What influences PhD graduate trajectories during the degree: a research-based policy agenda

During the past two decades, PhD graduate numbers have increased dramatically with graduates viewed by governments as a means to advance the knowledge economy and international competitiveness. Concurrently, universities have also invested in policies to monitor satisfaction, retention, and timely completion—and researchers have expanded the study of PhD experience. We, as such researchers, have increasingly received invitations from university decision-makers to present research evidence which might guide their doctoral programs. Their interest provoked us to do a qualitative systematized review of research on doctoral experience—seeking evidence of practices that influenced retention, satisfaction, and completion. The result contributes a synthesis of the critical research evidence that could be used to inform doctoral education policy. We also demonstrate the possibilities of such evidence by suggesting some potential recommendations, while recognizing that there is no direct relationship between research results and their transformation into particular institutional contexts in ways that enhance doctoral experience. We hope our initiative will be taken up and extended by other researchers, particularly the research gaps we note, so we can collectively support the use of research evidence to influence both doctoral policies and practices—with the goal to better prepare PhD researchers for their futures and better support their supervisors.


Context
Trajectories are built from the very beginning of the PhD (Mangematin 2000, 751) During the past two decades, there has been a large increase in the number of PhD researchers (OECD 2016) as governments have increasingly construed PhD graduates as needed to advance the knowledge economy and international competitiveness (Hancock et al. 2015). Alongside this growth, universities have increasingly invested in policies that originate in a desire for (a) shorter times to completion and (b) graduates who are better prepared for work outside the academy (Auriol et al. 2012).
Concurrently, there has been a tremendous growth of research into doctoral education. We, as such researchers, 1 have been invited to give presentations to university decision-makers who are seeking research evidence to guide their thinking (European University Association 2016) in supporting PhD researchers 2 and supervisors. Among these individuals are pro vice chancellors, vice principals, and deans of postgraduate studies and their teams and similar individuals at the faculty (associate deans) and departmental (directors of graduate studies) levels. Their interest in the transformation of research evidence into policy potential provoked us to undertake this systematized review.
At the same time, the outcomes of any initiative to implement reform will be influenced by specific constraints and affordances (Andres et al. 2015). For instance, institutional histories vary considerably, including local departmental practices, and these need to be taken into consideration in transforming research evidence to the local context. Further, as Ball et al. (2012) have argued individuals take on more active or passive stances in different elements of their jobs in relation to different policies and their roles. As a result, responses from different stakeholders can vary from professional pragmatism through symbolic compliance and formal instrumentality to rational resignation (Teelken 2012). Thus, policy changes do not necessarily lead to the desired practices. In other words, there is no certainty that the results of any policy change will produce the hoped for effects.

Purpose
Our goal was to systematically review the doctoral research in order to synthesize the critical evidence that could inform doctoral programs, particularly enhance retention, satisfaction, and completion, and from this synthesis suggest possible policies to support PhD researchers, but also of benefit to supervisors, especially those starting out. 3 We chose these three factors since they have been frequent reference points in the past, both in research and institutional policy. An early concern was PhD researcher attrition (dropout) and retention: two sides of the same coin. Retention incorporates the notion of persistence; it refers to researchers remaining in their 1 See https://www.researcher-identity.com regarding our European cross-national research on PhD and post-PhD researchers. 2 We refer throughout this paper to PhD researchers rather than students (the more common English term) as we want to highlight the shift in expectations from undergraduate or Master's to the PhD: with the PhD researcher expected to be agentive in progressing his/her own research. As well, the term, "student," is not used in some countries; of course, in some places, PhD researcher could be understood as a PhD graduate. 3 If you are an experienced researcher or manager in doctoral education, you may already have developed your own personal synthesis.
programs and, in some cases, remaining in their fields beyond graduation. Emerging from this focus on retention grew the interest in completion in various forms: first tracking time to completion and later timely completion against a standard. Satisfaction as regards aspects of the doctoral experience is a frequent tracking mechanism in understanding retention/ persistence and progress to completion as there is plentiful evidence that satisfaction is linked to engagement in research and more timely completion. Accordingly, satisfaction can be considered as a central indicator of study progress and positive doctoral experience. Moreover, it provides a link to PhD researcher well-being in terms of study engagement. This is particularly important since reduced levels of well-being have been identified as a significant problem among PhD researchers, that is, a significant number of them experience high levels of distress, even burnout during their studies.
We hope this synthesis and our suggestions for policy implications will (a) provide a starting point for researchers to engage decision-makers in similar discussions and, (b) most importantly, encourage other researchers to review and expand on our work. After summarizing the literature reviews from 2001 onwards to highlight recurring themes over the years, we describe the method we used and then the evidence we draw on to generate possible policy implications.

