Nagarjun's blog: How Often Do People Actually Fail a PhD?

A global look at PhD pass rates, failed vivas, attrition—and why the numbers are surprisingly difficult to compare

In my previous post, When Geniuses Nearly Failed Their PhDs, I looked at some remarkable stories from the history of science: Werner Heisenberg's disastrous oral examination, David Hestenes's actual failure of his doctoral oral, J. Robert Oppenheimer's intimidating examination, and several other memorable cases.

But those stories raise a natural question:

How often does someone actually fail a PhD?

The answer is surprisingly difficult to give.

One might imagine that universities routinely publish a simple statistic such as:

"92% of PhD students pass, while 8% fail."

In reality, there is no universally comparable international PhD pass rate. And there is a very important reason for this.

"Failing a PhD" can mean several completely different things.

A student can leave a doctoral programme before submitting a thesis. A candidate can fail a qualifying examination. A submitted thesis can be rejected by external examiners. A candidate can fail a viva. A thesis can be returned for major corrections or resubmission. And, in some countries, a doctorate that has already been awarded can subsequently be subjected to another layer of quality control.

These are not equivalent events.

Once this distinction is made, a fascinating pattern emerges:

The proportion of people who fail to complete a PhD can be substantial, while the proportion of candidates who actually reach the final examination and fail there is usually surprisingly small.

Three different meanings of "PhD failure"

Before comparing countries, it is useful to distinguish three statistics.

1. Attrition

This is the proportion of students who begin a PhD but leave without obtaining the doctorate.

This can happen for many reasons: academic difficulties, financial problems, changes in career plans, health or family circumstances, problems with supervision, loss of funding, or simply the realisation that a PhD is no longer what the student wants.

2. Completion rate

This is the proportion of students who eventually obtain the doctoral degree, usually measured within a specified period.

A student who takes nine years to finish may count as a successful completion in one dataset and as a non-completer in another if the study only follows students for six or seven years.

3. Final-examination failure

This is the number of candidates who actually reach the thesis examination or viva and are ultimately denied the doctorate.

This is the statistic most relevant to the terrifying stories in the previous article.

And it is generally much smaller than the attrition rate.

The United Kingdom: about 96% of viva candidates succeed

The United Kingdom provides one of the more useful datasets for understanding what happens at the final examination.

An analysis based on information obtained from 14 UK universities examined the outcomes of 26,076 doctoral candidates who sat their vivas between 2006 and 2017.

Of these candidates, 25,063 ultimately succeeded.

That is just over 96%.

Approximately 4% therefore did not receive an immediate successful outcome at the viva stage.

But even this number requires some caution.

A PhD viva in Britain is not normally a simple "pass/fail" event. Examiners can require corrections, major corrections, or resubmission.

Indeed, the most common outcome is not an immaculate pass.

It is:

"Yes—but please fix these things."

This is important because a candidate who leaves the examination room with several pages of corrections has not necessarily had a disastrous viva.

They may be on the normal path to receiving the degree.

But what about everyone who started the PhD?

This produces a dramatically different number.

The same analysis estimated an overall UK PhD completion rate of about 80.5%, with approximately 16.2% attributed to students leaving their programmes early and about 3.3% to students who failed the viva.

The precise figures should not be treated as an official national UK statistic—the dataset came from 14 universities—but the distinction is extremely useful.

It tells us something important:

Most PhD "failure" occurs before the final viva, not during it.

Sweden: when final rejection becomes extraordinarily rare

Sweden provides an even more striking example.

A study examined doctoral dissertation rejections in the humanities, law and social sciences across Swedish universities between 1984 and 2017.

The researchers identified only 18 rejected dissertations among 15,477 submitted dissertations.

That is approximately 0.12%.

The cases involved dissertations that had actually reached the examining stage and were rejected by the examining committees.

At first glance, a rejection rate of 0.12% seems almost unbelievable.

Does that mean Swedish doctoral students are nearly infallible?

Obviously not.

The explanation is more interesting.

Doctoral education contains multiple layers of quality control before the final examination. Supervisors, departments, research groups and faculties all have opportunities to identify serious problems.

A weak dissertation can therefore be stopped, delayed or revised long before it reaches the final public defence.

The Swedish study itself discusses this tension: the rarity of rejection may reflect effective quality control, but it may also raise the question of whether some dissertations that should have been rejected are being allowed through.

This leads to a general principle:

The stricter the filters before the final examination, the lower the apparent failure rate at the final examination.

Germany: substantial attrition, but a different examination culture

Germany illustrates another problem with international comparisons.

Doctoral education in Germany does not simply reproduce the British thesis-plus-viva model. The precise structure of the oral examination varies among universities and faculties.

Studies of German doctoral education nevertheless show substantial attrition during the doctoral process.

One longitudinal analysis, for example, found that approximately 71% of doctoral candidates had successfully completed, while around 13% had dropped out and another group remained unresolved at the end of the observation period.

