Skip to main content
NCBI home page
eLife logoLink to eLife
. 2023 Nov 23;12:e78706. doi: 10.7554/eLife.78706

The changing career paths of PhDs and postdocs trained at EMBL

Junyan Lu 1,‡, Britta Velten 1,†,§, Bernd Klaus 1,†,#, Mauricio Ramm 2,, Wolfgang Huber 1, Rachel Coulthard-Graf 2,✉
Editors: Peter Rodgers3, Peter Rodgers4
PMCID: PMC10666930  PMID: 37994719

Abstract

Individuals with PhDs and postdoctoral experience in the life sciences can pursue a variety of career paths. Many PhD students and postdocs aspire to a permanent research position at a university or research institute, but competition for such positions has increased. Here, we report a time-resolved analysis of the career paths of 2284 researchers who completed a PhD or a postdoc at the European Molecular Biology Laboratory (EMBL) between 1997 and 2020. The most prevalent career outcome was Academia: Principal Investigator (636/2284=27.8% of alumni), followed by Academia: Other (16.8%), Science-related Non-research (15.3%), Industry Research (14.5%), Academia: Postdoc (10.7%) and Non-science-related (4%); we were unable to determine the career path of the remaining 10.9% of alumni. While positions in Academia (Principal Investigator, Postdoc and Other) remained the most common destination for more recent alumni, entry into Science-related Non-research, Industry Research and Non-science-related positions has increased over time, and entry into Academia: Principal Investigator positions has decreased. Our analysis also reveals information on a number of factors – including publication records – that correlate with the career paths followed by researchers.

Research organism: None

Introduction

Career paths in the life sciences have changed dramatically in recent decades, partly because the number of early-career researchers seeking permanent research positions has continued to significantly exceed the number of positions available (Cyranoski et al., 2011; Schillebeeckx et al., 2013). Other changes have included efforts to improve research culture, growing concerns about mental health (Evans et al., 2018; Levecque et al., 2017), increased collaboration (Vermeulen et al., 2013), an increased proportion of project-based funding (Lepori et al., 2007; Jonkers and Zacharewicz, 2016) and greater awareness of careers outside academic research (Hayter and Parker, 2019). Nevertheless, many PhD students and postdocs remain keen to pursue careers in research and, if possible, secure a permanent position as a Principal Investigator (PI) at a university or research institute (Fuhrmann et al., 2011; Gibbs et al., 2015; Lambert et al., 2020; Roach and Sauermann, 2017; Sauermann and Roach, 2012).

Data on career paths in the life sciences have become increasingly available in recent years (Blank et al., 2017; Council for Doctoral Education, 2020), and such data are useful to individuals as they plan their careers, and also to funding agencies and institutions as they plan for the future. In this article we report the results of a time-resolved analysis of the career paths of 2284 researchers who completed a PhD or postdoc at the European Molecular Biology Laboratory (EMBL) between 1997 and 2020. This period included major global events, such as financial crisis of 2007 and 2008 (Izsak et al., 2013; Pellens et al., 2018), and also major events within the life sciences (such as the budget of the US National Institutes of Health doubling between 1998 and 2003 and then plateauing; Wadman, 2012; Zerhouni, 2006).

Results

EMBL is an intergovernmental organisation with six sites in Europe, and its missions include scientific training, basic research in the life sciences, and the development and provision of a range of scientific services. The organization currently employs more than 1110 scientists, including over 200 PhD students, 240 postdoctoral fellows, and 80 PIs. EMBL has a long history of training PhD students and postdocs, and the EMBL International PhD Programme – one of the first structured PhD programmes in Europe – has a completion rate of 92%, with students taking an average of 3.95 years to submit their thesis (data for 2015–2019). More recently, EMBL has launched dedicated fellowship programmes with structured training curricula for postdocs.

Data collection for this study was initially carried out in 2017 and updated in 2021. Using manual Google searches, we located publicly available information identifying the current role of 89% (2035/2284) of the sample (Table 1). These alumni were predominantly based in the European Union (60%, 1224/2035), other European countries including UK and Switzerland (20%), and the US (11%). For 71% of alumni (1626/2284), we were able to reconstruct a detailed career path based on online CVs and biographies (see Methods). EMBL alumni also ended up in a range of careers, which were classified as follows: Academia: Principal Investigator; Academia: Postdoc; Academia: Other research/teaching/service role; Industry Research; Science-related Non-research; and Non-science-related. We also collected data on different types of jobs within the last three of these career areas (Table 2).

Table 1. Career outcomes for 2284 EMBL alumni.

Career PhD alumni Postdoc Total
Academia: PI (AcPI) 215 (22.2%) 421 (32%) 636 (27.8%)
Academia: Other (AcOt) 102 (10.5%) 281 (21.4%) 383 (16.8%)
Academia: Postdoc (AcPD) 168 (17.3%) 76 (5.8%) 244 (10.7%)
Industry research (IndR) 153 (15.8%) 179 (13.6%) 332 (14.5%)
Science-related Non-research (SciR) 178 (18.4%) 171 (13%) 349 (15.3%)
Non-science-related (NonSci) 47 (4.9%) 44 (3.3%) 91 (4%)
Unknown 106 (10.9%) 143 (10.9%) 249 (10.9%)
TOTAL 969 1315 2284
Open in a new tab

See Table 2 for more information on the different jobs covered by Industry Research, Science-related Non-research, and Non-science-related. This classification is based on Stayart et al., 2020.

AcPI: includes those leading an academic research team with financial and scientific independence – evidenced by a job title such as group leader, professor, associate professor or tenure-track assistant professor. Where the status was unclear from the job title, we classified an alumnus as a Principal Investigator (PI) if one of the following criteria was fulfilled: (a) they appear to directly supervise students/postdocs (based on hierarchy shown on website); (b) they have published a last author publication from their current position; (c) their group website or CV indicates that they have a grant (not just a personal merit fellowship) as a principal investigator. AcOt: differs from Stayart et al., 2020 in that it includes academic research, scientific services or teaching staff (e.g., research staff, teaching faculty and staff, technical directors, research infrastructure engineers).

Table 2. Classification of Industry Research, Science-related Non-research and Non-science-related positions.

Job function PhD alumni Postdoc Total
Industry research
R & D scientist* 138 (14.2%) 167 (12.7%) 305 (13.4%)
Entrepreneurship† 6 (0.6%) 8 (0.6%) 14 (0.6%)
Postdoctoral 7 (0.7%) 1 (0.1%) 8 (0.4%)
Business development, consulting & strategic alliances ‡ 2 (0.2%) 3 (0.2%) 5 (0.2%)
Total 153 (15.8%) 179 (13.6%) 332 (14.5%)
Science-related Non-research
Administration and training 35 (3.6%) 35 (2.7%) 70 (3.1%)
Business development, consulting & strategic alliances 38 (3.9%) 20 (1.5%) 58 (2.5%)
Tech support and product development 20 (2.1%) 24 (1.8%) 44 (1.9%)
Science writing and communication 16 (1.7%0) 21 (1.6%) 37 (1.6%)
Data science, analytics, software engineering § 15 (1.5%) 13 (1%) 28 (1.2%)
Intellectual property and law 16 (1.7%) 10 (0.8%) 26 (1.1%)
Science education and outreach 10 (1%) 11 (0.8%) 21 (0.9%)
Clinical research management 8 (0.8%) 4 (0.3%) 12 (0.5%)
Regulatory affairs 5 (0.5%) 7 (0.5%) 12 (0.5%)
Clinical services/public health 4 (0.4%) 6 (0.5%) 10 (0.4%0)
Sales and Marketing 4 (0.4%) 6 (0.5%) 10 (0.4%)
Healthcare provider 1 (0.1%) 8 (0.6%) 9 (0.4%)
Other 4 (0.4%) 2 (0.2%) 6 (0.3%)
Entrepreneurship 2 (0.2%) 2 (0.2%) 4 (0.2%)
Science policy and government affairs 0 (0%) 2 (0.2%0) 2 (0.1%)
Total 178 (18.4%) 171 (13%) 349 (15.3%)
Non-science-related
Data science, analytics, software engineering § 18 (1.9%) 25 (1.9%) 43 (1.9%)
Business development, consulting & strategic alliances 16 (1.7%) 4 (0.3%) 20 (0.9%)
Other (inc retired) 7 (0.7%) 12 (0.9%) 19 (0.8%)
Entrepreneurship 5 (0.5%) 2 (0.2%) 7 (0.3%0)
Administration and training 1 (0.1%) 1 (0.1%) 2 (0.1%)
Total 47 (4.9%) 44 (3.3%) 91 (4%)
Open in a new tab
*

This function differs from the schema in Stayart et al., 2020; it includes alumni carrying out or overseeing scientific research in industry as group leaders, research staff, technical directors and non-directorship research leadership roles, including alumni who appear to be working in computational biology roles of a pharma, biotech, contract research or similar company regardless of job title (i.e. including data science roles that appear to be related to analysis of research-related data.).

†

Founders of companies whose primary focus is R&D (including contract research organizations).

‡

Includes director-level senior management roles overseeing the scientific direction & research of a company with R&D focus, e.g. CSOs in biotech start-ups.

§

Not including computational biology roles linked to R&D functions.

On average, the alumni in our sample published an average of 4.5 research articles about their work at EMBL, and were the first author on an average of 1.6 of those articles (Table S1 in Supplementary file 1). Overall, 90% of the sample (2047/2284) authored at least one article about their EMBL work, and 73% (1666/2284) were the first author on at least one article. The average time between being awarded a PhD and taking up a first role in a specific career area ranged from 4.2 years for Non-science-related positions to 6.8 years for a Principal Investigator (PI) position.

Most alumni remain in science

The majority of alumni (1263/2284=55.3%) were found to be working in an academic position in 2021, including 636 who were PIs, 244 who were in Academia: Postdoc positions, and 383 who were working in Academia: Other positions, which included teaching, research and working for a core facility/technology platform (Figure 1A). Just under one-sixth (332/2284=14.5%) were employed in Industry Research positons, and a similar proportion (349/2284=15.3%) were employed in Science-related Non-research positions, such as technology transfer, science administration and education, and corporate roles at life sciences companies. Around 4% were employed in professions not related to science, and the current careers of around 11% of alumni were unknown.

Figure 1. Career outcomes for EMBL alumni.

(A) Charts showing the percentage of PhD alumni (n=969) and postdoc alumni (n=1315) from EMBL in different careers in 2021 (see Table 1). (B) Charts showing percentage of PhD (left, n=800) and postdoc (right, n=1053) alumni in different careers five years after finishing their PhD or postdoc, for three different cohorts. Chart excludes 169 PhD students and 262 postdocs who have not yet reached the five-year time point. (C) Charts showing the percentage of PhD alumni from EMBL (blue column) in PI positions with the percentage of PhD alumni from Stanford University (grey column) in research-focused faculty positions (Stanford Biosciences, 2021). Detailed information about the comparison group can be found in Table S3 in Supplementary file 1.