What is the state of research on PhD education?
Over the past 15-20 years, there has been a dramatic growth in research on doctoral education. A series of reviews, each taking a somewhat different perspective, collectively provide a sense of the scope of this work: an account over time of more and less robust areas of research. You may note as you read that the three factors are mentioned: (lack of) completion (retention/ attrition) and time to completion nearly always, with satisfaction to a somewhat lesser extent. Generally, the studies explore specific aspects or processes of the experience, e.g., supervisorresearcher relationship and writing.
A pioneer review, Golde (2001) summarized the research on doctoral education in the USA from 1990 to 2001. Issues pertinent at that time included the following: lengthy times to completion, lack of completion (characterized as attrition), retention, lack of funding leading to debt, and supervision including the extent to which supervisory workloads might influence the potential to be "mentors." A further theme was satisfaction as regards the decision to do the degree, the supervisor, and overall experience. Finally, the number of PhD graduates "produced," even then, was higher than the available academic positions, suggesting therefore the need to prepare graduates for other careers.
Later, Leonard et al. (2006) explored the research on doctoral education published in the UK in order to name gaps for future research. Just over half of the studies focused on the departmental and institutional work study context in terms of provision but without any systematic comparison across institutions. Close to half looked at the overall program satisfaction. Similar to Golde (2001), there was a concern in one-third of the studies with completion times and rates (and briefly retention), and employment patterns, but little systematic information on causes of lack of completion. They similarly noted the focus on supervision as well as new foci, peer support such as research teams, researcher-led support groups, and doctoral writing. They concluded that much of the research had ignored individuals outside the social sciences and that very little research had been done on researchers' perspectives, such as their views of the doctoral experience. Driver (2006) was unusual in examining the literature in relation both to pedagogy and policy-likely due to his institutional role as a senior administrator. The review includes reports by national organizations of graduate education and guidebooks for researchers and supervisors. The core focus is completion as regards retention and progress towards completion. The literature is grouped as regards researchers, supervisors, and administrators. For each role, the common themes are noted; such themes for researchers include the following: (a) discuss expectations with supervisors (including non-program-based activities such as attendance at conferences, grant applications, research assistant (RA)/lab duties), (b) learn to manage time, establish goals, and follow deadlines, (c) initiate contact regularly with the supervisor, (d) learn to respect your supervisor's criticism, (e) recognize your supervisor may be busy, (f) have academic integrity and honesty, and (g) get involved in your department.
More recently, the reviews have burgeoned. John and Denicolo (2013) reviewed the literature on PhD experience in OECD countries (though with a UK emphasis) focusing on identifying gaps filled and remaining since Leonard et al. (2006). The review noted a large number of "how-to" books, reports on institutional workshops, and advice articles aimed at either the researcher or supervisor around the theme of improving aspects of this relationship. Research had grown on internationalization and the experience of international researchers. Also, researchers' accounts had increasingly been foregrounded, though with little inquiry into the differences in daily PhD events and practices. The authors suggested further investigation into researcher agency, the comparative specific experiences of doctoral researchers within and between institutions and countries, and inquiry into disciplines other than the social sciences (as in Leonard et al. 2006). They note an increase in attention to satisfaction, with variation as regards focus, e.g., supervision and overall experiences. Vos (2013) reviewed the literature on dissertations, notably on a broad scale, including theses from the PhD, Master's, and undergraduate levels. The review is of note since the intent was to explore the pedagogical implications of the studies. Vos grouped the literature into six categories each accompanied by commentary and recommendations. As in earlier reviews, the following themes emerged: (a) researcher-supervisor relationship with satisfaction linked to different aspects of the relationship; (b) researcher diversity, cultural background, and prior preparation for undertaking a dissertation; and (c) researcher progress through the PhD process, including ways to improve researcher motivation. Other themes included (a) teaching and learning research methods and (b) issues related to plagiarism and academic dishonesty. Vanstone et al. (2013) reviewed the literature on interdisciplinary doctoral supervision, given the increase in interdisciplinary PhDs. They argued that this integration of research fields could influence the (a) researcher-supervisor relationship, (b) process of forming and working with a supervisory committee, and (c) processes and outcomes of doctoral research. Their goal was to characterize the areas of research saturation and areas for future research. They reported that there is scant literature on the positive aspects of interdisciplinary research, given most research examines the potential challenges for researchers. The challenges were discussed directly or through suggestions for improvement. They concluded that while many claim that interdisciplinary work is growing at the PhD level, this assertion has still not been established on a national or international level. No reference was made to satisfaction, retention, or completion.
