Introduction

Since the beginning of the 21st century, the scale of doctoral education has steadily expanded (Gruzdev et al., 2020), becoming an increasingly important driver of social and economic development (Dericks et al., 2019). This expansion, coupled with the burgeoning of gender equality advocacy, has contributed to a rising proportion of female doctoral students worldwide. For instance, in the United States, women have outnumbered men in obtaining doctoral degrees for nine consecutive years since 2009. In 2017, a total of 78,778 doctoral degrees were awarded, with women accounting for 53% of the total (Okahana and Zhou, 2018). Similarly, according to data from China’s Ministry of Education, the proportion of female doctoral students has steadily increased over the years, reaching 39.27%, 40.37%, and 41.32% in 2017Footnote 1, 2018Footnote 2, and 2019Footnote 3, respectively. Furthermore, since 2000, the annual growth rate of female doctoral enrolment in the African region (excluding South Africa) has surpassed that of males, with rates of 20.2% and 16.9%, respectively (Cloete et al., 2015). Despite this growing participation, female doctoral students face many challenges during their studies compared to male doctoral students (Sverdlik et al., 2018; Quynh Phan, 2023; Wollast et al., 2023), which, to some extent, affects their satisfaction with pursuing a doctoral degree (Ching et al., 2021).

In higher education, doctoral satisfaction refers to the subjective perceptions of doctoral students regarding whether their doctoral experiences meet their personal expectations (Kahu, 2013). Dissatisfaction among doctoral students can have detrimental effects, both on individuals and on academia. Firstly, satisfaction level influences students’ academic productivity and mental health (Pyhältö et al., 2015). Female doctoral students, in particular, are more likely to experience mental health difficulties compared to their male counterparts (Gaete Sepúlveda et al., 2025), highlighting the need for a more nuanced understanding of their doctoral experiences. Indeed, in a gendered and unequal academic and research environment, women in academia are often the most vulnerable to adverse effects (Santos et al., 2021). Secondly, satisfaction positively influences doctoral students’ academic motivation and retention (Litalien et al., 2024; Dericks et al., 2019), with higher satisfaction being associated with a greater likelihood of successful completion. Moreover, given the crucial role the doctoral stage plays in academic career socialisation, doctoral students’ satisfaction also influences their career trajectories. Generally, those dissatisfied with their experiences are more likely to pursue non-academic careers (Gu et al., 2018), which can have adverse effects on the academic community as a whole.

Given the importance of doctoral students’ satisfaction, extensive research has been conducted on factors influencing satisfaction during doctoral studies, including supervision, academic engagement, interactions with faculty, and financial aid (Gruzdev et al., 2020; Shin et al., 2018; Van Der Haert et al., 2014). However, systematic research on gender differences in doctoral satisfaction remains limited. While some studies address gender, they typically focus on broad trends or challenges faced in male-dominated fields, such as academic marginalisation and gender bias (Kong et al., 2020; Tomassini, 2021; Hakiem, 2023), offering limited insights into the specific satisfaction experiences of female doctoral students. Although factors such as gender bias, role conflict, and family responsibilities have been identified as contributors to lower satisfaction among female doctoral students (Haynes et al., 2012; Humphries et al., 2021; Mbonyiryivuze et al., 2023), the underlying mechanisms leading to such dissatisfaction remain insufficiently explored. Furthermore, while existing research has offered valuable insights, there is a lack of rigorous empirical studies. The majority of the existing research relies on case studies or small samples, limiting the generalisability of findings. To address this, large-scale survey-based research is imperative to enhance the reliability and validity of the conclusions.

This study aims to address the existing research gaps by investigating female doctoral students’ satisfaction with their doctoral experiences. Using data from the 2019 Nature Global Doctoral Student Survey, this study empirically examines whether female doctoral students report lower satisfaction levels. If this is the case, it will identify the key mechanisms underlying this phenomenon. By filling these gaps, the study will offer a deeper understanding of gender dynamics in doctoral satisfaction and provide both theoretical insights and practical recommendations for improving the educational and developmental environment for female doctoral students. The findings are expected to offer valuable guidance to educational administrators and policymakers seeking to enhance the quality of doctoral education and better meet the diverse needs of doctoral students.

This study makes several key contributions. In terms of content, it addresses an underexplored area by focusing specifically on female doctoral students, filling a critical research gap. The empirical analysis examines gender differences in doctoral satisfaction, enhancing our understanding of gendered experiences in doctoral education and offering a fresh perspective on the issue. Furthermore, by exploring the heterogeneity of satisfaction among female doctoral students, this study refines existing literature and provides a foundation for more targeted policy interventions. Theoretically, this study extends the JD-R (Job Demands-Resources) model to doctoral education, empirically assessing its applicability in this context, contributing a novel theoretical framework for understanding the doctoral student experience and expanding the use of this theory in educational research. From a practical standpoint, the findings offer valuable insights for doctoral education policy and practice, particularly in fostering a more inclusive and supportive environment for female doctoral students. The study underscores the urgency of implementing stronger policies against gender discrimination and sexual harassment. Additionally, universities should provide faculty and supervisor training, enhance practical support for work-life balance, and adopt measures to manage workload expectations more effectively.