Figure 1—source data 1. Summary data plotted in Figure 1A, B.

Figure 1.

Figure 1—figure supplement 1. Sankey diagrams showing movement between different careers.

Figure 1—figure supplement 1.

(A) Sankey diagram showing that of the 539 alumni who have held an Academia: PI (AcPI) position at some time, 75.3% moved into their first AcPI position from an Academia: Postdoc (AcPD) position, and 20.6% moved from an Academia: Other (AcOt) positon; 93.1% of these alumni are still in an AcPI position. Abbreviations and percentages for a position are only shown for values of 10% or higher. (B) Similar Sankey diagram for the 477 alumni who have held an AcOt position at some time. (C) Similar Sankey diagram for the 415 alumni who have held an Industry Research (IndR) position at some time. (D) Similar Sankey diagram for the 364 alumni who have held a Science-related Non-research (NonRes) position at some time. (E) Similar Sankey diagram for the 131 alumni who have held a Non-science-related (NonSci) position at some time. Data are shown only for alumni for whom a detailed career path is available (n=1626). A preceding AcPD position includes entry direct from an EMBL postdoc and entry via a postdoc position held after leaving EMBL. If an EMBL PhD student became a bridging postdoc in the same lab, this is included in the Academia: PhD category. Diagrams were created in R and scaled manually so that the height is proportional to the number of alumni in the role.
Figure 1—figure supplement 1—source data 1. Summary data plotted in the simplified Sankey plots in Figure 1—figure supplement 1.
Figure 1—figure supplement 2. Career outcomes for EMBL alumni at five different time points for three different cohorts.

Figure 1—figure supplement 2.

Charts showing the percentage of PhD and postdoc alumni in different roles at five different time points for three different cohorts. See Table 3 for cohort sizes; alumni who have not yet reached a given time point are not included. (A) PhD alumni 1 year after EMBL (n=969). (B) Postdoc alumni 1 year after EMBL (n=1315). (C) PhD alumni 5 years after EMBL (n=800). (D) Postdoc alumni 5 years after EMBL (n=1053). (E) PhD alumni 9 years after EMBL (n=597). (F) Postdoc alumni 9 years after EMBL (n=791). (G) PhD alumni 13 years after EMBL (n=419). (H) Postdoc alumni 13 years after EMBL (n=578). (I) PhD alumni 17 years after EMBL (n=256). (J) Postdoc alumni 17 years after EMBL (n=369).
Figure 1—figure supplement 2—source data 1. Summary data plotted in Figure 1—figure supplement 2.
Figure 1—figure supplement 3. Comparing EMBL alumni with alumni from institutions in Canada and the United States.

Figure 1—figure supplement 3.

Charts comparing percentage of PhD alumni in PI or PI-like positions for EMBL and Stanford University, the University of California San Francisco (UCSF), the University of Chicago (Bioscience Division), the University of Michigan, and the University of Toronto (Life Sciences Division) for different cohorts at various time points. (A) Chart comparing PhD alumni from EMBL in PI positions (blue) and Stanford in tenure-track faculty positions (grey) for two cohorts; the Stanford data were collated in 2013 (Stanford IT&DS, 2020). (B) Chart comparing PhD alumni from EMBL in PI positions (blue) and PhD alumni from UCSF in tenure-track faculty positions (grey) a 2002–2006 cohort after 10 years and 5 years, and a 2007–2011 cohort after 5 years (UCSF Graduate Division, 2021) (C) Chart comparing PhD alumni from EMBL in PI positions (blue) and PhD alumni from Chicago in tenure-track faculty positions (grey) for two cohorts (University of Chicago, 2021). (D) Chart comparing PhD alumni from EMBL in PI positions (blue) and PhD alumni from Michigan in tenure-track faculty positions (grey) for two cohorts. (University of Michigan, 2018) (E) Chart comparing PhD alumni from EMBL in PI positions (blue) and PhD alumni from Toronto in tenure-track faculty positions (grey) for three cohorts; the Torotono data were collated in 2016 (University of Toronto, no date). Detailed information about the comparison groups, including cohort sizes, can be found in Table S3 in Supplementary file 1; percentages are the number alumni known to be in a PI or PI-like position at the relevant time point as a percentage of all PhD students in that cohort (including students whose position was unknown at that time point).
Open in a new tab

Of those who became PIs, 75.3% moved from a postdoc to their first PI position, with 20.6% moving from an Academia: Other position (Figure 1—figure supplement 1A). On average, PhD alumni became PIs 6.1 calendar years after their PhD defence, and postdoc alumni became PIs 2.5 years after completing their EMBL postdoc. Almost half of the postdoc alumni who became PIs did so directly after completing their EMBL postdoc (168 of 343). Other postdoc alumni made the transition later, most frequently after one additional postdoc (71 alumni) or a single Academia: Other position (56 alumni). 40 alumni held multiple academic positions between their EMBL postdoc and their first PI position, and eight had one or more non-academic positions during this period.

The career paths of those in other positions were more varied (Figure 1—figure supplement 1B–E). For example, for alumni who moved into Industry Research, 20.2% entered their first industry role directly from their PhD, 56.4% from a postdoc position, and 13.3% from Academia: Other positions. Moreover, 71.6% remained in this type of role long-term.

The wide variation in job titles used outside academia makes it difficult to assess career progression, but almost 60% (453/766) of alumni working outside academia had a current job title that included a term indicative of a management-level role (such as manager, leader, senior, head, principal, director, president or chief). For leavers from the last five years (2016–2020), this number was 45% (78/174), suggesting that a large proportion of the alumni who leave academia enter – or are quickly promoted to – managerial positions.

For further analysis, EMBL alumni were split into three 8 year cohorts. More recent cohorts were larger, reflecting the growth of the organization between 1997 and 2020, and also contained a higher percentage of female researchers (Table 3). When comparing cohorts, we observed some differences in the specific jobs being done by alumni outside academia 2021 (Table S2 in Supplementary file 1). For example, the percentage of alumni involved in ‘data science, analytics, software engineering’ roles increased from 2% (11/625) for the 1997–2004 cohort to 4% (37/896) for the 2013–2020 cohort. However, the absolute number of alumni for most jobs outside academia was small, so our time-resolved analysis therefore focussed on the broader career areas described above.

Table 3. Overview of PhD and postdoc cohorts.

Completion years PhD cohorts Postdoc cohorts All
1997–2004 2005–2012 2013–2020 1997–2004 2005–2012 2013–2020 All
n = 256 341 372 369 422 524 2284
n (%) known current role 225 (88%) 306 (90%) 332 (89%) 336 (88%) 364 (86%) 472 (90%) 2035 (89%)
n (%) detailed career path 220 (70%) 258 (79%) 413 (77%) 179 (60%) 271 (61%) 285 (79%) 1626 (71%)
n (%) female 85 (33%) 157 (46%) 173 (47%) 136 (37%) 149 (35%) 207 (40%) 907 (40%)
Open in a new tab

Percentage of EMBL alumni who become PIs is similar to that for other institutions

For all timepoints, the percentages of alumni from the 2005–2012 and 2013–2020 cohorts working in PI positions in 2021 were lower than the percentage for the 1997–2004 cohort (Figure 1—figure supplement 2). To assess whether this pattern was specific to EMBL, we compared our data with data from other institutions, noting that different institutions can use different methods to collect data and classify career outcomes. We also note that career outcomes are influenced by the broader scientific ecosystem and by the subject focus of institutions and departments, which may attract early-career researchers with dissimilar career motivations. Nevertheless, comparing long-term outcomes with other institutions allows us to interrogate whether the changes we observe for the most frequent, well-defined and linear career path – the PhD–>Postdoc–>PI career path – reflect a general trend.

A number of institutions have released data on career outcomes for PhD students. Stanford University, for example, has published data on the careers of researchers who received a PhD between 2000 and 2019 (Stanford Biosciences, 2021): Stanford has reported that 34% (145/426) of its 2000–2005 PhD alumni were in research-focussed faculty roles in 2018, and that 13% (63/503) of its 2011–2015 PhD alumni were in PI roles; these numbers are comparable to the figures of 37% (78/210) and 11% (25/234) we observe for EMBL alumni for the same time periods (Figure 1C). The EMBL data are also comparable to data from the life science division at the University of Toronto (Reithmeier et al., 2019; University Toronto, 2021): for example, Toronto has reported that 31% (192/629) of its 2000–2003 graduates and 25% (203/816) of its 2004–2007 graduates were in tenure stream roles in 2016; the corresponding figures for EMBL were 39% (52/132) and 28% (49/172).

We also compared our EMBL data with data from the University of Michigan, the University of California at San Francisco, and the University of Chicago, and found similar proportions of alumni entering PI positions for comparable cohorts (Figure 1—figure supplement 3). This is consistent with our hypothesis that the differences between cohorts are not EMBL-specific, and reflect a wide-spread change in the number of PhDs and postdocs relative to available PI positions.

We did not analyse the data for other career outcomes, as the smaller numbers of individuals in these careers made it difficult to identify real trends. Moreover, only a small number of institutions have released detailed data on the career destinations of recent postdoc alumni, and we are not aware of any long-term cohort-based data.

The proportion of EMBL alumni who become PIs has decreased with time

To estimate the probability of alumni from different cohorts entering a specific career each year after completing a PhD or postdoc at EMBL, we fitted the data to a Cox proportional hazards model. This is a statistical regression method that is commonly used to model time-to-event distributions from observational data with censoring (i.e., when not all study subjects are monitored until the event occurs, or the event never occurs for some of the subjects). In brief, we fitted the data to a univariate Cox proportional hazards model to calculate hazard ratios, which represent the relative chance of the event considered (here: entering a specific career) occurring in each cohort with respect to the oldest cohort. We also calculated Kaplan–Meier estimators, which estimate the probability of the event (entering a specific career) at different timepoints.

For both PhD and postdoc alumni entering PI positions, we observe hazard ratios of less than one in the Cox models when comparing the newer cohorts with the oldest cohort (Table S4 in Supplementary file 1), which indicates that the chances of becoming a PI have become lower for the newer cohorts. The Kaplan–Meier plots illustrate lower percentages of PIs among alumni from the most recent cohorts compared to the oldest cohort at equivalent timepoints (Figure 2A). Nevertheless, becoming a PI remained the most common career path for alumni from the 2005–2012 cohort (90/341=26.4% for PhD alumni) and (123/422=29.1% for postdoc alumni), and the most recent cohort of alumni appear to be on a similar trajectory.

Figure 2. Changes in career outcomes for more recent cohorts.

(A) Kaplan–Meier plots showing the estimated probability of an individual being in a PI position (y-axis) as a function of time after EMBL (x-axis) for three cohorts of PhD alumni (left) and three cohorts of postdoc alumni (right). Time after EMBL refers to the number of calendar years between PhD defence or leaving the EMBL postdoc programme and first PI position. (B–E) Similar Kaplan–Meier plots for Academia: Other positions (B), Industry Research positions (C), Science-related Non-research positions (D), and Non-science-related professions (E). Hazard ratios calculated by a Cox regression model can be found in Table S4 in Supplementary file 1.