Similar to Driver (2006), Bengtsen (2014) focused on literature beyond published research papers, examining eight often-cited handbooks for supervisors. The rationale was that handbooks synthesize the research in a way accessible to supervisors, that is, from a pedagogical perspective. The review highlights the common themes across authors as well as their distinct perspectives. Common themes include the following: (a) the entanglement of doctoral supervision within the institutional context; (b) enculturation within the specific disciplinary context, including tacit norms and values for good conduct; (c) academic craftsmanship including supporting the research project-especially given the more heterogeneous researcher population; (d) supporting and facilitating autonomy and emancipation; (e) the supervisor and researcher relationship in enhancing progress to completion; and (f) the supervisory dialogue which is related to (g) building an open and trusting dialogue. Perhaps, given the nature of the review, good supervisory support is referenced as a means to contribute to completion, with no reference to satisfaction or completion.
This characterization of the prior research on PhD experience is consistent with our view and makes clear the ways in which retention, completion, and satisfaction have been touchstones in examining doctoral experience. Overall, while one might argue there has been some progress as regards several issues reported in the reviews, the historical intractability of the concerns is striking, such as researcher-supervisor relationship, disciplinary-departmental issues, or the institutional context. From our perspective, what is of most concern is the silence since 2006 on the "surplus" of PhD graduates and the failure to prepare graduates for nonacademic post-PhD careers; such research is essential if we truly want to better prepare PhD researchers for their futures.
We would argue that the initial focus on researcher-supervisor relationship as critical to researcher success likely emerges from the historical nature of relatively unstructured PhDs with little sense of a program or a cohort. However, while the supervisor remains important, our view is that the role of the supervisor is now more institutionally structured and so may not play quite as powerful a role as previously. Specifically, (a) PhD programs are increasingly standardized including multiple forms of structured learning (e.g., modules, workshops, writing centers, and career services); (b) there is more institutional accountability (e.g., greater monitoring of PhD progress, more formal assessment procedures); and (c) other academics are increasingly involved in the support of doctoral progress, e.g., academics other than the supervisor involved in assessment and feedback.
Finally, we suggest future research should be directed at (a) a growing policy initiative, cosupervision; (b) related alternate models of supervision, such as non-academic co-supervisors; and (c) the impact of new technologies on doctoral and supervisory practices (McAlpine 2017). As well, while there is much reported on thesis writing, there is little on other genres of writing and publishing or on reading. Finally, the literature generally reports the PhD researcher as the object of attention which facilitates thinking about policy implications since it highlights departmental and institutional structures that might better support PhD researchers. However, this perspective pays little attention to the researcher as an active agent in the PhD process, an issue we address shortly.

Method
Framing what affects PhD graduate trajectories during the degree In approaching this analysis, we took a "nested contexts" perspective (McAlpine and Norton 2006), locating the influences on PhD experience in relation to embedded contexts. The framework integrated the previous research on doctoral education and remains useful today in differentiating the influences and interactions of different players in the process. The researcher and supervisor relationship rests at the center-nowadays sometimes embedded in research teams. 4 The supervisory relationship is situated within the local departmentaldisciplinary context which integrates (and interprets) both local practices emerging from institutional policies and disciplinary assumptions (which tend to be global) about disciplinary modes of research. The department is situated within the institutional context; this context creates the policies, the affordances, and constraints that influence the practices in departments related to, for instance, selection/admission, PhD funding, and program requirements. Given the interaction between the departmental and institutional contexts, we treat them together in our analysis. Finally, moving outward again, the societal-international context includes elements such as policies, both national (research funding) and international (global competitiveness), as well as economic trends (the strength of different labor sectors) which influence the other contexts. While this final context can act as a powerful influence on PhD programs, for this paper, we set it aside, given both a lack of research evidence and space limitations.

Approach
We chose to undertake a systematized review (Grant and Booth 2009). Such a review includes the same elements of a systematic review but not necessarily the same processes within each element: (a) comprehensive search; (b) assessment of the studies; (c) synthesis of the findings; and (d) analysis-what is known, recommendations for practice, what remains unknown, uncertainty around findings, and recommendations for future research.
We (a) set criteria for inclusion; (b) generated and integrated lists of pertinent studies; (c) analyzed them to generate pertinent research evidence in relation to completion, retention, and satisfaction-as well as gaps in research-and finally (d) generated potential policy implications.
