Saturday, May 31, 2014

A serious proposal for undergrad econ

In a recent post, I said that I think Econ 101 should have more empirics. Greg Mankiw, America's Econ 101 teacher (or "Ec 10", as Harvard calls it), thinks my suggestion is good but impractical:
I understand his argument, and have some sympathy with it, but I wonder if the substantial change he seems to be proposing is practical.  Economists usually do empirical work with statistical tools that most college freshmen have not yet learned.   
We teachers of introductory economics can and should explain where and why economists disagree. That is part of helping students develop their critical thinking skills.  But I doubt students are in a position to try to evaluate the competing empirical work that shapes the differing views. 
In the end, introductory economics is just that: an introduction to the economist's way of thinking.  That means giving students basic concepts--comparative advantage, supply and demand, market efficiency and market failure--that will make them more perceptive readers of the newspaper. 
I think Mankiw is right that my suggestion in Bloomberg was impractical - but not for the reason he says. Students are smart enough to do basic empirics by themselves, and even if those basic empirics don't always give satisfactory answers to economic questions, I think it's important for students to actually see why they don't give satisfactory answers, and therefore to understand how difficult it is to tell which theory to use to understand what they read in the newspaper.

The problem is that there's just no room. Econ 101 is already overflowing with material. So what I actually propose is to create an econ "lab sequence" just for empirics, similar to what they do in physics. This "lab sequence" would only be for econ majors, and would be a separate but linked course. It would teach R or some other equally useful data management tool (a more advanced second-year lab course would teach Matlab). If the only thing students learn in the class is how to manage data, run regressions, and use R, the time will not have been wasted.

I also think that there should be two versions of intro econ - one for econ majors, which is more advanced and mathematical, and one for people who just want to learn some economics ideas. This is also similar to how physics departments and math departments do things.

The basic idea is to make undergrad econ a bit more like an engineering major. Since a lot of America's future managers major in economics, a more technical focus for the entire econ major will give them the skills to succeed in today's increasingly technical business world.

So there's my real, full proposal.

Chris House also has a response to my post, which mostly veers off into a discussion of general issues in undergrad econ. He (and everyone) should know that I don't really think Econ 101 is "bad", as the Bloomberg headline asserts. I just think it's incomplete. Bloomberg, of course, wrote that headline to grab eyeballs (as all headlines everywhere are written).


Some people have been telling me on Twitter that their econ departments are already doing something very much along the lines of my suggestion - having a lab class just for econ majors. Nice.


  1. Argosy Jones3:06 PM

    "Econ 101 is already overflowing with material."

    But most of this material isn't very good. That's the real problem. Unverified or unverifiable stuff should be cut from the entry level courses to make room for your proposal.

    1. Actually I think most of the stuff in 101 has lots of empirical support in specific domains. It just doesn't apply everywhere. For example, there are lots of cases where supply and demand does a great job of explaining data...and a lot where it doesn't.

    2. Anonymous3:33 PM

      The analogy with math and physics doesn't work because of the social implications that the Econ 101 mentality has in political thinking.

    3. ^And there are no social implications from a nuclear bomb?

  2. Anonymous6:28 PM

    Noah, tune out the haters on Bloomberg. You Rock!!

    1. ?? I didn't notice any hatersz on Bloomberg, except for the people who were really upset about the Turkish conquest of Constantinople.

    2. Anonymous4:15 PM

      They were right because you called that the time of modernity. You also called Marc Andreessen the true inventor of the Internet.

      Now you advocating engineering style econ education. Something is wrong here.

    3. Guns are modern.

      The web browser is one of the key components of the internet.

  3. bjdubbs7:05 PM

    Much of what econ 101 teaches isn't specific to econ. Comparative advantage, opportunity cost, the power of incentives, all of these concepts apply to areas that are not normally considered economic. There's nothing uniquely *economic* about them. There are some people, like David Friedman, who think that econ is precisely about teaching these concepts. Ultimately I think that what Friedman is proposing is really a moral philosophy (or a systematic approach to prudence), but that would take a longer argument.

