The Quiet Crisis of PhDs and COVID-19: Reaching the nancial tipping point.

Before the COVID-19 crisis, existing high levels of nancial concerns amongst PhD students increased their vulnerability to disruptive events. Impacts from the pandemic have increased their nancial stress to the point that may result in many being forced to exit research studies. An exodus of doctoral students now would impact our future research capacity. The effects of the unfolding crisis on research students at a research-intensive Australian university was documented by a group of doctoral candidates who received 1,020 survey responses from their cohort. Here we show that the pandemic has severely affected research candidates and argue that these results have notable implications for a future research workforce. We found that 75% of students expect to experience nancial hardship as a result of the pandemic. Consequently, 45% report being pushed beyond their nancial capacities and expect to be forced to disengage from their research within six months. Comparative pre-COVID data from an Australian national survey, reports research student disengagement of 20% over four years. The 25% increase of expected disengagement, in a much-reduced period, signals a deeper PhD crisis. Swift intervention is required to avert substantial impacts to the pipeline of research talent.


Introduction
An April 2020 survey of 1020 research students on how the COVID-19 crisis impacted them provides a stark warning for institutes that rely on a robust research workforce. Pre-COVID-19 studies into the doctoral candidate experience show that many students were already experiencing nancial concerns (Lauchlan 2019;Universities Australia 2018;Sverdlik et al. 2018). In a 2017 Australian national survey, 57% (n=5874) of research students reported nances to often be a "source of worry" (Universities Australia 2018). New data reported on here (the Sydney COVID impact data) indicate 75% of research students are now expecting to experience " nancial hardship" as a result of the pandemic. Types of hardships respondents expected to experience included inability to pay for accommodation, medicines, utilities, meals, and carer responsibilities. A consequence of so many students unable to afford living necessities is that nearly half (45%, n=454) expect to be forced to disengage from research within the next six months. Such a severe disruption to the research community may have longer-term consequences for essential advances in innovation and knowledge development crucial to governments, corporations, and universities.
The primary data reported on here came from a survey conducted by several PhD candidates at a research-intensive university in Australia. They sought to document the research students' experience of a disruptor unique in the lived experience of research organisations. Our analysis of that data in context of pre-COVID-19 studies of the research student experience indicate the pandemic has pushed students to a tipping point that may cause a decline in the global research workforce.
Value of PhD candidates to the workforce. PhD students play a pivotal role in economic growth, innovation, and knowledge advancement (OECD 2019). Prior to the pandemic, the total scal impact from research and development (R&D) in the eight leading universities in Australia contributed $24.5 billion per annum to the economy (Conlon et al. 2018).
Productivity spillover of university research in Australia results in every AUD$1 invested contributing AUD$10 to the economy (Conlon et al. 2018). Spillovers include collaborations between universities and organisations, internationally published ndings, and contribution of highly-skilled graduates to a global research workforce (Conlon et al. 2018). The workforce outside of academia is increasingly employing doctoral holders and relies on a continual supply of graduates; by 2015 the United States saw more than half of PhD graduates employed by government and businesses (McCarthy and Wienk 2019). University research outcomes are reliant on the work contributions of PhD candidates (Horta et al. 2018); in Australia, research students make up 57% of university teams. As the world emerges from the pandemic, it will rely heavily on the work of researchers from many elds to develop strategies and innovations for rebuilding economies, health systems and social structures.
The PhD crisis waiting to happen.
Research students have been existing in a state of nancial precarity for years (Sverdlik et al. 2018;Guthrie et al. 2018;Evans et al. 2018). Precarity is characterised by living with persistent insecurities around employment, income, and living necessities; a state that increases vulnerability to major disruptive events such as a pandemic. Living in nancial precarity is similar living on the edge of a nancial cliff, replete with all the anxieties and health disorders that brings. Most studies into the experience of students in higher education are focused on the very different journey of coursework students3 despite large increases in the number of PhD candidates in the last two decades (OECD 2019; Universities Australia 2019; Norton et al. 2018). Speci c research into doctoral studies evidences experiences marred by isolation, high anxiety, severe depression, and other psychiatric disorders (Evans et al. 2018;Levecque et al. 2017;Virtanen et al. 2017;Woolston 2019) as well as increased physical health impacts (Kernan et al. 2011;Juniper et al. 2012). Less research has been conducted on the nancial instability of doctoral candidates and how that impacts their ability to complete their degree, though this work is growins (Sverdlik et al. 2018;Gururaj et al. 2010;Beasy et al. 2019;Ampaw and Jaeger 2012).
The issue of nancial precarity in doctoral candidates is complex due to varying program structures not just in different nations, but also due to the different research economies amongst science and humanities disciplines. The nancial situations of research students are tied to a range of factors including access to grants, scholarships, and loans; as well as supplemental employment opportunities available within academia (i.e. teaching) and externally. The ranging questions in studies of nancial impacts on doctoral progress, differing nancial foci, and diverse site contexts, does present challenges in direct comparisons; however, it's clear that increased nancial concerns negatively impact completions (Gururaj et al. 2010;Ampaw and Jaeger 2012;Mendoza et al. 2014).
Despite varying doctoral formats, nancial concerns are prevalent in research student cohorts globally. In the 2019 Nature survey of over 6,000 doctoral students, 75% of respondents ranked the statement "The di culty of maintaining a work/life balance" and 67% ranked "Financial worries after my PhD" in their top ve concerns (Lauchlan 2019); results consistent with the 2017 survey (Woolston 2019). No speci c question was asked in that study about current nancial concerns, but the ndings did show that of the 19% of respondents that indicated having an additional job, more than half (53%) said it was to make ends meet (Lauchlan 2019). The Nature report also stated that funding and work-life balance were key barriers to pursuing an academic career (Lauchlan 2019) In a 2017 Universities Australia survey, 54% (of 6,206 research students), indicated their nancial situation to be "often a source of worry" (Universities Australia 2018) (Fig.1). These similar results to the Nature survey suggest that data collected in Australia (an economically high-ranking OECD nation That PhD students have been living with nancial instability is not news to any academic involved in research training. What should be of interest to the broader community, is the extent of current pandemic impacts on this group. By understanding how the pandemic has affected doctoral candidates' nancial situation, we can see how this impacts research progress. Our ndings provide evidence that more students are reporting severe nancial stress, expecting to be forced to disengage, and consequently the pipeline of supply of new graduates into R&D may be disrupted.
A snapshot of PhDs at the peak of a pandemic.
During the peak of the pandemic in Australia, several PhD students at a large research intensive university became concerned at the impact on their cohort. These students decided to collect real-time data to report to their university management. The survey was advertised to all enrolled research students at the University of Sydney, but word of mouth spread quickly and additional doctoral students from many other Australian and international universities responded. Over 1,000 students responded within the rst 24 hours. A quarter of the total cohort of University of Sydney research students responded within 72 hours. That response rate carries a message itself: the students wanted to talk about their experience of COVID-19.
The resulting dataset and internal report to the university senior management clearly indicated a more serious impact was occurring than many had realised. The snapshot into the doctoral student experience of the pandemic provided a real-time insight into a shift that was occurring. The data showed that many research students had reached a tipping point and that a potentially large proportion of them were expecting to be forced to abandon their work due to nancial hardship. We realised the message needed to be communicated more widely within the research community, particularly to policy-makers in graduate research education and education ministers in governments. Our analysis looks not only at this snapshot in time during a major pandemic, but places it in context of pre-COVID studies of the doctoral experience. The shift caused by the pandemic highlights the precarity the students were already existing in.