Setting inclusion criteria Our starting point was that the studies of most use as regards institutional policy implications would be those that linked PhD experiences and practices to evidence of doctoral success. In other words, from an institutional perspective, monitoring and assessment procedures are linked to ensuring researchers are making satisfactory progress and completing in a timely fashion (e.g., University of Reading 2013). Thus, we set the following criteria for inclusion in our analysis: (a) timely through to lack of completion, (b) low through to enhanced retention, and (c) satisfaction (engagement in research, reduced burnout, and more timely completion) through to dissatisfaction, with the reverse effects (burnout, exhaustion, and lack of completion risk).
Generating and integrating lists While we are all experienced researchers in the area, we each have different scholarly and methodological roots. We took advantage of this by each reviewing our individual databases of doctoral studies to generate a list related to completion, retention, and satisfaction since 2000, since, as noted earlier, the past two decades have seen a growth in research on doctoral education. We collated our lists, and then to ensure we had not overlooked research evidence, we re-read the literature reviews from the same period (described previously) and added any studies that provided empirical evidence. Finally, to ensure that we had generated a comprehensive set of studies, we characterized each study as to (a) its disciplinary cluster, (b) its research methods (quantitative, mixed, and qualitative), and (c) the geographical location by continent in which the study took place-Europe (including the UK), Australasia, and North America. This information is noted immediately after the relevant citations in the Reference list. As well, Table 1 includes this information along with keywords to show the focus of the study and number of participants in the study. You will note that there is a relatively equal distribution of studies across continents, a relatively equal mix of qualitative or quantitative methods with a considerably smaller number of mixed methods. Social sciences are more prevalent than other disciplinary clusters (as we know); still, STEM and humanities studies are represented relatively robustly.
Analysis We then reviewed these studies. Our goal was to verify they did indeed address in meaningful ways features of doctoral experience that addressed completion, retention, and satisfaction.
Next, we grouped the citations thematically, noting any contrary evidence, in relation to the nested contexts of interest: the PhD researcher-supervisor and the departmental-institutional contexts. In many cases, there were a number of studies supporting a particular policy implication, e.g., writing support, which represented both different methods and geographical locations. However, given space limitations, we have included no more than two references for a given feature of PhD experience. These were chosen based on recency and robustness of the study.
Generating policy implications Based on the themes, we generated possible policy implications, while recognizing the following limitations. These studies were conducted in different countries with different national policies and pressures. Further, circumstances in the particular university settings must indeed have varied as do the circumstances in any university drawing on these results to inform policy. Thus, our policy implications must be viewed as suggested directions for a policy that need adjusting and embedding in local circumstances.

Researcher-supervisor interaction
We begin by highlighting specific personal practices that promote satisfaction and progress and end with policy implications, which can be implemented at departmental, faculty, or university levels.

What promotes satisfaction, retention, and completion
A range of studies suggest that specific practices by the PhD researcher and the supervisor can contribute to enhanced time-to-candidacy, degree completion, doctoral researcher well-being, and satisfaction with the overall doctoral experience (Ives and Rowley 2005;Pyhältö et al. 2015). We have grouped this evidence as follows: the researcher, the supervisor, and researcher-supervisor relations.
PhD researcher We begin with PhD researchers since, as we noted earlier, their agency has been largely absent from much of the research, yet the PhD researcher is the central player in the PhD journey-the one with the most invested.   1. Involved in supervisor selection: PhD researchers are often assigned supervisors on the basis of supervisor availability. Yet, since PhD researchers seek expertise and/or supervisor interest in them (Barnes et al. 2012), being involved in supervisor selection is a positive experience (Ives and Rowley 2005), related to good progress and being satisfied (see also researcher-supervisor relations). In contrast, being assigned a supervisor leads to less positive outcomes than choosing a supervisor (Barnes et al. 2012). 2. Expectations met: Individuals applying for PhDs hold a range of expectations of the nature of the learning expected. Not surprisingly, a match between anticipated expectations about both learning and doing research and the actual experience enhances satisfaction and completion (Holbrook et al. 2014)-whereas the opposite leads individuals to feel uncertain and dissatisfied with their fit for the PhD. In a similar fashion, the knowledge of what is expected of a PhD degree is lower for non-completers and higher for completers (Lovitts and Nelson 2006). (See #12, departmental leadership for more related to this.) 3. Taking ownership of the degree: PhD researchers vary in the extent to which they feel and act as agents (O'Meara et al. 2014), and studies have shown that variation in researcher agency during the degree has long-term positive and negative implications (McAlpine and Amundsen 2018). In a similar vein, only those researchers displaying an awareness of their own coping strategies and perceiving themselves as agents in their doctoral trajectory are able to engage effectively in overcoming supervision-related problems (González-Ocampo and Castelló 2018). Further, doctoral researchers applying active strategies in their researcher community suffer less anxiety, lack of interest in their studies, and entertain a lower risk of study abandonment (Pyhältö and Keskinen 2012). Lastly, variation in PhD researcher agency throughout the degree, e.g., choosing or not to publish during the degree, led to long-term differences in satisfaction, completion, and career success (McAlpine and Amundsen 2016, 2018). As well, feelings of lack of agency, i.e., no specific strategies to develop career knowledge, can contribute to apathy and anxiety (Thiry et al. 2015). 4. Personal and academic peer networks: This refers to the ability to develop and manage relationships with others who have the potential to support careers (Forret and Dougherty 2004) and to effectively engage and collaborate with others to achieve goals (Pilbeam and Denyer 2009). Developing and using extensive personal and academic peer networks is highly related to persistence, motivation, production, and professional success (Goller and Harteis 2013). Further, investing in building a network of support has long-term effects; it nearly doubles one's chances of immediate employment within 12 months of graduation (Jackson and Michelson 2015). At the same time, it appears that in any cohort, only a minority of PhD researchers perceive themselves as active agents as regards developing their networks (Pyhältö and Keskinen 2012).
