slightlyconfused1

Just in case anyone is still looking at this thread, I want to emphasize that this rumor indeed does not have any basis in fact. Over the past three years, placements of MIT job market candidates whose main interest is development have been: 2011: Daniel Keniston (Yale + Review of Economic Studies Tour), Simone Schaner (Dartmouth) 2010: Cynthia Kinnan (Northwestern + Review of Economics Studies Tour) 2009: Jim Berry (Cornell) Among candidates whose primary field is not exactly development, but who have a significant focus on development in their work, placements have additionally been: 2011: Marti Mestieri (Toulouse) 2010: Monica Martinez-Bravo (Johns Hopkins) 2009: Richard Hornbeck (Harvard + Review of Economic Studies Tour), Eric Weese (Yale + Review of Economic Studies Tour) I don't think that there is any program that matches MIT's record in development over the last several years, especially at the high end. In the last three years, fully 4 out of the 5 participants in the Review of Economic Studies tour---the top honor for graduating PhDs---with a focus on development economics came from MIT: Keniston, Kinnan, Hornbeck and Weese. The only exception was Dave Atkin from Princeton. If we're discussing LSE specifically, their placement record from previous years is visible at the bottom of this page. While I think LSE is certainly a great program, disregarding one extreme outlier (Dave Donaldson) their development placement hasn't been too hot over the past several years.

Let me follow up on this statement. When I said that these two students would both "be among the top job market candidates in the world this cycle", I wasn't kidding: one of them has an offer at Harvard econ, while the other was the clear star of the market and received offers at Stanford, Chicago, Princeton, etc., along with flyouts virtually everywhere. I think it is very revealing that the only two cases I could find where students ended up publishing their undergraduate theses in a top field journal were such overwhelming superstars. And again, even they did not have these publications before coming to grad school -- the processes wasn't done until several years after entering MIT. Looking back at candidates in the past, there are some other examples of publishing undergraduate work in a very good outlet. Mihai Manea published some of his undergraduate papers in top theory journals -- but again, even in his case, the publications only came after he was in grad school, and it's hard to think of many undergraduates as prodigious as Mihai was at Princeton. Raj Chetty, now one of the top public finance economists in the world, published his undergraduate thesis in the Review of Economic Studies... but only after he was already a faculty member at Berkeley. Jon Steinsson, now at Columbia, managed to publish his undergraduate thesis in the Journal of Monetary Economics three years after graduating, and he had Mike Woodford as his undergrad advisor! Maybe there is some hidden supply of economists who published successfully as undergrads and then slipped into obscurity, but from what I can see, publishing your undergraduate work is usually a sign of overwhelming superstardom. It's associated with people who are so outstanding in so many ways that they surely had no trouble getting into the programs of their choice. And even if they did, publications were no help, because they didn't have publications when they applied to grad school. (Sorry, I feel I have to reinforce this message through sheer repetition!) Again, while it is true that some people "publish" as undergrads, this universally involves (1) publication in a journal that is mediocre or worse or (2) sheer luck in collaborating with a faculty member whose publication timetable finished at exactly the right time. At least for more competitive programs, (1) doesn't mean very much; meanwhile, the only way to give yourself a chance at (2) is to find top professors and work with them, which is what you should be trying to do anyway regardless of the slight and mostly irrelevant chance of publication. So please, please... don't worry that you do not have "publications" when applying to grad school. Virtually no one else does either -- unless we define "publications" in a ludicrously generous way -- and the few who do are not necessarily good candidates along other dimensions.

