching through huge volumes of data, recognising ones it can do useful things with, and rejecting those it does not handle, would have value.

A big problem with neural nets, genetic algorithms et al, is that it is hard to trust them when you don't know what their limits are. To me intelligence, or even competence in an area must include the ability to say "I have no idea what this crap is, I'm not getting involved."

But like a lot of people I find myself drawn to AI in finance, indeed I tried to push this many years ago, and learned a really quite impressive set of negative things. As in "X1 does not work, X2, does not work, X3 does not work..."

At the risk of sounding like a buzzword generator I think a thesis of the form
Signal processing in high frequency markets using techniques from electronics, econometrics, and CS. (using C++ of course)
As I said recently in different QN thread, this is currently the most valuable collection of skills you might plausibly have.
But...
Whenever I give advice on PhDs I am exposing you to a big risk.
I have a reasonably good model of what is desired by employers, at this time, and of course what they have wanted in the past. If I was so smart I could predict the markets in 5 years time I would be doing very different things with my time :)

Signal processing is quite hot, and C++ is easily the dominant language.
But...
Two years ago, 95% of entry level quant jobs were C++ and/or VBA
That's dropped to the high 80s.
C# has grown big time and we expect this year that over 1% of quant jobs will ask for Java, and Java has now overtaken both Smalltalk and Fortran.
SP was extremely rate 3 years ago, and will never be the mass market, and my partner Paul Wilmott is always happy to share the fact that new cool techniques come and go, but PDEs are forever.

Ultimately, I would counsel you to honestly work out where you are better than other people.
A really good neural nets person will do better than a mediocrity in time series analysis.
From that set identify those things that you think will be demanded from banks.
 
Well, thank you, I suppose at one level I was hoping you'd come up with counterexamples. To be fair, it is not restricted to CS. I see physics PhDs who are essentially doing the same thing with large sophisticated, yet obscure equipment.
 
I agree about the selectivity of data sets, and it's a hard one to call.
Thanks for the reply and your time.

Sometimes it really painful to make a balance between "to study what you are better than others/what you are really interested in" and "to study what is obviously useful for finance". PDE/stochastic calculus are forever in quantitative finance, but they are not my interests ...

Some MFE program has a quantitative asset management track, which seems great to me, since statistics, machine learning and optimization might be the right skills for this track.
 
I'm glad you have the self knowledge to work out things you aren't great at.

Optimisation, especially classical methods has some demand, as does signal processing.
Multi factor optimisation in the domain of market impact models is also worth looking at.
 
Back
Top