The Economics Pre-Doctoral Fellowship: What is it, and should you apply for one?
Note: There was a recent discussion on economics Twitter about pre-doctoral fellowships and I will cite some of these tweets for anecdotal supporting evidence.
What is it?
Disclaimer: The discussion in this section is my analysis of the market. It may be incomplete or incorrect in part or in whole.
The short version is that it is a [paid] job, usually in an academic setting but also in a research focused government institution or think-tank organization, where the employee provides research assistance on economics research projects, and generally with the intent to go onto an economics PhD program, but without any guarantees. Basically, it’s just a Research Assistant position for economics, or economics-adjacent, faculty. So why don’t we just call it what it is, a Research Assistant (RA) job? Well, let’s explore how we got here.
In the early days of economics, through even up to the 1980’s, the focus of the discipline was largely on theory. To be sure, work done during this time was critical to the development and foundations of economics today; some notable names from this time were people like Jean Tirole who created the foundations of Industrial Organization, John Nash who developed fundamental results for Game Theory, and Milton Friedman who developed a lot of the basis for today’s Finance literature. However, there was not a lot of work on empirical aspects of economics such as testing some of these theories in the real world. This was due to a variety of factors: computation wasn’t as common place as it is today, data was not anywhere as easy to come by as today, and a lot of fundamental economic theory needed to be developed first, before testing of these theories would be fruitful. But things began to change in the 1990’s.
The 3 constraints we described above for the focus of economic work on theory began to loosen at this time: computing was becoming more accessible, governments had started collecting and making data more accessible, and a lot of the theory frameworks for thinking about economics had come into the literature. Now the field was prepped to make the transition into more empirical work evaluating theories and hypotheses in the real world. The graph below, lifted from Noah Smith’s blog on Bloomberg and originally presented by Daniel S. Hamermesh, shows the trend of economic research by type over time. This shift also represents a significant change in the nature of the work the majority of economists performed.
Theory is largely an individual pursuit. It’s probably well captured by the caricature of the genius professor standing alone writing on a blackboard with chalk. Sure, there may be some collaboration between theorists as they work across specialties or develop models together, but the work is not outsource-able; the theorist has to do all of the work him/herself because it is private knowledge trapped in the mind until it is written out for communication, i.e. the paper gets written. Empirical work, however, is quite different in its level of outsource-ability.
The empiricist, usually a professor or a PhD economist at a government institution, generally has the full responsibility for thinking about the problem, generating the hypothesis, acquiring the data, interpreting the results, and writing the paper, but the actual core computation work portion of the process, that can be outsourced. As more and more datasets become available, as the level of analysis required for a single paper to qualify for publishing increases, and as more and more papers need to be published for tenure, the opportunity cost to the empiricist of spending time toiling away on coding rises. This is where the RA position comes in.
Supply, generally, follows demand; and the RA position is no different. I mention this to distinguish between different ideas of why research positions in the form of “pre-docs” came into vogue; it was not that students wanting to do research created these positions, but rather the market dynamics described in the previous paragraph are the causal factors. Research Assistant positions came into vogue because empirically focused academics and researchers needed additional labor for computational work streams. The introduction of the “pre-doctoral fellowship” branding comes from subsequent market effects on the PhD application market, which we discuss later.
Who should do a pre-doctoral fellowship?
At the most fundamental level, only someone interested in pursuing a PhD in Economics should consider doing a pre-doctoral fellowship, i.e. become an economics research assistant. We will discuss this a bit further later, but the only unique benefit from this position relative to any other positions for which an economics RA might otherwise be qualified is an improved application to graduate school; the boost for PhD programs in particular is largest but applications for master’s programs, MBA programs, or law school can be improved relative to just directly from undergrad. Although if a program in this latter set is your target, there are other much more beneficial things one can do to improve an application. Assuming the economics PhD criterion is met, there are some additional concerns to address.