My mental model of why pedigree matters is that it's mainly about uncertainty/lack of information, but that this doesn't necessarily reflect risk aversion among admissions committees. It still matters even if they care only about expected values! (Of course, terms like "expected value" are a little fuzzy in situations like these, where there is no natural cardinal scale, but bear with me...) One model: Suppose that there is a "true" underlying skill variable X, which is distributed normally in the applicant population (in the same way for students from all undergraduate programs) with mean 0 and variance 1. Suppose there is also noise Y, which has mean zero and variance 1/2 for students from high-ranking universities and variance 1 for students from other universities. Ph.D. programs observe Z = X+Y, and are only willing to admit students for whom expected X is at least 2. Then their cutoff will be Z = 3 for students from high-ranking universities and Z = 4 for students from low-ranking universities. It will thus seem harder for students from low-ranking universities to get in, since they require higher Z; students with high underlying X will find admissions to be risky and difficult. (Almost nobody will have X=4, so everybody is relying on a positive Y draw to get in.) Needless to say, this is a toy model. The one dissatisfying part about it is that the "noise" component doesn't make it systematically harder for students from lower-ranked programs to get in -- indeed, in expectation, exactly the same fraction from both groups gets in, because although admission is more difficult from a low-ranked conditional on a particular Z, the higher noise means that more applicants get higher Zs. Another toy model: suppose now that X is distributed N(1,1) for applicants from high-ranking programs and N(0,1) for applicants from other programs. Y is now distributed N(0,1) for all applicants, and Ph.D. programs still admit students only if their expected X is at least 2. The Z cutoff for applicants from high-ranking programs will again be 3, while the Z cutoff for applicants from other programs will be 4. Now it's unambiguously harder for students from lower-ranking programs to get in -- even fewer are accepted than you would project from the underlying distribution of X, because their observed values Z are regressed to the mean. Of course, you can combine these toy models to get a richer depiction of the dilemmas at hand. The basic idea, however, is simple: either because of higher noise or a lower population average of X, Ph.D. programs will read applications from lower-ranking schools with a more skeptical eye, rationally being worried that they good qualifications they see are just noise (Y) rather than underlying ability (X). Sadly, this will make life difficult for applicants who actually have good underlying ability.

Edit: on second thought, I'll let this go through without providing my (100% accurate!) predictions. I certainly support some healthy skepticism of the prediction game...

I agree that these serve very different niches: your site is aimed at more general discussion while Stack Overflow-style sites have a more rigid question-and-answer format. I don't think that one format is inherently better than the other, but my sense is that they will generally serve very different audiences. In my experience, online forums aimed at general discussion of a topic tend to be substitutes for in-person discussion -- not mutually exclusive, of course, but substitutes to an extent. For instance, posters on TM are relatively less likely to come from elite undergrad schools that send a large number of students to grad school every year. This is presumably because such students have more opportunities to talk about graduate school with classmates and professors. Meanwhile, TM users are disproportionately likely to come from unconventional backgrounds where they've been out of school for several years -- in this case, TM is often the only place to obtain information, and naturally it's used a great deal. Similarly, I don't think most economics professors would be too interested in a forum for general discussion of economics. They already have plenty of colleagues and students available for that purpose. They might, however, be interested in a Q&A site more narrowly tailored to answering specific research questions. My goal is to harness the collective knowledge of the economics research community. (Frankly, to some extent this involves scaring nonacademics away -- philosophical discussions about capitalism are great, but they would make a research Q&A impossible to browse.) Of course, I still think that your forum is valuable -- but my guess is that it will be more successful catering to people who are not already deeply ensconced in academic economics, by providing an outlet to discuss economic ideas with like-minded individuals.

Hey TMers (and alumni!), You might be familiar with Q&A sites like Stack Overflow and Math Overflow, where experts congregate to answer questions and the best answers are voted to the top. They're incredibly effective at harnessing the collective intelligence of the internet, and the resulting Q&A pages are both pleasant and edifying to browse. There is not yet any analogue for economics, but that can change. It's possible to make a "proposal" on the StackExchange site, which will automatically be turned into a live forum once enough sample questions and commitments of support are provided. There is currently a fledgling proposal, but it will need contributions from the community before it amounts to anything. Although its target audience is somewhat different from that of a "research economics" forum, TM is still a natural place to ask for support -- after all, if you're trying to build a productive online community of academic economists, why not ask the one excellent community that already exists? Think of this proposal as "TM for grad students and faculty" -- a way to exchange questions and answers on the topics that will matter for the rest of your academic career. If any of you are interested, I strongly encourage you to visit the link above and "follow" the proposal. (It takes about 10 seconds to make an account.) And if you can think of a good sample question or two, that would be great! The infrastructure already exists to build a world class forum for economists to share knowledge -- let's make it happen!

