I submitted this link because I fervently disagree with the blog post. I started a Econ PhD but left soon after as I realized that much of the economic theory which is taught is not realistic and not even a good proxy for how economic systems work. Computer Science has allowed researchers to build much more realistic models using agent-based modeling. When looking for econ PhD programs, make sure this is something they will allow you to research.
The best skills you can get out of most traditional Econ PhD programs are econometrics and programming skills. Outside of that it is pretty much a waste of time. Yes, you can probably get a job quickly, but you will spend a lot of the time learning none-applicable skills.
I want people to read it and hopefully we can start a discussion about it on HN.
I agree completely. I studied Finance but had a work study with an econ professor and was amazed at how much hand waving existed in their economic models.
I felt like I was charting very simple things(gdp, oil prices, ect) and he would come up with his own suspect analysis off that limited info.
The worst part was that this guy had had a high ranking position in the presidents cabinet!
I don't really see the problem with that. The only problem I have is with people who do this and go around pretending that their numbers are neutral and speaking for themselves. From my standpoint Econ & Finance are social sciences where "reality" is always complexer than just the numbers.
UMich is not really emblematic of mid-level schools. While rankings don't mean a lot, UMich is still ranked #13 by US News & World Reports for graduate-level economics and thus would be either one of the lower top-tier schools or near the top of the mid-tier. It would probably be easier to get a good job from UMich than the vast majority of other mid-tier schools.
I think he is referring to abject failure. As he points out, failure to get a PhD, simply results in a MA or MS.
As for the work schedule, that 80 hours a week may be a case for a tenure track professor, but the bulk of economists are employed by the government. In government work, you rarely work beyond normal hours and getting a masters, let alone a PhD will automatically bump you up the GS payscale (GS11 for masters and GS14 for PhD, if I recall), to the point where you can get six figures simply working a 9 to 5 with ridiculous vacation and retirement benefits.
While it's true in his category of "PhD's that work" the most useful degree is Computer Science as it's closest to "data scientist" of all the available PhDs you still need domain knowledge to apply it. So while it's no coincidence that Physics, Computer Science and Stats/Maths postgraduates are all in high demand in the field of applied economics itself (hedge funds, high frequency traders and the like) it's also notable they aren't just trading by themselves.
I'm sympathetic but torn. I do time series econometrics, partly because macro is important but unrealistic like you say. But I've grown more sympathetic to macro over the last few years. For better or worse, it's a field that only really gets model validation during crises, and as a field it's been pretty willing to overhaul its methodology following a crisis: Sims's "Macroeconomics and Reality" was written in 1980, basically had the message, "macroeconomists suck," and is now entrenched methodology. The RBC stuff was written in the 80s and is now also entrenched. There's been a stretch of relative calm (people learned something from LTCM, etc, but since it didn't cripple the economy it didn't inspire quite the same urgency as the financial crisis).
So, as far as the agent-based modeling goes, now's the opportunity for people working on it to show that they can answer questions macroeconomists are interested in better than DSGE type models. I'm not very familiar with that literature, but I don't think that there's been much emphasis on things like predicting aggregate unemployment, output, etc. or understanding the effects of monetary policy, but there's been more emphasis on individual markets. VARs and DSGE models are ubiquitous because they outperformed the previous models, not because the underlying theory is especially beautiful or interesting.
But, yeah, I definitely agree that you should not get an Econ PhD unless you want to learn how to do research in Economics. 4-6 years of salary and job advancement is a lot to give up, even if you could be sure of getting the job you want after graduation. And the celibacy.
Edit: just to clarify what I meant by "as a field it's been pretty willing to overhaul its methodology." I don't expect that a lot of established people will radically change their research agenda, although some will. But I do expect to see a change in what grad students and assistant profs work on compared to what you'd see w/out the financial crisis. So the research will change, even if a lot of the same people keep putting out the same papers.
You'd probably like a book I just read: Debunking Economics by Steve Keen. He basically rips apart neoclassical economics, often using mostly-ignored findings of famous neoclassical economists. He advocates various more-realistic models.
An extra note: An undergrad degree in econ can be very useful in getting familiar with the terminology and the sorts of facets that finance and economics professions are interested in. Beyond the BA/BS in Econ, there isn't much more to learn in post-graduate economics. My suggestion for anyone wanting to learn economics is to learn it in undergrad and mix it with another discipline.