The important point is not the precise percentage, which depends heavily on the cohort and definition of completion.

It is that the major selection process occurs throughout the doctoral programme, rather than being concentrated entirely in one final examination.

The United States: the qualifying examination changes everything

The United States provides perhaps the clearest example of why "PhD pass rate" is an ambiguous expression.

American doctoral programmes frequently contain substantial examinations before the dissertation stage.

Depending on the university and discipline, students may have to pass:

coursework requirements;
qualifying examinations;
comprehensive examinations;
candidacy examinations;
proposal defences;
and finally the dissertation defence.

A student who fails a qualifying examination and leaves the programme is obviously a doctoral non-completer.

But that person never had the opportunity to "fail the PhD viva."

Consequently, American statistics on doctoral completion can look quite different from statistics on dissertation-defence outcomes.

This is why one should be extremely careful when comparing an American "PhD completion rate" with a British "viva pass rate."

They are measuring different stages of the process.

What about India?

This is where the statistics become much more difficult.

There does not appear to be a reliable, publicly available national dataset that allows us to say:

"X% of Indian PhD candidates fail their final viva."

That number should therefore not be invented.

India has enormous numbers of doctoral students and doctoral graduates, but national higher-education statistics generally provide information about enrolment, degrees awarded and disciplines rather than a comprehensive national breakdown of thesis-examination outcomes.

There is, however, something very important in the UGC regulations.

India has a major quality-control filter before the viva

Under the UGC regulations, the PhD thesis is evaluated by the research supervisor and at least two external examiners.

The viva is conducted only when the required external examination recommendations support acceptance of the thesis after incorporation of the suggested corrections.

If one external examiner recommends rejection, the thesis is sent to an alternate external examiner. The viva can proceed only if the alternate examiner recommends acceptance.

If the alternate examiner also does not recommend acceptance, the thesis is rejected and the candidate is declared ineligible for the PhD.

This produces a crucial observation:

In India, a candidate can effectively fail the PhD examination before ever reaching the viva.

Therefore, an Indian university's "viva failure rate" would not necessarily tell us how often doctoral candidates ultimately fail to obtain the degree.

China has something even more unusual: checking the thesis after the PhD has been awarded

China offers one of the most interesting contrasts.

The Chinese system has a formal national mechanism for post-award random inspection of doctoral dissertations.

Under the national regulations, approximately 10% of doctoral dissertations awarded in the previous academic year are randomly selected for inspection each year.

This is fundamentally different from the way most people imagine a PhD examination.

Imagine the following sequence:

The university examines the thesis.
The candidate successfully defends it.
The PhD is awarded.
The dissertation can subsequently be selected for national-level quality inspection.

The Chinese system therefore contains a form of post-award quality control.

The inspection is not simply a second viva. Selected dissertations are reviewed by external experts according to defined criteria.

If experts identify serious problems, the dissertation can be classified as a "problematic dissertation," triggering institutional consequences and remedial measures.

For example, in the 2018 inspection, the Chinese authorities reported that 6,572 doctoral dissertations were randomly inspected, representing approximately 10.4% of all doctoral dissertations awarded that year.

That is an enormous national quality-control exercise.

India and China therefore illustrate two different philosophies of quality control

The comparison is particularly interesting.

In India, the formal examination sequence places substantial emphasis on:

external thesis examiners;
acceptance of the thesis;
corrections recommended by examiners;
and the subsequent viva voce.

The UGC regulations explicitly prevent the viva from proceeding in certain circumstances where the thesis has not received the necessary external acceptance.

China, meanwhile, combines institutional examination with a national post-award sampling system in which roughly one in ten doctoral dissertations is independently inspected.

Neither system can simply be described as having a particular "PhD failure rate."

The quality-control mechanisms are distributed across different stages.

So does the PhD pass rate vary by country?

Yes—but not in a way that permits a simple league table.

Consider the following simplified picture:

Country/system	What we can reasonably say	Major quality-control stage
United Kingdom	About 96% of candidates in one large dataset succeeded at the viva	Thesis examination + viva
Sweden	Extremely rare formal dissertation rejection in one large humanities/social-science study	Strong pre-defence filtering + public defence
Germany	Substantial attrition during doctoral training	Supervision + institutional milestones + dissertation/oral examination
United States	Substantial doctoral attrition; final defence is only one of several filters	Qualifying/comprehensive exams + dissertation defence
India	No reliable national final-viva failure rate identified	External thesis examination + viva
China	No directly comparable national final-defence failure rate identified	Institutional examination + national post-award thesis inspection

The table therefore looks less like a ranking and more like a map of different doctoral systems.

Does it vary by discipline?

Almost certainly—but again, the distinction between attrition and final examination failure is critical.

Different disciplines have very different doctoral cultures.

A laboratory-based biomedical PhD may involve:

multiple experiments;
large datasets;
long periods of troubleshooting;
shared equipment;
animal or clinical work;
and dependence on funding and laboratory infrastructure.