In many other fields with quicker publication processes, this is more accurate. For instance, most papers in computer science are published in conference proceedings rather than journals. These papers are shorter and rapidly refereed, and I doubt a computer science Ph.D. without any such publications would do very well on the job market. The publication process in math, though more journal-oriented, is also quicker than in economics, and you see more hotshot students who are already at the research frontier when they enter grad school. Thus I imagine that publications are more important for math Ph.Ds going on the job market than economics students, even though math programs are a year shorter. And in many lab sciences, essentially every grad student contributing in the lab has her name put on the resulting papers, making it almost a given that graduates will have several publications to their name. Perhaps people with some awareness of the standards in these other fields falsely assume that economics is similar.

Yeah, I think you're right -- whenever there's a serious discussion on the topic, the consensus seems to be pretty reasonable. The problem is all the offhand mentions -- for instance, where someone says in a profile evaluation "try to take some graduate classes in econ and see if you can get a publication," which gives the highly misleading impression that publishing is both feasible and useful for a prospective applicant.

By the way, to get a realistic sense of the scores for typical admitted applicants, it's edifying to look at this ranking (derived from NRC data) of schools by average quantitative GRE score. (I'm assuming they're using some kind of trimmed mean, since certainly not all 9 programs at the top of the list never admit anyone with less than 800. Still, this ranking makes clear that one needs to dip down pretty far in the rankings to get in the range where 740 is standard. And competition has presumably become stiffer since this data was collected.)

At the top 20 program where I was an undergraduate, the director of graduate admissions told me that his secretary is directed to throw away all applications with We can certainly imagine wildly speculative scenarios where someone with a bad GRE score has a letter from Jean Tirole saying that he is (Jean Tirole)^2. In these situations, I'm sure that the applicant would still be admitted (IF the application ever reached the hands of an economics professor, which would probably only happen if Jean Tirole placed a phone call to make sure the application didn't get trashed). But this situation seems so improbable that it's really just a distraction. For all practical purposes, there is a cutoff at most top-ranked programs, and 740 doesn't pass it.

Rankings can sometimes be a useful starting point, but as people have mentioned here departmental strengths are always in flux. The best way to learn about the strengths of each department is to do the following. Step 1 Learn which journals are prestigious in your fields of interest. You can easily get a rough idea by looking at a journal ranking.* Step 2 Go to the departmental faculty listing and determine which professors specialize in your areas of interest. Often this will be listed next to their names in the directory -- if not, you'll have to quickly check each CV or list of papers. Step 3 Look at each relevant faculty member's publications list and see if she has published in good journals (either top field journals or solid general interest journals) within the last few years. This will give you a good sense of how many faculty in a department are actively contributing to research. ---- * Sometimes it isn't clear from the names alone which journals are in which subfield. My (very) rough guide follows---although it doesn't give an exact picture, it should be enough to form a general understanding of which professors are active at a high level in their fields: Top general interest journals: American Economic Review, Econometrica, Journal of Political Economy, Quarterly Journal of Economics, Review of Economic Studies. Very good general interest journals: Review of Economics and Statistics, Journal of the European Economic Association, International Economic Review. Also European Economic Review and Economic Inquiry (more marginal). Theoretical Micro: Journal of Economic Theory, Theoretical Economics, Games and Economic Behavior, American Economic Journal: Micro. Somewhat less prestigious journals include Economic Theory, International Journal of Game Theory, Journal of Mathematical Economics and Social Choice and Welfare. (though the latter two are each top in a certain specialization). Macro: Journal of Monetary Economics, Journal of Money, Credit and Banking, Review of Economic Dynamics, American Economic Journal: Macro, Journal of Economic Dynamics and Control, Macroeconomic Dynamics. Industrial Organization: Rand Journal of Economics is overwhelmingly the most prestigious. Others include the International Journal of Industrial Organization and Journal of Industrial Economics. Econometrics: Journal of Econometrics, Econometric Theory, Journal of Business and Economic Statistics, Journal of Applied Econometrics. Applied Micro: American Economic Journal: Applied Microeconomics (includes empirical labor, development, public finance, etc). Labor: Journal of Labor Economics, Journal of Human Resources. Less prestigious outlets include Industrial and Labor Relations Review and Labour Economics. Development: Journal of Development Economics, World Development. Public Finance: Journal of Public Economics is by far the best. The new American Economic Journal: Economic Policy may also be good, and there is also the National Tax Journal. For the health economics subfield, Journal of Health Economics is excellent. Environmental: Journal of Environmental and Economic Management is the best. Others include American Journal of Agricultural Economics, Environmental and Resource Economics, Land Economics and Ecological Economics. Growth: Journal of Economic Growth. Urban: Journal of Urban Economics. More marginal offerings include Regional Science and Urban Economics and Journal of Economic Geography.

