When the Academic Default Weakens
This chapter examines the changing share of economics PhDs who take academic first jobs in the leading U.S. programs studied by the IZA paper. The key task is to read that decline as a change in the allocation of early-career destinations—not as proof that academia has disappeared, become undesirable, or been replaced by one better alternative. The distinction matters because the same aggregate trend can contain very different prospects for graduates from different program tiers.
The academic share peaks, then recedes
For much of the period covered by the study, an academic job was the most common first destination for graduates of the selected programs. But the pattern is not a straight line downward. Academic placement follows an inverted U: it rises to a peak of about 67% in 2009 and then falls to 52% in 2021. The published figures use three-year moving averages, so they smooth year-to-year fluctuations rather than presenting each cohort as an isolated event.
That smoothing is useful for seeing the direction, but it also tells us what the number is not. A moving average should not be read as saying that every cohort in 2021 had exactly the same outcome, or that the decline began at one precise moment. The reported pattern is a sustained change in the composition of first destinations across neighboring cohorts. It is strong enough to motivate an explanation, but not fine-grained enough to identify a single hiring shock or turning point.
That shape changes the interpretation. A profession can experience a long period in which academic placement is strong, followed by a sustained reallocation, without the academic route becoming marginal. In 2021, academia still received a majority of the graduates in this sample. The meaningful change is that it no longer absorbs as large a share as it did around the peak. The question is what has grown alongside the decline and whether the new destinations offer comparable forms of opportunity.
Across the full period, an average of 42% of graduates find jobs outside academia, and that share increases over time. “Outside academia,” however, is a location in the original binary measure, not a description of one common experience. A public institution, a consultancy, a finance or industry employer, and a technology company may differ in work, pay, research freedom, geographic mobility, and the kinds of professional signals they create. Treating them as one remainder hides the very reallocation that makes the trend important.
The 42% figure also has a different role from the 52% figure at the end of the sample. One is an average across the observed period; the other is a smoothed value for a recent point. They should not be combined as if they were two pieces of one same-year accounting identity. The important conclusion is qualitative and supported by both measures: nonacademic work is a durable part of the first-placement market, and its importance rises as the academic share recedes.
The academic trend also differs by the rank of the PhD-granting institution. Graduates from top-10 programs have an academic-placement advantage ranging from roughly 3 to 15 percentage points over graduates of programs ranked 11–30. The divergence is visible not only in levels but in timing. Academic placement for top-10 graduates continues to increase into the late 2010s, while placement for graduates of programs ranked 11–30 begins declining in the early 2010s. In the most recent periods reported by the paper, the average academic-placement share is around 58% for top-10 graduates and 48% for those from programs ranked 11–30.
This comparison is not a claim that a program rank mechanically determines a person’s outcome. It is a within-sample description of different placement rates for candidates grouped by the rank of the institution awarding their PhD. The gap could reflect institutional prestige, differences in candidate selection or fields, advisor networks, employers’ expectations, or several factors at once. The value of the comparison is that it reveals heterogeneity hidden by the overall average; its limitation is that it does not identify the channel producing the heterogeneity.
This is a first indication that “the academic market” is not one market with one temperature. A falling overall share can conceal a persistent advantage at the top of the program hierarchy. It can also mean that the composition of academic destinations is changing: fewer graduates may enter academia, while the relative difference between program tiers remains. The aggregate trend opens the question of hierarchy rather than settling it.
The gender series adds another qualification. With the exception of the 2007–2012 period, the broad academic-placement patterns for men and women are similar in the study’s moving averages. In the candidate-level placement model, gender does not have a large effect across destinations; the notable differences are a higher association with consulting for women and, in the more recent sample, a lower association with finance and industry. This does not establish equal opportunity in the pipeline or in later careers. It says only that, conditional on the study’s measured variables and placement categories, the large aggregate decline in academic placement is not primarily a gender-specific trend.
The time series therefore gives us a useful but limited result. The academic default weakened across the observed cohorts. Program rank still separated candidates’ chances of entering academia. Other destinations grew. None of those observations tells us by itself whether the cause was weaker demand for academic labor, changing preferences among graduates, changes in fields, working conditions, macroeconomic shocks, or a combination of factors.
A first placement is a state, not a destiny
The diagram makes a distinction that a headline percentage tends to erase. The study observes an entry point and a first destination. It does not follow every graduate through subsequent moves. A person who starts at a university may later join a government agency or a company; someone who begins outside academia may return to a university. Those transitions are important, but they are not what the reported academic-placement share measures.