Figure 2.

Figure 2—figure supplement 1. Length of time taken to become a PI.

Figure 2—figure supplement 1.

(A) Box plot with overlaid dot plot showing the distribution of the length of time between PhD and first PI role for two cohorts of alumni who defended their PhD between 1997 and 2012 and became a PI within nine calendar years (and for whom we have a detailed career path; n=157). The mean value is indicated as a red cross and the p-values were calculated using Welch’s t-test. The difference between the mean for the two cohorts (5.2 years and 6.1 years) was statistically significant (P=0.01496). (B) Plots for length of time between completion of an EMBL postdoc and first PI role for two cohorts of alumni who completed their postdoc between 1997 and 2012 (n=218). The difference between the two cohorts was not statistically significant (P=0.192). (C) Plots for length of time between PhD and first PI role for the two cohorts of alumni who completed their EMBL postdoc between 1997 and 2012 (and for whom we know year of PhD and have a detailed career path; n=146; note these alumni completed their PhD somewhere else before starting a postdoc at EMBL). The difference between the mean for the two cohorts (5.3 years and 6.0 years) was statistically significant (P=0.0325).
Figure 2—figure supplement 1—source data 1. Summary data plotted in Figure 2—figure supplement 1.
Open in a new tab

Kaplan–Meier plots show increased proportions of the 2005–2012 and 2013–2020 cohorts entering Science-related Non-research and Non-science-related positions, compared to the 1997–2004 cohort for both PhD and postdoc alumni (Figure 2D, E). For the most recent (2013–2020) cohort, there was also an increased rate of entry into Industry: Research positions compared to alumni from PhD and postdoc cohorts from 1997 to 2004 and 2005–2012 (Figure 2C, Table S4 in Supplementary file 1). For Academia: other positions, the rate of entry was similar for all three PhD cohorts, though some differences between cohorts were observed for postdoc alumni (Figure 2B).

A small increase in time between year of PhD and first PI position

We decided to explore to what extent increasing postdoc length may contribute to the decreased proportion of alumni who are found as PIs in the years after leaving EMBL. In order to fairly compare alumni from different cohorts, we included only alumni for whom we had a detailed career path, who had defended their PhD at least nine years ago, and who had become a PI within nine years of defending their PhD. We chose a nine-year cut-off because this was the time interval between the last PhDs in the 2005–2012 cohort and the execution of this study; moreover, for PhD alumni from the oldest cohort (1997–2004), most of those who became PIs had done so within nine years (89/97=92%).

157 of the PhD alumni in our sample met these criteria, taking an average of 5.6 calendar years to become a PI (see Methods). There was a statistically significant difference in the average time from PhD to first PI position between the 1997–2004 cohort (5.2 years) and the 2005–2012 cohort (6.1 years; Figure 2—figure supplement 1A). 218 of the postdoc alumni in our sample met these criteria, taking an average of 2.5 calendar years to become a PI after leaving EMBL (see Methods). There was no statistically significant difference in time between EMBL and first PI role for the 1997–2004 and 2005–2012 postdoc cohorts (Figure 2—figure supplement 1B). However, the time between receiving their PhD and becoming a PI increased by from 5.3 calendar years for the 1997–2004 postdoc cohort to 6.0 calendar years for the 2005–2012 postdoc cohort (Figure 2—figure supplement 1C).

Gender differences in career outcomes

Many studies have reported that female early-career researchers are less likely to remain in academia (Alper, 1993; Martinez et al., 2007). Consistent with these studies, male alumni from EMBL were more likely than female alumni to end up in a PI position (Figure 3A and B; Table S5 in Supplementary file 1). However, for alumni from 1997 to 2012, there was no statistically significant difference in the length of time taken by male and female alumni to become PIs (Figure 3—figure supplement 1). Female alumni were more likely to end up in a Science-related Non-research position, and male alumni were more likely to end up in an Industry Research or Non-science-related position (Figure 3; Figure 3—figure supplement 2). However, female alumni were also more likely to be classified as unknown, and since it is more difficult to follow careers outside the academic world, it is possible that the number of women who established careers outside academia (in positions such as Industry Research, Science-related Non-research, and Non-science-related) is higher than our results suggest.

Figure 3. Gender differences in career outcomes.

(A) Charts showing the percentage of female (n=415) and male (n=554) PhD alumni, and female (n=492) and male (n=823) postdoc alumni, in different careers in 2021. (B) Kaplan–Meier plots showing the estimated probability of an individual being in a PI position (y-axis) as a function of time after EMBL (x-axis), stratified by gender for PhD alumni (left) and postdoc alumni (right). (C) Kaplan–Meier plots showing the estimated probability of an individual being in a science-related non-research position as a function of time after EMBL, stratified by gender for PhD alumni (left) and postdoc alumni (right). Kaplan–Meier plots for other career outcomes are shown in Figure 3—figure supplement 2. Hazard ratios calculated by a Cox regression model can be found in Table S5 in Supplementary file 1.

Figure 3—source data 1. Summary data plotted in Figure 3A.

Figure 3.

Figure 3—figure supplement 1. Length of time taken to become a PI for female and male alumni.

Figure 3—figure supplement 1.

(A) Box plot with overlaid dot plot showing the distribution of the length of time between PhD and first PI role for female alumni (left) and male alumni (right) who defended their PhD between 1997 and 2012 and became a PI within nine calendar years (and for whom we have a detailed career path; n=157). The mean value is indicated as a red cross and the p-values calculated using Welch’s t-test. The difference between the mean values for female and male alumni (6.1 years and 5.4 years) was not statistically significant (P=0.0719). (B) Plots for length of time between completion of an EMBL postdoc and first PI role for female alumni (left) and male alumni (right) who completed their postdoc between 1997 and 2012 (n=218). The difference between the mean values for female and male alumni (2.3 years and 3.1 years) was not statistically significant (P=0.0596). (C) Plots for length of time between PhD and first PI role for female alumni (left) and male alumni (right) who completed their EMBL postdoc between 1997 and 2012 and for whom we know year of PhD (and for whom we have a detailed career path; n=146). The difference between the mean values for female and male alumni (5.6 years and 5.7 years) was not statistically significant (P=0.778).
Figure 3—figure supplement 1—source data 1. Summary data plotted in Figure 3—figure supplement 1.
Figure 3—figure supplement 2. Rate of entry into different types of role by gender.

Figure 3—figure supplement 2.

(A) Kaplan–Meier plots showing the estimated probability of an individual working in Academia: Other (y-axis) as a function of time after EMBL, stratified by gender for PhD alumni (left) and postdoc alumni (right). (B, C) Similar Kaplan–Meier plots for alumni working in Industry Research (B) and in Non-science-related careers (C). Hazard ratios calculated by a Cox regression model can be found in Table S5 in Supplementary file 1.
Open in a new tab

Future PIs, on average, published more papers while at EMBL

Publication metrics have been linked to the likelihood of obtaining (van Dijk et al., 2014; Tregellas et al., 2018) and succeeding (von Bartheld et al., 2015) in a faculty position. In this study, alumni who became PIs had more favourable publication metrics from their EMBL work – for example, they published more articles, and their papers had higher CNCI values. (CNCI is short for Category Normalized Citation Impact, and a CNCI value of one means that the number of citations received was the same as the average for other articles in that field published in the same year; Figure 4A and B; Table S6 in Supplementary file 1). Using univariate Cox models for time to PI as a function of number of first-author research articles from EMBL work, we estimated that a postdoc with one first-author publication was 3.2 times more likely to be found in a PI position than a postdoc without a first-author publication (95% confidence interval [2.2, 4.7]), and a post-doc with two or more first-author publications was 6.6 times more likely (95% confidence interval [4.7, 9.3]; Figure 4C).

Figure 4. Publication factors are highly correlated with becoming a PI.

(A) Histograms showing the number of alumni who have 0, 1, 2, 3,... first-author articles from their time at EMBL and became PIs (bottom; n=662, excluding 23 outliers), and did not become PIs (top; n=1594, excluding 5 outliers). For clearer visualization, and to protect the identity of alumni with outlying numbers of publications, the x-axis is truncated at the 97.5th percentile. The mean for each group (including outliers) is shown as a red dashed line; alumni who became PIs have an average of 2.4 first-author articles from their time at EMBL, whereas other alumni have an average of 1.2 articles; this difference is significant (P<2.2 × 10–16; Welch’s t-test). (B) 1656 alumni had one or more first-author articles from their time at EMBL that had a CNCI value in the InCites database. For each of these alumni, the natural logarithm of the highest CNCI value was calculated, and these histograms show the number of alumni for which this natural logarithm is between –4.5 and –3.5, between –3.5 and –2.5, and so on; the bottom histogram is for alumni who became PIs, and the top histogram is for other alumni. A CNCI value of 1 (plotted here at ln(1)=0; vertical black line) means that the number of citations received by the article is the same as the average for other articles in that field published in the same year. The mean for each group is shown as a red dashed line; alumni who became PIs have an average CNCI of 5.7, whereas other alumni have an average CNCI of 3.1; this difference is significant (P<2.829 × 10–6; Welch’s t-test). (C) Kaplan–Meier plots showing the estimated probability of an individual becoming a PI (y-axis) as a function of time after EMBL (x-axis), stratified by number of first-author publications from research completed at EMBL, for PhD alumni (left) and postdoc alumni (right). Hazard ratios calculated by a Cox regression model can be found in Table S7 in Supplementary file 1. (D) Harrell’s C-Index for various Cox models for predicting entry into PI positions. The first seven bars show the C-index for univariate and multivariate models for a subset of covariates (which subset is shown below the x-axis), and the eighth bar is for a multivariate model that includes the covariates from all subsets. The subsets are time & cohort (multivariate, including the variables: cohort, PhD year (if known), start and end year at EMBL), predoc (ie PhD student)/postdoc (univariate), group leader seniority (univariate), nationality (univariate), gender (univariate), publications (multivariate: containing variables related to the alumni’s publications from their EMBL work; these are variables with a name beginning with “pubs” in Table S1 in Supplementary file 1) and group publications (multivariate: containing variables related to the aggregated publication statistics for all PhD students and postdocs who were trained in the same group; these are variables with a name beginning with “group_pubs” in Table S1 in Supplementary file 1). A value of above 0.5 indicates that a model has predictive power, with a value of 1.0 indicating complete concordance between predicted and observed order to outcome (e.g. entry into a PI position). Bars denote the mean, and the error bars show the 95% confidence intervals. A value of above 0.5 indicates that a model has predictive power, with a value of 1.0 indicating complete concordance between predicted and observed order to outcome (e.g. entry into a PI position). Bars denote the mean, and the error bars show the 95% confidence intervals.

Figure 4—source data 1. Summary data plotted in Figure 4A, B and D.

Figure 4.

Figure 4—figure supplement 1. Cox models for predicting entry into various careers.

Figure 4—figure supplement 1.