As far as I'm aware, only a few exceptional students manage to publish their senior theses, and even then the publication usually happens several years after graduation. I can think of two current MIT job market candidates who managed to publish their senior theses in top field journals -- needless to say, this is both very impressive and very rare. But neither of these students -- who both, I suspect, will be among the top job market candidates in the world this cycle -- came close to publishing their senior theses before the economics application process. I think they both got their papers accepted in their third year of the Ph.D. program. (And remember, these are exceptional cases. Most people can't even dream of publishing their senior thesis in a respectable outlet, at least without substantial modifications. I certainly can't! :) ) Part of what drives confusion on this issue, I think, is ambiguity about the word "publication". Sometimes overenthusiastic undergraduates will brag about "publishing" a paper in some barely-refereed outlet. If you happen to encounter such people, and don't yet know which journals are truly selective, you might get the sense that prospective Ph.D. applicants frequently publish their work. But this isn't really true: if we look at journals that actually mean something, no one manages to publish their senior thesis within even a year of graduation. If you don't believe me, read the CVs of job market candidates at Harvard and MIT and see if anyone had single-authored publications in serious journals before coming to grad school. Several posters have argued that the process of research that leads to a publication is beneficial in its own right. This is, of course, absolutely true! Everyone should try to form a relationship with faculty and be involved in research. But don't worry about not having publications before you apply -- virtually no one else does either.

I absolutely agree, and that's part of the reason why coauthored publications as an undergraduate aren't terribly relevant. If a student manages to get his name on a published paper, it's overwhelming likely that his contribution consisted of standard RA work, not deep empirical strategies or technical details. This isn't to say that RA work isn't important -- it is, and it's considered a useful formative experience for aspiring economists. I was exaggerating when I described an RA's work as "running regressions in Stata". But the key point is that "publication" itself is >99% determined by the timeline of the professors you happen to be assisting, and

If we're talking about single-authored publications in respectable journals for applicants coming out of college, "not everyone" should be "literally no one in the world". This is not an exaggeration. I'm sure that publishing in the QJE would indeed make an applicant stand out, but this is so extraordinarily unlikely that it's irrelevant for decision-making purposes. This is in contrast to the other option you mentioned, taking graduate level courses, which is a good strategy for many ambitious applicants. Papers with faculty are somewhat more plausible, but the mere fact of being published doesn't add much to the content of your letters, where a professor can detail your contribution to the project. (For instance, did you contribute substantial new ideas, or just run lots of regressions in Stata?) And even if coauthored publication turns out to be useful, it's not clear how to operationalize this as a distinct admissions strategy. The way to maximize your chances of coauthoring with faculty is to find active professors and work closely with them, which is generally what you should be doing anyway. Agonizing over the very unlikely possibility of publication is just a distraction. Again, as a worried applicant it's easy to let your imagination run wild about the backgrounds of accepted students. Trust me; I've been there. Yes, students accepted to top programs tend to be very accomplished. But they aren't omnipotent -- they can't somehow speed through the legendarily drawn-out refereeing process to obtain good publications before they even apply to grad school. It just isn't possible.

Often on this forum I see comments about how a prospective applicant should try to publish papers to boost his profile. I need to get this off my chest: both the value and the feasibility of publications are wildly exaggerated on this forum. If you don't have any publications, don't worry---hardly anyone else does! Why? The publication process in economics is simply too long for this to be practical. I don't think I've ever seen an economics graduate school applicant with a good publication that has not been coauthored with a professor. Certainly there is no one like this at MIT. There are a handful of people who manage to publish their undergraduate work in top field journals eventually, but I don't think I've ever seen this happen prior to the admission process. Moreover, the few cases that come close all seem to be in micro theory, which is the one area where it is possible to publish aggressively at an early stage; this is irrelevant, however, for anyone outside the far right tail of mathematical ability. Ultimately, the only way to publish in a decent journal so early is to coauthor with an accomplished professor. Yet in these cases, a publication is not a very useful signal: any professor who has collaborated with you will already be communicating the extent of your contribution in her letter of recommendation. The only situation where I can imagine a "publication" making a difference is when you are at a lower-ranked program and your professor is also unknown, and the fact that the work has been published makes the professor's letter more credible. But this case is almost irrelevant, since it's extraordinarily unlikely that both (1) you will be added as a coauthor to a paper and (2) that paper will be published in a decent journal before you enter graduate school. (Unknown professors, almost by definition, don't publish in selective outlets with such ease!) Of course, many people have "publications" in random, unselective journals, but this isn't a meaningful goal. Groucho Marx comes to mind: "I don’t care to belong to any club that will have me as a member." This is very true. No matter how good you are, it is virtually unthinkable that you will be able to finish the publication process in the short time available to you as an undergraduate (or even as a master's student) in an outlet that commands any respect, at least without a faculty coauthor doing most of the heavy lifting. (I admit that this is where my knowledge of the process frays a little; I'm mainly thinking about graduate programs in the top 20 range. If you're targeting less selective graduate programs, it's possible that they care about publications in low-tier journals. But ex ante, of course, you're almost certainly better off focusing on advanced classes and your undergraduate research, without the distraction of trying to push a paper through the refereeing process.) Moral of the story: if you see people chattering on this board about publications, please don't freak out. Only a sliver of the incoming students at even the most selective Ph.D. programs have publications, and none of these students produced the published research themselves. In 2009 and 2010, even the top candidates on the academic job market (Dave Donaldson and Alp Simsek) didn't have publications in economics, and they got tenure-track jobs at MIT and Harvard. When you're an undergraduate worried about grad school, it's easy to let your imagination run wild and envision hordes of superstars who took functional analysis as freshmen and wrote their first Econometrica articles as juniors. But such people do not exist.

I don't know exactly how programs abroad work, but my impression is that while virtually all international students here have a master's degree, generally they did not "transfer" from within a Ph.D. program. My impression is that many master's programs abroad frequently lead into PhD programs -- if you get a master's degree at X university and do well, you are quasi-automatically invited to continue onto the PhD track. Perhaps this creates some semantic confusion about whether people are "transferring" or merely continuing on to a different school after their master's degree. I would say, however, that generally they're not transferring in the sense we would normally use the term. Most elite PhD programs in economics seem to admit most of their international students from roughly the same set of feeder programs in Europe and Latin America, and the usual track in these cases tends to involve a master's degree. (This stands in contrast to admissions from the US, where candidates almost never have a master's degree.) That said, I would still encourage you to apply -- this all depends a lot on where you're from, and there's certainly no hard-and-fast rule about degrees required from applicants.

Sorry, I probably didn't phrase that very well. I didn't mean to say that Oregon is doing badly, just that Harvard and MIT verifiably place a higher percentage of their Ph.Ds in academic jobs than 36.6% and 39.6%, as a quick look at their placement record will indicate. (For instance, in 2009, only 3 out of 19 placements that I see for MIT were in non-academic jobs.) Perhaps the NRC underestimated academic placement across the board and Oregon would still have come out on top using the correct numbers -- I have no idea. But these errors suggest some serious problems in the data used to create the ranking. Of course, as you suggest, the raw fraction of graduates that go into academia isn't a particularly useful metric anyway. For students hoping for a career at a high-ranked research institution, the most important feature of a Ph.D. program is its ability to place students at such institutions -- past a certain level in the graduate program rankings, the ability to get some academic job is pretty much guaranteed, and programs cannot meaningfully be distinguished on that basis. Yet this is something that the NRC didn't even really try to measure. Fortunately, there are some studies that examine placement -- see, for instance, this. (Inevitably, though, all these rankings are a little dated, and it's best to look at recent placements directly.)

