Introduction

The circumstances during the COVID-19 pandemic severely disrupted many academics’ research activities. In its early phase especially, lockdowns of universities and schools, work-from-home orders, social distancing measures, and travel restrictions forced them to deviate from their routines. Most academics had to invest time in switching teaching to online formats and many had to spend more time on household duties and care responsibilities (Deryugina et al., 2021). Restricted access to research facilities and fieldwork sites directly interrupted research processes (Peetz et al., 2022; Staniscuaski et al., 2021), which induced shifts within academics’ portfolio of activities (Sawert & Keil, 2021), for instance, from work in the laboratory to work on manuscripts. In various disciplines, academics furthermore paused some of their research lines and directed their efforts to dealing with the causes and consequences of the pandemic (Riccaboni & Verginer, 2022; Walker et al., 2022).

Already early into the pandemic, it became clear that not all academics were affected equally (see, for instance, Cui et al., 2022; Deryugina et al., 2021; Myers et al., 2020). Understanding these differences has become an important endeavor and may yield insights into influences behind research productivity more generally. In a highly competitive, tournament-like system such as academia, even seemingly minor differences in research productivity, which remains the main criterion for career advancement, can amplify over time and lead to systematic inequalities. Gauging and addressing the consequences of the pandemic requires identifying those academics who were particularly susceptible to disruptions of their research productivity. In addition, identifying factors associated with academics’ susceptibility to disruptions can contribute to debates on research productivity differences outside the context of the pandemic as well. As exceptional as the pandemic has been, academic work always includes handling various demands and coping with disruptions. The time of the pandemic may thus serve as a magnifying glass for identifying factors associated with academics’ susceptibility to such disruptions, which were hitherto difficult to observe (see also King & Frederickson, 2021; Oleschuk, 2020).

Against this backdrop, we explore empirically which factors shaped the pandemic’s impact on the research activities of professors with a focus on differences at the intersection of academic discipline and gender. The analysis uses the outcomes of an online survey for 1891 university professors in Germany, which we conducted in October and November 2020. In addition to comprehensive information on professors and their jobs, the survey data comprise information on how professors perceived the impact of the pandemic during its early phase (i.e., March to June 2020) on several academic activities and the time spent on care responsibilities. The analysis covers a comprehensive set of factors identified based on the literature, which may have shaped the pandemic’s impact. In a first step, we estimate multivariate models for the sample as a whole including all those factors. This allows us to identify the relevance of various factors while accounting for the influence of a comprehensive set of other factors, irrespective of specific group differences. In a second step, we follow up on disciplinary and gender differences using sample splits and an econometric decomposition technique. Via these analyses, we can, first, assess and quantify how the relevance of the factors of interest varies among groups of academics and, second, obtain a more detailed understanding of the reasons underlying differences in academics’ susceptibility to disruptions of their research productivity. We complement this main analysis by several Appendices that investigate alternative explanations for our findings and inform the interpretation of our results.

The paper contributes to the research literature in three main ways. First, we extend prior research on the pandemic’s impact on research activities by covering factors that have not been investigated systematically so far. This includes factors pertaining to academics’ research engagement, features of the institutional environment, and the negative impact of the pandemic on teaching activities and administrative tasks. Second, we contribute to the same strand of the literature by providing detailed results on group differences, especially as they concern academic disciplines and genders. Third, our detailed investigation of group differences as regards academics’ susceptibility to disruptions of their research productivity contributes to long-standing debates on research productivity differences among academics.

Our findings yield implications for policy and management efforts directed at academics’ research productivity both within and outside of the context of the pandemic. Insights into the differential impact of the pandemic have triggered various activities aimed at mitigating the inequalities caused (see King & Frederickson, 2021; Malisch et al., 2020; Oleschuk, 2020). The effectiveness of these activities depends on identifying those academics who were particularly susceptible to the disruptions caused by the pandemic (see, for instance, Expert Group on the COVID-19 Impact on Gender Equality in R&I, 2023; Malisch et al., 2020). Our findings highlight a great complexity in this regard, which renders a careful differentiation of cases necessary. The findings also stress the pervasiveness of the pandemic’s negative impact for female academics. In addition, they highlight that this pervasiveness extends beyond the stronger constraints associated with care responsibilities and household duties, which have often been in the focus of debates. The combined insights obtained with regard to the pandemic furthermore provide reference points for promoting equal conditions for research productivity more generally.