The paper is structured in the following manner. In the subsequent section, the relevant literature review will be presented. Immediately following this is the empirical design, empirical and mechanism analyses. Finally, the conclusion and policy implications are put forward in the final section.

Literature review

Satisfaction and academic performance among female PhD students

According to Self-Determination Theory (Ryan and Deci, 2000) and Social Support Theory (Cohen and Wills, 1985), there is a positive correlation between doctoral students’ satisfaction with their academic experience and their academic performance. Specifically, when students are satisfied with their doctoral experiences, their intrinsic motivation is significantly enhanced, which, in turn, contributes to improvements in their academic performance. The Social Support Theory further emphasises that various forms of societal support—including supervisor guidance—enable students to cope more effectively with academic stress, strengthening their self-efficacy, and ultimately improving their academic performance (Wilcox et al., 2005).

Academic performance is generally defined as students’ achievements in academic domains (Richardson et al., 2012) and is often measured through quantitative indicators such as grade point average (GPA), research accomplishments (e.g., publications), and participation in academic activities. At the doctoral level, academic performance is more closely aligned with research outcomes, commonly evaluated through retention rates, the quantity and quality of papers published in high-impact journals, and the completion of high-quality dissertations. Previous research indicates that doctoral student satisfaction contributes to students’ retention in academic programmes (Dericks et al., 2019). Moreover, research has shown that female doctoral students’ satisfaction is positively associated with their academic performance. Not only are the satisfaction levels of female PhD students lower than those of male students, but their publication output and citation rates are also generally inferior to those of their male counterparts (García-Aracil, 2009; Sarrico and Rosa, 2014; Corsini et al., 2022). Furthermore, significant gender disparities persist in the scientific innovativeness of doctoral dissertations. Female students are less likely to publish groundbreaking papers, and their dissertations tend to exhibit lower levels of innovation and interdisciplinary research than those of their male peers (Liu et al., 2024; Liu et al., 2023). Consequently, we put forward Hypothesis 1 as follows:

Hypothesis 1. Female doctoral students have lower satisfaction with their doctoral experiences than male doctoral students.

To enhance the academic performance of female doctoral students, it is of utmost importance to investigate their satisfaction with their doctoral experience, identify the factors influencing their satisfaction, and create an inclusive and equitable academic environment that can enhance their satisfaction.

Factors contributing to satisfaction and dissatisfaction with doctoral experiences among female PhD students

Academic factors and female PhD students’ satisfaction

In academia, supervisory support plays a pivotal role in the satisfaction of female students. Supervisors who provide supportive, approachable, and constructive guidance tend to positively influence female PhDs’ academic experiences. In contrast, inadequate or absent supervision can lead to dissatisfaction. Gender bias also presents significant challenges, particularly in male-dominated fields such as STEM (Paksi, 2015; Kong et al., 2020), where women often struggle to adapt to a male-dominated culture and face marginalisation. Furthermore, supervisory relationships represent a key area where gender differences become apparent. Leonard (2001) noted that societal stereotypes led professors to view female doctoral students as less academically capable than their male counterparts, a perception that can contribute to gender discrimination. Besides, O’Connor et al. (2020) found that male doctoral students are more likely than their female counterparts to receive guidance and support from supervisors or senior male academics. This suggests that professors often harbour preconceived notions that female PhD students are inferior to their male peers, simply because of social stereotypes and prejudices, which can be regarded as a form of gender discrimination.

In addition, female doctoral students often struggle with social integration within informal networks, which are vital for both academic and emotional support. Mason et al. (2009) argued that academic systems have historically been structured around “masculine” ideals and lifestyles, which conflict with traits such as “passivity”, “dependence” and “family care”, often associated with women (Krefting, 2003). This mismatch between gendered expectations and academic norms can significantly hinder the academic progress of female doctoral students and thus affect their overall satisfaction. Langin (2018) observed that female PhD students had greater difficulty in building connections with other scholars and are more vulnerable to harassment. Some studies have identified that female graduate students experience significantly higher rates of sexual harassment compared to male students (Sapiro and Campbell, 2018; Karami et al., 2020), often as a result of the gender imbalance in senior academic positions (Rosenthal et al., 2016). Consequently, we put forward Hypotheses 2 and 3 as follows:

Hypothesis 2. In the process of pursuing a doctoral degree, female doctoral students are more prone to encountering gender discrimination, which consequently results in a lower level of satisfaction.

Hypothesis 3. In the course of pursuing a doctoral degree, female doctoral students are more liable to experience sexual harassment, which subsequently leads to a lower level of satisfaction.

Family-related factors and female PhD students’ satisfaction

During their doctoral training, female doctoral students come across diverse challenges, some of which are unique to their gender and not commonly encountered by their male counterparts. Work-life balance is a critical factor influencing the dissatisfaction of female doctoral students. Married female PhDs, in particular, are confronted with the “double burden” of academic responsibilities and household chores. Brown and Watson (2010) identify balancing family and academic commitments as a major source of stress, especially for women who are endeavouring to juggle the dual roles of wife and mother while striving to make progress in their academic pursuits. This enduring challenge continues to affect women’s academic careers, with many female PhDs experiencing compounded difficulties due to family-related responsibilities (Szalma et al., 2020).