As long as we're doing a little trash-talking... If MIT's penance for having better placement than Harvard is coming second in a study of questionable methodology, I'm more than happy with the tradeoff. ;)

At first, I was a little disappointed to see Harvard "outrank" MIT -- in the same superficial sense that I'm disappointed when my alma mater ranks lower in the AP coaches' poll. Then I took a closer look at the Chronicle site. According to the "% of PhDs with academic jobs" subranking, NRC seems to think that only 36.6% of MIT Ph.Ds and 39.6% of Harvard PhDs get academic jobs. This is as compared to the survey leaders, the University of Oregon and Florida State University, at 56.5% and 53.3% respectively. As much as this Duck fan would like to see the University of Oregon doing well, this is obviously not at all credible -- a much higher percentage of MIT and Harvard students get academic jobs than that. When the data errors are so obvious, it's hard to treat this ranking with any respect.

I wish I was paid by ETS -- it would supplement the grad student stipend quite nicely! In all seriousness, I don't want to freak everybody out about GRE math scores below 800. Students can be (and are) accepted into virtually every program with non-perfect GRE Q all the time. But for the very narrow subset of people I have in mind -- students from lesser-known programs who are extremely ambitious and are applying to the top 8 -- I think that it's important to keep in mind that the primary barrier to admission is an information failure, and every universally understood credential that helps remedy this information failure (like the GRE Q) is relevant to your chances.

You're probably right that MIT is not a good example -- I naturally gravitate toward mentioning it because it's the school where I have the most knowledge, but in most cases the signal required for someone from an unknown program to be admitted is so strong that the GRE adds very little value. There are, however, a few exceptions. Winning the NSF is one of the best ways for people without top econ programs on their resumes to earn a shot at admission; conditional on the NSF, your other accomplishments don't have to be nearly as absurd to draw attention. In this environment -- where financially, your admission is essentially costless to the department -- I can imagine GRE Q scores playing a role. A better example would be someplace like Yale or Northwestern where it is more plausible for students from lower-ranked programs to be admitted. Suppose that you've taken a large set of advanced math and economics classes and have great letters from unknown professors at your institution. Conditional on that background, someplace like Yale might guess that your chance of success in the program is about 50% (I'm making these numbers up.) The 50% chance of failure arises from the chance that your stellar background is essentially a fluke -- the grading policy is unusually generous, the classes are not as advanced as they seem, your advisor happens to write really good recommendations, and so on. If Yale thinks that a 50% chance of success is too low, they might pass on the application despite its promise. Suppose that in all cases, the chance that you'll pass the lower-bound GRE cutoff is virtually 100%. Suppose further that in the case where you're a "fluke", you have a 50% chance of scoring below an 800 on the GRE, while in other case you have a only a 20% chance of scoring below 800. Conditional on scoring 800, then, your estimated success probability increases to 61.5%. This doesn't seem like a huge change, but admissions decisions at the margin hinge often hinge on very minor distinctions between applicants, and it's plausible that the extra signal offered by an 800 on the GRE math would make the committee feel more secure about a promising but hard-to-pin-down background.

This is definitely one possible explanation. It's not clear to me how much predictive value the GRE has at the high end -- whether it's exclusively useful as a culling device to avoid costly reviews of applicants who will never make it, or whether it has utility beyond that. My instinct is to say that there is some value -- if only because I've never known anyone who was really good at math who failed to get an 800 -- but as you point out, then we have to ask whether that mathematical ability will show up on an academic record anyway. (making the signal from the GRE essentially superfluous) My guess is that it depends on (A) where you are applying, and even more importantly (B) the undergraduate (or master's) institution you attended. Unfortunately, it is very difficult to get into top 5 institutions from lesser-known universities. This isn't because there aren't talented students at these universities; it's simply because the signal-to-noise ratio is so low that programs have trouble identifying who the best students really are. For instance, among domestic students in the MIT first-year class, there is a grand total of 2 (!) coming from economics programs outside the top 20. One of those 2, one was an RA for an MIT professor, while the other won an NSF fellowship and had an extremely good record as an undergraduate and as an RA at an elite think tank. I am absolutely confident that there are many, many students at programs outside the top 20 who are just as talented as people who are accepted to MIT, but unfortunately the noisiness of information coming from these programs makes it very hard to send a credible signal. This is a central issue in admissions. When admissions committees like these evaluate someone coming from an institution with a low signal-to-noise ratio, they will put relatively more weight on uniform metrics like the GRE. Hence if you are an aspiring applicant from a school without a history of sending students to top programs, I suspect that it is very important to get the highest GRE math score possible (at least if you're aiming for the top tier). On the other extreme, I'm sure that someone who aced Math 55 at Harvard will have no trouble getting admitted with a 780Q; in this case, the signal from coursework is so strong that the GRE pales in comparison. Given the undergraduate backgrounds of people at MIT, I doubt that minor distinctions in GRE scores played much of a role, precisely because most of them came from institutions with long histories of sending students to top programs -- or received letters from famous economists whose judgment the committee trusted. But my impression is that this forum consists of students from a much wider variety of backgrounds, in large part because it has higher value as a resource to people who don't have peers with similar plans. So... for many of the people on this forum, I think that even minor GRE distinctions may play a larger role than is commonly recognized.