For International Students with a Non-US Bachelor’s Degree:
Applicants in this category face a different challenge as compared to American students graduating from US undergraduate programs, or other international students graduating from US undergraduate programs. Specifically, admissions committees are not 100% sure about the context of the grades on transcripts from a foreign institution. To get around this, many international students intending to go onto a US economics PhD program will get a US master’s degree before applying to remove the noise on their undergraduate grades — doing well in US master’s coursework can validate strong performance on the undergraduate transcript.
Richard Startz and Garrison Schlauch describe in their exploratory paper “The Path to an Economics PhD” various observable characteristics of the graduating cohort of Economics PhD students in 2016–2017, presumably for students entering/applying around 2010–12. They state, about master’s degrees, in response to data collected about the fraction of incoming PhD students holding a graduate degree at time of enrollment from another study: “The difference presumably reflects international students being much more likely than Americans to have a master’s degree.” I hypothesize that the reason is as described above.
For Students with a US Bachelor’s Degree or US Master’s Degree:
The answer is somewhat that it depends. If one is targeting a top 5 program to do applied work, then almost certainly yes. If the target isn’t necessarily top 20 then perhaps it is possible to do without it, assuming good grades on the undergraduate transcript. If primarily interested in theory, then also maybe not. In a general sense my answer to this question would have been that it makes sense if one is a non-traditional applicant such as someone who wants to transition from industry back into academia, or has no formal economics exposure but has a strong technical background. However, due to competitive pressures, my answer now is more along the lines of: do an RA job if you want to be in a top economics department and you can actually get a position.
The Market for Research Assistant Positions
As I begin this section, I will call back to the motivating factor for the rise of research assistant positions in economics: labor demands for programming/econometrics talent. As with all labor markets, this market has minimum qualifications expected for the skills necessary to get a position and it is competitive, i.e. the more qualified market participants find jobs on the market at higher rates than less qualified market participants.
Note: While there are more policy oriented RA positions such as those at the Federal Reserve, this article focuses on academic RA positions and the market discussed here is limited to the same.
Let’s start with what the typical application requirements looks like:
- Bachelor’s degree in Economics, Computer Science, or other quantitative discipline.
- Strong Quantitative Background.
- Strong Programming Skills. Stata is usually required with Python and/or R being asked as an additional “preferred” requirement. However, some positions will accept R in lieu of Stata.
- Academic transcripts. Some advertisements will explicitly state strong/good grades required.
- Previous Research Experience required or preferred.
- 2–3 References often required.
- Some programs require unsponsored ability to work in the United States but this is becoming less and less of a requirement with time. The trend is briefly discussed later.
- A skills test is usually required after the first review of applications.
To provide a quantitative sense of how common some of these requirements are, the following is a graph showing the proportion of various skills asked for on predoctoral fellowship advertisements taken from the NBER Non-NBER jobs page on July 12th 2020.
Following on from the requirements listed above, Stata is asked for in almost all applications. Sometimes this is asked for as a requirement and sometimes it is asked for as an example of a programming language in the “Strong Programming Skills” requirement. In practice however, this may be an absolute requirement for most microeconomics focused positions. To provide some additional context, a screenshot of a position at UC Berkeley is shown below. The programming skills section includes Stata as an example rather than a requirement but when the skills test was sent it was mandatory to complete it in Stata.
Why is this? Here is a brief aside about the history of computing in Economics.
Stata is proprietary software created by the Stata Corporation. It was launched in 1985, well before the creation of today’s popular data processing languages like Python and R (these would come about in the 1990’s). At this time in the computing world there were only a few “modern” programming languages around which could handle computation such as C++, Objective-C, Perl, etc. There was also a fairly strong separation between general programing, maybe more known today as Software Engineering, and computation focused fields like economics. Specifically, the level of training required to get up to speed for programming in something like C++ was somewhat prohibitive when combined with the additional training necessary to become an economist; there was no common place internet, YouTube, or Coursera at this time and these low level languages required significant training in Computer Science to use. There was a niche in the market for computational software that was specialized for just the needs of economists, such that it was ok if limited for general programming but be easy to get started on for just econometric computation. The Stata Corporation filled this niche with Stata.