Supervisor We know that supervision has an impact as regards completion time and employment post-submission (Platow 2012). So, here, we look more closely at two different supervisory influences.
5. Supervisor expertise: Researcher degree completion is linked to the supervisor's knowledge and expertise in the research field (Donald et al. 1995;Holdaway et al. 1995). The latter study recommends restricting supervision only to active PhD researchers with expertise in the researcher's area of research. Further, researcher satisfaction and completion are enhanced when supervisors are senior academics or have greater experience in research supervision (Ives and Rowley 2005). 6. Number of supervisors: Co-supervision is growing as a new form of supervision, enabling supervisors to discuss how to handle emerging issues and researchers to feel better protected (Spooner-Lane et al. 2007). Thus, PhD researchers with more than one supervisor are more satisfied than those with only one (Pearson 1996) (though Heath (2002 found no difference). Further, Sala-Bubaré and Castelló (2016) noted the value of more than one supervisor for researchers who feel isolated or discouraged regarding their doctoral process and outcomes. Still, while co-supervision is perceived positively by supervisors if they meet jointly with the researcher, this process is time consuming and may not be recognized institutionally (Olmos-Lopez and Sunderland 2016). Further, if responsibilities are not clearly articulated, potential problems include, for instance, fragmented supervisory responsibilities (Pole 1998).
PhD researcher-supervisor relations While well researched, this area is paradoxically the most variable given differences in personal histories, preferences, and motivations between individual PhD researchers and supervisors.
7. Availability and meetings: The evidence relates to both availability and frequency of formal meetings. Donald et al. (1995) reported one of the two most important factors in supervision satisfaction was supervisor availability. Zhao et al. (2007) reported availability strongly correlated with researcher satisfaction; a more recent study reported the same (Sverdlik et al. 2018). In a similar vein, Main (2014) reported that the frequency of communication affected program duration and suggested communicating at least semi-monthly. As regards frequency of meetings, another study (Heath 2002) supported this view-though noted that if meetings were at least once a month, satisfaction was not substantially reduced. Further, to enhance researcher progress, completion, and satisfaction, meetings should provide guidance on (a) topic definition, (b) research design and data analysis, and (c) literature to be reviewed. 8. Good working relationship: Lovitts and Nelson (2006) and Ives and Rowley (2005) have linked good progress, completion, and satisfaction with clearly understood expectations. Further, a good PhD researcher-supervisor fit is shown to be associated with positive PhD experience (Golde 2005). More specifically, Hyun et al. (2006)) showed that graduate researchers reporting more functional supervisory relationships are less likely to report mental health issues, whereas those reporting ambiguous expectations were more likely to report such issues. Yet, shared expectations cannot be assumed, so unfortunately, differences may be implicit and lead to misunderstandings (Woolhouse 2002). 9. Feedback: A frequently overlooked aspect of feedback is researcher progress. Zhao et al. (2007) reported satisfaction was correlated with regular and direct reviews of progress and purposefully helping PhD researcher progress in a timely manner. The latter point is supported by Main (2014) who linked good progress and completion with supervisors who explicitly encouraged finishing quickly.