A theoretical mathematics PhD may have a completely different structure.

A humanities doctorate may take substantially longer and may involve a very different relationship between supervisor, candidate and thesis.

Engineering, computer science, social science and the natural sciences each have their own patterns of funding, publication, supervision and completion.

These differences can strongly influence attrition and time to degree.

But we have much less evidence that the probability of outright failure at the final examination differs dramatically between disciplines.

That may be because the final examination is itself the end product of a long selection process.

The great paradox of the PhD pass rate

We can now understand an apparent contradiction.

On the one hand, a substantial fraction of people who begin PhDs never receive a doctorate.

On the other hand, once someone actually reaches the final examination, outright failure can be remarkably uncommon.

There is no contradiction.

It is the consequence of selection.

Think of the PhD as a series of filters:

Admission
↓
Coursework / qualifying examinations
↓
Research progress
↓
Thesis submission
↓
External examination
↓
Viva / defence
↓
PhD

By the time a candidate reaches the last box, many of the people who would have failed earlier have already disappeared from the population being examined.

Therefore:

A 96% viva pass rate does not mean that 96% of people who start PhDs get doctorates.

It means that approximately 96% of the people who survived all the previous filters and reached the viva passed it in that particular dataset.

And this changes how we should interpret Heisenberg's story

Return to Werner Heisenberg.

He did not simply stroll into a university after an undergraduate degree and almost fail a random exam.

He had already completed an advanced research project under Arnold Sommerfeld.

He had demonstrated extraordinary mathematical ability.

He had already established himself as an exceptional young theoretical physicist.

And yet, during the final oral examination, Wilhelm Wien identified a genuine weakness so serious that he reportedly wanted to fail him.

That is precisely why the story is so remarkable.

Heisenberg's examination demonstrates that:

A candidate can be extraordinarily good overall and still have a serious, examination-relevant weakness.

Conversely, the statistics tell us something equally important:

A genuinely catastrophic final PhD examination is unusual because candidates have already passed through multiple filters before reaching it.

What should count as a serious PhD-viva failure?

This distinction is especially important for examiners.

A candidate who cannot remember a minor fact is not necessarily demonstrating doctoral-level incompetence.

A candidate who cannot reproduce a formula from memory is not necessarily unfit for a doctorate.

And a candidate who says "I don't know" to an unexpected question may actually be demonstrating scientific honesty.

Much more serious are situations in which a candidate cannot explain:

the central methodology of the thesis;
why a particular analysis was performed;
what an important parameter means;
the assumptions underlying a major conclusion;
the limitations of the principal results;
why the data support the conclusions;
or how the work differs from previous research.

These are not merely memory failures.

They can indicate that the candidate does not have sufficient intellectual ownership of the research.

That is fundamentally different from failing to answer a difficult peripheral question.

The examiner is also part of the experiment

There is another lesson from comparing countries.

A PhD examination is not a purely objective measurement like determining the melting point of a chemical.

The outcome depends on:

the candidate;
the thesis;
the supervisor;
the examiners;
institutional rules;
disciplinary culture;
and the examination system itself.

That is why two equally strong candidates could potentially experience very different doctoral examinations in different countries—or even in two departments within the same country.

A particularly important safeguard is therefore the use of multiple examiners.

India's regulations, for example, explicitly provide an additional examiner when one external examiner rejects a thesis.

China's national sampling system similarly introduces an additional layer of external scrutiny after the degree has been awarded.

These mechanisms acknowledge an uncomfortable truth:

Examiners can make mistakes too.

So what is the "real" PhD failure rate?

There isn't one.

And anyone who gives you a single worldwide percentage without defining exactly what they mean is probably comparing apples with oranges.

A more honest summary is:

PhD attrition can be substantial.
Completion rates vary considerably by country, discipline, institution and cohort.
Final thesis/viva failure is generally much rarer than PhD attrition.
Countries differ substantially in where they place the major quality-control filters.
India does not currently appear to have a robust national statistic for final-viva failure.
China has an unusually strong national post-award dissertation inspection system.
International comparisons are difficult because "PhD examination" means different things in different systems.

The surprising conclusion

Perhaps the most interesting conclusion is that the terrifying PhD viva is actually the least likely place for most doctoral journeys to fail.

The candidate has already survived years of research.

The supervisor has approved the thesis for submission.

The institution has accepted it for examination.

External experts have evaluated it.

Only then does the candidate walk into the room for the final defence.

That is why stories such as Heisenberg's are so memorable.

They represent an unusual event: a candidate who has survived the entire doctoral pipeline suddenly encountering a weakness that one examiner considers serious enough to threaten the degree.

And that is also why a difficult viva should not automatically be interpreted as a failed PhD.

A difficult viva may be exactly what a doctoral examination is supposed to be.

The real question is not:

"Did the candidate answer every question?"