Background and objectives

Academics’ research productivity in the context of the pandemic

Major impacts on research activities

The circumstances during the COVID-19 pandemic had a broad negative impact on academics’ research activities. The time available for research declined for many academics as the time required for household duties and care responsibilities, for shifting to online teaching, and for administrative tasks increased (Deryugina et al., 2021; Gao et al., 2021; Myers et al., 2020; Peetz et al., 2022). The closure of universities, the need to work from home, and limited access to fieldwork sites directly interfered with research processes (Peetz et al., 2022; Sawert & Keil, 2021; Staniscuaski et al., 2021; Walters et al., 2022) and induced shifts in academics’ work portfolios, for instance, away from project implementation to work on manuscripts (Sawert & Keil, 2021). In addition, academics faced difficulties with initiating new research projects (Gao et al., 2021) and the reduction in in-person meetings (such as conferences and seminars) may have reduced the possibilities for discussing research ideas and results. However, a reduction in commuting time and in the number of in-person meetings may also have freed up time for research. These counteracting effects may explain why the picture as regards research output is mixed so far: Whereas some studies find that publication output in the form of pre-prints (Abramo et al., 2022) and journal publications (Madsen et al., 2022) has fallen in conjunction with the pandemic (see also Rusconi et al., 2020), other studies actually report an increase (see Ali, 2022; Aviv‑Reuven and Rosenfeld, 2021; Cui et al., 2022; King & Frederickson, 2021).

Differences in impact between academics

The pandemic’s impact on research activities was not uniform across academia. A first relevant characteristic in this regard is the academic discipline. Prior studies suggest that the disruptions caused by the pandemic were particularly severe in disciplines where research processes are dependent on access to research facilities such as laboratories or fieldwork sites (Myers et al., 2020; Sawert & Keil, 2021; Walters et al., 2022). Secondly, more negative effects were observed for academics at earlier career stages (Cui et al., 2022; Sawert & Keil, 2021). This has been attributed, among other reasons, to greater pressures to publish for career advancement. Thirdly, strong differences were also related to socio-demographic characteristics, especially parenthood and gender. Several studies have shown that academics with (young) children were affected more severely (Gao et al., 2021; Myers et al., 2020; Sawert & Keil, 2021; Staniscuaski et al., 2021) due to additional time demands from household duties and care responsibilities. Pronounced disadvantages have also been observed for female academics, especially at the beginning of the pandemic (Andersen et al., 2020; Gao et al., 2021; Lerchenmüller et al., 2021; Liu et al., 2022). This concerns self-reported time budgets and productivity (Deryugina et al., 2021; Myers et al., 2020; Peetz et al., 2022; Rusconi et al., 2020; Sawert & Keil, 2021; Staniscuaski et al., 2021) and the initiation of new research projects (Gao et al., 2021; Peetz et al., 2022), as well as research output in form of submissions (Squazzoni et al., 2021), working papers (Cui et al., 2022; King & Frederickson, 2021; Muric et al., 2021; see, however, Abramo et al., 2022), and journal articles (Lerchenmüller et al., 2021; Liu et al., 2022; Madsen et al., 2022; Muric et al., 2021).

Moreover, there is evidence that group differences did intersect, especially as far as gender is concerned. For instance, disadvantages for female academics have been found to be stronger among more junior academics (Cui et al., 2022; Madsen et al., 2022). Variation in the strength of gender differences has also been observed between academic disciplines (King & Frederickson, 2021; Lerchenmüller et al., 2021; Madsen et al., 2022) and countries (Abramo et al., 2022; Cui et al., 2022; Lerchenmüller et al., 2021; Liu et al., 2022; Madsen et al., 2022; Muric et al., 2021).

Research productivity differences beyond the pandemic

Research on the pandemic’s varying impact has examined various factors, but not all of those identified as pertinent in studies investigating disparities in research productivity among academics in a broader context. This strand of research (see Fox, 2005; Mayer & Rathmann, 2018; Nygaard et al., 2022; Stack, 2004 for overviews) has underscored the significance of the aforementioned characteristics in shaping researchers’ productivity. There are, however, two pertinent factors in this literature that have not been investigated systematically in relation to the pandemic: first, the size of research groups or departments, even though evidence for its association with productivity is mixed (see Mayer & Rathmann, 2018), and, second, academics’ stance on their job such as the relevance ascribed to research. Academics’ stance on certain parts of their job has been shown to matter for how they prioritize activities, for instance, research (see Mayer & Rathmann, 2018) and engagement with society (Püttmann et al., 2023).