Overall, the evidence suggests that factors such as those above can reduce burnout and lack of completion risk and contribute to more engagement in research, greater productivity and satisfaction, and more timely completion. The contrary will produce the reverse effects, hindering progress and satisfaction. For instance, PhD researchers perceiving a misfit between themselves and their working environment in terms of problems with supervisors (and the department) may decide to leave (Golde 2005). Here, we suggest some possible policy recommendations at the personal (PhD researcher and supervisor) levels. Of course, the effects of these policies are cumulative as a constellation of factors interacts in providing a supportive environment for doctoral success. We suggest ensuring these policies are easily accessible online, so they can begin to have effect "on the ground." → Ensure policies are consistent in emphasizing the responsibilities of the PhD researcher in advancing through the degree.
& Incorporate this responsibility in all documents, notably, program descriptions, application processes, and codes of practice so that applicants understand the shift in status that is required & Incorporate PhD researcher agency in different PhD processes, e.g., researcher involvement in the supervisory selection, researcher report on initiatives to grow network in annual report, and researcher participation in program evaluation and improvement. & Ensure applicants can access and mobilize this information, for instance, through links to program/scholarship application pages, references in the application process, and encouraging questions about it at interview.

Departmental-faculty-institutional contexts
As noted earlier, we have integrated these contexts since (a) individuals in all three environments have the ability to create policies and (b) institutional change can cascade both up and down. In this section, we attend not just to supporting PhD researchers but also to supervisors. While support of the latter has often been overlooked in doctoral policies, our view is that we need to invest as much policy effort in the support of the supervisor as the researcher and, in other words, view supervision as a collective responsibility, shared by many other individuals and units besides the named individual (McAlpine 2013). A recent review noted that all relationships and structures involving individuals, resources, and institutions beyond the PhD researcher, e.g., supervision, the department, and financial support opportunities, can directly or indirectly impact progress and completion (Sverdlik et al. 2018).

What promotes satisfaction, retention, and completion
A range of studies suggest that specific departmental through institutional practices can contribute to greater satisfaction, research engagement , and productivity (Scaffidi and Berman 2011) and reduce risk of burnout and lack of completion Jairam and Kahl Jr 2012). We have grouped the evidence as follows: institutional leadership, departmental leadership, and institutional-departmental shared leadership.
Institutional leadership There is minimal research at this level, but what there is provides some policy direction.
10. Institutional structure: Studies have shown that external pressures can lead to a greater institutional focus on reform. For instance, Humphrey et al. (2011) looked at the influence of three UK-required PhD policies within one university on PhD completion within 4 years. The three elements were (a) completing training assessment, (b) completing project approval form, and (c) having team supervision. Having all three experiences, while not the most frequent situation (only 29%), led to the highest probability of completion in 4 years: 70%. Having only two components, a project approval form and team supervision, was the most frequent situation (39%), with the probability of completion dropping to 49%. Similarly, Lee (2013) in an Australian study reported that a change in institutional receipt of government funding to after PhD completion rather than earlier led universities to set clearer milestones for completion, do more careful PhD recruitment of those who would complete, and attend to satisfaction during the program so researchers would not withdraw or transfer. Of course, such initiatives can be undertaken regardless of external funding. Yet, as Golde (2005) has noted, doctoral programs are often a hodgepodge of inherited requirements, norms, and traditions. So, institutional investment in doctoral reform requires clarity about the purpose and a clear plan with robust support structures and monitoring (Driver 2006). Such intentional planning often includes a curricular component, collective forms of supervision, and strong intellectual climate and may be structured as a doctoral school (Nyhagen and Baschung 2013). 11. Human and infrastructure resources (access and awareness): access to equipment, work space, and library resources is influential in doctoral satisfaction (Harman 2002). Still, Golde and Dore (2001) reported researcher awareness of institutional resources was not particularly high (except for teaching development), and only two-thirds of those who knew of the resources used them. Nevertheless, resources should include career information and training since researchers have consistently reported both the desire for this support (Golde and Dore 2001;Thiry et al. 2015) and a lack of support and advice from supervisors-perhaps due to a lack of knowledge (Sauermann and Roach 2012). Interestingly, Jackson and Michelson (2015) demonstrated that using career services more than triples the likelihood of employment within 12 months of graduation. Finally, universities should ensure good PhD counselling services since PhD researchers are high users and such use can reduce emotional and stress-related problems (Hyun et al. 2006) Departmental leadership At issue here is not just the creation of departmental policies but the extent to which institutional policies are actively and accurately implemented and practiced within departments in disciplinary appropriate ways.