The dissatisfaction is further intensified by reproductive demands, as many female doctoral students find themselves simultaneously managing marriage, childbirth, and academic research. These overlapping responsibilities disrupt academic engagement more significantly for women than for their male counterparts (Mbonyiryivuze et al., 2023). Moreover, societal norms often impose additional expectations on female doctoral students, such as raising children and overseeing household chores. Cultural expectations regarding female passivity and prioritisation of family responsibilities amplify this pressure, thereby creating tension between social relationships and academic performance (Carter et al., 2013). These obligations drain time and energy, leaving many women struggling to balance competing priorities and, consequently, dissatisfied with their doctoral experience (Gür and Bozgöz, 2023). Hence, we put forward Hypothesis 4 as follows:

Hypothesis 4. Due to gender stereotypes, women have to shoulder more family responsibilities. Therefore, female doctoral students find it more difficult to balance work and life during their studies, which can lead to lower satisfaction.

In conclusion, drawing upon the analyses presented in Sections “Academic factors and female PhD students’satisfaction” and “Family-related factorsand female PhD students’satisfaction”, it is our contention that gender discrimination, sexual harassment and the work-life balance are highly likely to be the factors contributing to the relatively low satisfaction among female doctoral students.

Empirical design

Theoretical framework

Demerouti et al. (2001) proposed the JD-R (Job Demands-Resources) model, which emphasises that job demands and job resources are central elements of work that impact an individual’s physical and mental health as well as their behaviour patterns. Job demands refer to the various physical, psychological, and social requirements of work that require sustained physical, cognitive, and emotional effort. These negative factors deplete an individual’s energy, often resulting in exhaustion. In contrast, job resources are regarded as supportive elements that assist individuals in achieving work objectives, alleviating job demands, and facilitating growth and development (Bakker and Demerouti, 2007).

A core assumption of the JD-R model is the dual-path hypothesis, which posits that work can have both depleting and enhancing effects on employees (Demerouti et al., 2001; Lewig et al., 2007; Schaufeli et al., 2009). The first path, the depletion process, occurs when excessive job demands coupled with insufficient job resources give rise to negative outcomes. When job demands (negative factors) persist at high levels without sufficient job resources (positive factors), an individual’s energy is drained, negatively affecting both the individual and the organisation. The second path, the motivation process, is triggered by ample job resources, leading to high levels of work engagement and positive outcomes.

According to the depletion pathway of the JD-R model, job demands drain an individual’s physiological and psychological resources (Demerouti et al., 2001; Bakker and Demerouti, 2007), negatively impacting their emotional well-being. Workload is a typical example of a job demand, referring to the intensity of labour required, often measured quantitatively, and is a significant predictor of job satisfaction (Weigl et al., 2016). Although doctoral students are not formal employees, their academic responsibilities, such as literature reviews, experimental work, and data analysis, constitute their ‘workload’. Like employees, doctoral students often encounter academic tasks that exceed their capabilities, leading to excessive demands (Sang et al., 2015).

Academic work does not adhere to the traditional nine-to-five model (Sang et al., 2015); rather, it requires long-term commitment and substantial task burdens. Cha and Weeden (2014), in their study on overwork and gender differences, employed working hours of greater than or equal to 50 h as an indicator of overwork. Building on existing research, the present study defines overwork based on the time doctoral students spend over 50 h on coursework. According to the 2019 Nature Global Doctoral Survey, 50% of doctoral students work more than 50 h a week. In practice, female doctoral students often struggle to balance their studies with personal life (Fuhrmann et al., 2011). They usually work in a demanding academic environment, where leaving the lab before sunset is often discouraged, making work-life balance even more challenging (Evans et al., 2018). This balance is vital for well-being and directly affects academic satisfaction. When female doctoral students fail to manage this balance effectively, it leads to increased stress and dissatisfaction with their academic experience. Excessive work demands (e.g., discrimination, harassment, work-life imbalance) and a lack of supportive resources (e.g., adequate supervision or social support) can diminish motivation and engagement, ultimately lowering their satisfaction with doctoral education.

Empirical studies centred around the JD-R model have predominantly concentrated on job satisfaction, and the subjects typically encompassed employees, teachers, as well as healthcare workers. Future research endeavours ought to investigate the model’s applicability within diverse sectors and among different groups, thereby further refining its theoretical underpinnings.

Sources of data

This study utilises data from the 2019 Global Doctoral Student Survey from NatureFootnote 4, one of the world’s oldest and most prestigious scientific journals. The survey encompasses a wide range of doctoral students’ personal background, training process, study satisfaction, learning gains, career development, etc. Given the potential existence of cross-cultural and regional disparities, the Nature team carried out exploratory interviews to ensure the consistency of the questionnaire questions among respondents from different cultures before formulating the official questionnaire and commencing the survey. In addition to English, the survey was translated into Chinese, Portuguese, Spanish, and French, with the aim of enhancing the representativeness of the sample. The total valid sample of the survey was 6812. Specifically, it comprised 1857 (27.26%) respondents from Asia, 214 (3.14%) from Australia, 2487 (36.51%) from Europe, 165 (2.42%) from Africa, 1866 (27.39%) from North or Central America, and 223 (3.28%) from South America. After cleaning missing values and outliers, the current study obtained a total sample of 6372 doctoral students from 108 countries. Among this sample, 50.5% were female (including transgender individuals), and 49.5% were male (including transgender individuals). Overall, the sample is extensive and well-represented.