    Most econ courses today are probably some mix of the two approaches. But what if a course took seriously the "facts about the economy" approach and taught economics like astronomy? First an astronomer needs to understand the differences among planets, stars, solar systems, black holes, galaxies, meteors, etc. The equivalent in economics would be to begin with an inventory of the different kinds of goods (services, consumer durables, natural resources, etc) and different kinds of economic actors (firms, households, gov't)? Beginning in that way, it would start with the firm, because only the firm is unique to economics. And in order to understand the firm, a basic grasp of accounting would be essential. Assets equals liabilities plus shareholder equity is pretty much the F=ma of accounting. No astronomy course begins with "how to think like an astronomer," instead it begins with a little astronomy. Why would econ courses be any different?

    1. Anonymous12:15 PM

      Have you taken a college stronomy course? It's nothing like what you describe, for the obvious reason: college courses that just teach you a bunch of facts are a waste of time and money--if you have a scholar in front of you, you shouldn't spend your whole time learning what you could learn from reading a book. Similarly, you can read what a durable good is from a book and you'll get the same out of it as if a lecturer told you.

  4. Noah,

    Late to this, but if I might give my take:

    1) You seem pretty sure that today, at least, micro and macro 101 does actually let students know about monumentally important flaws to the basic classical market model, including externalities, natural monopoly power, asymmetric and just poor information (and it should be expertise too), and so on. But this certainly wasn't true a generation ago, and I'm not so confident as you it is true today. That it's in the econ 101 text means very little. These texts are like 1,000 pages long and there's nowhere near enough time to cover everything in them well, or even close, in a mere three unit course. Professors could regularly consider market problems relatively important extra material – amazingly look at the attitude of Chris House – as if ignoring pollution externalities, monopoly, and asymmetric information wouldn't really result in much harm; they're just "exceptions to the rule". Oh, just a cesspool for a planet, with devastated health and quality of life that's decades shorter, monopolies strangling the economy and advancement, little basic scientific or medical research, consumers having to do a dissertation on any purchase to make sure it won't kill their family,… It should be obvious that these "exceptions" are as important as the rule, and that today there's far more harm and inefficiency incurred from erring on the "exceptions" than from erring on the rules.

    2) There's this attitude of teaching econ 101 for econ majors. So you math them up with the classical model and figure later they'll learn the problems and other important things. But for 99% of students – voting citizens – there is no later. So this is a horribly costly approach for society. 101 and 102 should be 100% made to teach the most important intuitions for society's citizens in two three-credit courses. So you teach only the most necessary math for intuition, get to the intuitions behind the classical model, then get to the intuitions behind all the most important market problems, and yes some empirics to quantify their magnitude and level of certainty. You do this, when all you have is 6 units of student time, and you've done awesome. I would be very wary of putting in much more, as it will take time and effort away from this crucial core, and may really hurt this crucial learning, making students rush through it.

    3) If there's enough time to do the above very well, I'd add the intuition of what good modeling is, and where many economists go wrong, as in this crucial paper from Stanford's Pflieder, which should be required for anyone getting an econ degree at any level, any school:

    It basically says what I've always said, "A model is only as good as its interpretation.", but with great specifics and detail, and, of course, crucially, a big name saying it.

  5. Cactus3:51 AM

    Perspective from a German undergrad university student:

    We have the typical Econ 101, split into a couple of lectures (Intro to Economics, Micro I, Macro I), and then building on that with Micro II (IO and Game Theory), Macro II and beyond.

    In semesters 1 and 2, there are lectures for statistics I/II (descriptive/inferring). Now, in semester 4, we have a lecture on Econometrics, and a class that teaches how to do that in R (why would one use Matlab?).

    Econ 101 mentions that, e. g. elasticities are oftentimes estimated via OLS (if the data is available) and in statistics 1 this is done manually (1 independent variable), but on a very basic level. Econometrics is much more empirical, tapping into datasets aquired from Datastream etc - you know, the fun stuff.

    I think this approach works well, IF the first semester students are made aware that there's more to the shiny graphs in the textbooks than meets the eye. I'm glad I did not have to worry about model specification in 101, but knowing that it is both not straightforward AND will be covered in a later class made me happy enough.

  6. In Econ 101, theoretic of supply and demand are taught.

    Then, we embark on ad campaign to increase demand because some good looking gal with no econ background likes it!

    Managing demand is the driver in the business. And, Econ 101 does not teach that; instead it tries to teach elasticity, demand shift, etc which then become math class without any axioms!

    Noah, you could still go back to physics!