12. Application process: Research on doctoral application processes is virtually absent, though Maher and Barnes (2010) demonstrated that traditional measures are unreliable. Further, we know that one influence on lack of completion can be inappropriate PhD researcher expectations due to poor application processes (Golde 2005). Notably, 80% of researchers have reported the first year overwhelming (Cluett and Skene 2006) and 64% found the transition difficult (West 2012), likely due to poor understanding of expectations. The Maher and Barnes (2010) study demonstrated the value of a more systematic approach to assessment during selection; assessors found the increased rigor provided a more precise measure of previously ill-assessed critical thinking skills leading to more appropriate admissions decisions. 13. Prompt completion: A number of studies have shown departmental leadership and expectations influence completion rates. For instance, completion rates are higher in departments encouraging prompt completion rather than thesis refinement (Main 2014).
As well, programs with better supervising and clear requirements reduce lack of completion probabilities (Ehrenberg et al. 2006). Finally, O'Meara et al. (2014 demonstrated that departmental leadership can also influence researcher agency as regards planning for post-PhD careers. 14. Supervisory change: Here, the evidence is negative rather than positive, highlighting the need to reduce and buffer such changes as much as possible. Supervisory change slows progress (Ives and Rowley 2005) yet can be relatively frequent, up to 25% of researchers (Golde and Dore 2001). Unfortunately, researchers experiencing a disruption or change in supervision feel exceptional, isolated, and upset (McAlpine et al. 2012) since they are not aware that such changes are not exceptional and are guided by institutional procedures. Interestingly, these changes are often described by researchers as unexpected or unplanned, suggesting the change could result from supervisor absence, e.g., moving university and taking family leave (Ives and Rowley 2005). Co-supervision would help alleviate such upsets. 15. Departments, research teams, and intellectual/social climate: Researcher program satisfaction has been linked to a collegial climate (autonomy in academic planning and freedom to express opinions) as well as good relations with peers and supervisor (Shin et al. 2018). In other words, engaging intellectually and socially with other researchers promotes a positive experience (Gardner 2010) and greater satisfaction and reduces the risk for study abandonment (Pyhältö et al. 2009;Schmidt and Hansson 2018). Faculty can offer role modelling and intellectual excitement (Gardner 2010) and PhD peers social support. Satisfaction with academic involvement within the program enhances researcher completion (Gittings et al. 2018). But, a perceived mismatch between researcher values and those attributed to the departmental culture may lead researchers to abandon their studies (Gardner 2009). Similarly, researchers in research teams can display higher motivation, experienced support, reduced stress levels and emotional problems, and enhanced success rates (Jairam and Kahl Jr 2012;Stubb et al. 2011). However, a research team does not necessarily have these effects, if, for instance, the researcher is uncomfortable with the policies, values, and atmosphere within the team (Castelló et al. 2017;Gardner 2009). So, support needs to be conceived broadly, to include informal groups and individual relationships.
More formal support mechanisms can include (inter)departmental research teams and disciplinary networks (e.g., conferences) which also enable sharing of positive experiences, alleviation of negative ones, and enhanced research engagement . Formal support also includes guides for self-direction by feedback and mentoring or funding information (Gardner 2010;Jairam and Kahl Jr 2012).
Institutional-departmental shared leadership Some issues are best addressed collectively, that is, in institutional ways for consistency and efficiency, and departmental ways to address disciplinary variation.
16. Researcher writing support: Support for publishing would appear essential, with publishing a determining factor in academic advancement (Bergeron et al. 2014), alongside PhD researchers' difficulties in this regard (Aitchison et al. 2012). Such support, for example, workshops, can result in increased confidence, higher levels of productivity, and self-efficacy (Badenhorst et al. 2014). Yet, researchers report struggling to write scientific articles and lack opportunities to learn to write (Sala-Bubaré et al. 2018). This is unfortunate since such activities increase publication rates post-intervention (McGrail et al. 2006) and might also reduce the risk of abandonment (Castelló et al. 2017). While some of these initiatives may best be disciplinary-based, since they incorporate disciplinary approaches and enhance intellectual climate, there is also evidence that crossdisciplinary or disciplinary cluster initiatives are powerful (Sundstrom 2014;Tardy 2005). 17. Supervisor writing support: As for supervisors, providing constructive PhD researcher support can be hindered by perceptions of writing, in that the feedback they provide will be based on their own writing perceptions (González-Ocampo and Castelló 2018). For instance, they may perceive their support through the lens of writing as using a set of technical skills (Lee and Murray 2015;Kamler 2008) rather than as creating knowledge, that is, developing and clarifying research ideas ). The former stance means they may have difficulty understanding the complexity of writing feedback . Further, writing can generate fear and anxiety in some academics (Lee and Boud 2003). This may be especially true for new supervisors struggling to develop their own writing (Maher and Say 2016). Reports of writing support designed specifically for supervisors are limited, but Lee and Boud (2003) demonstrated that institutionally funded and faculty-situated writing workshops for academics created a shared language for describing the writing process and increased confidence and productivity. As with PhD researchers, disciplinary cluster (Lee and Boud 2003) and cross-disciplinary initiatives (McGrail et al. 2006) are useful. 18. Funding: Having funding has a positive influence on satisfaction (Harman 2002) and completion (Jackson and Michelson 2015). Not surprisingly, lack of funding leads to individuals being less involved in the departmental intellectual climate. Interestingly, the same is true of scholarships, since while they carry recognition value, they do not engage holders in research or teaching with faculty and researchers. Rather, it is research (RA) and teaching assistant (TA)-ships that draw individuals into the department (Lovitts and Nelson 2006), though TA-ships may take time away from doctoral progress (Gardner 2009). Further, those with RA-ships have the highest likelihood of degree completion compared with other forms of financial support (Ampaw and Jaeger 2012). This evidence again suggests the need for a joint approach with departments having considerable freedom in how institutional financial support for researchers is distributed.