Stata’s legacy still affects economics today, despite the emergence of strong data facilities in Python (Pandas by Wes McKinney) and R (Tidyverse by Hadley Wickham). Economists trained in the 1980’s through the 2000’s were trained on the use of Stata in their computational work. After becoming full fledged economists themselves, they became somewhat immune to the traditional competitive dynamics that would dictate they learn new technologies; economists’ work streams are fairly isolated from others and they are the ones who dictate what technologies they use. This meant that as economists advised students coming into economics PhD programs and taught coursework for computing in economics, they entrenched what they know, i.e. Stata. As a consequence, Stata is the defacto common language of computational economics. Some of the more enterprising new economists are learning Python and R, particularly if they are interested in new methods like NLP and machine learning, but the large pool of economists are still monolingual and, especially if the predoctoral position is to be spread across more than one faculty member, it is very likely that Stata will be a silent, or explicit, requirement for the position. Now back to the RA market.
Expanding beyond Stata, there are a number of other programming languages that can be in demand on the market. A number of positions will also ask for R as preferred, or may accept in lieu of Stata but this is limited. Among “preferred” languages, Python is probably the most common language asked for. I speculate this is because of its machine learning and web scraping libraries. C and C++ will generally be asked for in macroeconomics focused positions since they use models which are very computationally expensive. SAS will tend to be asked for positions dealing with healthcare data since this became the industry standard for industries with healthcare data such as the health insurance and medical industries. SQL is trending upward in roles that work with large datasets since the data is likely stored in a relational database which uses SQL as the interface.
As mentioned there is also a skills test portion of the application if successful at the resume review stage. The test generally involves a provided data set, some type of data manipulation and/or cleaning, and then an econometric analysis with discussion. This is probably best shown rather than described so please navigate over to my GitHub for an example of a task and the analysis. I was successful in moving onto the interview stage for this skills test so it is a reasonable example of what is required. The skills tests vary based on position. Microeconomics positions will often test some type of reduced form model, whereas macroeconomics might test a time series model.
By this point in the article you might be thinking: “this is all great, but where can I get a position?” The following graph shows the distribution of universities in the sample I pulled from the NBER website for analysis.
This is somewhat representative of the distribution of positions. The universities listed do not differentiate between specific schools at the university, e.g. Harvard includes both the Harvard Economics Department and positions at Harvard Business School. That said, the largest recruitment pools do tend to be at Chicago, Harvard, and Yale. There is also sizable recruitment at Stanford, Columbia, and MIT. The others, less so but they do recruit. Perhaps the largest, and I believe the first pioneer of the Research Assistant program, is Harvard Business School. They easily take in the most of any school and they recruit all year round (as compared to the others that mostly have summer start dates). However, their positions are also much more diverse than other programs; not all are intended for transitioning to graduate school. You can find out more about that on their website.
Finally, a very important question for the market portion of predoctoral fellowships is how much do they pay? Well, the biggest compensation is the letter of recommendation for graduate school applications. This isn’t even a joke. If a letter, or letters, for applications isn’t very valuable to you, this is not the job you want, because frankly the pay is not very good compared to what someone with the skills to actually get one of these positions would make in industry right out of their bachelor’s degree program.
Of the advertisements sourced, only a few of them actually stated the salary on the ad. As you can see, the pay is generally around USD$50,000. I can confirm this from personal experience and from talking to other RAs. The outlier on the left of the graph is the London School of Economics which advertises its salary in Pounds Sterling. This converts to around $39,000 as of July 18th 2020, still quite a bit lower than the US based positions but perhaps this is comparable when adjusted for cost of living in London. The salary is generally non-negotiable, so you either take it or leave it, and given the number of people who would happily accept the position for the possibility of getting into an Economics PhD program, faculty don’t really have any incentive to negotiate. We will discuss how this affects potential diversity and equity in the Economics field in another article. There is, however, one exception to this $50,000 standard.
