How to explain a gap in your CV, and what it costs not to

There is a stretch of time that your employment history does not account for, and you are about to send an application that makes it visible. The advice you have been given is almost certainly some version of do not draw attention to it. Use years rather than months. Say nothing. Hope the conversation gets far enough that you can explain it to a person.

Three field experiments have sent thousands of fictitious applications to real employers and counted who got called back. Between them they bear on the explain-it-or-hide-it choice, on the popular alternative of filling the gap instead, and on whether its length matters. The one that isolated the choice most directly points the other way from the advice: most of what a gap cost was attached to the silence rather than to the time off.

What was actually measured

In 2016, Sheryll Namingit, William Blankenau and Benjamin Schwab sent 3,771 applications to 1,257 job advertisements across the fifteen largest cities in the United States: sales, customer service, clerical and accounting assistant roles, none requiring more than six years of experience.¹

Each advertisement received three applications, matched on everything the researchers could hold constant. All three applicants were out of work. The first had left their last job within the previous two months. The second had a gap of seven to twelve months for most of the sample, explained in one phrase in the cover letter as a medical illness followed by a complete recovery, with a cancer survivors' support group listed among their interests on the CV. The third had a gap of the same length and said nothing about it. A coding error gave about 5% of those two groups a gap of twenty to twenty-two months instead, which becomes relevant further down.¹

The callback rates came out in that order. Newly unemployed applicants heard back 27.4% of the time, applicants with the explained gap 25.6%, and applicants with the unexplained gap 23.3%.¹

The 2.3 percentage points separating the explained gap from the unexplained one sounds like very little. Set against the overall callback rate of 25.5%, it is a reduction of about 9%, which is the authors' own way of putting it.¹ The regression is where the shape of the result shows. Against the newly unemployed baseline, an explained gap cost 1.9 percentage points and an unexplained gap cost 4.2, and a test of whether those two effects are the same returns p = 0.038.¹ Relative to that 27.4% baseline, the explained gap reduced callbacks by roughly 7% and the unexplained gap by 15%.¹

So the gap was not free. Nothing here says a break in your employment history costs you nothing. What it says is that a bit over half of the measured penalty went away when the applicant supplied a reason, and that the difference between the two was unlikely to be noise.

The part that does not generalise

It would be convenient to stop there and conclude that any explanation beats none. The evidence does not support that, and the study that shows why is worth more than the one that gives the encouraging answer.

Between May and July 2015, Stijn Baert and four colleagues sent 288 trios of applications to real vacancies in Belgium. The structure was almost the same: one recently unemployed candidate, and two with a full year out of work, one of whom attributed it to former depression and one of whom gave no reason.²

Disclosing the depression reduced the probability of an interview invitation by about 34% compared with the recently unemployed candidate. And the stigma attached to it was "not significantly higher than the stigma effect of a year of unexplained unemployment".²

That sentence is easy to over-read, so it is worth being precise about. The study did not establish that disclosing depression and saying nothing are equivalent. It failed to detect a difference between them, which is a weaker claim, and at 288 trios this is a much smaller experiment than the American one: a benefit the size of the one found there could plausibly sit inside its margin of error undetected. What it does show is that the clear advantage measured in the first study did not turn up in this one.

Put the two results beside each other and the useful distinction is not honesty. It is what the employer infers from the reason. The mechanism the first team proposes is that an unexplained gap invites a downward revision of your ability, and a gap explained by something that strikes people regardless of how good they are at their job does not.¹ An explanation only pays if it is read that way, and whether it is read that way is a fact about employers rather than about you. Cancer, in the American study, cleared that bar by a measurable margin. Depression, in the Belgian one, did not do so detectably.

Which sets the boundary of what you can take from this. Caring for a relative, a redundancy that came with a relocation, a period of full-time study, waiting on a visa: each fits the same logic, and none of them was tested. The logic extends further than the evidence does, and it is worth being clear about which one you are relying on.

Filling the gap can be worse than having it

The most common practical advice about a gap is to end it. Take something, anything, so that the most recent line on your CV has a current date on it.