I am surprised Abnormal Returns picked up this link.
I majored CS and with minors in Math and Econ. The economics minor was one of the better decisions I made in college and sometimes I think I use it more than the CS stuff. For example, things like opportunity cost I use almost every day, while it's rare I analyze the "big O" run time of code I'm writing.
From what I saw of the higher level econ classes, you're completely correct. For the most part they're not learning much new stuff, they're learning the same stuff in more detail and investigating why. It's interesting, but has limited applicability to real life. The low level courses are also interesting but have real life applications all over the place.
A good undergrad minor will certainly give you the more readily transferable skills: a grounding in econometric/statistical techniques you might actually use, a style of reasoning that covers things like optimal pricing and game theory and an overview of what the main economic debates of the twentieth century were and their policy implications. But if you like a challenge the PhD is generally regarded as an order of magnitude harder than undergrad economics, more so than in other social sciences.
I'd recommend instead learning this stuff through Coursera or CFA or actuarial exams instead. Way cheaper and there's no benefit to the classroom for this material.
I looked at an actuarial textbook once, and almost all the material seemed so specialized that I don't see why anyone would recommend learning it unless they were actually going to be doing actuarial stuff.
A popular actuarial textbook like "Actuarial Mathematics" is extremely specialized (especially to life insurance) and would only relate to the 3rd exam (MLC) and beyond. However, the first two exams are much more generic to statistics and finance respectively.
Econ aims to capture important aspects of real world systems, not model them perfectly. You can compare it with CS: big O notation is not useless even though in principle the constants in the big O notation could be huge. Turing machines are a useful abstraction even though in real life it is impossible to build one.
The criticism by mainstream economists of agent based models, is that they are so flexible you can create almost any kind of behavior you like. That said, the constraints that rationality impose on models don't prevent this either: with enough work you can get any kind of phenomenon you like in an econ model.
At the end of the day, economics is hard because people, institutions and technology are complex, much more complex than atoms. If you stuck with it, I'm sure your viewpoint would have softened somewhat. Once you try to actually do economics, you'll see why people made so many assumptions that seemed stupid before.
That's not my impression of the criticism of agent based models and I don't think there's a criticism of the models per se. Computationally intensive models with heterogeneous agents are used in parts of macro (an office neighbor of mine in grad school was using these models to look at efficient taxation, for example).
They're not part of the core of macroeconomics because AFAIK they haven't been shown to do better than existing approaches at addressing the fundamental macro issues: the business cycle, monetary policy, forecasting, etc. (I'm not that familiar with the literature, so if there are key papers I'm missing I apologize).
I'm repeating this because HN seems to have a disproportionately high number of people working on agent-based models. If you're working on that stuff and you want to take over macro, give new insights on monetary policy. That will do it.
The issue is that there just isn't that much data. You collect a bunch of macro level time series variables. Ok now what? No matter how cool or sophisticated your model is, there is just too little data to tell which models actually fit reality.
The criticism I describe is what most economists tell me when I bring up agent based models. In practice they mostly just ignore the field.
That's probably a reasonable description for why most macroeconomists don't work on these models and wouldn't advise a student to start working on them, but if the models were giving new insights or understanding, people would pay attention.
Because you're massively missing his point, which is just talking about career prospects. Noah's criticisms of what's taught in grad school econ echo your own:
You are right, thanks for showing me this. I agree with his ideas about undergraduate information versus graduate information. Undergraduate programs contain a lot more history, and yet still mathematical. However, I still disagree with the current post. You don't get autonomy in a graduate program; you are tied to a faculty and the programs currently available. It can be difficult to breakout on your own. This was my experience. This mixed with the time spent learning material that is at its best, interesting; and at worst, incorrect makes it seem like a poor choice. You can get the same correct information by studying other graduate programs. Intellectual fulfillment should come from discovering how these systems actually behave.
Our criticisms about graduate economic programs are similar, but our praise about their benefits differ very much.
There are some truths in this article -- for example, it is generally true that economics PhDs don't have a hard time finding jobs. Don't want or can't get an academic job? No problem, there are plenty of consulting/industry/government jobs to go around for economics PhDs.
I'm an economics PhD dropout (after 2nd year) from a top ~40 program. Unlike the authors school, my former PhD program did not have the prestige of his top 10-20 school. I left my program because I didn't have a research match with my professors. Now I'm a computer programmer at a top ~1-5 ranked school and I'm infinitely happier than I was in graduate school. I may eventually go back for a PhD, but it will most likely be in a business school rather than an economics department because I enjoy applied work more than theory.
The main ridiculousness in the current coursework taught in most economics PhD programs is the macro course (e.g. freshwater/Minnesota macro). DSGE (Dynamic Stochastic General Equilibrium) is just stupid. For the uninitiated, here is why DSGE exists and what it means: A while back there was this thing called the Lucas critique, which was the observation that the models used in microeconomics and macroeconomics are very dissimilar -- that is, micro models say 1 agent will make a certain decision, so why isn't it that all agents will make that decision? Macro theorists took this criticism way too seriously and decided that their models needed to have micro foundations, hence giving rise to the macro modeling technique known as DSGE. How does the typical/introductory DSGE model work? It goes something like this: at the beginning of time, everyone in the world meets in a parking lot. At the meeting, they decide on the price of all K goods in the economy at all T periods of time by assigning a present value to each good at each period of time. Once all the prices have been decided, the world starts. That is the gist of this model, but if you want to read more about how silly it is, look up the idea of a representative agent DSGE model.
Anyways, my point here is that depending on what you want to do with an economics PhD, it may or may not be a waste of time. For example, if you study micro and econometrics and emphasize in industrial organization + game theory + experimental, you will make a great candidate for a cushy research scientist job at one of the big tech companies. If you study macro, well, you'll make a good candidate for a job.... doing macroeconomics?
TL;DR If you're going to do economics, don't do macro.
"If you study macro, well, you'll make a good candidate for a job.... doing macroeconomics?"
There's a damned good living to be made as a macroeconomist in D.C. The bar for intellectual rigor is pretty low, and you're basically a lobbyist/speechwriter masquerading as a "policy advisor." It's the sale of your soul, but the price can be attractive.
Alternatively, there are some very handsomely paying "life of the mind" roles available at the think tanks. Of course, this stuff is politics at its most base and animalistic, and so your advancement is 100% predicated on playing the game of thrones. (The modern-day think tank is a lobbying organization given a euphemistic name.)
* Macro's different than micro in that it's the predictions of the models, rather than their intellectual coherence, that matters.[1] They're obviously simplified because otherwise they're intractable. Even now, the emphasis isn't going to be and shouldn't be to make the models "more realistic," it's to make the models more accurate so that they can explain interactions between the economy and the financial sector better. The person or people who improve these models will probably win a nobel prize in ~30 years, so there are strong incentives w/in the discipline to get there.
* The Lucas Critique is the macroeconomist's way of saying "correlation is not causation" -- patterns that you observe under one policy regime may not hold up if you change the policy. It has very little to do with going from one agent to many or partial to general equilibrium.
* The private sector version of "doing macroeconomics" is "work for a bank" or "work for a hedge fund." I've heard it pays pretty well.
[1] Yes, their predictions before and during the financial crisis were shit. That's a valid but separate criticism.
Isn't your criticism of DSGE a little too easy? I don't see the silliness.
Here's my understanding (I am not an expert): DSGE is designed to give an approximate response to exogenous shocks for NL models. For that, it needs to start from an equilibrium.
You seem to say that shocking from an equilibrium is silly. Can you explain me why? I wonder what alternative solution you have in mind that could improve DSGE.
The underlying reason I think most DSGE models are absolute hogwash is that they do economics backwards. Economics is supposed to test theories against data. Instead, in the world of calibration in macroeconomic models, the creator of the model is now testing data against theory by tuning parameters to values they think are good. It is completely backwards.
While I appreciate the attempt to make macroeconomics more computational, I believe DSGE goes about it in the wrong direction. In an ideal world, I'd like to see models like the Leontief Input/Output model come back to fruition. In Leontief's model (which is often given as an example in many undergrad linear algebra classes), the economy is divided into many sectors. Data is organized on each sector to estimate its influence on other sectors. In an age where data is so vast, I just don't understand how one can decide that building deeper macro theories is a good idea. We need better empirical models, not better theoretical models (we have enough of those).