Harrell’s C-Index for various Cox models for predicting entry into Academia: Other (A), Industry Research (B), Science-related Non-research (C), and Non-science-related careers (D). As in Figure 4D, the first seven bars show the C-index for univariate and multivariate models containing subsets of variables, the eighth bar is for a multivariate model containing all variables, and a value of above 0.5 indicates that a model has predictive power.
Figure 4—figure supplement 1—source data 1. Summary data plotted in Figure 4—figure supplement 1.
Figure 4—figure supplement 2. Entry into various careers and number of first-author publications.

Figure 4—figure supplement 2.

Kaplan–Meier plots showing the estimated probability of an individual being in various careers (y-axis) as a function of time after EMBL (x-axis), stratified by number of first-author publications from research completed at EMBL for PhD alumni (left) and postdoc alumni (right). Hazard ratios calculated by a Cox regression model can be found in Table S7 in Supplementary file 1.
Open in a new tab

Publication factors are highly predictive of entry into a PI position

To understand the potential contribution of publication record in the context of other factors – including cohort, gender, nationality, publications, and seniority of the supervising PI – we fitted multivariate Cox models. To quantify publication record, we considered a range of metrics including journal impact factor, which has been shown to statistically correlate with becoming a PI in some studies (van Dijk et al., 2014) and has been used by some institutions in research evaluation (McKiernan et al., 2019). It should be stressed, however, that EMBL does not use journal impact factors in hiring or evaluation decisions, and is a signatory of the San Francisco Declaration on Research Assessment (DORA) and a member of the Coalition for Advancing Research Assessment (CoARA).

To evaluate the predictive power of each Cox model, we used the cross-validated Harrell’s C-index, which measures predictive power as the average agreement across all pairs of individuals between observed and predicted temporal order of the outcome (in our case, entering a specific type of position; see Methods). A C-index of 1 indicates complete concordance between observed and predicted order. For example, for a model of entry into PI roles, a C-index of 1 would mean that the model correctly predicts, for all pairs of individuals, which individual becomes a PI first based on the factors included in the model. A C-index 0.5 is the baseline that corresponds to random guessing. Prediction is clearly limited by the fact that we could not explicitly encode some covariates that are certain to play an important role in career outcomes, such as career preferences and relevant skills. Nevertheless, the C-index for models containing all data were between 0.61 (entry to Industry Research, Figure 4—figure supplement 1B) and 0.70 (entry into PI positions, Figure 4D), suggesting that the factors have some predictive power.

To investigate which factors were most predictive for entry into different careers, we compared models containing different sets of factors. Consistent with previous studies, we found that statistics related to publications were highly predictive for entry into a PI position: a multivariate model containing only the publication statistics performs almost as well as the complete multivariate model, reaching a C-index of 0.69 (Figure 4D). The publications of the research group the alumnus was trained in (judged by the aggregated publication statistics for all PhD students and postdocs who were trained in the same group) was also predictive, with a C-index of 0.61.

Cohort/year, gender, and status at EMBL (PhD or postdoc) were also predictors of entry into a PI position in our Cox models, with C-indexes of 0.59, 0.57 and 0.55, respectively (Figure 4D). This is consistent with our observation that alumni from earlier cohorts/years (Figure 1B), male alumni (Figure 3A) and postdoc alumni (Figure 1A) were more frequently found in PI positions. Models containing only nationality or group leader seniority were not predictive.

For Academia: Other positions, the factors that were most predictive were those related to publications of the research group the alumnus was trained in (Figure 4—figure supplement 1A). It is unclear why this might be, but we speculate that this could reflect publication characteristics specific to certain fields that have a high number of staff positions, or other factors such as the scientific reputation, breadth or collaborative nature of the research group and its supervisor. The group’s publications were also predictive for Industry Research and Science-related Non-research positions.

Time-related factors (i.e., cohort, PhD award year and EMBL contract start/end years) were the strongest prediction factors for Industry Research, Science-related Non-research, and Non-science-related positions (Figure 4—figure supplement 1B–D), and more recent alumni were more frequently found in these careers (Figure 2C–E).

Overall, statistics related to an individual’s own publications were a weak predictor for entry into positions other than being a PI (Figure 4—figure supplement 1; Figure 4—figure supplement 2; Table S7–S11 in Supplementary file 1). For example, for Industry Research, a model containing statistics for an individual’s publications had a C-index of only 0.53, compared to 0.61 for the complete model, and there were no differences in likelihood of a PhD alumnus with 0, 1 or 2+publications entering an Industry Research position.

Changes in the publications landscape

Reports suggest that the number of authors on a typical research article in biology has increased over time, as has the amount of data in a typical article (Vale, 2015; Fanelli and Larivière, 2016); a corresponding decrease in the number of first-author research articles per early-career researcher has also been reported (Kendal et al., 2022). For articles linked to the PhD students and postdocs in this study, the mean number of authors per article has more than doubled between 1995 and 2020 (Figure 5A). The mean number of articles per researcher did not change between the three cohorts studied (Figure 5B; the mean was 3.6 articles per researcher), but researchers from the second and third cohorts published fewer first-author articles than those from the first cohorts (Figure 5C). However, more recent articles had higher CNCI values (Figure 5D). The proportion of EMBL articles that included international collaborators also increased from 47% in 1995 to 79% in 2020.

Figure 5. Publications are increasingly collaborative.

Figure 5.

Open in a new tab

(A) Mean number of authors (y-axis) as a function of year (x-axis) for research articles that were published between 1995 and 2020, and have at least one of the alumni included in this study as an author (n=5413); the winsorized mean has been used to limit the effect of outliers. The mean number of authors has increased by a factor of more than two between 1995 and 2000. (B) Boxplot showing the distribution of the number of articles published per researcher for three cohorts. The mean is indicated as a red cross; the circles are outliers. No statistically significant difference was found between the cohorts; the p-value of 0.1156 was generated using a one-way analysis of variance (ANOVA) test of the full dataset (including outliers); the p-value excluding outliers is 0.26. (C) Boxplot showing the distribution of the number of first-author articles published per researcher for three cohorts; the two most recent cohorts published fewer first-author articles than the 1997–2004 cohort; the p-value (excluding outliers) was 6.5x10–7; see Table S12 in Supplementary file 1. (D) Mean CNCI (y-axis) as a function of time (x-axis) for research articles that were published between 1995 and 2020, and have at least one of the alumni included in this study as an author (n=5413). Recent articles have higher CNCI values. For clearer visualization, and to protect the identity of alumni with outlying numbers of publications, the y-axis in (B) and (C) is truncated at the 97.5 percentile.

Figure 5—source data 1. Summary data plotted in Figure 5.

Discussion

Many early-career researchers are employed on fixed-term contracts funded by project-based grants, sometimes for a decade or more (OECD, 2021; Acton et al., 2019), and surveys suggest that early-career researchers are concerned about career progression (Woolston, 2020; Woolston, 2019). We hope PhD students and postdocs will be reassured to learn that the skills and knowledge they acquire during their training are useful in a range of careers both inside and outside acaemica.

Further changes to the career landscape in the life sciences are likely in future, not least as a result of the long-term impacts of the COVID-19 pandemic (Bodin, 2020). It is essential, therefore, that early-career researchers are provided with opportunities to reflect on their strengths, to understand the wide range of career options available to them, and to develop new skills.

The provision of effective support for PhD students and postdocs will require input from different stakeholders – including funders, employers, supervisors and policy makers – and the engagement of the early-career researchers themselves. At EMBL, a career service was launched in 2019 for all PhD students and postdocs, building on a successful EC-funded pilot project that offered career support to 76 postdocs in the EMBL Interdisciplinary Postdoc Programme. The EMBL Fellows’ Career Service now offers career webinars and a blog to the whole scientific community as well as additional tailored support for EMBL PhDs and postdocs including individual career guidance, workshops, resources and events. Funders and policymakers may also need to reassess the sustainably of academic career paths, and to review how funding is allocated between project-based grants and mechanisms that can support PI and non-PI positions with longer-term stability. These measures will will also support equality, diversity and inclusion in science, particularly if paired with research assessment practices that consider factors that can affect apparent research productivity such as career breaks, teaching and service activities.

Factors related to publication are highly predictive of entry into PI careers, and one challenge for an early-career researcher hoping to pursue such a career is to balance the number of articles they publish with the subjective quality of these articles. The trend towards fewer first-author articles per researcher likely reflects a global trend towards articles with more authors and a greater focus on collaborative and/or interdisciplinary approaches to research. Working on a project that involves multiple partners provides an early-career researcher with the opportunity to develop a range of skills, including teamwork, leadership and creativity. Such projects also allow researchers to tackle challenging biological questions from new angles to advance in their field of research, something viewed very positively by academic hiring committees (Hsu et al., 2021; Clement et al., 2020; Fernandes et al., 2020); however, multi-partner interdisciplinary projects can also take longer to complete. It is therefore important that early-career researchers and their supervisors discuss the potential impact and challenges of (prospective) projects, and what can be done to reduce any risks. For example, open science practices – including author credit statements, FAIR data, and pre-printing – can make project contributions more transparent and available faster (Kaiser, 2017; McNutt et al., 2018; Wilkinson et al., 2016; Wolf et al., 2021).

Limitations

The limitations of our study include that its retrospective, observational design limits our ability to disentangle causation from correlation. The changes in career outcomes may be driven primarily by increased competition for PI roles, but they could also be influenced by a greater availability or awareness of other career options. EMBL has held an annual career day highlighting career options outside academia since 2006, and many of our alumni decide to pursue a career in the private sector, attracted by perceptions of higher pay, more stable contracts, and/or better work-life balance. Likewise, early-career researchers with an interest in a specific technology might, for example, prefer to work at a core facility.

Additionally, we cannot exclude the possibility that other factors may also affect the differences we see between cohorts (such as variations in the number of alumni taking up academic positions in countries that offer later scientific independence). Finally, although comparisons with data from the US and Canada suggest that the trend towards fewer alumni becoming PIs is a global phenomenon, it is possible that some of the trends we observe are specific to EMBL.

Next steps

We plan to update our observational data every four years, and to maintain data on the career paths of alumni for 24 years after they leave EMBL. This will help us to identify any further changes in the career landscape and to better understand long-term career outcomes in the life sciences. Silva et al., 2019 have also described a method for tracking career outcomes on a yearly basis with estimations of the time and other resources required. We encourage institutions to consider whether they can adapt our methods, or Silva’s method, to the administrative processes and data-privacy regulations applicable at their institutions.

Future studies should also ideally include mixed-method longitudinal studies, which would allow information on career motivations, skills development, research environment, job application activity and other factors to be recorded. Combining the results of such studies with data on career outcomes would allow multifactorial and complex issues, such as gender differences in career outcomes, to be investigated, and would also provide policymakers with a fuller picture of workforce trends. Such studies would, however, require multiple institutions to commit to supplying large amounts of data every year, and coordinating the collection and analysis of such data year-on-year would be a major undertaking that would require the support of funders and institutions.