Variable selection

Explained variables

In this research, satisfaction during the doctoral programme serves as the explanatory variable. Drawing inspiration from Douglas et al. (2014) and Dericks et al. (2019), student satisfaction herein is regarded as a general perception regarding the overall educational experience. Additionally, due to the heterogeneity of individuals, the initial satisfaction levels of doctoral students upon commencing the programme tend to vary. Considering this issue, the present study takes into account the changes in satisfaction during doctoral students’ enrolment by referring to Naumann et al., (2022) approach and constructing the indicator of “satisfaction decline”. Therefore, the explained variables of this study were based on the questions in the questionnaire: “How satisfied are you with your PhD experience?” and “Since the very start of your graduate school experience, would you say your level of satisfaction has improved or worsened”. The first question adopts a 7-point Likert scale, where values ranging from 1 to 7 signify from “not at all satisfied” to “very satisfied”. For the second question, a value of 1 is assigned if the respondent selects “worsened”, and 0 is assigned otherwise.

Core explanatory variables

Referring to Turner’s (2023), Asmar’s (1999) and Goldin’s (2014) research design, the use of gender as an explanatory variable is a prevalent approach in existing studies. In this paper, the core explanatory variable is the gender of the respondents. It was operationalised as a dichotomous variable, which was determined based on the doctoral students’ responses to the gender question within the questionnaire. Specifically, if the response is Female (including trans female), it is assigned a value of 1, whereas if the response is Male (including trans male), it is assigned a value of 0.

Control variables

It has been acknowledged that multiple factors, including individual characteristics and university support, can exert an influence on doctoral students’ satisfaction (Barnes, Randall (2012); Shin et al., 2018). In light of this, with the aim of mitigating omitted variable bias, and by drawing on the existing empirical evidence based on this data, certain personal circumstances were selected as control variables in this study (Li and Horta, 2022). These personal circumstances include age, whether or not the students are undertaking a doctoral degree in their home country, the time dedicated to the course, part-time study status, academic career intention, the weekly time spent communicating with their supervisor, the relationship with the supervisor, the availability of funding, as well as the organisation’s support and attitude towards their career. The definitions and descriptive statistics of the above key variables are shown in Table 1.

Table 1 Variable definitions and descriptive statistics.
Full size table

Empirical modelling

Since the dependent variable OS is an ordered categorical variable, we use an Ordered Probit (Oprobit) model for empirical analysis. The specific model is shown in Eqs. (1) and (2):

$${S}_{{ij}}^{* }=\alpha +\beta {{Gender}}_{{ij}}+\gamma {{\bf{Control}}}_{{\bf{ij}}}+{\delta }_{j}+{\varepsilon }_{{ij}}$$
(1)
$${S}_{{ij}}=F\left({S}_{{ij}}^{* }\right)=\left\{\begin{array}{l}1\quad{if}\quad{{S}_{{ij}}^{* }\le C}_{1}\\ 2\quad{if}\quad{C}_{1}\, < \,{{S}_{{ij}}^{* }\le C}_{2}\\ 3\quad{if}\quad{C}_{2} < {{S}_{{ij}}^{* }\le C}_{3}\\ \qquad\qquad\ldots \ldots \\ 6\quad{if}\quad{C}_{5} \,< \,{{S}_{{ij}}^{* }\le C}_{6}\\ 7\quad{{if}}\quad{C}_{6} < {S}_{{ij}}^{* }\end{array}\right.$$
(2)

where Sij indicates the satisfaction level of Doctoral Student i in Country j, \({S}_{{ij}}^{* }\) indicates the explained variable in the form of an ordered dummy variable. Genderij is the core explanatory variable indicating whether the gender of Doctoral Student i in Countryj is Female (including trans female) or not. c1, c2, c3, … c6 and c7 are the discontinuity points. Controlij is a vector of control variables including individual and doctoral training unit characteristics. α, \(\beta\), and \(\gamma\) are the parameters to be estimated; δj is the country fixed effect and εij is the error term.

In addition, the other dependent variable, SD, is a binary dummy variable, so we use a Probit model for estimation. The model is specified as follows:

$${\rm{Prob}}\left({\rm{SD}}_{{ij}}=1\right)=\Phi \left(\alpha +{{\rm{\beta }}}_{1}{\rm{Gender}}_{{ij}}+{{\beta }}_{2}{{\bf{Control}}}_{{\bf{ij}}}+{\delta }_{j}+{\varepsilon }_{{ij}}\right)$$
(3)

where SDij indicates whether the satisfaction level of Doctoral Student i in Country j has declined. In Eq. (3), the other variables share the same meanings as those in the previously specified Oprobit model.

Empirical results and analysis

Benchmark regression

Table 2 reports the results of the baseline regression on the effect of females on satisfaction with doctoral study. Columns (1) and (2) show that the effect of females on overall satisfaction is negatively significant at the 1% statistical level, regardless of the inclusion of relevant control variables such as career intention, indicating that females exhibit significantly lower levels of satisfaction with their PhDs in comparison to males. Since the estimated coefficients of the Oprobit model possess limited explanatory power, this research primarily concentrates on the direction and significance level associated with females. Further, the results presented in columns (3) and (4), which utilise the average marginal effects reported by the Probit model, demonstrate that females have a statistically significant positive effect on the decline in satisfaction at the 1% level. Specifically, when compared to males, females are 3.88% more likely to experience a decline in satisfaction with their PhD study. These empirical results offer a preliminary response to the concerns addressed in our study.

Table 2 Results of baseline regression.
Full size table

A number of control variables also exert significant influences on doctoral satisfaction. Firstly, the satisfaction of doctoral students who study in their home country tends to decline over time. However, those doctoral students who pursue a doctoral programme out of academic motivation, maintain long-term contact with their supervisors and possess a favourable mentor–student relationship exhibit higher levels of satisfaction with their doctoral studies and do not encounter a decrease in satisfaction as time elapses. In addition, students who want to pursue academic careers after graduation do not experience a decrease in satisfaction with their doctoral studies over time. Doctoral students who have access to funding and are able to receive career training as well as support from their organisations are more satisfied with their doctoral studies. These results are consistent with previous findings of Kulikowski et al. (2019) and Shin et al. (2018). The study also found that age does not affect satisfaction level with doctoral study; members’ attitude toward non-academic career does not affect doctoral students’ satisfaction.

Robustness check

This paper further utilises multiple approaches for robustness tests with the aim of ensuring the accuracy and reliability of the estimation results presented previously.

Propensity score matching (PSM) method

Due to the specificity of the core explanatory variable in this paper, it is not feasible to construct a gender-based instrumental variable for instrumental variables (IV) estimation. Therefore, we primarily employ the propensity score matching (PSM) method for a robustness check. In empirical studies, selective and mixed biases tend to seriously interfere with estimation results. The ideal test for inferring an actual causal relationship between two variables is a randomised experimental method that fully controls for covariates. Mixed bias implies that the treatment and control groups differ not only in the intervening factor but also in other individual differences. As a result, it becomes impossible to determine whether the difference in the variables is caused by the intervention simply by directly comparing the outcomes of the two groups.

PSM addresses the problems of sample selection bias and endogeneity due to observable variables. By matching “similar” samples in a set of treatment and control groups and estimating the average treatment effects on treated (ATT) of the treatment groups, the net effect of the core explanatory variables can finally be obtained. In this study, female (including trans females) was designated as the treatment group, while male (including trans males) was set as the control group. Moreover, three matching methods were employed, namely nearest-neighbour matching (with n = 4), calliper matching (where calliper = 0.01), and kernel matching.

The results of the balance test in Table 3 show that the standardised deviations of the variables Age, Motivation, Career Support, and Career-promoting were all significantly greater than 10% before the sample was matched, and were significant at least at the 5% statistical level. This indicates that some of the unmatched variables exhibited significant differences between the treatment and control groups. Consequently, the results obtained would lack credibility if a direct comparative analysis were conducted. After matching, the standardised deviations of the majority of the covariates were substantially reduced, with %bias generally less than 5%, and t-tests for all variables were not significant. The above results indicate that PSM effectively reduces the systematic differences between the two groups of samples, particularly the differences in individual characteristics. This conforms to the requirements of randomised experiments, thus making the results of data analysis subsequent to the matching process more dependable.

Table 3 Results of the balance test.
Full size table

By using the PSM method, the results presented in Table 4 reveal that, regardless of whether nearest neighbour matching, caliper matching, or kernel matching is utilised, the ATT for all three matching methods pass the significance test, thereby verifying the robustness of the conclusions drawn previously. All results are significant at the 1% statistical level, with the exception of the result of nearest neighbour matching for doctoral satisfaction, which is significant at the 5% statistical level. The results of the PSM show that female doctors are 0.109 less satisfied and have a 4.3% higher probability of experiencing a decline in satisfaction in comparison to their male counterparts.

Table 4 Results of the PSM method.
Full size table

Variable adjustment and trimming

Given that the explained variable in this research, namely the satisfaction with the doctoral experience, constitutes an ordered variable spanning from 1 to 7, we implement trimming procedures thereon to mitigate the impact of outliers. The estimation results presented in Table 5 demonstrate that regardless of whether we trim the top and bottom 1% as depicted in column (1), or the top and bottom 3% as shown in column (2), the estimated results are in line with the benchmark regression and remain significantly negative.