Overall, the above evidence suggests that policies should be directed: (a) institutionally at structures and infrastructure, including career and counselling services; (b) departmentally at prompt completion, supervisor change, research team engagement, and intellectual climate; and (c) collectively at writing support and funding. Such policies will reduce burnout and completion risk and contribute to more engagement in research, greater productivity, more timely completion, and greater satisfaction with doctoral studies (see, e.g., Sala-Bubaré et al. 2018). In contrast, lack of such supportive environments increases dropout risk among doctoral researchers (e.g., Gardner 2010) and has been related to isolation, stress, burnout, and abandonment (Castelló et al. 2017;Peltonen et al. 2017).
Our policy suggestions at the institutional level: → Commit institutionally to structural PhD review and reform.
& Agree on a clear purpose for the reform and create an institutional plan for the reform that allows for departmental adjustments within the requirements. & Consider the value of creating policies that move from single to co-supervision. & Create efficient monitoring systems that enable proactive management of potential problems. & Monitor and consult on program/departmental implementation and progress for new policies/practices → Ensure university services are well resourced and that information systems enable PhD researchers to easily learn and access them.
& Invest in institutional resources for PhD researchers that are more efficiently offered at the institutional level, e.g., career counselling. & Offer writing support for PhDs, negotiated jointly with supervisors and departments/ faculties. & Ensure evidence-based training for doctoral supervisors to enable them to understand the core determinants of positive doctoral experience. & Make sure specific counselling services are prepared to help PhDs. & Ensure career services adequately address post-PhD careers.
→ Departmentally, review selection processes, supervisory change procedures, and funding "packages" & Funding: Incorporate departmental responsibilities in scholarship funding; create mixes of TA-and RA-ships, without overloading doctoral researchers with additional tasks, in order to promote researcher community integration; ensure pay guidelines meet institutional, funding council, and national standards. & Selection process: Increase rigor in assessing applicants for their preparedness. Admissions should be based on tasks that are similar to the actual work carried out by doctoral researchers and not undergraduate grades. Further, tweak admissions processes to engage applicants to consider career opportunities. → Departmentally, review support for alternate careers & Consider engaging alumni with PhDs as potential mentors for junior researchers that aspire to careers outside the academia. & Consider external internships or placements of some kind. & Rethink the role of theses (monograph or publications; also types of publications) to ensure researchers have opportunities to become active agents of their future communities (academic and non-academic).

What do we recommend to policymakers?
We undertook this review to generate evidence-based policy implications that might be useful for directors of graduate studies, department heads, and institutional senior management teams. In our view, the starting point for the policy is ensuring that individuals accepted into PhD programs experience satisfaction and progress through their degree. While these researchers will likely focus on the personal relationships they have with those in their program, as Driver (2006) argued, such individual aspects of supervision can best be supported and enhanced not only through policy creation but also through institutional monitoring and oversight of policy implementation. Such oversight means that administrators have the potential to track implementation with diligence and clarity so that the personal relationship can grow.
Yet undertaking change is not easy. Institutional and local leaders are needed to deal with the range of institutional constraints. Also, affordances need to be used effectively. For instance, providing increased financial and/or human resources can be a motivator. Robust tracking systems are also needed to ensure changes are occurring; in this instance, they should, we believe, include career tracking post-graduation. The results of the monitoring can be disseminated and robust practices rewarded.
We hope that this synthesis will be taken up by stakeholders, policymakers, and also other researchers and that we can collectively become change agents-at least in our own institutions-for educational reform that could better prepare PhD researchers for their futures and supervisors for their important role in the doctoral journey. At the same time, our review highlighted gaps in the research which we hope will be taken up in future research efforts. There is still much we do not know.

Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of interest.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.