  7. Why R or Matlab rather than a menu driven program like Stata (recall David Card's perhaps apocryphal comment, "If Stata can do it, I can do it")? That would allow students to focus on the analysis rather than how to code.

    (Sorry if this is a repeat; the first posting of this comment appears to have disappeared into the ether).

    1. If one is trying to teach good research habits, like a course aimed specifically at economics majors should, then a big focus should be repeatability, then using a menu to do your work is terrible. There's also a lot of more advanced analysis that you can do in R, that you can't do in Stata, and a lot that Stata actually does incorrectly. R generally forces you to actually think about your estimator implementation, which is useful in a pedagogical setting.

      I think the matlab is more for the sake of teaching students how to code, which is useful.

  8. Anonymous4:47 PM

    My two cents. I think only experimental evidence should be taught. Two reasons, first you do not need to know econometrics. Just a high-school level statistics. Second, and most important, this is beyond doubt this is serious science not some correlation that supports an ill defined theory.

  9. If you want to start with F=ma like a physicist, you should start with some basic accounting with cash flows, assets and liabilities. Physics 101 consist so of drawing those wretched little diagrams of inclined planes showing gravity, friction, normal forces and so on. The horrible thing is that you can get your physics PhD and spend years working as a physicist and at some point every advanced topic you are exposed to reduces to those little diagrams, though now redrawn in Hamiltonian perspective.

    Every physics class usually throws in a reminder or two, for those who haven't heard of Einstein, Hawking or Heisenberg, that what is being taught in 101 is only valid for certain ranges of energy, velocity and so on.

    I like the idea of a lab, but I think it makes more sense to start with accounting. You can build supply and demand on top of it. In fact, you can build all of conventional economics on top of it, and, as a bonus, you can build all sorts of exotic economies e.g. Asian peasant economics, bumblebee foraging economics, on top of it as well.

    1. Economics isn't the same as accounting. Accounting is just a technique to account for measuring economic units, it doesn't give theoretical underpinning.

  10. "Econ 101 is already overflowing with material. So what I actually propose is to create an econ "lab sequence" just for empirics, similar to what they do in physics. This "lab sequence" would only be for econ majors, and would be a separate but linked course."

    Ok, sorry, but I just read this, even though I earlier left a long comment. Why leave a long comment before you've read everything in the post (and referenced posts)? Well, it often comes down to that or nothing, and sometimes a promising comment without reading the whole thing is better than nothing, because with the speed of blogs if you wait a week until you can read everything, everyone moves on.

    So… that said, I have now read everything, including the referenced posts and articles by Noah! With regard to the quote above, the humongous problem is this doesn't help the vast majority of voting citizens, and people making important business and personal finance decisions, who don't get a degree in economics (for business and personal finance, key ideas like fixed, variable, and sunk cost). There's still the question of how to make econ 101 (micro 101 and macro 101) much better for the non-econ majors.

    Also, with regard to the idea of having one 101 for econ majors and another for the non-majors, the big worry I have with this is it seems like a lot of people who start as econ majors, change. Then they would never learn so much important material that was taught in good survey 101 and 102 for the general public, as I outline in my comment above. And for econ majors, a good 101 and 102, that really focused on an important overview of the most valuable intuitions, would be probably something you'd want them to start out with anyway. What's best for the non-econ majors in 101 and 102 may be pretty similar to the best introduction to the field for the econ majors anyway, and then they delve into specifics and less core material from there.

  11. Well, lots to say, let's just throw in one more thing:

    My econ 101 and 102, a generation ago, were terrible. Learn the classical model, and problems with it are for intermediate, which few people ever take. In intermediate I finally learned about the income and substitution effects and the backward bending labor supply curve.

    Now, when one of our two major parties makes it's supreme goal taxes as low as possible for the rich, then it's important, obviously, for citizens to understand this issue well. So obviously, for compelling civics and utilitarian reasons, we should expect all college graduates to understand the income and substitution effects and the backward bending labor supply curve – and in intuitive terms – You might work 45 hours/week instead of 40, if your pay went from $8/hour to $12, but if it went from $8/hour to $8 million/hour, you easily might go from 45 hours/week to 45 hours/year, and spend the rest of the year on your island. Higher pay is a double edged sword in incentive to work.