Methods

Data collection and analysis

The study includes individuals who graduated from the EMBL International PhD Programme between 1997 and 2020 (n=969), or who left the EMBL postdoc programme between 1997 and 2020 after spending at least one year as an EMBL postdoctoral fellow (n=1315). Each person is included only once in the study: where a PhD student remained at EMBL for a bridging or longer postdoc, they were included as PhD alumni only, with the postdoc position listed as a career outcome.

For each alumnus or alumna, we retrieved demographic information from our internal records and identified publicly available information about each person’s career path (see Supplementary file 2). Where possible, this information was used to reconstruct a detailed career path. An individual was classified as having a "detailed career path" if an online CV or biosketch was found that accounted for their time since EMBL excluding a maximum of two one-calendar-year career breaks (which may, for example, reflect undisclosed sabbaticals or parental leave). Each position was classified using a detailed taxonomy, based on a published schema (Stayart et al., 2020), and given a broad overall classification (see Supplementary file 2). The country of the position was also recorded. For the most recent position, we noted whether the job title was indicative of a senior or management level role (i.e., if it included "VP"; "chief"; "cso";"cto"; "ceo"; "head"; "principal”; "president"; "manager"; "leader"; "senior"), or if they appeared to be running a scientific service or core facility in academia.

We use calendar years for all outcome data – for example, for an individual who left EMBL in 2012, the position one calendar year after EMBL would be the position held in 2013. If multiple positions were held in that year, we take the most recent position. We use calendar years, as the available online information often only provides the start and end year of a position (rather than exact date).

An EMBL publication record was also reconstituted for each person in the study. Each of their publications linked to EMBL in the Web of Science and InCites databases in June 2021 were recorded. The data included publication year and – for those indexed in InCites – crude metrics, such as CNCI, percentile in subject area, and journal impact factor. EMBL publications were assigned to individuals in the study based on matching name and publication year (see Supplementary file 2 for full description). When an individual was the second author on a publication, we manually checked for declarations of co-first authorship. Aggregate publication statistics for individuals with the same primary supervisor were also calculated.

The names and other demographic information that would allow easy identification of individuals in the case of a data breach were pseudonymised. A file with key data for analysis and visualisation in R was then generated. A description of this data table can be found in Table S1 in Supplementary file 1, along with summary statistics.

Statistical model

A Cox proportional hazards regression model was fitted to the data in order to predict time-to-event probabilities for each type of career outcome based on different covariates including cohort, publication variables and gender. Multivariate Cox models were fitted using a ridge penalty with penalty parameter chosen by 10-fold cross-validation. Harrell’s C-index was calculated for each fit in an outer cross-validation scheme for validation and analysis of different models, with 10-fold cross-validation.

Acknowledgements

We thank Monika Lachner and Anne Ephrussi for their critical reading of the manuscript and strong support of this project. We also acknowledge the instrumental support of the Alumni Relations, DPO, HR, SAP, Library, International PhD Programme and Postdoc Programme teams at EMBL. We also thank Edith Heard, Brenda Stride, Jana Watson-Kapps (FMI), and the Directorate, SAC, SSMAC and Council of the EMBL for discussion. The work was supported by: EMBL (JL, BK, MR, WH, RCG) and the EMBL International PhD Programme (BV). RCG is employed by EMBL’s Interdisciplinary Postdoc Programme, which has received funding from the European Union’s Horizon 2020 programme (Marie Skłodowska-Curie Actions).

Biographies

Junyan Lu, Genome Biology Unit, European Molecular Biology Laboratory, Heidelberg, Germany

Britta Velten, Genome Biology Unit, European Molecular Biology Laboratory, Heidelberg, Germany

Bernd Klaus, Genome Biology Unit, European Molecular Biology Laboratory, Heidelberg, Germany

Mauricio Ramm, EMBL International Centre for Advanced Training, European Molecular Biology Laboratory, Heidelberg, Germany

Wolfgang Huber, Genome Biology Unit, European Molecular Biology Laboratory, Heidelberg, Germany

Rachel Coulthard-Graf, EMBL International Centre for Advanced Training, European Molecular Biology Laboratory, Heidelberg, Germany

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Contributor Information

Rachel Coulthard-Graf, Email: rachel.coulthard@embl.de.

Peter Rodgers, eLife, United Kingdom.

Peter Rodgers, eLife, United Kingdom.

Funding Information

This paper was supported by the following grants:

  • Horizon 2020 Framework Programme 664726 to Rachel Coulthard-Graf.

  • Horizon 2020 Framework Programme 847543 to Rachel Coulthard-Graf.

  • European Molecular Biology Laboratory to Britta Velten, Bernd Klaus, Mauricio Ramm, Wolfgang Huber, Rachel Coulthard-Graf, Junyan Lu.

Additional information

Competing interests

No competing interests declared.

Author contributions

Data curation, Formal analysis, Visualization, Methodology, Writing – review and editing.

Data curation, Formal analysis, Methodology, Visualization, Writing – review and editing.

Data curation, Formal analysis, Visualization, Methodology, Writing – review and editing.

Investigation, Methodology.

Supervision, Methodology, Writing – review and editing.

Conceptualization, Data curation, Formal analysis, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing.

Additional files

MDAR checklist
Supplementary file 1. Supplementary tables (Table S1–11).
elife-78706-supp1.docx (147.5KB, docx)
Supplementary file 2. Supplementary methods.
elife-78706-supp2.docx (20.4KB, docx)

Data availability

The data were collated for the provision of statistics, and are stored in a manner compliant with EMBL's internal policy on data protection. This policy means that the full dataset cannot be made publicly available (because the nature of the data means that sufficient anonymisation is not possible). Summary statistics for the main data table can be found in Supplementary file 1 (Table S1). Rmarkdown documentation of the analysis and figures can be found here and is available on GitHub (copy archived at Coulthard and Lu, 2022).