Research interest

Our primary research focus encompasses two key aspects. First, we investigate the pandemic’s disparate effects on the research activities of academics. Second, we seek to gain a broader understanding of the factors that underlie variations in their research productivity. Even though prior research has identified various factors along the lines of which the pandemic’s impact differed, some potentially relevant factors have not been investigated systematically so far. We thus consider the most important factors already investigated (see the “Differences in impact between academics” section) and complement them with additional factors from the wider literature on research productivity differences (see the “Research productivity differences beyond the pandemic” section). Moreover, prior research has unveiled selected interactions between factors associated with the pandemic’s impact, but there is a lack of systematic insights in this area. We thus also investigate whether the factors identified as relevant differ in their relevance between groups of academics, in particular, by academic discipline and gender. These findings allow to derive more general implications concerning academics’ susceptibility to disruptions of their research activities, which is one of the influences on academics’ research productivity.

Empirical setting and data

Data basis

The empirical analysis uses results of an online survey of professors in Germany conducted in October and November 2020, which were complemented with secondary data (see Appendix B for details). The sample comprises 1891 professors who work at universities governed by the state or religious institutions and who are engaged in both research and teaching. The survey data cover respondents’ perceptions of the impact of the COVID-19 pandemic during its early phase (i.e., March to June 2020) on their professional and private lives as well as information on respondents and their academic occupation. These data were complemented with secondary data from the German Federal Statistical Office, which cover certain conditions of academic work in respondents’ disciplines and institutions.

Sample composition

Our survey data cover the German university system comprehensively. The sample includes professors at different career stages, from all academic disciplines and from more than 90% of the universities in Germany. Almost all professors (more than 90%) are tenured, which is a first notable difference when comparing the sample to the population of professors (see Appendix Table A.1 ). There is furthermore an overrepresentation of full professors and an underrepresentation of assistant professors. Pressure to publish in order to stay in the academic system, which may have been an issue for academics at other career stages during the pandemic, thus does not apply to most of the professors in our sample. Another notable difference is the underrepresentation of professors from the natural sciences. However, all groups of professors based on the characteristics of interest (e.g., gender and discipline) are sufficiently represented in our data. Hence, our data allow us to generate robust insights into the group differences that constitute the main focus of our analysis.Footnote 1

Main variables

The main outcome of interest is the pandemic’s impact on respondents’ research. This variable is based on the survey question whether the circumstances during the initial phase of the COVID-19 pandemic (i.e., March to June 2020) had a negative impact, no impact, or a positive impact on research activities (see Appendix Figure A.2  for the exact wording of the question and Appendix Tables A.2 and A.3 for descriptive statistics). The question thus refers to a time when a nationwide lockdown, which encompassed educational institutions like kindergartens, schools, and universities, and stringent contact and travel restrictions were enforced. As only a minority of professors (9.1%) reported a positive impact, we group these professors together with professors who reported no impact (22.6%). The analysis then uses a dummy variable indicating whether professors perceived a negative impact or not.Footnote 2

We consider five groups of independent variables: (1) socio-demographic characteristics, (2) characteristics of respondents’ academic occupation, (3) respondents’ research engagement, (4) the environment provided by respondents’ institutions, and (5) the impact of the circumstances during the pandemic on respondents’ academic activities other than research (see Appendix B for details, Appendix Table A.2 for descriptive statistics, and Appendix Table A.3 for descriptive statistics by gender). Two variables based on the survey comprise socio-demographic characteristics: gender (male, female) and whether respondents had to increase the time spent on care responsibilities due to the pandemic (no, yes). Three variables cover characteristics of respondents’ academic occupation. The first two variables are based on the survey: holding a full professorship (no, yes) and seniority (measured in decades since the first appointment as professor). The third variable is based on secondary data from the German Federal Statistical Office and assigns respondents to three groups (lowest third, middle third, highest third) according to the average total expenditures per professor in the 31 academic disciplines to which respondents could assign themselves (see Appendix B for the full list of disciplines). This variable approximates the dependence of respondents’ research processes on access to research facilities and equipment. Three variables based on the survey pertain to professors’ research engagement: (1) the comparative relevance ascribed to research and teaching (teaching higher, no difference, research higher), (2) the number of prestigious third party-funded research projects, and (3) an index measuring professors’ publication prolificness in the form of the number of books and peer-reviewed articles and contributions they published (see Appendix B for details). Two variables based on secondary data from the German Federal Statistical Office cover the institutional environment: (1) department or faculty size based on the number of professors in the form of two groups (below median, equal to or above median) and (2) the size of institutions (measured as the absolute number of professors). Two variables based on the survey concern the pandemic’s impact on activities other than research: (1) whether respondents perceived a negative impact on teaching (no, yes) and (2) whether respondents perceived a negative impact on administrative tasks (no, yes).