This boundary also explains why the study separates top-50 academic jobs from other academic jobs. The distinction is about position within the academic hierarchy, not a universal verdict on job quality. A university outside the top 50 may offer better working conditions, a stronger fit with a person’s field, or a more promising local academic environment. At the same time, separating the two categories lets the researchers ask whether the forces associated with reaching the profession’s most prestigious institutions differ from those associated with academic employment more generally.
The difference between a state and a destiny matters for career decisions. A first job can affect the references a graduate receives, the collaborators they meet, the research agenda they can pursue, and the skills that later employers recognize. Those effects may compound, but the initial placement does not mechanically determine the rest of a career. In practical terms, a declining academic share can mean that candidates are facing a more varied set of starting points, not that every person who takes a nonacademic job has left research permanently.
The national Survey of Earned Doctorates provides a separate way to see this measurement issue. The SED asks doctorate recipients about postgraduation plans at the time they complete the survey. Its “definite commitment” means accepting employment or postdoctoral study in the coming year, or returning to predoctoral employment. A recipient still seeking employment may find a position after responding. The survey therefore provides a current graduation snapshot rather than a completed account of the labor market.
Its categories also have their own boundaries. Academic employment includes four-year colleges and universities, medical schools, university-affiliated research institutes, community or two-year colleges, and foreign educational institutions. Industry or business includes for-profit organizations and self-employment. Postdoctoral study is separate from non-postdoc employment. These definitions make the SED useful for a national benchmark, but they do not reproduce the IZA study’s top-50 distinction, exact employer records, or advisor links.
The two sources should consequently answer different questions. The IZA study can tell us how first placements are distributed within a selected group of leading economics programs and how those placements relate to program and advisor characteristics. The SED can tell us how a much broader population reports its plans at graduation. Neither source alone tells us the complete long-term career path, and comparing their percentages without checking denominators would turn a difference in measurement into a false trend.
The SED’s field taxonomy adds another reason for caution. Beginning in 2021, the survey changed how it collects and aggregates fields of doctorate. It provides a trend taxonomy for historical comparisons and a newer detailed taxonomy for recent single-year tables. Even the number of economics doctorates reported for 2024 differs between those systems. A careful reader should therefore ask not only “What percentage is academic?” but also “Which population, time point, field definition, and denominator produced that percentage?”
What the decline does not explain
A descriptive trend is an observation about what happened, not a ready-made mechanism for why it happened. The placement study’s falling academic share could reflect changes in the demand for faculty, but it could also reflect graduates’ preferences for other work, the relative compensation or stability of different sectors, the supply of academic positions across countries, changes in research fields, or shifts in the composition of candidates. The study’s cohort and program information help organize those possibilities, but they do not convert the time series into a causal estimate.
There is a similar danger in reading the rank gradient too quickly. The higher academic-placement share for top-10 graduates may reflect the value of institutional prestige, differences in candidate preparation, advisor networks, field composition, or employers’ assessment of program signals. The evidence shows that rank is associated with different placement rates; it does not isolate one channel. Later analysis of advisors and networks will make that problem more visible, but the basic rule already applies here: a predictor in a placement pattern is not automatically its cause.
Nor should a weaker academic default be treated as an unambiguous improvement. More options can increase autonomy and reduce dependence on a single career model. They can also expose graduates to different trade-offs. A company job may offer higher compensation but less control over research questions. A public-sector role may widen policy influence but be constrained by institutional priorities. An academic job may preserve research independence while offering uncertain location or promotion prospects. The available evidence does not rank these outcomes for every person; it shows that the allocation among them is changing.
The conclusion could change with better evidence. Longitudinal records following the same graduates would show how often first placements persist or reverse. Comparable data from European, Chinese, and other doctoral systems would test whether the U.S. elite-program pattern travels. More direct measures of candidate preferences, job quality, and employer demand would help distinguish supply from demand. Until such evidence is available, the defensible claim remains a description of changing first destinations, not a complete theory of the academic labor market.
The most useful conclusion at this point is modest but substantial: in the elite-program sample, academia is no longer the only destination that can plausibly describe a large share of new economists’ first jobs. Yet the weakening of the default does not tell us whether the alternatives are broadly accessible, who benefits from them, or whether one new sector is doing most of the absorbing. Those questions cannot be answered by treating “outside academia” as a single box.
Technology is the clearest test. Its growth is large enough to change the overall map, but its internal structure may differ from the public, consulting, and finance routes. If it recruits across program tiers and candidate groups, it could represent a genuine widening of opportunity. If its placements are concentrated among a few employers, the apparent opening may be narrower and more fragile than the sector label suggests. The next question is therefore not simply where graduates go when they leave academia, but how the new technology door is built.