References

  1. Acton SE, Bell AJ, Toseland CP, Twelvetrees A. A survey of new PIs in the UK. eLife. 2019;8:e46827. doi: 10.7554/eLife.46827. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Alper J. The pipeline is leaking women all the way along. Science. 1993;260:409–411. doi: 10.1126/science.260.5106.409. [DOI] [PubMed] [Google Scholar]
  3. Blank R, Daniels RJ, Gilliland G, Gutmann A, Hawgood S, Hrabowski FA, Pollack ME, Price V, Reif LR, Schlissel MS. A new data effort to inform career choices in biomedicine. Science. 2017;358:1388–1389. doi: 10.1126/science.aar4638. [DOI] [PubMed] [Google Scholar]
  4. Bodin M. University redundancies, furloughs and pay cuts might loom amid the pandemic, survey finds. Nature. 2020 doi: 10.1038/d41586-020-02265-w. [DOI] [PubMed] [Google Scholar]
  5. Clement L, Dorman JB, McGee R. The Academic Career Readiness Assessment: Clarifying hiring and training expectations for future biomedical life sciences faculty. CBE Life Sciences Education. 2020;19:ar22. doi: 10.1187/cbe.19-11-0235. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Coulthard R, Lu J. EMBL-Career-Analysis. swh:1:rev:f82a94f925fb44aa59383e0afc167b4f4f0c67bdSoftware Heritage. 2022 https://archive.softwareheritage.org/swh:1:dir:8e764aae6f23b2d2a6a4ecb3322ada777e604388;origin=https://github.com/Huber-group-EMBL/EMBL-Career-Analysis;visit=swh:1:snp:66c550eefce2ed4462b612ac31ee4d219cc97f6c;anchor=swh:1:rev:f82a94f925fb44aa59383e0afc167b4f4f0c67bd
  7. Council for Doctoral Education Tracking the Careers of Doctorate Holders. 2020. [September 23, 2023]. https://eua-cde.org/component/attachments/attachments.html?id=2988
  8. Cyranoski D, Gilbert N, Ledford H, Nayar A, Yahia M. The PhD factory. Nature. 2011;472:276–279. doi: 10.1038/472276a. [DOI] [PubMed] [Google Scholar]
  9. Evans TM, Bira L, Gastelum JB, Weiss LT, Vanderford NL. Evidence for a mental health crisis in graduate education. Nature Biotechnology. 2018;36:282–284. doi: 10.1038/nbt.4089. [DOI] [PubMed] [Google Scholar]
  10. Fanelli D, Larivière V. Researchers’ individual publication rate has not increased in a century. PLOS ONE. 2016;11:e0149504. doi: 10.1371/journal.pone.0149504. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Fernandes JD, Sarabipour S, Smith CT, Niemi NM, Jadavji NM, Kozik AJ, Holehouse AS, Pejaver V, Symmons O, Bisson Filho AW, Haage A. A survey-based analysis of the academic job market. eLife. 2020;9:e54097. doi: 10.7554/eLife.54097. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Fuhrmann CN, Halme DG, O’Sullivan PS, Lindstaedt B. Improving graduate education to support a branching career pipeline: Recommendations based on a survey of doctoral students in the basic biomedical sciences. CBE Life Sciences Education. 2011;10:239–249. doi: 10.1187/cbe.11-02-0013. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Gibbs KD, McGready J, Griffin K. Career development among American biomedical postdocs. Cell Biology Education. 2015;14:ar44. doi: 10.1187/cbe.15-03-0075. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Hayter CS, Parker MA. Factors that influence the transition of university postdocs to non-academic scientific careers: An exploratory study. Research Policy. 2019;48:556–570. doi: 10.1016/j.respol.2018.09.009. [DOI] [Google Scholar]
  15. Hsu NS, Rezai-Zadeh KP, Tennekoon MS, Korn SJ. Myths and facts about getting an academic faculty position in neuroscience. Science Advances. 2021;7:eabj2604. doi: 10.1126/sciadv.abj2604. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Izsak K, Markianidou P, Lukach R, Wastyn A. Impact of the Crisis on Research and Innovation Policies. Study for the European Commission DG Research by Technopolis Group Belgium and Idea Consult. 2013. [September 23, 2023]. https://www.technopolis-group.com/wp-content/uploads/2020/02/The-impact-of-the-financial-crisis-on-research-and-innovation-policies-in-EU-Member-States.pdf
  17. Jonkers K, Zacharewicz T. Research Performance Based Funding Systems: A Comparative Assessment. JRC Publications Repository; 2016. [DOI] [Google Scholar]
  18. Kaiser J. Are preprints the future of biology? A survival guide for scientists. Science. 2017;1:aaq0747. doi: 10.1126/science.aaq0747. [DOI] [Google Scholar]
  19. Kendal D, Lee KE, Soanes K, Threlfall CG. The great publication race’ vs ‘abandon paper counting’: Benchmarking ECR publication and co-authorship rates over past 50 years to inform research evaluation. F1000Research. 2022;11:95. doi: 10.12688/f1000research.75604.1. [DOI] [Google Scholar]
  20. Lambert WM, Wells MT, Cipriano MF, Sneva JN, Morris JA, Golightly LM. Career choices of underrepresented and female postdocs in the biomedical sciences. eLife. 2020;9:e48774. doi: 10.7554/eLife.48774. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Lepori B, van den Besselaar P, Dinges M, Potì B, Reale E, Slipersæter S, Thèves J, van der Meulen B. Comparing the evolution of national research policies: What patterns of change? Science and Public Policy. 2007;34:372–388. doi: 10.3152/030234207X234578. [DOI] [Google Scholar]
  22. Levecque K, Anseel F, De Beuckelaer A, Van der Heyden J, Gisle L. Work organization and mental health problems in PhD students. Research Policy. 2017;46:868–879. doi: 10.1016/j.respol.2017.02.008. [DOI] [Google Scholar]
  23. Martinez ED, Botos J, Dohoney KM, Geiman TM, Kolla SS, Olivera A, Qiu Y, Rayasam GV, Stavreva DA, Cohen-Fix O. Falling off the academic bandwagon. Women are more likely to quit at the postdoc to principal investigator transition. EMBO Reports. 2007;8:977–981. doi: 10.1038/sj.embor.7401110. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. McKiernan EC, Schimanski LA, Muñoz Nieves C, Matthias L, Niles MT, Alperin JP. Use of the Journal Impact Factor in academic review, promotion, and tenure evaluations. eLife. 2019;8:e47338. doi: 10.7554/eLife.47338. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. McNutt MK, Bradford M, Drazen JM, Hanson B, Howard B, Jamieson KH, Kiermer V, Marcus E, Pope BK, Schekman R, Swaminathan S, Stang PJ, Verma IM. Transparency in authors’ contributions and responsibilities to promote integrity in scientific publication. PNAS. 2018;115:2557–2560. doi: 10.1073/pnas.1715374115. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. OECD . Reducing the Precarity of Academic Research Careers. OECD iLibrary; 2021. [DOI] [Google Scholar]
  27. Pellens M, Peters B, Hud M, Rammer C, Licht G. Public investment in R&D in reaction to economic crises - A longitudinal study for OECD countries. SSRN Electronic Journal. 2018;1:3122254. doi: 10.2139/ssrn.3122254. [DOI] [Google Scholar]
  28. Reithmeier R, O’Leary L, Zhu X, Dales C, Abdulkarim A, Aquil A, Brouillard L, Chang S, Miller S, Shi W, Vu N, Zou C. The 10,000 PhDs project at the University of Toronto: Using employment outcome data to inform graduate education. PLOS ONE. 2019;14:e0209898. doi: 10.1371/journal.pone.0209898. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Roach M, Sauermann H. The declining interest in an academic career. PLOS ONE. 2017;12:e0184130. doi: 10.1371/journal.pone.0184130. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Sauermann H, Roach M. Science PhD career preferences: Levels, changes, and advisor encouragement. PLOS ONE. 2012;7:e36307. doi: 10.1371/journal.pone.0036307. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Schillebeeckx M, Maricque B, Lewis C. The missing piece to changing the university culture. Nature Biotechnology. 2013;31:938–941. doi: 10.1038/nbt.2706. [DOI] [PubMed] [Google Scholar]
  32. Silva EA, Mejía AB, Watkins ES. Where do our graduates go? A tool kit for tracking career outcomes of biomedical PhD students and postdoctoral scholars. CBE Life Sciences Education. 2019;18:le3. doi: 10.1187/cbe.19-08-0150. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Stanford Biosciences Alumni Career Outcomes by Cohort. 2021. [September 10, 2021]. https://biosciences.stanford.edu/prospective-students/alumni-career-outcomes-dashboard/alumni-career-outcomes-by-cohort/
  34. Stanford IT&DS The Stanford PhD Alumni Employment Project. 2020. [January 17, 2020]. https://tableau.stanford.edu/t/IRDS/views/StanfordPhDAlumniEmployment/StanfordPhDAlumniEmploymentDashboard?%3Aembed_code_version=3&%3Aembed=y&%3AloadOrderID=0&%3Adisplay_spinner=no&%3Adisplay_count=n&%3AshowVizHome=n&%3Aorigin=viz_share_link
  35. Stayart CA, Brandt PD, Brown AM, Dahl T, Layton RL, Petrie KA, Flores-Kim EN, Peña CG, Fuhrmann CN, Monsalve GC. Applying inter-rater reliability to improve consistency in classifying PhD career outcomes. F1000Research. 2020;9:8. doi: 10.12688/f1000research.21046.2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Tregellas JR, Smucny J, Rojas DC, Legget KT. Predicting academic career outcomes by predoctoral publication record. PeerJ. 2018;6:e5707. doi: 10.7717/peerj.5707. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. UCSF Graduate Division PhD Program Statistics. 2021. [September 9, 2021]. https://graduate.ucsf.edu/program-statistics
  38. University of Chicago Career Outcomes of PhD Alumni. 2021. [August 20, 2021]. https://biosciences.uchicago.edu/after-uchicago/outcomes
  39. University of Michigan Rackham Doctoral Program Statistics. 2018. [November 9, 2018]. https://tableau.dsc.umich.edu/#/site/UM-Public/views/RackhamDoctoralProgramStatistics/ProgramStatistics
  40. University Toronto Employed and Engaged: Career Outcomes of Our PhD Graduates. 2021. [September 10, 2021]. https://www.sgs.utoronto.ca/about/explore-our-data/10000-phds-project/
  41. Vale RD. Accelerating scientific publication in biology. PNAS. 2015;112:13439–13446. doi: 10.1073/pnas.1511912112. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. van Dijk D, Manor O, Carey LB. Publication metrics and success on the academic job market. Current Biology. 2014;24:R516–R517. doi: 10.1016/j.cub.2014.04.039. [DOI] [PubMed] [Google Scholar]
  43. Vermeulen N, Parker JN, Penders B. Understanding life together: a brief history of collaboration in biology. Endeavour. 2013;37:162–171. doi: 10.1016/j.endeavour.2013.03.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. von Bartheld CS, Houmanfar R, Candido A. Prediction of junior faculty success in biomedical research: Comparison of metrics and effects of mentoring programs. PeerJ. 2015;3:e1262. doi: 10.7717/peerj.1262. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Wadman M. A workforce out of balance. Nature. 2012;486:304. doi: 10.1038/486304a. [DOI] [PubMed] [Google Scholar]
  46. Wilkinson MD, Dumontier M, Aalbersberg IJJ, Appleton G, Axton M, Baak A, Blomberg N, Boiten JW, da Silva Santos LB, Bourne PE, Bouwman J, Brookes AJ, Clark T, Crosas M, Dillo I, Dumon O, Edmunds S, Evelo CT, Finkers R, Gonzalez-Beltran A, Gray AJG, Groth P, Goble C, Grethe JS, Heringa J, ’t Hoen PAC, Hooft R, Kuhn T, Kok R, Kok J, Lusher SJ, Martone ME, Mons A, Packer AL, Persson B, Rocca-Serra P, Roos M, van Schaik R, Sansone SA, Schultes E, Sengstag T, Slater T, Strawn G, Swertz MA, Thompson M, van der Lei J, van Mulligen E, Velterop J, Waagmeester A, Wittenburg P, Wolstencroft K, Zhao J, Mons B. The FAIR Guiding Principles for scientific data management and stewardship. Scientific Data. 2016;3:160018. doi: 10.1038/sdata.2016.18. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Wolf JF, MacKay L, Haworth SE, Cossette M-L, Dedato MN, Young KB, Elliott CI, Oomen RA. Preprinting is positively associated with early career researcher status in ecology and evolution. Ecology and Evolution. 2021;11:13624–13632. doi: 10.1002/ece3.8106. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Woolston C. PhDs: the tortuous truth. Nature. 2019;575:403–406. doi: 10.1038/d41586-019-03459-7. [DOI] [PubMed] [Google Scholar]
  49. Woolston C. Uncertain prospects for postdoctoral researchers. Nature. 2020;588:181–184. doi: 10.1038/d41586-020-03381-3. [DOI] [PubMed] [Google Scholar]
  50. Zerhouni EA. NIH in the post-doubling era: Realities and strategies. Science. 2006;314:1088–1090. doi: 10.1126/science.1136931. [DOI] [PubMed] [Google Scholar]

Decision letter

Editor: Peter Rodgers1
Reviewed by: Barbara Janssens2, Sarvenaz Sarabipour3, Reinhart Reithmeier4

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Decision letter after peer review:

Thank you for submitting your article "Meta-research: The changing career paths of PhDs and postdocs trained at EMBL" to eLife for consideration as a Feature Article. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by the eLife Features Editor (Peter Rodgers). The three reviewers have agreed to reveal their identity: Barbara Janssens; Sarvenaz Sarabipour; Reinhart Reithmeier.

The reviewers and editors have discussed the reviews and we have drafted this decision letter to help you prepare a revised submission. Please also note that eLife does not permit figures in supplementary materials: instead we allow primary figures to have figure supplements. I can suggest how to accommodate the figures in your supplementary materials if your article is accepted.

The eLife Features Editor will also contact you separately about some editorial issues that you will need to address.

Summary:

This paper deals with career outcomes for some 2284 PhD graduates and post-doctoral fellows from the European Molecular Biology Laboratories, a prestigious research institute that attracts top talent from around the world – the first of its kind. The paper is rich in data beyond simple career outcomes over time and includes gender, publications and collaboration. The methodology involved internet searches like other studies and was enhanced by robust statistical analyses. This important and timely study fills a gap in our knowledge, highlights the important role that institutions like EMBL play in training the next generation of researchers and innovators, and may stimulate other universities and research institutes to do the same. However, there are a number of points that need to be addressed to make the article suitable for publication.

Essential revisions:

1. It is a great achievement to show career destinations for 89% of the 2284 searched for. However the question is in which cohort the 249 "unknown" belong: it could be more transparent to keep those numbers included also in the detailed cohort analysis. It is also unclear, whether the cohort sizes reflect the actual number of researchers who left EMBL in that time period or whether data were lacking. The cohorts reported here increased about 33% from 2004 to 2012 and another 9% by 2020 (Table 2). Could it be, that for more recent cohorts more data are available – for example due to the fact that younger researchers can be found on social media like LinkedIn?

2. The main and interesting conclusion in the abstract is that of the 45% of alumni not continuing in academic research, one third does industry research and one third is in a science-related profession. Other interesting take-home messages are, that a large proportion of alumni changing sectors enter – or are quickly promoted to – managerial positions, that of those who entered a research position in industry one in ten returned to academia and that female Alumni were found less frequently in PI roles. It would be interesting to know, whether there is a difference in these numbers between cohorts – e.g. if more alumni return from industry to academia in recent cohorts and whether female researchers stay longer in postdoc roles (which could influence the total number of female PIs at a given time point).

3. In the time-resolved analysis the authors claim that the probability of being found as a PI in Academia diminished by about 15% after the first cohort (Figure 2C). However the question is whether the absolute number of PIs also decreased. This could be clarified with some reference numbers in the supplementary materials. When calculating the % of the given cohort sizes at different institutes (supp. Table 4) some differences (increase vs decrease) in the number of PIs can be found. Even if the hypothesis that the differences between cohorts are not EMBL-specific, and reflect a wide-spread change in the number of PhDs and postdocs relative to available PI positions seems valid, it would be good to clarify this issue.