Empirical results

Descriptive statistics show the broad negative impact of the circumstances during the pandemic on professors’ research activities, but also reveal marked variation that warrants further analysis. Slightly more than two-thirds of professors (68.3%) perceived a negative impact on their research activities (see Appendix Table A.2 ). Even more emphasized was the pandemic’s impact on teaching, which a markedly higher share of professors (86.5%) perceived as negative. In addition to confirming how disruptive the pandemic has been, these figures show that slightly less than one-third of all professors were nevertheless able to maintain their research productivity.

Given that our main research interest lies with the immediate consequences of the pandemic, we consider the variable covering an increase in the time spent on care responsibilities and not, for instance, a variable pertaining to the presence of underage children in respondents’ households. We nevertheless conducted an analysis of the associations between the presence of underage children in the household, the presence of other care responsibilities, and respondents’ perceived need to increase the time spent on care responsibilities (see Appendix E). This analysis shows that even when controlling for the presence of care responsibilities of any sort, female professors were more likely to report such an increase. An on average more pronounced uptake of additional care responsibilities among female academics compared to male academics has been observed repeatedly within the context of the COVID-19 pandemic (Deryugina et al., 2021; Myers et al., 2020; Staniscuaski et al., 2021) as well as outside of this context (Morgan et al., 2021). This pattern applies to our sample as well.

Factors associated with the pandemic’s impact

To investigate which factors were associated with the pandemic’s impact on research, we estimate a linear probability model. We regress the binary variable indicating a perceived negative impact on the five sets of explanatory factors (see the “Main variables” section for details) using the full sample. The estimation furthermore considers indicators for the federal states of institutions to control for regional differences. The analysis allows us to identify the influence of each of the factors considered, while accounting for the role of a comprehensive set of other factors. We thus interpret the coefficients as ceteris paribus within the context of the respective model and, given the comprehensiveness of the factors considered, assume that factors with a significant coefficient are relevant for the pandemic’s impact. An advantage of the linear specification lies in the simplicity of interpretation: The coefficients represent the marginal effect on the probability that professors perceived the impact of the pandemic as negative, that is, a percentage point change if multiplied by one hundred. Their sum, taken together with the constant, is the average probability for a specific group of professors.

Our results confirm that socio-demographic characteristics and related consequences of the pandemic play an important role (see column 1 in Table 1). In line with prior studies, we find that professors who had to increase the time spent on care responsibilities are significantly more likely, by 5.2 percentage points, to report that their research activities were negatively affected. Female professors are at an additional disadvantage, even when other factors such as changing time demands for care responsibilities are taken into account. Their probability to report a negative impact is 9.0 percentage points higher than that of their male colleagues.

Table 1 Estimates of linear probability models for a negative impact of the pandemic on research activities for the full sample and by science branch
Full size table

Characteristics of professors’ academic occupation matter as well (see column 1 in Table 1). Whereas the formal career stage (i.e., holding a full professorship) appears to make no difference, longer tenure is associated with a lower probability to report a negative impact. Each additional decade in seniority decreases this probability by 5.2 percentage points.Footnote 3 Furthermore, the dependence on access to research facilities, as measured by the average overall expenditures per professor in each academic discipline, has a strong and systematic influence on the probability to have witnessed a negative impact of the pandemic. The difference in this probability between the lowest and highest dependence amounts to 11.8 percentage points, on average.

Professors’ research engagement does not appear to matter, in contrast to certain facets of the institutional environment (see column 1 in Table 1). We observe no associations for the differential relevance professors ascribe to research and teaching, neither for their actual research engagement. From the conditions of the institutional environment considered, the size of professors’ departments turns out to be relevant (see Appendix Table A.4 for the disciplinary affiliation of professors by department size). Professors located in larger departments are 6.8 percentage points less likely to report a negative impact of the pandemic on their research.

The strongest associations we observe concern the pandemic’s negative impact on different areas of academic activity (see column 1 in Table 1).Footnote 4 Professors who perceived a negative impact on teaching are more than 20 percentage points more likely to also report a negative impact on research. A similar albeit weaker association can be observed for a negative impact on administrative duties.