Table 5 Robustness test.
Full size table

Moreover, we have constructed a novel variable, OS1. In this regard, individuals who possess a satisfaction level ranging from 5 to 7 regarding their doctoral experience are deemed to be relatively satisfied and are thus assigned a value of 1. Conversely, those with a satisfaction level falling between 1 and 4 are considered to be less satisfied and are accordingly assigned a value of 0. The estimation results of the Probit model in column (3) of Table 5 suggest that, in comparison to men, women exhibit a 2.72% higher probability of being dissatisfied with their doctoral experience. This finding once again corroborates the robustness of the benchmark regression within this paper.

Mechanism analysis

Mechanism testing

In order to validate the research inference of the three influence mechanisms presented in Sections “Academic factors and female PhD students’satisfaction” and “Family-related factorsand female PhD students’satisfaction” (gender discrimination; sexual harassment; work-life balance), this research employs the Probit and Oprobit models for mechanism testing. Since there is a potential endogeneity problem between the core explanatory variables and the explained variables, this study further adopts the PSM method to overcome this problem.

Firstly, the estimation results presented in column (1) of Table 6 show that the effect of females on gender discrimination is positively significant at the 1% statistical level and the average marginal effect is 0.1235. This indicates that females are 12.35% more likely to encounter gender discrimination compared to males. The estimation results based on the three matching methods of PSM in Table 7 show that the t-values are all significantly greater than the critical point of 2.58 at the 1% statistical level, so the effect of females on gender discrimination remains significant at the 1% level. Meanwhile, the results of ATT show that, after effectively reducing the standardisation bias of the samples in the treatment group and its corresponding control group, the probability of females being subjected to gender discrimination is still 11.9% higher than that of males. This corroborates the research findings of Leonard (2001) and O’Connor et al. (2020). Owing to social stereotypes and biases, there prevails a preconceived gender discrimination against female doctoral students within the academic community, which results in their lower satisfaction.

Table 6 Mechanism testing results: probit and OProbit models.
Full size table
Table 7 Mechanism testing results: propensity score matching method.
Full size table

Secondly, the estimation results in column (2) of Table 6 show that females are 6.81% more likely to experience sexual harassment than males. The estimation results in Table 6, which utilise the three matching methods, show that the effect of females experiencing sexual harassment is still positively significant at the 1% level. The ATT results show that, after matching the samples of the treatment and control groups that are “similar” on the other factors, the probability of experiencing sexual harassment is still 5.5% higher for females in the treatment group than for males in the control group. This is consistent with the research of Sapiro and Campbell (2018) as well as Karami et al. (2020), suggesting that female doctoral students are indeed more liable to experiencing sexual harassment.

Finally, the estimation results in column (3) of Table 6 show that the effect of females on work-life balance is negatively significant at the 1% level 1, with a regression coefficient of −0.0678. This implies that females face greater difficulties in balancing work and life. As for the estimation results reported by the PSM in Table 7, while the results using kernel matching are not significant, the remaining two matching methods, namely nearest-neighbour and calliper, are both statistically significant at the 5% level, and the direction of the effect is negative. Therefore, the estimation results of PSM still prove the robustness of the estimation results in the third column of Table 6.

Undoubtedly, experiencing gender discrimination, suffering from sexual harassment, and not being able to balance work and life during the course of doctoral studies significantly reduces an individual’s satisfaction level (Rosenthal et al., 2016; Leonard, 2001; Gür and Bozgöz, 2023). In conclusion, in comparison with males, females are more likely to experience gender discrimination and sexual harassment and tend to struggle with work-life balance during their doctoral studies. As a consequence, this results in relatively lower satisfaction levels among female doctoral students.

Heterogeneity analysis

Upon clarifying the influencing mechanism, this paper will conduct further research and aims to carry out a heterogeneity analysis with the intention of identifying “which groups of female doctoral students are more affected by the mechanism and, consequently, exhibit lower levels of satisfaction”. Drawing on existing research designs and in conjunction with the survey data from Nature, this paper centres on two aspects, namely overwork and country.

Overwork is generally defined as a longer period of overtime and over-intensive employment. Table 8 reports the corresponding heterogeneity test estimates. In both the non-overworked and overworked groups, the effect of females’ satisfaction with doctoral studies is negatively significant at the 1% statistical level, and the effect of the decline in females’ satisfaction is also positively significant at the 1% statistical level. This, to some extent, reaffirms the accuracy of the benchmark regression. By comparing the corresponding estimated coefficients and the average marginal effects, it can be observed that the satisfaction level of overworked female PhDs is significantly lower than that of non-overworked female PhDs as well as male PhDs within the overworked group. Furthermore, overworked female PhDs have a higher probability of experiencing a decline in satisfaction. This may be due to the fact that overworked women allocate more time to their studies, encounter greater difficulties in balancing work and life, and consequently exhibit a lower level of satisfaction during their doctoral studies. Currently, female doctoral students still predominantly assume the responsibilities of family and child-rearing. The time and energy devoted to household chores will, to some extent, influence women’s perception of their careers (Mayrhofer et al., 2008). This conclusion further extends Paksi’s (2015) finding that not only female PhDs in STEM disciplines but also all female PhDs are confronted with work-life balance challenges.