4. Interesting findings is also the significant change in time between PhD and first PI position of 0.8 calendar years between the 2004 and 2012 cohorts. It is not surprising that publication factors are highly correlated with entry into PI positions: indeed all ECRs who became PIs published well, but, not all ECRs who published well became PIs! Publications have become more collaborative over the last decade (the number of coauthors has doubled and the number of first author publications per ECR diminished). Another relevant observation is the lack of correlation with group leader seniority or nationality. Group publications were also predictive for other research and science-related careers. Finally a strong observation is that 45% of leavers from the last 5-years who were found to be working outside of academia held senior or management-level roles. These findings can be reassuring for ECRs and the authors could consider to clearly state these in their conclusions.

5. Regarding the career tracking method used for this study: doing google searches for 2284 Alumni is a plausible effort and has probably been time consuming. A question for other research institutes and universities would be, whether this method of career tracking is scalable and/or feasible to continue as a regular task, or whether the authors see this as a one-time effort. If so, what kind or extent of career tracking would the authors recommend to continue sustainably? Performing career tracking is quite relevant, as institutes worldwide now start to be asked to deliver such data to governments and funding agents.

6. In Discussion and future work, it would be valuable to briefly discuss/aspire that institutions such as EMBL compile and publicly report on this type of data/records analysis together with surveys (what authors call mixed methods) of career intent and research environment.

i. With surveys one could have the number of EMBL trainees that actually applied for PI jobs.

ii. The fact that women were found less frequently in PI jobs does not reveal if (1) women apply less frequently or (2) search committees offer PI jobs less often to women or (3) combination of the two.

iii. Surveys together with the data presented in this work can examine the role of lab environment during training and job application.

7. An observation that authors have in the manuscript is that women are less represented as AcPIs (academic PIs). But it's not possible to claim it's an active mechanism. It would be valuable if authors plot the timeline by gender so readers could see the noise.

8. In various panels of Figures 2-4, please clarify if "Time after EMBL" (the label on the x-axis) means "Time after leaving EMBL" or "Time after arrival at EMBL".

Also, it appears that regardless of the cohort, postdocs have increased chances to become a PI only years after they leave EMBL, why? did they go on to do a second postdoc?

9. Please discuss supplementary table 4 in the text, and highlight any common findings from these studies.

10. The authors may also wish to comment on how some members of faculty recruitment committees may need to be trained to recognize bias in relying too heavily on citation indices and first author publications in hiring decisions rather than the scientific contribution of highly-qualified candidates to collaborative projects.

eLife. 2023 Nov 23;12:e78706. doi: 10.7554/eLife.78706.sa2

Author response

Essential revisions:

1. It is a great achievement to show career destinations for 89% of the 2284 searched for. However the question is in which cohort the 249 "unknown" belong: it could be more transparent to keep those numbers included also in the detailed cohort analysis. It is also unclear, whether the cohort sizes reflect the actual number of researchers who left EMBL in that time period or whether data were lacking. The cohorts reported here increased about 33% from 2004 to 2012 and another 9% by 2020 (Table 2). Could it be, that for more recent cohorts more data are available – for example due to the fact that younger researchers can be found on social media like LinkedIn?

Thank you for these comments, we are happy to clarify here and in the manuscript:

The cohort sizes reflect the number of all PhD students and postdoctoral fellows who left in that time period, according to the official institutional administrative records. The organization grew during the time period included in the study (1997-2020), and the number of PhD students and postdocs in the later cohorts is therefore greater. It excludes individuals who were marked as deceased in our alumni records at the point the data was originally shared with us (June 2017 for 1997-2016 leavers and April 2021 for 2017-2020 leavers; or whose death we learned of during the update of our career tracking data in summer 2021 (through an updated alumni record, or if we found an online obituary)).

For each PhD or postdoc cohort, the percentage of alumni whose current position is unknown is between 9% and 14%, with no consistent trends with time. However, for the oldest cohort we less frequently found a complete online CV or biosketch that was detailed enough to confirm the type of position held for the full-time span since EMBL, particularly for postdoc alumni. Fewer of the older cohort therefore have a detailed CV/career path and there is a higher percentage of unknowns for specific timepoints after EMBL.

Changes to manuscript:

We have added an additional row with the number and percentage of alumni with detailed career paths for each cohort in Table 2 – new row = n(%) detailed career path

We have also clarified that the increased cohort size is due to growth in sentence that refers to this table: “More recent cohorts were also larger (Table 2), reflecting growth of the organization between 1997 and 2020.”

Column charts showing type of position by cohort (Figure 1B and Figure 1 —figure supplement 2 as 1B, but for other time-points) now include all alumni, not just those with a full career path available.

2. The main and interesting conclusion in the abstract is that of the 45% of alumni not continuing in academic research, one third does industry research and one third is in a science-related profession. Other interesting take-home messages are, that a large proportion of alumni changing sectors enter – or are quickly promoted to – managerial positions, that of those who entered a research position in industry one in ten returned to academia and that female Alumni were found less frequently in PI roles. It would be interesting to know, whether there is a difference in these numbers between cohorts – e.g. if more alumni return from industry to academia in recent cohorts and whether female researchers stay longer in postdoc roles (which could influence the total number of female PIs at a given time point).

‘if more alumni return from industry to academia in recent cohorts’

This is difficult to assess due to the different career lengths and small numbers transitioning from one type of career to another. For example, of 415 alumni who we have a career path for and had at least one Industry role, just 22 returned to a faculty position. Twelve of these were from the oldest cohort (of 117 who held an industry role), compared to 7 (of 178) for the most recent cohort. Similarly, for PI to industry, 21 transitions were observed from the 539 career paths – 16 from the oldest cohort (from 231) and 1 (from 128) in the newest. Given that the propensity to transition may also change with career length, it is difficult to make comparisons or detect meaningful trends from these small numbers.

“whether female researchers stay longer in postdoc roles (which could influence the total number of female PIs at a given time point).”

We did not observe a statistically significant difference in length between PhD and becoming a PI, but agree that his is interesting and that it should be included in the manuscript.

To add this to the manuscript we made three changes: Additional figure supplement showing the average PhD to PI length for male vs female alumni [as previous figure comparing 2005-2012 and 1997-2004 cohorts, but now comparing female vs male alumni] (Figure 3 —figure supplement 1) – this suggests that male and female researchers spend similar times in postdoc roles as the differences are not statistically significant.

We have now included Kaplan Maier plots by gender, which also illustrate the entry into PI (and other) roles with time (as Figure 3B -C and Figure 3 —figure supplement 2 additional plots for AcOt, IndR, NonSci).

We also expand discussion of these data in the main text – see below with new detail italicised. (note: to allow more detailed discussion without requiring repetition of the information, we have moved this section after the sections on changes in career outcomes, where the Kaplan-Meier and PhD to PI length are first discussed).

Gender differences in career outcomes

Many studies have reported that female ECRs are less likely to remain in an academic career (44, 45). Consistent with these previous studies, we observed that male alumni were found more frequently in PI roles (Figure 3A-B; Table S5 in Supplementary information). Figure 3A-B; Table S5 in Supplementary information. There was no statistically significant difference in the time to obtain a PI role between male and female alumni for alumni from 1997-2012, who became PIs within 9-years (Figure 3—figure supplement 1). The difference in career outcomes is therefore unlikely to be explained by different career dynamics for male and female alumni.

Female alumni were more frequently found in science-related non-research roles than male alumni (Figure 3A). In our Cox models, there was also a statistically significant difference between genders in entry into science-related non-research roles for postdoc alumni [p = 0.016] (Figure 3C; Table S5 in Supplementary information).

We more frequently found male alumni in industry research and non-science-related roles than their female colleagues (Figure 3A; Figure 3—figure supplement 2B-C). However, a higher percentage of female than male alumni could not be located. As academics are usually listed on institutional websites, often with a historical publication list that allows unambiguous identification, we expect that most alumni who were not located are employed in the non-academic sector. This means that, considering the higher percentage of female alumni with unknown career paths (where non-academic careers are likely over-represented), the true percentage of female alumni in industry research and non-science-related roles is likely higher, and possibly comparable with the percentage of male alumni in these roles.

3. In the time-resolved analysis the authors claim that the probability of being found as a PI in Academia diminished by about 15% after the first cohort (Figure 2C). However the question is whether the absolute number of PIs also decreased. This could be clarified with some reference numbers in the supplementary materials. When calculating the % of the given cohort sizes at different institutes (supp. Table 4) some differences (increase vs decrease) in the number of PIs can be found. Even if the hypothesis that the differences between cohorts are not EMBL-specific, and reflect a wide-spread change in the number of PhDs and postdocs relative to available PI positions seems valid, it would be good to clarify this issue.

We have added information on the absolute number of PIs for each group in the supplementary table that collates the data published from other institutions and comparable EMBL data (in original manuscript table 4; now Table S3 in Supplementary file 1). We have also expanded discussion of this table the text in response to comment 9 (see comment 9 below) and include the absolute numbers.

Additional clarification

In the datasets, the number of PhDs trained per year has increased with time at all institutions. The cohort size – and how much this has increased for more recent cohorts however varies – for example, Stanford’s 2011-2015 cohort was just 18% larger than its 2000-2005 cohort (503 vs 426), whilst the University of Toronto’s 2012-2015 is 96% larger than its 2000-2003 (1234 vs 629). For EMBL, the PhD cohort sizes increased 45% from 256 for the 1997-2004 cohort, to 372 for 2013-2020. Therefore, the absolute number of PIs with time is difficult to compare between institutions. We therefore feel that the percentage of alumni entering different career options is the most pragmatic measure for comparing how career outcomes are changing with time. It can be viewed as the ‘chance’ of a ECR from a specific programme of entering that career area. If an institution continues to train the same absolute number of PIs per year, but trained many more scientists, it nevertheless saw a big difference the career outcomes of its alumni, with more alumni entering non-PI roles.

4. Interesting findings is also the significant change in time between PhD and first PI position of 0.8 calendar years between the 2004 and 2012 cohorts. It is not surprising that publication factors are highly correlated with entry into PI positions: indeed all ECRs who became PIs published well, but, not all ECRs who published well became PIs! Publications have become more collaborative over the last decade (the number of coauthors has doubled and the number of first author publications per ECR diminished). Another relevant observation is the lack of correlation with group leader seniority or nationality. Group publications were also predictive for other research and science-related careers. Finally a strong observation is that 45% of leavers from the last 5-years who were found to be working outside of academia held senior or management-level roles. These findings can be reassuring for ECRs and the authors could consider to clearly state these in their conclusions.

We have made the following changes to the manuscript to emphasise the positive aspects of our career findings (but balance the editorial comment that “If possible please shorten the first paragraph of the discussion and avoid any unnecessary repetition of material from earlier in the article.”).