Methods
The Sydney COVID Impact data reported here were collected from 6th-9th April 2020 via an online survey. The survey received 1076 replies from research students,;1020 responses from University of Sydney students, and 56 from students at other universities in Australia and abroad. The 1020 University of Sydney students provided their ID numbers and a detached list of student numbers only were sent by the doctoral students collecting the data to the University for veri cation. After veri cation it was determined that the response number represented 24% of the total research student cohort as at April 2020.
Respondents were self-selecting research students, and broadly demographically representative, incorporating 300 international and 713 domestic students, as well as 843 full-time and 177 part-time students. The original survey and de-identi ed data is available at http://doi.org/10.25910/5ed06563d9d79 In our analysis, we considered only the 1020 responses from the University of Sydney students as they were the only respondents that had been veri ed. Some questions were not mandatory, as such not every respondent answered all questions. Where that is the case, the number of responses to the question is reported else the percentage of the entire 1,020 veri ed Sydney students is given.
In questions where respondents were allowed to indicate multiple selections, for example in the candidature impact question, we counted all instances of a negative selection of a single respondent and converted that to either a 1 (at least one negative response given) or 0 (no negative response given). That means that when calculating the overall disengagement number each respondent is only given a value of 1 in 1,020 regardless of how many negative responses they indicated. The reason was to understand the proportion of the cohort expecting to disengage from studies on a whole.
In our discussion we have compared the Sydney COVID Impact data with pre-COVID data on doctoral students. Whilst the exact questions asked are not the same, therefore no statistical tests can be run, comparisons can be made.
The data were collected in the early phase of the COVID-19 pandemic in Australia. Government mandated social distancing restrictions had been rolled out progressively over the previous three weeks, with many organisations severely impacted from 26th March. There was a global, heightened sense of anxiety during the survey period which may have affected some responses.