Absolutely. My musings were probably a little out of place in this thread; I just wanted to push back a little against the overwhelming consensus here that "small" variations in quantitative score are meaningless, not specifically say that the OP should retake the test. If there is no compelling reason to suppose that he has a high chance of getting an 800 on a rest -- and there is no certainty that he won't do significantly worse -- then he absolutely shouldn't retake. My advice only applies to a relatively small subset of the applicant population. (Then again, if we are willing to go a little further with the economic analysis, we could pose it as a game theory problem and arrive at somewhat different conclusions. If an applicant retakes the GRE whenever his expected score is higher than the score he received on his first attempt, then failure to retake indicates that his expected score is below the score he received. But then applicants in the top half of this bottom group will want to retake to prove that they aren't in the bottom fourth of the population, and so on... you see unraveling until virtually everyone retakes the test. In the real world, of course, this doesn't happen, partly because there are nonzero costs to taking the GRE (so there needs to be a bigger disparity between the expected score and received score to make a retest worthwhile) and partly because both parties in the process aren't fully rational. Still, it's interesting to consider whether failure to retake might be viewed as a sign that the applicant didn't expect a much higher score, and whether inferences are adjusted accordingly.)

I don't have any conclusive answers here, but I'd like to inject a little doubt into the discussion. It's been the consensus on TM since time immemorial that a 780Q or 790Q is virtually as good as an 800Q for admissions purposes, but it's not clear to me whether this is based on any hard evidence. There is certainly some reason to think this might be the case: the primary role of the quantitative GRE at most programs is to act as an initial filter, and 780 and above is good enough to pass the first cut essentially everywhere. At the very least, then, there is a steep decline in the marginal benefit from GRE score improvements once you pass 780 or so. But do we really know that the marginal benefit is zero once you pass 780 -- or at least close enough to zero that retaking the test is never justifiable? This isn't clear to me at all. I've heard some suggestions from faculty involved in admissions that the GRE score is meaningless past the cutoff, but I've also heard statements that suggest that GRE scores carry some weight at every stage. My best guess is that attitudes are simply heterogenous: professors read applications in different ways. Some will care about the difference between a 780 and an 800 (or even a 790 and an 800) and some won't, based on their own analysis of whether that difference is meaningful. I hear a lot of attempts to intuitively justify why the difference between a 790 and an 800 shouldn't matter because it's so "small", but these don't strike me as very compelling. Let's frame it as an econometrics problem with censored data. Your "true GRE score" y* is the sum of your mathematical ability 'b' and a noise term 'e'. Your "reported GRE score" y is the minimum of y* and 800. Depending on the distributions of 'b' and 'e', the expectation of 'b' conditional on y=800 may be very different from the expectation of 'b' conditional on y=790. That said, I'm not arguing that the OP should retake the test -- the typical person in his situation almost certainly should not, unless there is very strong reason to think that he will do better on the second try. But I don't want to be quite so glib in saying that there is no difference between a 790 and an 800, and I can imagine a few situations where someone with a 790 might want to retake -- if, say, he had finished every practice test with 15 minutes to spare and an easy 800, and every indication was that the 790 was a massive fluke.

Honestly I don't think that this applies to the vast majority of applicants who talk about recommenders who "know them well". I certainly didn't have any professors inviting me to Thanksgiving! I think "know well" in this context really just means that you've had either a great deal of meaningful one-on-one conversation, or some one-on-one conversation combined with sustained engagement in a small group setting (i.e. seminars, a small class, as an RA on a project, etc).