Harvard Business School will pay you more depending on your qualifications. As mentioned above, HBS has a much more diverse set of research positions that they recruit for. In order to deal with the diversity in requirements for these positions they have a more nuanced compensation scheme. I have an MBA from the Yale School of Management and they priced that at an annual $70,000 salary. Another RA here at HBS has a master’s degree in economics and they priced him/her at around $62,000. Yet another RA has an MBA from HBS itself and MBB consulting experience pre-MBA: they priced him/her at $92,000. I’m not sure what their compensation scheme is but if I had to speculate, they don’t incorporate experience or skills, but rather brand names and degree levels. Is it fair? I’m not sure that it is, but with so many people trying to work at HBS, they don’t have any incentive to change it. While HBS may pay more based on your qualifications, there is one notable downside to HBS’s positions: they do not offer visa sponsorship for international applicants.
As promised above, here is a brief discussion about visas and trends for international applicants. In prior years most of these programs did not sponsor visas for international applicants. If an international student had OPT (still a thing as of 2020) available then they could obtain a position but otherwise they were not eligible. This is largely not the case anymore and most of the larger programs do offer visa sponsorship. As far as I can tell, UChicago was the pioneer in making visa sponsorship available for their RA positions and lots of schools followed suit over time. As of this writing I can confirm UChicago, Yale, and Stanford will provide sponsorship, and I believe the same is true for Columbia and Wharton but they do not advertise it. I can confirm that currently Harvard and MIT do not. It may be that they do not feel the need to do so as they can likely leverage the strength of their brands to secure the top candidates from the US market.
Competitive Dynamics in the Market and Spillover to PhD Applications
The answer to this question is the same as the answer to another: Do you want to get into a top economics PhD program?
As we start this section, I call back again to the motivating factor for the rise of research assistant positions in economics: labor demands for programming/econometrics talent. I do this to emphasize the difference between this and something like a school education program. This is a labor market, schools/faculty pay for the research assistant to do work. This is not an apprenticeship program or a training program (although there may be some elements of these), and it is not a university education program. This distinction matters because the direction of compensation highlights something these positions would not generally be selecting on: academic/training potential or diversity. Education programs can select on these things because the compensation goes from the student to the school and the goals are different, namely educating/training students to do well. Diversity and academic/training potential fit these goals; labor market positions select primarily on existing skills and experience, not potential to learn skills, although this may be selected for as a secondary aspect. With that in mind, what are the competitive dynamics for applications in this market?
We make 2 assumptions for this question:
- The market is demand constrained. In other words, the number of research positions faculty are looking to fill is (significantly) smaller than the number of applicants looking for these positions.
- Admissions committees strongly value research experience in PhD applications. There may be a similar conclusion if it is only weakly valued but I think reality more closely resembles a strong valuation.
A demand constrained market usually implies that the demand side has more market power with regards to having their preferences met, and potentially leads to having stricter preferences. It’s not possible to disentangle whether preferences are stricter than would otherwise be the case with the data here, but that is also not the purpose of this article so we simply take preferences as given. To recap, what are the most salient requirements advertised?
- Required programming experience, often specifically with Stata
- Transcripts and/or good grades
- Prior research experience
- References
- Skills test
Since the primary value proposition of the research assistant position is an improved application for PhD programs, I discuss the competitive dynamics in the market and the effects on PhD applications together.
When I discussed who should do a pre-doctoral fellowship, my answer was that I would previously have said someone who wants to apply to a PhD in economics but is non-traditional, either without formal economics training or transitioning from industry. A research assistant position would help get additional economics exposure and build relationships with professors that would convert into letters of recommendation. This may even have applied to students who had weaker applications immediately after their bachelor’s programs and wanted to build a stronger application. The point being, it would have been a good stepping stone for potential PhD applicants who either couldn’t get into a PhD program or wanted to aim for higher ranked programs than they could get into at the time. This is where our second assumption comes in; this strategy makes sense only if admissions committees strongly value research experience, at least enough to offset the other weaker parts of the application that were causing problems in the first place.