Results
Living and nancial impacts of COVID-19 The Sydney COVID impact data shows that 75% (of 888 respondents) expect to experience nancial hardship as a direct result of the pandemic (Fig. 1). Almost a fth of respondents (154 out of 853 respondents) indicated the pandemic had already resulted in them unable to afford necessities such as utilities, internet connection, and medicines, or they anticipated this impact within 6 months. Additionally, 11% (of 849 respondents) had already gone without meals or expected to in the next six months as a result of the pandemic. Of those with carer responsibilities, 19% (of 844 respondents) indicated they could no longer meet the nancial cost of supporting others or expected to experience this within 6 months.
Much of the nancial impact was driven by changes to supplementary employment, that is income in addition to scholarship stipends. Changes included job losses (18% of 996 respondents), work and income reduction (19%), and workload increases without increased pay or even less pay (Fig. 2). Many also said they had undertaken unpaid teaching work to bring courses online for coursework students.
Some respondents said they had become entirely reliant on their scholarly stipends or they were living on their partner's income.
In Australia, most universities cap extensions for centrally funded scholarships at 6 months after the initial 3 years, but 66% of respondents indicated they would need an extension of funding beyond that due to COVID-19 impacts on their research. Scholarship holders were concerned their stipend would cease long before they had completed their degree due to pandemic related delays to their research.
Many (29% of 858 respondents) had already been unable to meet accommodation costs or expected to face this within 6 months. An additional 30% (of 1008 respondents) had already, or expected soon, to be forced to nd alternative accommodation. Notably, 5% (of 983 respondents) were already, or expected to be soon, experiencing homelessness as a result of the crisis, with 24% unsure if they may face this. If this 5% is indicative of the whole Australian research student cohort, potentially over 3,300 doctoral candidates may face homelessness in the coming months.

Impact to candidature
Many respondents indicated they may be forced to disengage with their studies due to the impacts of the pandemic. Disengagement ranges from blocks of extended leave, through suspension of studies, to complete withdrawal from candidature (Fig. 3). Actual or potential disengagement from candidature as a result of the pandemic was indicated by 45% of all 1,020 Sydney respondents. The extent of expected disengagement was similar among science-based and arts-based students. It is important to remember that most students do not move directly from active enrolment to withdrawal; when students take suspension of studies or leave of absence, not all return and some end up ultimately withdrawing. Perhaps the most telling gure is that only 18% (of 899 respondents) agreed with the statement "I do not believe that the COVID crisis will have any impact on my candidature".

Discussion
The quiet crisis.
In most research endeavours, generation of new knowledge and technologies are frequently dependent on inexpensive, yet highly-skilled contributions from doctoral candidates. An Australian national thinktank noted that PhD students are frequently a source of "unpaid or lowly-paid research-labour" (Norton et al. 2018). Studies show that research students often feel compelled to undertake unpaid or low-paid work outside of their core research (Cressey 2012), to improve their chances of gaining increasingly rare academic positions (Beasy et al. 2019). The result of systematically facilitating nancially precarious behaviour results in a contribution of free or inexpensive labour to the research community. In times of pandemic induced austerity measures, cost-e cient approaches to producing research outcomes such as utilising doctoral candidates for skilled work, may be more frequently employed by organisations struggling to survive deep economic impacts. We have unfortunately become normalised to doctoral candidates often living on minimal incomes, thus it is unsurprising that the additional impacts of the pandemic have resulted in the PhD crisis falling below the radar.
To live in a state of precarity is to be vulnerable to widespread disruptions. When students' concerns around nancial stability shift toward expectations of nancial hardship, it should be seen as warning.
Every year, doctoral students around the world disengage from research candidature for a wide range of reasons. Our ndings suggest the level of disengagement, and the short time-frame this disengagement is expected to occur in, are a signal that something deeper and systemic is happening (Fig. 4).
The 2019 Nature PhD survey (Lauchlan 2019) reported nancial related statements in two of the top ve concerns of doctoral candidates. The 2017 Universities Australia ndings (Universities Australia 2018) indicated just over half of research students nd nances to often be a source of worry. The 2020 Sydney COVID impact data shows three in four students expect to experience nancial hardship as a result of the pandemic. Comparing these ndings suggests that the intervention of the COVID-19 crisis has increased the number of students suffering severe nancial anxieties.
The consequence of so many doctoral candidates expecting to experience nancial hardship is a potentially higher than usual impact on candidature progress. The Australian Department of Education, Skills and Training (DEST 2020) reported that from 2010 to 2014, 20.2% of research students had dropped out or disengaged during the standard 4 year duration. Our ndings suggest a much higher disengagement rate during the next six months (Figure 4). If 45% of doctoral candidates at one of Australia's leading universities do disengage within the coming months, it is possible that many will not return. We cannot know the real impact for another four years, thus the ndings presented here should be taken as a warning not a prediction.
What will be the long reach of the PhD crisis?
Here we have provided evidence of the doctoral candidate experience of the pandemic. Our ndings indicate that the crisis has exacerbated the nancial precarity of doctoral candidates, with many now feeling they must abandon research. We may be witnessing the beginning of a wider systemic disruption to the research workforce that drives our knowledge-based economies. It seems prudent to consider the current experience of doctoral students, and the long-term implications of allowing the current impacts to students to go unchecked. If we do nothing, the potential disruption to research capacity may weaken our societal, public health, and economic recovery. Governments, universities, learned societies, and industry need to produce coherent strategies to address the PhD crisis before it diminishes our research capacity.

Figure 1
Comparison of Australian research students' attitudes to their nances prior to and during the COVID-19 pandemic.

Figure 2
Negative impacts on employment arrangements of research students at the University of Sydney as a result of the COVID-19 pandemic.

Figure 3
Disengagement with candidature as a result of the COVID-19 pandemic.