Differences between academic disciplines

Following up on disciplinary differences, we investigate whether the relevance of the factors of interest varies between disciplines by repeating the previous analysis using subsamples obtained by splitting the sample along disciplinary lines. With a view to retaining sufficient statistical power, we use only two subsamples. These consist of two broad branches of science, namely, the humanities and social sciences and the natural and engineering sciences (see Appendix B for the full list of disciplines). There are, of course, notable differences within these two groups, such as differences in research methodologies, which should be relevant for the pandemic’s impact. The groups should nevertheless provide a sufficiently strong contrast that allows us to gain valuable insights, even though at a higher level of aggregation. Descriptive statistics show that this distinction indeed separates disciplines in terms of the pandemic’s impact. Whereas only 63.6% of professors in the humanities and social sciences perceived a negative impact, this share increases to 72.3% in the natural and engineering sciences. For the further analysis, we estimate the linear probability model for each of the two subsamples. We use the same specification as before with one exception: we do not include the variable covering the average expenditures per professor in the academic disciplines, because its variance is markedly reduced by the sample split, but include dummy variables for the 31 academic disciplines to which respondents could assign themselves as controls instead.Footnote 5

The results reveal certain similarities between the two science branches (see columns 2–3 in Table 1). Most of the factors that were not systematically associated with the pandemic’s impact in the initial analysis remain so after the sample split, for instance, the factors related to professors’ research engagement. Furthermore, there appear no substantial differences when it comes to the role of an increase in the time spent on care responsibilities as the effect size remains similar in both subsamples (even though the smaller number of observations in the subsamples leads to larger standard errors). The strong associations between the pandemic’s impact on different academic activities also hold for both science branches.

However, there are also notable differences (see columns 2–3 in Table 1). It turns out that the role of gender is much more pronounced within the humanities and social sciences. The increase in the probability to report a negative impact is stronger for females within this science branch (where it amounts to 12.2 percentage points) and not even statistically significant in the natural and engineering sciences. Likewise, the characteristics of professors’ academic occupation appear more relevant in the humanities and social sciences. This group of professors drives the effect associated with higher seniority observed in the previous analysis, with a 7.4 percentage point decrease associated with an additional decade of holding a professorship. The institutional environment, by contrast, seems to be more influential in the natural and engineering sciences. The effect of department size observed previously is visible only in this group, where working at a large department is associated with a 7.7 percentage point decrease in the probability to report a negative impact.

Differences between genders

Taking into account the role of academic disciplines, we analyze the marked gender differences observed in more detail (see also Appendix Table A.3 for descriptive statistics for the main variables differentiated by gender). For this purpose, we estimate linear probability models for the two science branches further distinguished by professors’ gender in a first step. The estimation results allow us to quantify basic differences in the associations of the factors considered between genders. In a second step, we use Oaxaca–Blinder decompositions (Blinder, 1973; Oaxaca, 1973), which can be considered the generalization of an analysis technique developed by Kitagawa (1955; see also Oaxaca & Sierminska, 2023), of the gender gap in the perception of a negative impact, which we estimate using the subsamples consisting of the two science branches.

The overarching objective of the decompositions is to provide additional insights into how exactly the factors of interest relate to the difference between male and female professors when it comes to the pandemic’s impact. Specifically, they distinguish analytically two separate mechanisms leading to the gender gap. The first mechanism, the so-called differences in characteristics, concerns the influence of differences in the levels of the factors considered. This mechanism thus concerns observable differences in the composition of the two groups in terms of the factors investigated. For instance, female professors may be more likely to report a negative impact because their level of seniority is on average lower, given that higher seniority is associated with a lower probability to report a negative impact. In other words, the first mechanism concerns how the difference between genders would change if male and female professors had the same observed characteristics. The second mechanism, the so-called differences in coefficients, concerns the influence of differences in how the factors translate into the pandemic’s impact. This mechanism thus concerns differences in the associations between the factors investigated and the pandemic’s impact as the outcome. For instance, female professors might be less likely to report a negative impact because they benefit more from seniority than their male colleagues. In other words, the second mechanism concerns how the difference between genders would change if the association between the factors considered and the pandemic’s impact (i.e., the coefficients) were the same for male and female professors. The decompositions thus provide additional insights into the mechanisms underlying the differential probability to report a negative impact of the pandemic between genders.

The results of the linear probability models reveal three pronounced differences in the role of the factors considered between genders (see Table 2). First, the relevance of characteristics of the academic occupation observed within the humanities and social sciences is mainly driven by female professors. Whereas an additional decade of seniority is associated with a 14.8 percentage point decrease in the probability to report a negative impact of the pandemic among female professors, the effect is not statistically significant for male professors. Second, the lower probability to perceive a negative impact associated with working at a larger department observed for the natural and engineering sciences is driven by male professors, who have an 8.2 percentage points lower probability to report a negative impact compared to an insignificant coefficient for their female colleagues. Third, we observe an increase of this probability associated with working at larger institutions among female professors in these disciplines, which we did not observe in the previous analysis. The remaining associations, in particular those between the pandemic’s impact on different academic activities, hold, by and large, across genders and disciplines.