This now reads:

“Many ECRs are employed on fixed-term contracts funded by project-based grants, sometimes for a decade or more (52, 53), and surveys suggest that ECRs are concerned about career progression (18-22). We hope ECRs will be reassured by the results of our time-resolved analysis that indicate that the skills and knowledge developed as an ECR can be applied in academia, industry and other sectors. Within academic research, service and teaching, we observed a marked reduction in the percentage of alumni entering PI roles; nevertheless, academic careers continue to be an important career destination. The percentage of alumni entering career areas outside academic research, service and teaching has increased, and our data suggest that ECRs’ skills are valued in these careers; 45% of leavers from the last 5-years who were found to be working outside of academic research and teaching held senior or management-level roles.”

5. Regarding the career tracking method used for this study: doing google searches for 2284 Alumni is a plausible effort and has probably been time consuming. A question for other research institutes and universities would be, whether this method of career tracking is scalable and/or feasible to continue as a regular task, or whether the authors see this as a one-time effort. If so, what kind or extent of career tracking would the authors recommend to continue sustainably? Performing career tracking is quite relevant, as institutes worldwide now start to be asked to deliver such data to governments and funding agents.

We have expanded the Discussion section ‘Future career studies’ to include a recommendation as follows:

“Evaluating the outcomes of training programmes, and making these data transparently available is a valuable exercise that can provide information to policymakers, transparency for ECRs, and help institutions provide effective career development support. We plan to update our observational data every four years, and maintain data on the career paths of alumni for 24 years after they leave EMBL. This will help us to identify any further changes in the career landscape and better understand long-term career outcomes. Silva et al. (2019) (73) have also described a method for completing career outcome tracking on a yearly basis with estimations of the time and other resources required. We encourage institutions to consider whether they can adapt our, or Silva’s method to the administrative processes and data-privacy regulations applicable at their institutions.”

6. In Discussion and future work, it would be valuable to briefly discuss/aspire that institutions such as EMBL compile and publicly report on this type of data/records analysis together with surveys (what authors call mixed methods) of career intent and research environment.

i. With surveys one could have the number of EMBL trainees that actually applied for PI jobs.

ii. The fact that women were found less frequently in PI jobs does not reveal if (1) women apply less frequently or (2) search committees offer PI jobs less often to women or (3) combination of the two.

iii. Surveys together with the data presented in this work can examine the role of lab environment during training and job application.

Changes to manuscript: We have re-written the ‘Future career studies’ section of the discussion to incorporate these points. This now reads:

“Future studies should also ideally include mixed-method longitudinal studies. This would allow career motivations, skills development, research environment, job application activity, and other factors to be recorded by surveys during ECR training. Correlating these factors to career and training outcomes would allow investigation of multifactorial and complex issues such as gender differences in career outcomes, and provide policymakers with a fuller picture of workforce trends. Such studies will, however, require large sample sizes from multiple institutions and would need significant time investment and coordination over a long time period. The commitment and the support of funders and institutions would therefore be critical.”

7. An observation that authors have in the manuscript is that women are less represented as AcPIs (academic PIs). But it's not possible to claim it's an active mechanism. It would be valuable if authors plot the timeline by gender so readers could see the noise.

Changes to manuscript: We have generated Kaplan Maier plots by gender for entry into each career area, and include these in Figure 3 and Figure 3 —figure supplement 2 (see discussion of reviewer comment 2, above). The hazard ratios, 95% confidence intervals and p values are provided in a supplementary table in file 1 so that the confidence can be judged.

8. In various panels of Figures 2-4, please clarify if "Time after EMBL" (the label on the x-axis) means "Time after leaving EMBL" or "Time after arrival at EMBL".

We have added a clarification in the figure legend (“Time after EMBL refers to the number of calendar years between defence of an EMBL PhD and first PI role (for PhD alumni)) – or number of calendar years between leaving the EMBL postdoc programme and first PI role (for postdoc alumni)”, or clarified this in the figure labels, for each figure.

Also, it appears that regardless of the cohort, postdocs have increased chances to become a PI only years after they leave EMBL, why? did they go on to do a second postdoc?

We have expanded on this in the Results section ‘EMBL alumni contribute to research and innovation in academic and non-academic roles’, adding the following text:

“On average, PhD alumni became PIs 6.1 calendar years after their PhD defence. For postdoc alumni, the start year of the first PI role was on average 7.3 calendar years after their PhD and 2.5 years after completing their EMBL postdoc. Almost half of EMBL postdoc alumni who became PIs did so directly after completing their EMBL postdoc (168 of 343 alumni with a detailed career path available). Other postdoc alumni made the transition later, most frequently after one additional postdoc (71 alumni) or a single academic research / teaching / service position (56 alumni). Some alumni held multiple academic (40 alumni), or one or more non-academic positions (8 alumni) between their EMBL postdoc and first PI role.”

And to provide balance / to avoid focusing only on academic careers, we also include an additional sentence in the subsequent paragraph on non-academic areas:

“The average time between being awarded a PhD and the first industry research, science-related or non-science-related role was 5.0, 5.3 and 4.2 calendar years respectively.”

9. Please discuss supplementary table 4 in the text, and highlight any common findings from these studies.

We have expanded the discussion of this in the Results section ‘The percentage of EMBL alumni becoming PIs is similar to data released by North American institutions for both older and more recent cohorts’. [note that due to rearrangements, supplementary table 4 is now ‘Table S3 in Supplementary file 1’]. This now reads:

“A number of institutions have released cohort-based PhD outcomes data online or in publications (32, 44-49). Of these, a recent dataset from Stanford University offers the longest career tracking, reporting outcomes for PhD graduates from 20 graduation years (2000-2019) (44). In this dataset, Stanford University reported that 34% (145/426) of its 2000-2005 PhD alumni were in research-focussed faculty roles in 2018. This is comparable to the 37% (78/210) we observe for the EMBL alumni for the same time period. For 2011-2015 graduates, comparable percentages of Stanford (13%, 63/503) and EMBL (11%, 25/234) graduates were also in PI roles in 2018 (Figure 1C). Figure 1 – Supplement 3 plots data from five other datasets alongside equivalent data from EMBL (further details of each dataset are available in the Table S3 in Supplementary file 1). This includes a published dataset from the University of Toronto (32, 49), which reported that 31% (192/629) of its 2000-2003 life science division graduates and 25% (203/816) of its 2004-2007 life science graduates were in tenure stream roles in 2016. The equivalent EMBL data is comparable at 39% (52/132) and 28% (49/172) of graduates in PI roles respectively. Across all six datasets, EMBL and the other institutes generally have a similar proportion of alumni entering PI roles for comparable cohorts. This is consistent with our hypothesis that the differences between cohorts are not EMBL-specific, and reflect a wide-spread change in the number of PhDs and postdocs relative to available PI positions.”

10. The authors may also wish to comment on how some members of faculty recruitment committees may need to be trained to recognize bias in relying too heavily on citation indices and first author publications in hiring decisions rather than the scientific contribution of highly-qualified candidates to collaborative projects.

We have expanded this section of the discussion (third paragraph of ‘Addressing ECR career challenges’ in the discussion), also mentioning the trend to narrative CVs that has accelerated recently. This now reads:

“Research assessment and availability of funding play an important role in determining the career prospects of an academic. Therefore, it is also vital that factors that may lead to misperception of the productivity of ECRs, such as involvement in large-scale projects, career breaks, or time spent on teaching and service, are considered in research assessment during hiring, promotion and funding decisions. Initiatives such as the San Francisco Declaration on Research Assessment (DORA) and Coalition for Advancing Research Assessment (CoARA) have been advocating for an increased focus on good practice in evaluating research outputs, and many institutions and funders have reviewed their practices. Cancer Research UK, for example, now asks applicants to its grants to describe three to five research achievements, which can include non-publication outputs (64) and narrative CV formats that allow candidates to put their achievements in context are also becoming more common (61). The impact of the coronavirus pandemic on research productivity of researchers with caregiving responsibilities makes such actions imperative (65-67). We welcome this increased focus on qualitative assessment of scientific contribution, rather than reliance on publication metrics. To ensure successful implementation and to avoid unintended consequences (such as introducing new biases), it will be important for funders and institutions to provide appropriate guidance and/or training to evaluators and to carefully monitor implementation. Other initiatives that may help ECR involved in ambitious projects to demonstrate their contributions include more transparent author contribution information in publications (68, 69) and promotion of "FAIR" principles of data management (70). The increasing use of preprints (41, 71) is also likely to have a positive effect on the careers of ECRs involved in projects with longer time scales (72).”

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Supplementary Materials

    Figure 1—source data 1. Summary data plotted in Figure 1A, B.
    elife-78706-fig1-data1.xlsx (10.3KB, xlsx)
    Figure 1—figure supplement 1—source data 1. Summary data plotted in the simplified Sankey plots in Figure 1—figure supplement 1.
    elife-78706-fig1-figsupp1-data1.xlsx (10.9KB, xlsx)
    Figure 1—figure supplement 2—source data 1. Summary data plotted in Figure 1—figure supplement 2.
    elife-78706-fig1-figsupp2-data1.xlsx (11.9KB, xlsx)
    Figure 2—figure supplement 1—source data 1. Summary data plotted in Figure 2—figure supplement 1.
    elife-78706-fig2-figsupp1-data1.xlsx (9.8KB, xlsx)
    Figure 3—source data 1. Summary data plotted in Figure 3A.
    elife-78706-fig3-data1.xlsx (44.8KB, xlsx)
    Figure 3—figure supplement 1—source data 1. Summary data plotted in Figure 3—figure supplement 1.
    elife-78706-fig3-figsupp1-data1.xlsx (9.8KB, xlsx)
    Figure 4—source data 1. Summary data plotted in Figure 4A, B and D.
    elife-78706-fig4-data1.xlsx (20.6KB, xlsx)
    Figure 4—figure supplement 1—source data 1. Summary data plotted in Figure 4—figure supplement 1.
    elife-78706-fig4-figsupp1-data1.xlsx (10.9KB, xlsx)
    Figure 5—source data 1. Summary data plotted in Figure 5.
    elife-78706-fig5-data1.xlsx (12.8KB, xlsx)
    MDAR checklist
    elife-78706-mdarchecklist1.docx (102.1KB, docx)
    Supplementary file 1. Supplementary tables (Table S1–11).
    elife-78706-supp1.docx (147.5KB, docx)
    Supplementary file 2. Supplementary methods.
    elife-78706-supp2.docx (20.4KB, docx)

    Data Availability Statement

    The data were collated for the provision of statistics, and are stored in a manner compliant with EMBL's internal policy on data protection. This policy means that the full dataset cannot be made publicly available (because the nature of the data means that sufficient anonymisation is not possible). Summary statistics for the main data table can be found in Supplementary file 1 (Table S1). Rmarkdown documentation of the analysis and figures can be found here and is available on GitHub (copy archived at Coulthard and Lu, 2022).

    Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd

    RESOURCES