3

This appears to be a difficult thing to search for, given the plethora of stuff out there on, say, the maximum number of universities to apply for a PhD, the person with the most PhDs, etc., when I search for things like, "What is the highest number of PhD attempts for one person?" - and so on.

There doesn't seem to be anything like this question here on Academia SE.

The Question:

What is the distribution of the number of PhD attempts among PhD holders?

Clarification:

I mean, does it fall on a bell curve? What's the average? What is the most number of failed PhD attempts for a PhD holder?

What classifies as a failed attempt? To me, starting but not obtaining the PhD, for whatever reason.

A "PhD attempt" here, I suppose, would be enrolment on a PhD programme.

If this data could be sorted by country, subject, etc., then that would be a great; I'm just looking for the overall statistics though.

This reminds me of the popular claim that the average millionaire goes bankrupt at least 3.5 times. Something like that, but in PhD terms, might serve to motivate students like me.

I'm on my second attempt.

13
  • 6
    The vast majority of people make zero attempts. A few make 1 attempt; of those, a fair number succeed (how many will depend on field, country, institution). Rarely, people make more than 1 attempt; their individual situations are too unique to analyze as a group.
    – Bryan Krause
    Commented Jun 23, 2023 at 23:06
  • 4
    How do you define "failed attempt"? Being in grad school without getting a degree and leaving? Or showing up for a defense and failing it? Commented Jun 23, 2023 at 23:09
  • 5
    Like the millionaire statistic you give (and whether or not it's actually true), the way you've framed it is actually a very misleading way to approach the question, especially if you're trying to figure out whether bankruptcy is good or bad. It would make more sense to ask how many people who are in a PhD program will complete it successfully. And perhaps also worth asking whether that rate is different for people who have left or failed out of a program previously.
    – Bryan Krause
    Commented Jun 23, 2023 at 23:14
  • 13
    It can't be bell-shaped because the vast majority of counts (surely the median/mode/rounded mean) will be the minimum possible number of "1" (as per current question title). Commented Jun 24, 2023 at 1:17
  • 4
    Certainly an outlier, but here's my dismal history: left a Ph.D. program after 4 semesters (did not take any qualifying exams), then completed Masters in 4 semesters at a Masters-only (at that time) department, then left another Ph.D. program after 3 semesters (took and did not pass qualifying exams twice, which resulted in dismissal from the program), then worked for 1.5 years (community college part-time, substitute HS teaching, full-time summer HS teaching, 1 full-time year HS teaching, and some "lesser" jobs), then successfully completed Ph.D. in 4 years at a 4th university. Commented Jun 24, 2023 at 12:03

2 Answers 2

5

As a partial answer, consider this other question:

Is it true that almost everyone who starts a PhD and sticks around long enough can get one?

The selected answer there boils down to, "this is pretty much the case". Top answers present the following statistics:

  • Success rate in the UK is around 80% (Times Higher Education), with the expectation that the rate is similar throughout Europe.
  • Success rate in the US is around 50% (Chronicle of Higher Education).

If we put these together then the success rate across PhD programs in the Western world must be above 50%. And this implies that the mode, the median, and the rounded mean (whatever flavor of "average" you like) will be "1"; and that the distribution will be necessarily right-skewed.

To the extent that in the cited statistics, some people are being double-counted on the negative side (failing multiple programs), then this would increase the individual success rate (i.e., a higher proportion with the modal count of "1"). On the other hand, I assume that the number of people being double-counted on the positive side (getting multiple PhDs) is negligible.

3

I am not aware of any existing empirical research on the number of "attempts" within a PhD program, but I'll give some conextual information that is relevant for anyone planning to undertake this type of research, and the types of barriers and issues that may arise. It is noteable that there is existing data on the time-to-complete for successful PhD candidates, and this could be useful as a rough proxy for what you want to know.

At most universities in most countries, the PhD candidature involves a set of milestones and progress checks throughout the program, followed by submission of a dissertation that is reviewed against a set of possible outcomes ("scores") that say whether it is acceptable for the degree and whether the dissertation requires any revisions. The scoring rubric may differ between universities, but it would typically use at least four categories:

  • 1 - Acceptance: Dissertation is acceptable without further revisions, so degree should be awarded now. (Referees might make suggested revisions that are non-compulsory.)
  • 2 - Revision with internal review: Dissertation requires minor revisions and re-review that can be done internally at the university (withour sending back to referees), so degree should not yet be awarded.
  • 3 - Revision with external review: Dissertation requires revisions and subsequent review by the referees, so degree should not yet be awarded.
  • 4 - Rejection: Dissertation is not of sufficient quality for the award, even with revisions.

As noted, different insitutions (and different countries) may have different requirements for their PhD degree and may have differences in their scoring system. Some will use a more detailed scoring system than the above, but usually this will include an outcome for acceptance, outcome for revise-and-resubmit (with internal or external review), and an outcome for rejection.


What would research on this require? If one were to undertake empirical research on PhD "attempts", a substantial part of this research would be looking at the number of submissions for the dissertation before it is accepted and the outcomes of these submissions under the relevant scoring rubrics. (I note that you are conditioning on award of the PhD, so you are not interested in failed attempts.)

These submission scores are private information held by the university (much like class marks), so it would not typically be accessible by an outside party. Consequently, in order to obtain this information for research you would need to undertake an ethics review, make a successful request for the information to the university, and you would need to demonstrate appropriate safeguards in the research which would make the data safe and the published results non-identifiable. Since you would want data from a substantial number of universities (which is unlikely to be held centrally), this would be a large data-sourcing exercise.

Moreover, because the scoring systems at universities may differ, you might need to account for different pathways to acceptance at different institutions. This could add a layer of complexity to the research, since it could potentially make the number of "attempts" non-comparable across different universities. Finally, there may also be other requirements of different PhD programs in addition to submitting a successful dissertation. Other requirements may also differ between universities and could potentially make it difficult to count or compare "attempts".

2
  • 5
    While the OP has asked a possibly unanswerable question, this unfortunately seems to address a different issue. It looks like you address "number of submissions for the dissertation", while the OP seems to be asking about number of PhD programs entered, which is very different (and possibly even harder to gather statistics for). Commented Jun 24, 2023 at 1:22
  • 2
    @DanielR.Collins: Okay, it looks like I misunderstood his question. Thanks for bringing that to my attention. (Yes, that does seem even harder to get statistics for.)
    – Ben
    Commented Jun 24, 2023 at 1:25

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .