Not Waving but Drowning
Depression, Anxiety and Suicidality among Economics PhD students
I was recently asked to deliver a keynote speech on the “hidden curriculum” of economics at a mental healthcare conference. I usually consider the “hidden curriculum” as “all the stuff in economics they don’t teach in grad school but which is critical to success”. For an applied microeconomist like myself, this includes things like language agnostic programming skills, version control, empirical workflow and publishing.
But those topics did not seem especially relevant when speaking at a conference on the economics of mental health, so I decided to instead discuss the mental health of PhD students. In today’s substack, I would like to do that by talking about two studies: one on the mental health of all PhD students, and another one on the mental health of economics PhD students. The keynote video is embedded at the bottom of this substack entry.
The title of this substack, “Not Waving but Drowning”, comes from a poem of the same name by 20th century British poet, Stevie Smith.1 Here I include the poem. I encourage you to read the poem aloud when you are somewhere alone as that is a good way to immerse yourself in the sounds and lyrics of any poem.
“Not Waving but Drowning” by Stevie Smith
Nobody heard him, the dead man,
But still he lay moaning:
I was much further out than you thought
And not waving but drowning.
Poor chap, he always loved larking
And now he’s dead
It must have been too cold for him his heart gave way,
Oh, no no no, it was too cold always
(Still the dead one lay moaning)
I was much too far out all my life
And not waving but drowning.
There are a few things I’d like to point out in this poem, some of which I will frame in a way that will be relevant to economists’ way of thinking. The first is that the imagery, while tragic, is beautiful. A man, far enough out to sea so as to be visible, is swimming. He is waving to the people on the beach. You can almost imagine them smiling and waving back.
But this is a story of asymmetric information and pooling. The individuals on the shoreline are witnessing a tragedy but do not know it because they do not know the man’s type (Harsanyi 1967; 1968). There are essentially two types of swimmers: a man drowning and a man larking. Given these two types, witnesses have different best responses: swim out and rescue the drowning man if he is, in fact, drowning, or laugh and wave if he is larking. They ultimately chose to wave, probably in no small part because rescuing is higher cost and made less likely by bystander effects. The drowning man knows his type, but the witnesses do not, and so they do not stop the tragedy even though they could have.
This leads to my third point: the pooling equilibrium. Pooling equilibrium happen in bayesian games when signals are cheap for all types. While the drowning man signals his dangerous situation (i.e., he waves for help), the signal is not cost discriminating because the cost of sending it is the same as the cost that larking swimmers incur. Without cost discriminating signals, observers often cannot update their beliefs about what is truly going on.
It is not that the witnesses on the shore were bad people. They were not necessarily indifferent to the man’s predicament. Rather, the witnesses simply did not understand the message they received. Thus despite communicating to the shore, the man drowns because people misunderstood the signal. I would suggest to you that this is the metaphor to consider when thinking about the struggles of our graduates all around us, and perhaps even more people than that.
Two New Articles
There are two recently published articles that I would like to bring to your attention related to the mental health struggles of PhD students in our programs. The first is a new paper coauthored by me and 12 others entitled “Systematic Review and Meta-Analysis of Depression, Anxiety Suicidal Ideation Among PhD Students” (Satinsky, et al. 2021). The purpose of our study was to summarize what had been published on mental health struggles among PhD students using a meta-analysis methodology. Our abstract reads (with emphasis mine):
“University administrators and mental health clinicians have raised concerns about depression and anxiety among Ph.D. students, yet no study has systematically synthesized the available evidence in this area. After searching the literature for studies reporting on depression, anxiety, and/or suicidal ideation among Ph.D. students, we included 32 articles. Among 16 studies reporting the prevalence of clinically significant symptoms of depression across 23,469 Ph.D. students, the pooled estimate of the proportion of students with depression was 0.24 (95% confidence interval [CI], 0.18–0.31; I2 = 98.75%). In a meta-analysis of the nine studies reporting the prevalence of clinically significant symptoms of anxiety across 15,626 students, the estimated proportion of students with anxiety was 0.17 (95% CI, 0.12–0.23; I2 = 98.05%). We conclude that depression and anxiety are highly prevalent among Ph.D. students. Data limitations precluded our ability to obtain a pooled estimate of suicidal ideation prevalence. Programs that systematically monitor and promote the mental health of Ph.D. students are urgently needed.”
To gauge these magnitudes, consider the following benchmark: using American data, two surveys show that the prevalence of major depressive disorder to be between 13 and 15% for young people ages 18 to 29 (the age range of PhD students). The prevalence of generalized anxiety disorder, for young adults of the same age range, is 4%. Thus the prevalence of depression and anxiety among PhD students is considerably higher than what is observed in the general population.2
Yet, while these data are very worrisome, they may not be representative of what the typical economics PhD student experiences. After all, one of the positive features of a PhD in economics is that our students will graduate into job with very low unemployment rates, often reporting satisfaction with their jobs upon graduation. The combination of low unemployment, relatively high compensation and the possibility of pursuing intellectually fulfilling research topics are only three of the perks of being a professional economist.3
Insofar as contemporaneous graduate school experiences are hard, then economics PhD students may unnecessarily struggle even with such bright professional futures. This is where a new paper entitled “Graduate Student Mental Health: Lessons from American Economics Departments” forthcoming in the Journal of Economic Literature by Valentin Bolotny, Matthew Basilico and Paul Barreira documenting the unique struggles of economics PhD students with original survey evidence is particularly insightful for thinking about this problem.
I would like to summarize the paper along three dimensions: a discussion of the prevalence of mental health symptoms among PhD students in top 10 programs, correlations between those symptoms and other statements in their survey that suggest possible mechanisms and suggestive policies for us to consider.
Prevalence of Mental Health Symptoms
The authors in this study fielded a survey of PhD students in 8 of the top-ranked economics departments in the US. The authors did not diagnose mental health disorders in this survey, as that can only be done by a licensed professional through in-depth interviews. Rather the authors documented mental health symptoms using the PHQ-9 survey for depression and suicidality, the GAD-7 for anxiety, the SBQR for suicidality as well as other instruments used to measure loneliness. The authors find that the “prevalence of poor mental health in these economics PhD programs is substantial.”4
Depression and Suicidal Ideation
The PHQ-9 survey instrument used to measure the prevalence of depression symptoms includes nine questions. These nine questions are related to the primary symptoms of Major Depressive Disorder such as mood, sleep, interest, guilt, energy, concentration, attention, psychomotor slowing, and suicidality. Note that the ninth symptom in this list of symptoms is suicidality, which is one of the main outcomes the authors study in addition to depression more generally.
Similar to our meta-analysis, the authors report that 17.7% of students are experiencing moderate to severe symptoms of depression, which they note is 2 to 3 times the national prevalence. This is the ballpark of what others have found, including ours. One of the more problematic findings is that only 25.2% of these students are receiving treatment. Suicidal ideation is 1.5 times higher than survey respondents of young people in the same age range.
Loneliness and isolation
Loneliness and anxiety are also high among economics PhD students. One of the more interesting comparison groups in this paper is to that of retired American workers. They write:
“Loneliness is common among our survey respondents, with the average student finding himself or herself considerably lonelier than the average retired American. The mean score on the UCLA 3-item loneliness scale was 5.2 with a standard deviation of 1.8. For a sample of over 2,000 retired Americans in 2002, that score was 3.9, with a standard deviation of 1.3” (my emphasis).
Students also report high levels of isolation and while isolation is not identical to loneliness, it can be it independently contribute to mental health struggles too. 16% of survey respondents say they regularly experience isolation from others, or at least feel they do. How high is 16%? Again, consider as a benchmark the general population: 11% of Americans report feeling isolated from others. 17% of our PhD students also say they felt they lacked to companionship compared to 13% of Americans overall.
The GAD-7 is a commonly used survey instrument for studying anxiety and using it, the authors find that anxiety symptoms among economics PhD students is substantially higher than the general population. Whereas around 6-7% of Americans experience anxiety symptoms, around 17-18% of economics PhD students do. Economics PhD students, in other words, experience anxiety symptoms that is on average 3x worse than the average American in the same age range.
These levels of anxiety are unequally distributed across subpopulation. Females, for instance, have slightly higher rates of anxiety than male students (19% versus 16%). While US and international students experience roughly the same anxiety rates, I was surprised to learn that anxiety rates were slightly better for minorities than the average student.
The one disparity that really jumped out for me was that of LGBTQ students. Their anxiety rates were 22% which is higher than the 17-18% of all economics PhD students noted earlier.
In the age of the credibility revolution, we have not always valued descriptive studies. This is unfortunate because describing empirical regularities is often the starting point causal analysis. I applaud, therefore, the authors for fielding this survey as it allows us understand both the prevalence of mental health struggles of students, as well as to consider potential causal mechanisms even if those mechanisms are purely speculative. Such suggestions are valuable as they provide researchers with possible avenues of future research, as well as concerned faculty members with potential mechanisms they might consider utilizing in the meantime. So what might be some possible mechanisms explaining these high rates of mental health symptoms?
Potential Causal Mechanism I: Event Time
These high rates of mental health symptoms have at least two broad explanations: they are either caused by selection into the programs or they are caused by our PhD programs themselves. If the former, it is not as clear what mechanisms we should entertain. But if the latter, then we might consider policies targeting the department experience itself. The answer appears to be, not surprisingly, a bit of both selection and the programs.
Consider the degree to which symptoms both exist at entry as well as increase as students progress through their PhD studies. Whereas 21% of first years experience moderate to severe depression symptoms, 37% do so in years 6+. Suicidal ideation also shows progressively worse symptoms with time in the program: 8% of first years reporting suicidal ideation within the last two weeks versus 23% in their 6+ year.5 There are other signs that this is not merely due to selection. Around 20-30%, for instance, received a diagnosis during their program while 10-13% had one before their program.
One of the things that may be happening is students are experiencing never before seen negative shocks that increase in frequency over the course of their graduate life. Many students in economics PhD program are positively selected from earlier schools. Put a different way, they are and have always been some of the smartest people in school. As such, the accumulation of stress and failure over the life cycle of their programs may take a substantial toll on their psyche and mood in part because it is so unfamiliar.
Potential Causal Mechanism II: Social Interactions
Two of the most important relationships one has in graduate school is the main adviser and other students. While other faculty are likely very important, the adviser is unique because they serve on the front line and at a frequency that their encouragement — or discouragement — can be very impactful on the student’s mental health. This is, again, where paradoxes can be spotted in the data.
First some good news. The prevalence of friendship is high. 73% say they have very good friends in the department. Complete social isolation is rare: only 3% say they never turn to someone when they have problems or worries, and 6% report having zero people in their lives they can talk to about their innermost feelings.
Nevertheless, there are weird signs that students nonetheless feel disconnected. Consider seminars for instance. Around 29% say they are comfortable speaking up in a seminar; 77% say they would only speak up if they were absolutely certain that their comments were good ones. Women feel an especially high barrier in seminars: 19% of women are comfortable voicing opinions versus 35% of men.
Potential Causal Mechanism III: Advisers
The frequency of meetings with advisors does not appear to be a problem: 96% met with their adviser at least once in the last 2 months. Even meetings with second and third advisers are high. So if it isn’t the frequency of meetings with advisers, what might it be? It may be what students feel their advisers think about them as people. Keep in mind — feelings do not have to be true for them to be real. The important point is what students feel about their advisers’ opinions, not what their advisers opinions actually are.
Many students do not feel that they are properly supported by their advisers. 19% think their advisers do not care at all about the success of their research. 27% to 35% of women and men, respectively, say their advisers do not care about them as a person. This worried me because a person is our fundamental identity. It is far more closely aligned to other longterm and deeply held identities such as the more salient category of student or economist. So to feel rejected on a fundamental level by advisers, who most need and look up to, might be as challenging as the rejection a child feels from a neglectful or judgmental parent.
Role models are one of the main ways that some of us navigate and understand our historical lives, and many students say they have trouble finding suitable ones. 18% of economics students say they do not have a professional role model, and again, there are disparities by sex and citizenship with women and international students experiencing it even less so.
But perhaps this is optimal? Maybe the reason students cannot find role models is because they don’t want or need them, at least not in the profession. Maybe the lack of concern they feel from their advisers is no big deal because they get those needs met elsewhere. Through a series of interesting questions, the authors show that this is probably not the case because there is a difference between what students would like to be the case and what they feel is the case.
They experienced these gaps in a number of directions such as when considering non-academic options (presumably with students feeling advisers would be less open to this and thus students being less likely to discuss it openly), when preparing for the job market, the progress of their research, issues they have with other advisers, coauthorship issues with other faculty and students, their presentation and refereeing experiences and skills, teaching success, mental health, decisions about starting a family, and personal issues more broadly. They even experience gaps in whether they have been let into the program in the first place. All signs point to the fact that these gaps in what they want to be true about open communication with advisers and what is true are pervasive and not desired.
Potential Mechanism IV: Sexual Harassment
The AEA has over the last couple of years taken seriously the problem of sexual harassment within the profession. But one of the things I found particularly surprising in this survey was that most of the sexual harassment that students experience comes not from faculty, but from other students.
Around 16% of respondents report ever experiencing sexual harassment with higher rates among women (22%) than men (13%), and higher rates among US students (22%) than international students (9%). These cases are considerably more likely to come from other students (63%) than faculty (19%). Only 10% came from a stranger.
These forms of sexual harassment include sexual remarks, jokes, or insulting and offensive stories; inappropriate comments about body, appearance and sexual activities; crude or gross sexual comments used to try and pry unwanted sexual conversations out of the student; requests for dates, drinks and sex despite being told “No”; email, text, phone calls, and DMs with various forms offensive sexual remarks.
Who doesn’t love it when at the close of a seminar, that one person in the audience raises their hand and asks “what do you think are the policy implications of this study?” Feel free to ask it now. The authors suggest a few possible policy implications that I think warrant our attention. I will mix my own thoughts throughout as well.
Suggestive Solutions I: Treatment
With such high rates of depression, anxiety, loneliness and suicidality, you would hope that at minimum students are engaged with mental healthcare resources, such as outpatient meetings with licensed therapists or psychologists, but that number appears too low to me. But only 15% of students are in treatment for some issue which rises over time. Treatment rates rise from 8% in year 1 to 19% in year 5 and 32% in year 6+. Those with moderate or severe mental health symptoms see higher rates of treatment ranging from 14% in year 1 to 27% in year 6+.
Part of the under utilization of mental healthcare resources may be a result of not well understood barriers such as stigma. While 87% know about these resources, only 55% say that they would be likely to use them. One very alarming finding is that these numbers are even lower for suicidal thoughts. Future research should try to understand why a cohort of extremely talented PhD students can simultaneously know about mental healthcare resources, experience substantial symptoms that might benefit from utilization and yet not do take it up. Policy experimentation to try and match struggling with students mental healthcare resources should also be given serious thought.
Solution II: Socializing
Having good friends in the program is positively correlated with student mental health, as is good relationships with one’s adviser. It is likely that supportive advisers and peers can be a potential mechanism by which at the margin student mental health could improve.
Speaking from experience, and acknowledging that I don’t know the counterfactual, I believe that my positive and supportive relationship with my adviser (and eventually coauthor, mentor and friend), Christopher Cornwell, was the source of a particularly positive and healthy PhD experience. But what does one do when the set of potential advisers may not be particularly good at the kind of supportive relationship that students prefer or need. After all, economists don’t have the best reputation of being all that approachable when it comes to the harder stuff. In fact, if anything, their outside reputation is that they may be a bit insular with a tendency to skew towards intellectualism at the expense of more supportive connections.
One of the things to remember about honest and open social relationships is that oftentimes they require frequent interactions of a particular kind. The psychological sense of community, after all, is largely a function of small groups of people with a common purpose meeting frequently. In other words, openness with one another might be partly mathematical and deterministic in nature. Graduate coordinators might want to consider organizing fairly normal, even fun, social experiments like pizza night together or other group activities such as game night.
As these kinds of activities require time, money and other resources, departments will need to prioritize it so as to justify allocating scarce resources to it. Departments may want to consider subsidizing such events as they can function like platforms matching students to other people. When accompanied by responsible leadership, frequent social events outside of the seminar experience can become an important input in the creation of open and supportive friendships. Small groups and one-on-one time together are also mechanisms which can match students with others and thus departments may want to keep that in mind as a possible goal when allocating scarce resources.
But doing these sorts of group activities is also a fine line to walk given the absolute importance that advisers and students maintain strictly professional boundaries with one another for the sake of graduates student safety and overall health. I have heard it said that faculty and students cannot be friends, and while I won’t make strong statements about that, I fully acknowledge that the potential for power imbalances and exploitation could be an issue when inappropriate intimacy is introduced into the relationship. Faculty may want to get advice from others about where the lines should be before creating overly personal entanglements with students if such lines are not intuitively known as they are risky, not only to the student, but to the faculty themselves.
Finally departments may want to consider the role that physical space can play on social interactions. Jane Jacobs, in the phenomenal classic The Death and Life of Great American Cities hypothesized that sidewalks, short city blocks and the mixture of residential and commercial uses can have significant effects not only on reducing neighborhood mishap and bad behavior but also in the coordination and creation of community. Perhaps resources should be allocated so that students are naturally and frequently forced to see and interact with one another. Such experiences can have the byproduct of creating friendship through the construction of public spaces like designated locations that guarantee seeing and interacting with one another.
Solution III: Encouragement and Self Care
Related to Solution II is simply the importance of encouragement. Students with low mental health scores report lower perceived encouragement from faculty, so it may be worthwhile that faculty regularly encourage students. We should hold students to high expectations and believe that they can attain them. Students need to regularly hear that we believe in them, are proud of them, and that they are doing a good job.
We should also encourage students to invest in their own self care. This requires publicly and privately encouraging students to take care of themselves as it can help prompt them to think about personally optimal work-life balance. This has both future benefits, but more importantly present ones as well.
Skills-based mental health treatment, or what is often terms cognitive behavioral therapy, should also be on the table. Though economists have studied its efficacy in a few papers, its actual practice has not really been touted as much as it probably should be within the profession. Mindfulness exercises, meditation, exercise, art, movies, or whatever students may deep down love to do should be actively and regularly encouraged by our faculty. Likewise investment in skills that help them tolerate distressful feelings as well as identify various cognitive errors common under stress and dissatisfaction can all be addressed fruitfully with widely available technologies such as cognitive behavioral therapy and life coaching.
Solution IV: Field-Specific Research Groups
The authors recommend that departments institute policies that can help identify potentially drowning students, as opposed to simply encouraging the private actions of faculty. Policies lead to institutions and institutions may help students not swim so far out or alone at all. These policies can take on different forms, but departments may want to consider some that simultaneously increase the psychological sense of community while also keeping students on the track to success. Research groups, for instance, that meet regularly with both one another and faculty can help students feel their work is meaningful while helping them with their research and presentation skills.
These types of activities will always require leadership from faculty participants, and it’s possible that that leadership should come from those faculty who have the comparative advantage in providing it. Departments might consider, therefore, experimenting with various models, designs and incentives, including compensation, to reward such invisible yet valuable forms of service. Attention should also be given to the overburdening of women with such tasks, too, as they as a demographic have traditionally been disproportionately asked to serve in service work, most of which is often unrewarded in the department.
Solution V: Meaningful Research, Job Satisfaction and Careers
Students show signs of diminished job satisfaction. Whereas 77% of economics faculty said that they found their jobs meaningful, only 37% of students in the sample did. 26% reported feeling their work was useful versus 70% for faculty. 63% of people in the general population for the same working ages, or more than twice that of the PhD students in their sample, said they were satisfied with their work.
Part of the struggle here may be due to mismatch — not so much between the student and the job as much as the students’ talents and the jobs required. Only 40% of students felt they had opportunities to fully use their talents versus 85% of faculty and 53% of the population. The authors write:
“The economics PhD program thus appears to be distinct from the average occupation and from the economics professorship in the rarity with which one experiences satisfaction, usefulness and meaningfulness.”
One of the ways that students may have fewer negative feelings and opinions about their own work’s meaningfulness is simply to encourage them to work on projects that they do find meaningful. One of the things my adviser was particularly good at doing for me was that when he saw I was excited about particular projects, he tended to nudge me towards those with the highest upside conditional on that enthusiasm. Perhaps we can and should be doing more of the same.
Encouraging students to be open to a diversity of possible career options might also be a way to address the negative feeling students feel about their present work’s meaningfulness. Many students do not share the same talents or preferences as their advisers. Such differences may inadvertently cause students to sort into work they do not like simply because they believe their advisers want this (whether or true or not).
Departments may also want to consider, therefore, bringing alumni back on a regular basis from a diverse set of careers to talk about their jobs, what they like, what they don’t. In addition to seeing future possibilities, this may also help create attachment to role models which are so often missing.
Students are our in our care and therefore we bear a certain amount of responsibility to these mental health struggles. It is likely that these mental health struggles are endogenous, not only to the students own choices, but to our personal and collective choices as well. They are young, they are working hard, and their struggles are often unknown to the very people, like advisers and friends, who might be in a position to help. The first step is to learn more about the pervasiveness of these problems. I encourage you read these two papers with curiosity and an open mind.
But positive analysis, without policy experimentation, can easily become virtue signaling, posturing and unproductive navel gazing. Ultimately we seek to create a good society, one that addresses and minimizes the unnecessary harms to Earth’s inhabitants. Students are our neighbors, and if they are suffering, we are most likely obligated as people as well as educators to search out and answer their calls for help.
I encourage all readers to think and discuss these issues, both with other faculty, but perhaps more importantly with their students themselves. Talking to students about these papers could help identify unknown problems, and knowing is half the battle because if we do not know about these threats, then we cannot address them. Having identified them, perhaps concerned faculty will take initiatives to create policies in their departments aimed at improving the mental health of our students. They deserve it.
Stevie Smith lived from 1902 to 1971. She struggled with depression her entire life. The specific manifestation of those depressive symptoms was associated with her intense shyness and sensitivity according to Wikipedia, which many readers can attest is a particularly painful cocktail of emotions to drink.
Data limitations on the prevalence of suicidal ideation kept us from being able to make definitive statements.
Economists have themselves contributed to the efficiencies enjoyed by graduates from the hosting of centralized platforms used for first round interviews like the annual ASSA meetings and the introduction of a signaling mechanism in 2010. Further readings can be found on the AEA website.
The authors are careful in this study by taking into consideration non-response bias and including bounds based on it. I will discuss their average prevalence in this paper, but even when taking in lower bounds, the prevalence of these symptoms are still higher than the general population which is obviously extremely problematic for those of us worried about our students.
Obviously it goes without saying that if those with lower mental health struggles graduate earlier, then the higher prevalence over the life cycle of the PhD program may simply be a reflection of survivor bias. I personally suspect that that explanation is a stretch, though. Still, feel free to collect the panel data to test it.