Table 2 Estimates of linear probability models for a negative impact of the pandemic on research activities by science branch and gender
Full size table

The results of the decompositions confirm the patterns concerning the intersection between academic disciplines and genders observed previously (see Panel A of Table 3). The difference in the probability to report a negative impact of the pandemic between male and female professors is statistically significant only in the humanities and social sciences. Within this science branch, female professors are 13.1 percentage points more likely to report such an impact. The results furthermore show that differences in the levels of the factors considered are not the main drivers of the difference between genders. Beyond that, the decomposition results provide several detailed insights into the mechanisms behind the difference between male and female professors.

Table 3 Estimates of Oaxaca–Blinder decompositions of gender differences as regards a negative impact of the pandemic on research activities for the full sample and by science branch
Full size table

Even though the differences in characteristics turn out to be less relevant overall, they still point toward noteworthy differences between male and female professors (see Panel B of Table 3). The on average lower levels of seniority of female professors turn out to be a disadvantage as far as the pandemic’s impact is concerned, especially in the humanities and social sciences, where this difference explains 2.2 percentage points of the gender gap. At least in the natural and engineering sciences, the same applies to the fact that female professors are more likely to report a negative impact of the pandemic on administrative activities. That is, the difference between genders would be 1.4 percentage points lower if male professors in these disciplines would be as likely to report this consequence of the pandemic as female professors. The opposite holds true for the pandemic’s negative impact on teaching. In this regard, female professors are actually less likely to report a negative impact. This has the effect that the gender difference within the full sample is 1.5 percentage points lower than it would be otherwise.

The differences in coefficients reveal three noteworthy differences between male and female professors that extend our prior findings (see Panel B of Table 3). First, the decrease in the probability to report a negative impact associated with higher seniority is much stronger for female professors in the humanities and social sciences than their male colleagues. If this was not the case, the gender gap within these disciplines would be 13.9 percentage points larger. Second, prioritizing research over teaching is more strongly associated with a negative impact of the pandemic among female professors. If maintaining such a prioritization would have the same association with reporting a negative impact of the pandemic for male and female professors, the difference between genders observed would be 2.4 percentage points smaller. Third, female professors in the natural and engineering sciences derive particularly strong disadvantages from working at large institutions. For female professors in these disciplines, the increase in the probability to report a negative impact associated with the size of universities is much steeper than for their male colleagues.

Discussion

Despite the challenging circumstances during the early phase of the COVID-19 pandemic, its impact on research productivity varied markedly between academics. Our analysis confirms prior findings that gender, changing time demands for care responsibilities, and seniority were important factors influencing academics’ susceptibility to the disruptions caused. Our analysis furthermore identifies features of the institutional environment and the relation between the pandemic’s impact on different academic activities as relevant factors. Working at a university as part of a larger group of professors in the same discipline turns out to be advantageous as far as the impact of the pandemic is concerned. A possible explanation for this finding is that reacting to disruptions becomes easier when there is access to more resources and possibilities for a division of labor (see also Mayer & Rathmann, 2018). For the size of institutions as a whole, we find an association in the opposite direction, which is also somewhat weaker. This association could be the result of greater difficulties for larger institutions to react quickly to changing circumstances and to adapt the reactions to differing needs. The strongest associations with the pandemic’s impact on research activities in our study appear for the pandemic’s impact on teaching and administrative tasks, which suggests strong mutual reinforcements between those impacts to us. Even though these associations may be partly driven by unobserved factors that shape the pandemic’s overall impact, this finding also attests to the difficulties that emerge when several areas of activity are disrupted at the same time.