According to work done by T. Aldrich Finegan, there have been roughly around 1100–1200 graduating economics PhD candidates per year since 1997 across all US PhD programs. This number has not shown any significant growth. Assuming no underlying changes to the overall rate of conversion from enrollment to graduation of PhD candidates, this means the number of available positions in economics PhD programs has been fairly consistent since at least 1997. Harvard states on their website: “The Harvard Economics Department typically receives over 600 applications and admits between 40–44 students each year.” Harvard is probably an outlier on the ratio of admits to applicants, but other schools show similar ratios. Columbia Business School received 256 applicants to their Finance and Economics program and enrolled only 6. The number they admitted was probably somewhere around 12–20 assuming yield of 30%-50%. This suggests to me that this market is demand constrained as well. In a competitive fixed demand market that is also demand constrained, the supply will compete, strongly.
With our assumption about admission committee’s strong preference for research experience, and the difficulty to adjust other parts of the application such as transcripts, the easiest way to compete would be to get additional research experience as part of a post-graduate research assistant position. As people begin competing in this way, and research assistant positions grow in demand/availability, more and more applicants have a year or two doing full-time research and the letters to support it in the application. Eventually, this becomes the norm to be competitive.
This norm then trickles down to the prior step in the process: the research assistant position. Competition for these positions also increases as PhD applicants who would have previously been competitive in PhD applications without postgraduate research experience now also have to enhance their applications with significant research experience to be competitive. This extension of competition from the point of the PhD application backward to the application for the research assistant position effectively shifts the point of entry into economics. The research assistant position is no longer an alternative path to a PhD program for non-traditional applicants or applicants who want to strengthen their applications, it is the de-facto path, with the traditional path now being the exception.
This explains some of the requirements stated on the advertisements for research assistant positions. Why would great or excellent grades/transcripts necessarily be required in addition to a skills test? Surely, only one of these would be necessary if the purpose of the research assistant position was to do computational work. As a point of comparison, industry relies only on the skills test to validate ability to perform the tasks related to a job, no request for transcripts is made. My hypothesis is that what faculty are really selecting for with the research assistant position nowadays is not just someone to do the computational work for their projects, but rather they are selecting for candidates likely to get into a top economics PhD program.
There are 3 pieces of evidence that support my hypothesis:
- Faculty do have to compete at least somewhat for top talent in the pool. The way they seem to do this is by highlighting where previous RAs have gone onto for graduate school. If they have the best ranked schools on that list, this is likely a more appealing solicitation for applications than if they do not.
- There is some level of investment in training the RA in research methods and developing their interest in economics. One of the common requirements stated on the advertisement is “a long term interest in pursuing an Economics PhD”. This differs from industry where there is not any expectation from the company that someone hired for a job is to progress in a specific direction and where the company is invested in that progression.
- econjobmarket.com, the main market place for economic jobs matching, has added a section specifically for pre-doctoral job advertisements suggesting that they have some expectation that these pre-doctoral opportunities are a significant part of the economics pathway.
Finally, we get to the question we asked way back at the beginning of this article, why don’t we call pre-doctoral fellowships what they are, i.e. research assistant positions? The branding has changed because the competitive pressures at the PhD application level propagated backward to research assistant positions, and now these positions are a fully formed feature of the pathway to academic economics, much like the post-doctoral fellowship. The term “Pre-Doctoral Fellowship” solidifies the shift in the pathway to an economics PhD and captures the information that the 1–2 (or more) years of full-time post-graduate research is indeed a common part of a full application to a PhD program.
So, should you apply for one? The answer to this question is the same as the answer to another: Do you want to get into a top economics PhD program?
