Ph.D. stipends and program placements
Motivation
It is a truth universally acknowledged, that a Ph.D. student in possession of a stipend, must be in want of a raise.
Ph.D. programs vary widely between countries and disciplines, so it is hard to generalize. But in many programs, Ph.D. students are funded: the university covers the cost of their tuition and pays a modest stipend to cover the student’s living expenses. This stipend may require the student to do some work outside their academic studies, such as serving as a teaching assistant for courses, or it may pay them to conduct their thesis research. Nominally, this pay is for about 20 hours of work a week, so it is not a full-time job even though the student may spend much more than 40 hours a week on their academic work.
Whatever the case, Ph.D. stipends are paid for from university funds, research grants, fellowships, and other sources, and they are rarely generous. Stipends are almost always lower than wages for jobs in the field, and are often below what is considered a living wage. Students thus often argue for higher stipends. There are many possible arguments—based on wages for comparable jobs, on living wage standards, on the effect of low pay on mental and physical health, and so on—but one possible argument is that improved pay may improve outcomes:
This analysis looks at graduate student pay from another angle: that of program outcomes. Setting aside the debate over whether students deserve higher wages, there may be another reason for which student pay should be higher. If better wages lead to greater success, then this will be reflected in program outcomes, and higher-paying departments can reap reputational benefits as a result. (Warner and Stecuła, 2024)
This data comes from a study that aimed to answer this question. The authors gathered data on 85 Ph.D. programs in political science in the United States, including their number of graduates, program ranking, location, and outcomes: how many graduates got tenure-track jobs. Stipends are often not listed publicly, so they gathered self-reported data from PhDStipends.com, where students can report the stipends they receive or were offered by Ph.D. programs.
Data
The data is split into two files. One file contains one row per Ph.D. program, giving metadata about the program. The other file contains one row per stipend report on PhDStipends.com. Each program may have more than one reported stipend, if more than one student reported it or if students reported stipends in different years.
Data preview
phd-stipends.csv
phd-programs.csv
Variable descriptions
| Variable | Description |
|---|---|
| program | Ph.D. program |
| stipend | Stipend amount (dollars) |
| coaiadj_pay | Ratio of stipend amount to county-level cost of living, as estimated by MIT Living Wage Lab (less than 1 = lower than cost of living) |
| acyear | Academic year stipend was reported |
| pryear | Student’s year in their Ph.D. program |
| stipendnotes | Additional comments listed by the student in their submission |
| Variable | Description |
|---|---|
| program | Ph.D. program |
| link | Link to the program’s web page |
| listed | Whether the program’s outcome data was available (1 = yes, 0 = no) |
| grads | Ph.D. graduates in the 2019-2021 academic years (if a list was publicly available) |
| ustenure | How many of those graduates obtained tenure-track jobs at US institutions during the 2019-2021 hiring cycles, as of January 2022 |
| alltenure | Same but for all institutions, not just US |
| non_tt | How many of those graduates obtained non-tenure-track academic jobs during the 2019-2021 hiring cycles, as of November 2023 |
| size | Number of Ph.D. students listed in the program’s directory |
| rank13 | The program’s rank in the 2013 US News and World Report program rankings (or 100 if the program was not ranked) |
| top_subfield | If the program was ranked in one of the political science subfields in the 2013 US News and World Report (1 = yes, 0 = no) |
| union | Whether the graduate students were unionized as of 2018 (1 = yes, 0 = no) |
| metro_rating | The size of the metro area in which the program is located (2 = top 20 metro areas by population, 1 = top 100, 0 = other) |
| endowment_fte | University’s endowment divided by the number of full-time enrolled students |
| type | Whether the university is public or private |
| region | US Census region of the university |
Questions
- Determine how to merge the program and stipend information. How much do stipends vary within a program? Should you simply take the average stipend per program? Merge the data.
- Conduct EDA. What is the typical Ph.D. stipend? Is it related to the program’s rank, the cost of living, whether the university is public or private, or the region? Do universities with higher endowments (per student) pay better stipends? Devise plots or tables to address these questions.
- The core question is whether better-paying programs have better outcomes, in terms of the fraction of students who get academic jobs. Select a model to answer this question. Can you find evidence for it?
- The authors of the original study claim that according to their results, the stipend amount “clearly and meaningfully impacts their ability to thrive on the highly competitive job market”. That is, they regard the association as indicating that stipends cause an increase in placement. Under what conditions can we make a causal claim? Are there confounding variables, or other variables we would need to measure?
References
Warner, S. B., & Stecuła, D. A. (2024). PhD stipends and program placement success in political science. Research & Politics, 11(2). https://doi.org/10.1177/20531680241241131
Warner, Seth & Dominik Stecuła (2024). Replication Data for: PhD Stipends and Program Placement Success in Political Science. Harvard Dataverse. https://doi.org/10.7910/DVN/FO7JBZ