Throughout the analysis, academics’ seniority turns out to be a crucial factor. Specifically, it appears that, at least among professors, it is not the formal status, but primarily the time of holding a professorship that matters. We have conducted a series of additional analyses (see Appendix G for details), which ascertain that this finding is not driven exclusively by cohort effects due to which academics with different levels of seniority exhibit basically different characteristics. Among other things, we investigated whether professors’ age group, a better indicator for the cohort to which they belong, features the same association with the perception of the pandemic’s impact as did professors’ seniority, which turned out not to be the case. One explanation for the finding may thus be that there are basic differences in academics’ work portfolios, for instance, a stronger involvement of more senior professors in management activities and services to the academic profession. This would make temporary disruptions of research processes less notable. It may even provide them with more time for research when other activities are disrupted. However, as also suggested by our additional analyses (see Appendix G for details), this mechanism likely does not explain the full associations observed either. We could not detect pronounced differences in, for instance, the relevance professors ascribe to different academic tasks in line with their seniority. It thus seems that academics build up resilience against disruptions of their research over time. This resilience could result from, for instance, acquiring relevant competences and developing routines, having access to well-established networks, developing broader research portfolios that render disruptions of some research strands less relevant, and becoming better at drawing on useful resources, including those provided by the institutional environment.

Despite some similarities, noticeable differences in patterns emerge among academic disciplines. These differences can be attributed to fundamental disparities in the research methodologies employed, especially the resulting reliance upon access to research facilities (see also Myers et al., 2020; Walters et al., 2022). Notably, the greater dependence on such access within the natural and engineering sciences not only resulted in a more pervasive negative impact of the pandemic but also influenced the relevance of the factors we have examined. Due to the greater uniformity, individual characteristics such as gender and seniority hardly matter, on average, among natural and engineering scientists when it comes to perceiving a negative impact or not. This could also explain why features of the institutional environment are more relevant in these disciplines. Here, only coordinated and targeted efforts drawing on a large pool of resources may have been sufficient to actually avert the disruptions caused by the pandemic. In the humanities and social sciences, by contrast, individual characteristics do matter. Apparently, a less straightforward connection between the pandemic and research processes, coupled with a more pronounced variability in the pandemic’s impact, created a greater opportunity for the advantages or disadvantages associated with different genders and levels of seniority to manifest.

Differences between genders are pronounced as well and indicate that female academics experience greater challenges with shielding their research activities from disruptions such as those caused by the pandemic. Even though we also find that the pandemic’s implications for care responsibilities put female academics at a disadvantage (see also Deryugina et al., 2021; Myers et al., 2020; Staniscuaski et al., 2021; Walters et al., 2022), our findings clearly show that the pronounced gender differences in the pandemic’s impact cannot be reduced to this mechanism (see also Rusconi et al., 2020; Sawert & Keil, 2021). A first pertinent finding in this regard is that prioritizing research over teaching seems to be more difficult for female academics than their male colleagues. Among other reasons, this could owe to a comparatively stronger emphasis on and engagement in teaching and service activities (see also Bam et al., 2023; Guarino & Borden, 2017; King & Frederickson, 2021; Madsen et al., 2022; Peetz et al., 2022; Squazzoni et al., 2021; Walters et al., 2022). The finding that female academics were actually less likely to report a negative impact on teaching activities also points in this direction. It furthermore appears that female academics benefit more from higher levels of seniority (see also Squazzoni et al., 2021), especially in the humanities and social sciences.

Differential cohort effects by gender are likely not the main driver of our findings concerning seniority. To substantiate this conclusion, we conducted a complementary analysis imposing a median split based on professors’ seniority for the full sample and assigned professors to four groups based on the median split and their gender (see Appendix G for details). We then investigated whether the four groups of professors differ in their attitudes toward certain academic activities and in their engagement with actors from outside of academia. The results of these analyses provide no indication for differential cohort effects by gender. Part of the explanation for the finding that female professors benefit more from seniority could thus be that it takes more time for female academics to build up a diversified, resilient research portfolio, for instance, due to greater demands related to non-research activities at earlier stages of the academic career (see also Guarino & Borden, 2017). Another part of the explanation could be that it also takes more time for them to obtain access to resources that can be leveraged to mitigate disruptions, for instance, due to an overall weaker integration into departments (Fox, 2010; Kyvik & Teigen, 1996) or other types of collaboration networks (Larivière et al., 2011). In addition, the more intricate challenges female academics face with prioritizing research activities might render having access to relevant resources more beneficial. This may also explain why female academics, especially those in the natural and engineering sciences, are more susceptible to the downsides of larger institutions and their potentially less discriminate policies (see also Expert Group on the COVID-19 Impact on Gender Equality in R&I, 2023). All in all, it thus appears that female academics are on average more susceptible to disruptions of their research productivity, but that they can also benefit more from conditions that support resilience against these disruptions.