Table 8 Heterogeneity analysis: categorised by overwork.
Full size table

The enhancement of the education system constitutes a long-term process, within which the economic development level of a country plays a pivotal role. It not only exerts an impact on the hardware support for the doctoral study environment (such as funding and resources), but also has a profound influence on the learning and research experiences of doctoral students through soft factors (including culture and career development). In accordance with the World Bank’s classification of countries based on their economic development levels, we categorise the countries from which the doctoral student samples are sourced into four groups: low-income countries, lower-middle-income countries, upper-middle-income countries, and high-income countries. Specifically, there are 41 high-income countries, 26 upper-middle-income countries, 30 lower-middle-income countries, and 11 low-income countries.

Table 9 presents the estimation results across different country groups. Owing to the relatively small sample size of low-income countries, they are combined with lower-middle-income countries. After conducting the group regression, the regression results in columns (1)–(6) are all significant at least at the 10% statistical level. By comparing the regression coefficients and marginal effects, it can be observed that female doctoral students in lower-middle-income and low-income countries display significantly lower satisfaction with their doctoral studies and a higher likelihood of experiencing a decline in satisfaction. This empirical result can be accounted for from the perspective of the influence mechanism mentioned previously. In lower-middle-income and low-income countries, traditional gender concepts remain prevalent, the progress towards gender equality is sluggish, and the protective mechanisms against gender discrimination or sexual harassment are yet to reach maturity.

Table 9 Heterogeneity analysis: categorised by national income.
Full size table

Conclusions and discussion

Significant relationship between gender and satisfaction

This study, drawing upon the data from the 2019 Nature Global Doctoral Student Survey, undertakes an in-depth exploration into the satisfaction of female doctoral students regarding their doctoral experiences. Through a series of empirical analyses, the following principal conclusions are derived:

Female doctoral students exhibit significantly lower satisfaction during their doctoral studies in comparison to their male counterparts and are more prone to experiencing a decline in satisfaction. This finding is corroborated in the benchmark regression as well as in various robustness tests, such as the PSM method, thereby thoroughly demonstrating the reliability of this conclusion.

In terms of the mechanism, female doctoral students are 12.3% more likely to encounter gender discrimination than male doctoral students. This exposes them to greater amounts of unfair treatment within the academic environment, consequently impacting their satisfaction levels. Besides, female doctoral students are 6.81% more likely to endure sexual harassment than male doctoral students, and this adverse experience exerts a negative influence on their doctoral study experiences. Moreover, they face more substantial difficulties in balancing their studies and personal lives, which leads to lower satisfaction levels.

As for heterogeneity analysis, firstly, the impact of workload variations on the satisfaction of female doctoral students is significant. Female doctoral students who are overworked (working more than 50 h per week) display lower satisfaction, and this situation becomes even more conspicuous in the state of severe overwork, i.e., working more than 60 h per week. Female doctoral students, when confronted with high-intensity academic pressure, find it more arduous to balance work and life in the same way as male doctoral students, thereby affecting their satisfaction with the doctoral experience. Additionally, there are disparities in the satisfaction of female doctoral students across countries with varying levels of economic development. In low- and middle-income countries, the satisfaction of female doctoral students is significantly lower than that in high-income countries. This might be associated with the prevalence of strong traditional gender concepts, the sluggish progress in gender equality, and the absence of a comprehensive protection mechanism against gender discrimination and sexual harassment in these countries.

Discussion

Reinforcement of existing frameworks and theories

Drawing upon the survey data of 6372 doctoral students from 108 countries and via a relatively rigorous empirical design, we have furnished robust empirical evidence in support of the existing hypotheses. The findings of this study align with existing frameworks in higher education, reinforcing the notion that universities continue to function as both gendered and gendering institutions (Rosa et al., 2020). While reaffirming established conclusions, this study also offers new insights into gender disparities in doctoral education. Firstly, our results are consistent with prior research (Leonard, 2001; Rosenthal et al., 2016; Haynes et al., 2012; Humphries et al., 2021; Mbonyiryivuze et al., 2023), which suggests that female doctoral students generally report lower satisfaction than their male counterparts. Furthermore, our analysis highlights that female doctoral students face heightened risks of gender discrimination, sexual harassment, and difficulties in achieving work-life balance—findings that corroborate prior research (e.g., Sapiro and Campbell, 2018; Karami et al., 2020). However, this study advances the literature by systematically examining the heterogeneity within the female doctoral student population. Specifically, our findings indicate that female doctoral students who experience excessive workloads, as well as those from lower-middle-income and low-income countries, report significantly lower satisfaction levels. By integrating factors such as labour intensity and socioeconomic status into the analysis, this study deepens the discourse on gendered experiences in doctoral education, providing a more nuanced understanding of the structural and contextual factors that shape female doctoral students’ academic trajectories.