    And the empirics of this economic concept are, of course, crucial too. My professor spent about 15 minutes just quickly saying what the consensus in the field was on the empirical evidence. It basically was this quote from MIT's Jonathan Gruber:

    "Changes in tax rates appear to have relatively modest effects on total gross income; the total amount of income actually generated through work or savings does not respond in a sizable way to taxation."

    – "Public Finance and Public Policy", 2nd edition, 2007, Page 734

    Now, this kind of thing is very valuable, and it doesn't require a lot of time learning the massive math of econometrics. It's analogous to global warming. I'm never going to be able to learn enough about climate science to really evaluate the literature in this field. But when pretty much 100% of the top climate scientists in the world say this is real and a catastrophic risk, that's overwhelming evidence that this is a profound risk to take seriously and try to insure against (Do you not buy fire insurance for your home bacause the odds of it burning down are significant, but still not 100%?). But you can survey this evidence pretty quick, and give the consensus of the top experts, and how strong it is. It's like Jonathan Chait said recently (paraphrasing), I'm not a chemical scientist, man, but I don't drink Draino, because the overwhelming consensus of chemical scientists is that this would be catastrophic.

  12. Anonymous5:41 AM

    I think important in education, and this applies to economics and everything else, that people question. I do not think making economics look like engineering is all together healthy thing. More important than learning the mathematics of econometrics and computer programmes is understanding the problems in interpreting data. What it means, how it was gathered, how is it defined. And theories are only a guide. The more you have as reference points, the better. And context should override theory.

  13. And it's important to add that formal statistical direct evidence is hardly the only kind of valuable evidence. I could have a theoretical model that says if someone fires a cannon that hits a small plate glass window dead center, it will break. Now, is there no evidence for this model until someone fires a cannon at 1,000 windows, records the results, and runs formal statistical tests? Of course not, there's completely compelling evidence from our a priori information on the world combined with unbroken logic.

    To say the same thing in words from someone with a name: from the must-read paper I quote above by Paul Pfiederer of Stanford:

    Obviously we don’t in practice conduct our research as if we have these agnostic or uniform priors.
    We routinely take these priors into account. Now it is true that people don’t have exactly
    the same priors and “reasonable people can disagree,” but this doesn’t open the door for
    taking seriously models whose assumptions are severely athwart what we see in the real
    world and ignore factors we know are of first order importance. A model that claims that
    “high” leverage is optimal for banks while ignoring almost everything that we know is
    important is a bookshelf model. It is not a model that can pass through any reasonable
    real-world filter because it is at odds with all of the background information we have
    about the world, information that we use to form reasonable priors.

    -- pages 33-34, at:

    1. Well, the quote's not exactly the same, rushed. The idea is the use of "background information we have about the world". It can be important evidence, combined with unbroken logic. If I have time later, I'll see if I can find a more directly related quote.

  14. Noah,

    Here is a headline of an article in today's Bloomberg:

    Unstoppable $100 Trillion Bond Market Renders Models Useless

    After reading, it should be obvious that there is really no comparison between models used in Econ and hard sciences, where models are based on exceptionally well verified laws and range of its applicability is also well known. Econ is arm-waving.

  15. Noah, virtually all econ majors have to take Econometrics as part of their core requirements. To the extent that the instructors for that course chooses to emphasize applied econometrics in addition to the theory (insisting that students run regressions), then your proposal does not really amount to anything new. Indeed, since economists work more with observational rather than experimental data, this is more-or-less the "lab sequence" that you're proposing. Now, there are two alternative ways to bring what you said forward: (1) that the econometrics course should be expanded to allow for experimental material. That's a wonderful idea, especially since experiments are increasingly regarded as a legitimate part of the economist's toolkit. The concern is that since many econ professors (including econometricians) may not themselves have taken an experimental econ class, then we run into ideals versus feasiblity; (2) that the lab/econometrics sequence run *in tandem* with Econ 101. This is again laudable, but would impose significant costs on those taking the course (given, as you already alluded to, the amount of material already in Econ 101). Better t incorporate experiments/regressions directly alongside the standard curriculum. Bergstrom-Miller is an excellent example of this, although their text doesn't really go into regression analysis. The ideal would be a text that incorporates elements of Bergstrom-Miller classroom expeirments, basic applied regressions using bundled software, and economic theory. Perhaps you should be writing that text ;-p

  16. I wonder if the lack of empirics has to do with the mathematical preparation of the average undergrad? I know I didn't even take statistics until I got to college, and non-math majors aren't typically taking calculus their freshman year. Maybe econ departments are afraid of hurting their passing ratios if the program requires more math.