>in the world of calibration in macroeconomic models, the creator of the model is now testing data against theory by tuning parameters to values they think are good. It is completely backwards.
This is a little inaccurate, the purpose of these macroeconomic models is either to make future predictions or run simulations to see what happens when exogenous shocks occur. They're not "testing data against theory", the data is used for parameter estimation and then verifying the accuracy of the models. It's actually very similar to the way certain AI models are developed and trained.
I do agree that these models are usually pretty inaccurate and somewhat useless though.
The main issue with DSGE models is that their simplifying assumptions make them useless for prediction: they replaced essentially atheoretical time-series models with hard, unrealistic theoretical assumptions about agent behaviour (over)fitted to time series and gained only a veneer of sophistication. Arguably no representative agent is better than a badly specified agent.
As an intellectual exercise to show that (for example) menu costs or nominal wage rigidities or technology shocks alone could have the economy's accounted for a shift from a [purely theoretical] equilibrium over a time period they're very interesting. For identifying which aspect contributed most towards economic change they offer little, and as a policymaking tool they're actually worse than useless since the models are generally built on the economists' assumptions about agents' response to policy and fitted to the data to justify those assumptions, rather than using the data to understand how responsive, rational and optimizing agents actually are. As such, their forecasting performance isn't very good either...
(this is the closest I could find to a self-contained link, sorry). This stuff's not great, but I don't know what you mean by "useless for prediction". And, given a model, it's pretty trivial to figure out which shocks contributed when; this is the whole point of Impulse Response Functions, variance decompositions, etc.
I've said this elsewhere on the page, but I'll repeat it here: this stuff isn't popular because it's logically airtight and compelling, but because it seemed to have done pretty well empirically. So the criticism that dsge models are unrealistic isn't very interesting; everyone knows that already, especially the people who use the stuff (for the most part. I'm sure there are some dsge truthers too). If anyone has an approach that works better empirically at addressing core macro questions, especially the newly important interplay between the real economy and credit markets, this would be a great time to put it out there. There are a lot of people paying attention.
2) I don't necessarily disagree with this. I know there used to tons of cushy analyst jobs and world bank/IMF/UN positions that hired Econ PhDs. Not sure if the world has changed in this regard.
3) One of the things that make life easy for Econ PhDs is the lack of a publishing treadmill. In CS, if you don't get 3 [magic number alert!] papers at a decent conference, few respectable advisors let you graduate. In Econ, my understanding is that exceedingly few people publish a journal paper by the time they graduate (please don't argue about Econ grads that write conference papers ... Econ Journal papers are equivalent to CS Conference paper when it comes to metrics like credit). Many people can write a hundred pages that their advisor+committee is happy with. It takes an order of magnitude more effort to get 100 pages that make your advisor+committee happy and also three random reviewers in your field.
An important, ending thought: the reason I didn't pursue a career in Economics was that I didn't get to build stuff. This is how I get my kicks in life. A CS PhD isn't designed to make you a better builder but it often does! Frankly, my biggest problem is that I need to stop building and get into the higher-level business of managing others to build, and higher value algorithmic stuff. The love of building is what holds me back I fear :(
Former economic consultant here. Speaking as a lowly analyst with a BA, I got to work with many PhDs in Economics/Finance.
Companies are always looking for ways to back up their economic damages/projections with data provided by a third party aka an economic consulting firm. I got to work on some very cool projects like defending a pricing fixing case in the consumer electronics industry and calculating the number of jobs created if the company injected $35B in the US economy.
These reports were not cheap and PhDs could bill ranging from $600 to $1,200 per hour. A 15 min phone call costed $150 to $250 and we billed EVERYTHING. As long as you had a good reputation and played both sides of the fence (Plaintiff/Defense) you always had a job.
I could have gone on to get my PhD, but I wanted to dream bigger. It's still a salary job and the hours you put in is what you get out.
Do you have to stick with neoclassical macroeconomics, or can you go into heterodox approaches and still do well? (Like, say, dynamic modeling, or agent-based models.)
I don't think an econ PhD is inherently the right thing nor the wrong thing for people. But this post is just a mix of weak arguments and factual incorrect assertions.