Taken together, the insights obtained suggest that several factors shape academics’ resilience against disruptions of their research activities and that these factors are interrelated in complex ways. Our explorative approach focused on investigating basic associations for a set of factors identified as potentially relevant based on prior empirical investigations and on interpreting these associations based on comprehensive complementary analyses and the recourse to the pertinent literature. Even though this approach generated robust results, it does not allow us to infer causality (see also Traag & Waltman, 2022). Nevertheless, the model that emerges when combining our findings and interpretations (see Fig. 1) can serve as a valuable starting point for further causal analyses for at least two reasons: First, it highlights factors particularly relevant to investigate. Second, it provides various reference points for specifying models suitable for causal inference.

Fig. 1
figure 1

Factors influencing academics’ resilience against disruptions of their research activities. Notes: The figure provides an overview on our findings concerning the factors that influence academics’ resilience against disruptions of their research activities. The black arrows indicate assumed causal relationships, whereas the gray arrows indicate assumed moderating relationships

Full size image

Three limitations of our study that point to additional avenues for further research are worth noting. First, our measure of the pandemic’s impact on professors’ research activities was restricted to an initial set of three categories, which we furthermore transformed into a binary variable for the empirical analysis. Whereas additional analyses (see Appendix D) show that the dichotomization used for the main analyses does not distort our results, the initial differentiation of merely three categories may have affected the size of the estimated effects as well as the magnitude of group differences (in particular, as far as nuanced differences in the pandemic’s negative impact are concerned). Investigations with a more differentiated measure for disruptions of academics’ research activities could thus yield more detailed insights. Second, our investigation focused on the early phase of the pandemic, even though restrictions affecting academics’ work such as the need to teach at a distance continued afterwards. Extending the investigation to the subsequent periods of the pandemic would thus be relevant to gauge better its long-term consequences. Third, despite the comprehensive set of factors investigated by us, several additional factors may be worth investigating, especially as far as gender differences are concerned. Further research could, for instance, investigate the role of the size and gender composition of academics’ working groups (see Mayer & Rathmann, 2018) and particularities of academics’ collaboration networks (see Cui et al., 2022; Ductor et al., 2023; Rusconi et al., 2020).

Conclusion

Understanding the impact of the circumstances during the COVID-19 pandemic on academics’ research activities requires dealing with great complexity. Our research has revealed that multiple factors played a pivotal role in determining whether academics could (or could not) mitigate significant disruptions of their research productivity. What is more, the picture changes markedly depending on which groups of academics are considered. Differences among academic disciplines and gender, as well as the interplay between these factors, turn out to be crucial in this regard. This does not only concern the prevalence of the pandemic’s negative impact but also the relevance of factors shaping this impact. In terms of academic disciplines, socio-demographic characteristics and characteristics of academics’ occupation are particular relevant factors in the humanities and social sciences, whereas features of the institutional environment matter more in the natural and engineering sciences. These factors are furthermore, by and large, more relevant among female academics, who faced greater challenges than their male colleagues, which can by no means be reduced to differential demands associated with care responsibilities. Instead, our findings suggest that female academics are generally less likely to be able to effectively shield their research productivity from disruptions.

Our empirical findings also provide points of reference for debates about research productivity differences in academia more generally. As extreme as the situation of the pandemic was, unexpected disruptions of research processes are a common feature of academics’ work. Consequently, our findings concerning academics’ differential susceptibility to disruptions have implications of broader relevance. A first noteworthy finding in this regard is that the factors associated with academics’ susceptibility to disruptions are not uniform across disciplines, but apparently depend on the particularities of the underlying research processes such as the relevance of access to research facilities. A second, particularly relevant finding concerns the gender differences observed. Whether it is greater challenges with developing diversified research portfolios, with building large and strong networks, with obtaining access to useful resources, or with maintaining a prioritization of research activities, there are systematic gender differences in academics’ susceptibility to disruptions, on average, worth considering with regard to research productivity differences.

Our findings furthermore have implications for policy and management efforts to mitigate the inequalities caused by the pandemic and to address research productivity differences more generally. A crucial precondition for efforts to promote equality in relation to the pandemic (see, for instance, King & Frederickson, 2021; Oleschuk, 2020) and outside of this context is to adapt these efforts to different levels of disadvantages. Given the complexity unveiled by our analysis, this requires careful consideration in each case of relevant differences among, for instance, disciplines, levels of seniority, and genders. As regards strategies for preventing that inequalities arise in the first place, individual conditions such as diversified research portfolios and features of the institutional environment such as policies tailored to diverse needs appear promising levers for reducing imbalances in academics’ susceptibility to disruptions of their research productivity. Efforts in this direction would be particularly relevant for addressing differences among male and female academics, which still appear to be pronounced in this area.