    I think that econ is still transitioning from being primarily a business/social science field into an empirical/computational one. A lot of people picking econ as their major aren't necessarily looking for a super scientific job after graduation. But, it seems like the more rigorous analysis of data is a big part of, at the risk of sounding like Steve Levitt, "thinking like an economist."

    That being said, my school has started to focus on this, because that's where most of the jobs are now. Yay data!

  17. Cactus3:32 PM

    "non-math majors aren't typically taking calculus their freshman year"

    Oh, that's a thing. Here, calculus is normally taught in high school (integrals/derivatives) and expanded in freshman year at university (Lagrange, Kuhn-Tucker). I sometimes stop and wonder why american textbook authors go to great lengths to introduce work-arounds (e. g. for game theory problems) that could be much more accessible.

  18. So very strange. There's nothing you can say about the real world that doesn't require calculus? And yet they teach special relativity in high school. Somebody is wrong and I don't think it's Einstein.

  19. Well, we have the modeling tools, think Sim City or whatever, with the ability to change assumptions. The aim of a starting lab course should be to show how changing assumptions and input data change outcomes, the "model" in the middle for a 101 course can be a black box without any loss of value and the mantra can be in the next four years we will show you how to build the black box.

    Also, go to the social scientists and have them build labs about surveys and marketing to show a) how they work b) their limitations and c) how they can be used.

    Do-able project.

  20. Anonymous11:21 AM

    I took lower and upper division macro. Not until upper division were correlations performed and it was very simplistic, look at consumption a few quarters ago and compare with current income. There were other variables you could look at (eg interest rate), but the TAs did most of the project for a large fraction of the student population. And we used EViews! My suggestion would be to make students take a programming class before they start taking required. Then as they get into the upper division they can tackle the "cooler" problems. Also, what sort of econ classes are you going to offer the general public (non-majors) who might not have an R background. Maybe two econ classes one for majors and for non-majors. The one for majors make it a pre req to have R.

  21. Anonymous4:04 PM

    I would love to see if Mankiw's 10 commandments can survive lab evidence. It would be fun to watch.

  22. R? Matlab? Ehhhh. We should be teaching Econ undergraduates programming and data analysis in Python from day one. Python is both easier to learn and generally more powerful than either R or Matlab. Python is also used more in industry than either R or Matlab (so far as I am aware) and therefore looks better on CVs/resumes.

  23. First, Im not sure why you want to keep the current econ 101 for non-majors. Basically they try to teach some non-mathy version of "how to think like an economist" without regard to any real world data. I think you have it backwards--this might be useful for people who want to become economists, and therefore need to learn to think like one, but less useful for non-majors. I don't think it's too much to ask of econ 101 students to think about correlations and whether or not they are causal. We should focus on empirics for non-majors, more so than for actual econ majors.

    Second, why do you consider Matlab for the more advanced class? R can do everything matlab does. We should stick to open source platforms as much as absolutely possible. Anything less violates science's mandate to be reproducible.

  24. Anonymous8:35 AM

    My hobby horse: there's a rift in economics. At the BEA and the BLS and Eurostat and the ECB people actually measure the economy (production, income, expenditure, employment, prices, stocks and flows of money and credit), at universities people hardly know that this is happening (and costs mighty amounts of money). Students of economics should learn about the quadruple entry methods and the surveys used to do this, as well as about the discussions about for instance estimating house price inflation, U1-U6 unemployment (why do we have 6 estimates of unemployment anyway, how decides about this (google 'ILO')) or about the SNA manuals. Every science teaches its students how to actually measure or observe, most Nobles are awarded for new (methods of) observation(s). Economics is the outlier, for no good reason. For bad reasons, in fact. Very bad reasons. The national accounts (including financial accounts, estimates of money, credit and employment and unemployment etcetera, are the real macro-model, fully specified and estimated. But not understood by the majority of academic economists, who prefer to quibble about badly defined esoterics like utility (why in the hell do people assume that social indifference curves have a 'cobb-douglas shape' - what is a social indifference curve, anyway, conceptually).

    Merijn Knibbe

    Merijn Knibbe