Henry Farber, Dan Silverman and Till von Wachter tested that. Their audit study sent 12,224 applications to 4,594 office support vacancies across eight American cities between 2012 and 2014. On a random half of the applications from round three onwards, the applicant had started, in the month before applying, a low-wage job at a chain restaurant, a big box store or a grocery store: work the paper describes as "quite different from the career work on the rest of the resume".³

Callbacks fell. Applicants with no interim job heard back 9.8% of the time, applicants with one 8.5%. The difference of 1.3 percentage points is a reduction of about 15%, and it is statistically significant at p = 0.038.³

Note what was tested, because the finding is narrower than "never take a job". These were deliberately mismatched roles, held right now, at the top of a CV otherwise full of office work, in applications for more office work. That is the configuration that lost callbacks. It says nothing about interim work related to what you are applying for, and nothing about the money, which is usually the actual reason people take these jobs and is not a consideration any callback study can weigh.

Nobody agrees about how much the length matters

You would expect a longer gap to cost more, in proportion. The experiments do not deliver that cleanly.

In the same Farber study, unemployment duration was randomised across 0, 4, 12, 24 and 52 weeks, and the authors found no relationship at all between duration and the callback rate, a result they note differs from earlier work finding declines concentrated in the first six or seven months.³ In the Namingit study, the 5% of applications that the coding error gave gaps of twenty months or more did worse by a further 2.3 percentage points, but with too few cases to call it precise, and the authors say so.¹

Two things follow, one of them reassuring and one of them not. Nothing in this evidence supports the assumption that an eighteen-month gap is three times the problem of a six-month one. And the fact that the studies disagree is a reason to hold any specific number loosely, including the ones above.

What to actually do

The intervention that was tested is small, but it is not quite one sentence. The explained-gap application carried a phrase in the cover letter and a corroborating detail on the CV, the survivors' support group sitting among the interests. Those two moved together, so the experiment cannot tell you how much of the effect came from the letter and how much from the CV, or whether the letter would have done it alone. What the result supports is the pair: a brief explanation where you are writing prose, and, if you happen to have something that corroborates it naturally, no reason to strip it off the CV.

Write the explanation as a statement of fact in the past tense, with an end on it. What it was, and that it is finished. The version in the study was a phrase rather than a paragraph, and everything else in the application was doing the ordinary work of an application. Do not apologise for it, and do not build the letter around it. If the reason is genuinely private, you are choosing between disclosure and the unexplained-gap penalty, and that is your call to make rather than an optimisation problem; the Belgian result is a warning that not every disclosure buys you a measurable return.

Two things not to do. Do not let the explanation contradict your dates, which is a good argument for leaving the dates legible rather than blurring them to years. And do not manufacture a reason, which is a bad idea for reasons that have nothing to do with callback rates.

If the blank page is the obstacle rather than the wording, JobCraftly will draft a cover letter from your CV and the job you are applying for, which gives you a sentence to edit instead of one to compose.

What these numbers are not

Every figure here is a callback rate: the probability that an employer responded to an application at all. None of these studies followed anyone into an interview, an offer or a job, and the effect of a gap could shrink or grow at every stage after the one that was measured.

The settings were narrow. Entry-level sales and clerical roles in fifteen American cities in 2016; office support roles in eight American cities between 2012 and 2014; a few hundred Belgian vacancies in 2015. The fieldwork is now between ten and fourteen years old. The interim-job and duration results also come from an NBER working paper, which NBER's own cover page says has not been peer-reviewed, although a four-page version of it was later published in the American Economic Review.⁴ In the Namingit study, the applicants with gaps were also, by the way the CVs were constructed, six to ten months older than the ones without, and the authors flag it.¹ Their headline difference between explained and unexplained gaps also does not reproduce as a significant one inside most of their subgroups, which is what usually happens when a sample is split, and is a reason to treat the pooled result as the finding rather than any particular slice of it.¹

What survives all that is modest and still worth having. Employers who could see a gap treated it as information about the applicant, and gave a different answer when they were told what it was. Whatever you write, you are choosing between two versions of the same application, and the one that says nothing is not the neutral option it looks like.