Regarding point 1, one reason most econ PhD's get jobs is that many advisors won't sign off on someone until they have a job (because it looks bad for the advisor). This results in students sticking around for years when they would be better served by having graduated.
Reason 3: "Do it for intellectual fulfillment." If you want intellectual fulfillment, study something you personally enjoy. Maybe that's econ, maybe it isn't. A blogger you've never met is in no position to advise you about this.
Reason 4: Low risk of failure
It's true that you get the consolation masters if you fail out of an econ PhD. That's worth something.
On the other hand, it's a field with an extremely high failure rate. Where I did my PhD, many people invested 4+ years before determining they'd never get their PhD. The completion rate was about 30%. If anything, the risk of failure is exceptionally high.
Not to agree or disagree with the conclusion. Just surprised to see the use of such flimsy arguments.
It is inherently wrong in the sense that the models can either misdirect policymakers into doing things that are not helpful for economic growth or that are easy to manipulate, creating false proofs. If one desires to do so, they should study economics. It could be good, because they may discover what things are wrong with the discipline and how they might be changed.
There's a well known saying by statistician George Box that "All models are wrong, but some of them are useful."
Economic models are like maps, physical models, and every other type of models in that they are gross simplifications of reality.
Like every other type of model, those simplifications can be cause a decision maker to reach false conclusions. When that occurs, we should try to figure out how to improve the models so they are more useful.
The fact that economic models aren't perfect isn't news to the economics profession, but it isn't a realistic expectation either.
But they aren't useful because they are very bad at predicting. There are other tools that are more suited to model the complex and emergent properties of individual actions leading a market place and what its conditions may be.
"Econ is one of the few places in our society where overtly racist and sexist ideas are not totally taboo"
Maybe because economics doesn't care about taboo, but rather understanding and seing the payoff. Yes some ideas and concepts are overly simplified - but one has to make shortcuts to reason. Mainstream economics provides reasoning, and the means to test if it works against real life data (econometrics) - that's much more than most social sciences.
I am saying that as someone who is deeply interested in the "logical" underpinnings for racism and sexism. I love complex things - mainly to understand how they works.
Here, if these behaviours were illogical and had a negative payoff in all cases, they may not have existed in the first place- or persisted. But they do. Why?
To the best of my current understanding, they are related to tribalism (as in parochialism). If you prefer, there is a nash equilibrium with cooperative and non cooperative behaviour.
The external characteristics of the agent are used by the other agents to guess whichever is more likely - being cooperative or not - and to play accordingly.
Basically, in both cases the external characteristic of an agent are used in a "best guess" function, which becomes self reinforcing, yet there seems to be more than self reinforcement, or these ideas would not be so prevalent. (So the interesting theories about say how agriculture helped the development of modern sexism are then developed, and tested in regressions)
I find that very cool - at a "theory of knowledge" coolness level. Different paradigm coexist (because there are more than one explanation) so you can pick up your "school" (Chicago, Austrian) and see how well it works, ie test it against new concepts in an experimental way - like bitcoin and check its predictive power. That's challenging, especially when it is counterintuitive in your personal biais (ex: I'm so surprised Keynesian theory works well)
Anyway, should you care about avoiding taboo (omg the n word! run!) or should you understand why and how it works?
EDIT: They are not mutually exclusive, but when you study a concept, the taboo associated with it quickly goes away. Yet it remains in other persons, which makes it funny and interesting
> Here, if these behaviours were illogical and had a negative payoff in all cases, they may not have existed in the first place- or persisted. But they do. Why?
In my opinion, that is not a meaningful question because the answer is simple: people are not always totally rational, they do not make all of their decisions based on logical analysis. In fact, many of the opinions they hold are not even based on conscious choices.
I agree that racism, sexism, and bigotry in general are probably rooted in the distant past and that, in some cases, they provided evolutionary advantages. But it is partially for this reason that their persistence needn't be logical or advantageous.
A child socialized by its parents to be racist doesn't usually face a conscious decision later on whether to become (or continue to be) a racist. This is, in my very humble opinion (MA in econ, take from that what you will) one of the weaknesses of the field: economists seem to think people and social dynamics are icky.
Why accept the obvious explanation that parents teach their kids racism and some of the kids, for various psychological or social reasons, continue to be racists? It is much more interesting to make a bunch of crazy assumptions about people so that you can plug the whole thing into a nice clean mathematical model.
PhD in economics here from a top 15 uni, completed a couple of years ago. No regrets, even though I realized, after a few years, that I would probably not enjoy a career as an economist (in academia or policy). I successfully 'converted' into data science. Most of my cohort found jobs they love. Some, though, had to settle for post-docs (sometimes for terrible pay) and are still in academic limbo.
There is always the question of the opportunity cost, of the road not taken. What would I have done of those 5/6 years? Perhaps I would have learned more, perhaps not. Many PhD students who end up outside of academia may have made better use of those years. I think the bottomline is, for every PhD, that you should only start it if you are passionate about the field and would like to work as a researcher (in academia or, say, IMF / World Bank). That may no longer be the case at the end of the PhD, but it probably should be at the start.
I'm finishing a PhD in Materials Science at a top 5 US institution with mostly experimental work. I don't do someone else project, I do my own project and I direct it, this is true for most of my peers.
It's funny, because my advice is basically the opposite for getting an Econ PhD. You get paid very little during grad school; it seems to exacerbate depression, etc. and is bad for relationships (no formal data to back those up, just my experience and observations); you work an ungodly amount if you want a prestigious job, and peer pressure will probably convince you that you do no matter what you believe entering the program; most likely you're not as smart or as good as you believe; etc...; an MA/MS in econ is nearly worthless, since it's widely known to be a consolation prize for failing out of a PhD program (and most top programs don't even offer a terminal masters); etc.
And the only benefit is that you get to spend all of your time learning about economics (or, in my case, econometrics), and, if you're lucky, get a job where you're expected to do that for the rest of your life.
So, that had better be one big fucking cherry to compensate for what you're giving up. And my thought process is, if my discouragement is enough to stop someone from going to grad school, they probably shouldn't go. Which is why I tell everyone not to go to grad school, even though I'm very grateful that I had 6 years to learn about and think about econ and wouldn't trade it for any other job experience.
I thought a long time about whether to do economics or a computer science PhD, eventually deciding on CS / machine learning. The usefulness of both is very appealing, and economics is full of interesting problems, a wonderful analytic method to approach thinking about society, and fascinating ideas and techniques you don't see much of in CS or statistics. The smartest economists are really smart and great at just thinking about the world.
But economics is also saddled with crazy non-empirical theories they force everyone to learn in the first two years, and an anemic journal-centric research culture that requires years to get an article published. I mean, from the OP:
And after you pass the prelims, there is little risk of not finishing a dissertation; unlike in most fields, you do not have to publish to graduate.
That is insane. Talk about tyranny of low expectations.
I wonder if their centrally-planned job market hurts academic productivity. It lessens the risk to spending years with mediocre research -- everyone gets a job eventually.
It's a python implementation of a bunch of classical economics models I wrote as a way to avoid actually doing the exercises in Caltech's Economics for Scientists at Coursera. I think having the models as functional code does indeed provide an interesting perspective and understanding, and it simplifies exploration. Haven't had time to publish the docs, though.
"Econ is not as intellectually deep as some fields, like physics, math, or literature."
I took 2 years of Econ in my undergraduate, along with courses in each of those other fields. My memories of economics was that do the work, you get an A, but its so incredibly boring (especially in contrast to the 3 fields mentioned) that the biggest challenge was staying awake in class.
There is some truth in this article. But, I know many physics PhDs, myself included, that had no real trouble getting jobs outside of academia or industry research. Of course, most of us are also programmers and now do R&D related to simulations, machine learning, etc, because our research prepared us for that.
Don't get an economics PhD. As some have pointed out, a lot of the "status quo economics" is just hand-wavy nonsense, and there might not be much demand for that stuff five years from now.
I wish I could share your optimism when it comes to hand-wavy nonsense.
I am convinced that this is hard wired in human kind, and no matter if it was back then when were were sitting around the fireplace talking about out last animal hunt or today listening to talk radio/tv, humankind will always enjoy subjective perspectives and interpretations of reality.
The best skills you can get out of most traditional Econ PhD programs are econometrics and programming skills. Outside of that it is pretty much a waste of time. Yes, you can probably get a job quickly, but you will spend a lot of the time learning none-applicable skills.
I want people to read it and hopefully we can start a discussion about it on HN.