At the theoretical level, this study not only supports the depletion pathway of the JD-R model but also expands its application to doctoral education. While the JD-R model has traditionally been employed in workplace settings (Demerouti et al., 2001; Lewig et al., 2007; Schaufeli et al., 2009), our findings demonstrate its relevance in academic environments. In particular, major job demands such as gender discrimination, sexual harassment, and work-life balance challenges deplete female doctoral students’ resources, heightening their susceptibility to adverse academic and psychological outcomes and ultimately diminishing their overall satisfaction. Although our results generally align with the JD-R model, they also highlight the need to refine its assumptions. The traditional JD-R framework posits that similar demands exert uniform effects across individuals (Demerouti et al., 2001); however, our findings reveal significant gender differences. Female doctoral students encounter distinct forms of discrimination, harassment, and work-life balance struggles that may not affect male students to the same extent. These findings suggest that the JD-R model could be further developed to incorporate gender-specific variations.

Implications for doctoral education practice and policy

The findings of this study have important implications for doctoral education policy and practice, particularly in fostering a more inclusive and supportive academic environment for female doctoral students. First, universities must address gender-related challenges by implementing strong policies to prevent discrimination and harassment, ensuring a safe and equitable academic space. The persistent prevalence of these issues underscores the need for clear reporting mechanisms and strict enforcement. Beyond policy measures, raising faculty and supervisor awareness of gender bias is essential. Universities should also provide practical support, such as childcare services and flexible study arrangements, to help female doctoral students balance academic and family responsibilities.

Additionally, the study highlights the detrimental impact of excessive workloads on female doctoral students’ satisfaction, underscoring the importance of managing workload expectations. Universities should take proactive steps to prevent overwork, ensuring that doctoral students are not burdened with unnecessary or exploitative tasks. Supervisors and doctoral training institutions should regularly assess students’ workloads, particularly for female doctoral students, making necessary adjustments to promote a sustainable and productive research environment.

From a broader policy perspective, addressing gender disparities in doctoral education requires targeted efforts, particularly in lower-middle-income and low-income countries. Policymakers should strengthen gender equity initiatives, enhance legal protections, and safeguard female doctoral students’ rights. Establishing dedicated funding programmes can help mitigate financial pressures, ensuring equal access to academic resources, mentorship, and career development opportunities. Such measures are critical to fostering a more equitable academic landscape and improving doctoral satisfaction.

Reflections on theoretical significance

This study not only deepens our understanding of the unique challenges faced by female doctoral students but also provides critical theoretical insights into the intersection of gender and doctoral satisfaction. From a gender perspective, it challenges the limitations of gender-neutral research frameworks, which often overlook the distinct barriers women encounter in academia, particularly in male-dominated fields. Gender-neutral policies frequently fail to account for the structural disadvantages that shape female doctoral students’ experiences. Our findings underscore the necessity of incorporating gendered perspectives into higher education theory. Moreover, while existing research has examined the challenges of female doctoral students (Sverdlik et al., 2018; Quynh Phan, 2023; Wollast et al., 2023; Gaete Sepúlveda et al., 2025) and the broader factors contributing to lower satisfaction levels (Haynes et al., 2012; Humphries et al., 2021; Mbonyiryivuze et al., 2023), our study extends this work by highlighting the exacerbating effects of excessive workloads and the national economic situation disparities. The intersection of gender, work intensity, and socioeconomic conditions complicates the issue of academic gender inequality, reinforcing the need for more nuanced policy interventions. These findings challenge the assumption that doctoral education follows a universal model, instead advocating for frameworks that acknowledge the special experiences of female doctoral students—particularly those shaped by gender and socioeconomic conditions.

Implications and limitations

This study furnishes empirical evidence for a deeper comprehension of the doctoral learning experiences of female doctoral students. Nevertheless, it does possess certain limitations that warrant further investigation. Although we have empirically addressed the question of “Are female doctoral students less satisfied with doctoral experiences?” and analysed the underlying mechanisms, our study design is somewhat restricted due to data limitations. As discipline, which plays a highly significant role in doctoral studies, was not encompassed within Nature’s investigation, subsequent research could be carried out either to control for the discipline factor or to explore the heterogeneity existing within different disciplines. Besides, due to the cross-sectional nature of the data, it is infeasible to dynamically track the changes in the satisfaction of female doctoral students.

Future research holds significant potential for advancing our understanding of the satisfaction levels among female doctoral students in the context of doctoral pursuits. Firstly, collecting longitudinal data could be considered, as it would enable a more comprehensive disclosure of the changing trends of female doctoral students’ satisfaction as well as its influencing factors. This would provide a more in-depth and dynamic perspective on how various elements interact with and impact their satisfaction over time. Furthermore, given that there are differences in the satisfaction of pursuing a doctoral degree between genders across countries, as evidenced by the data from Nature which shows that in certain countries such as Sweden and Finland, female doctoral students are actually more satisfied than their male counterparts, it becomes crucial to explore the underlying causes of these country-specific differences. Understanding what leads to such variations across different national contexts could offer valuable insights into the multifaceted nature of the factors influencing satisfaction.

In addition, building upon our existing findings, further research adopting a more targeted approach is essential. Specifically, it is necessary to conduct a more meticulous scrutiny in subsequent studies to determine whether the observed dissatisfaction among female doctoral students stems from the omission of variables or systematic gender discrimination. Such investigations would not only help to clarify the root causes of the satisfaction disparities but also contribute to formulating more effective strategies to address potential issues and enhance the overall doctoral experience for female students.