Seen from inside the paper, a rejection is a bad day. Seen from the head of a ten-person group it is something else: a scheduling problem that repeats twelve or fifteen times a year and that nobody is counting. This article tries to put defensible numbers on it, which is harder than it sounds, because half the figures that circulate on this topic do not survive a close look.
A warning before we start: everything here is a number I can point at. Where no published figure exists, I say so and make the argument without it. If somebody sells you a round number of the type "each rejection costs your institution 4,700 euros", ask for the source; it is almost always an estimate built on three other estimates.
The cost you can actually measure: months
The most useful data on this comes from an analysis of 3,500 review experiences reported by authors themselves on the SciRev platform (Huisman & Smits, 2017). Their figures, by field:
- Psychology: 14 weeks on average to a first response, and 20 weeks of total review duration for accepted papers. Only 11% of authors heard anything within the first month.
- Social sciences: 17 weeks to a first response and 23 weeks in total. There, only 8% heard anything within a month.
- Across all fields: 13 weeks on average for the first review round (ranging from 8 to 18 weeks between disciplines) and 17 weeks of total review.
And on desk rejection, which you would expect to be immediate: the mean is 12 days, but one third of journals take more than two weeks to tell you they are not sending your paper out for review, and one sixth take more than four.
Do the arithmetic with your own group. A psychology manuscript rejected after review takes roughly fourteen weeks before you even find out, plus the reformatting for another journal, plus the next wait. Three papers a year in that situation is close to a full year of group calendar displaced, without anybody having done worse work.
The myth of the lost APC
Worth knocking down because it comes up in every department meeting: a rejection does not cost you the article processing charge. The open access fee is paid once the article has been accepted, not when it is submitted. If you are rejected, you pay nothing. What can happen is that the journal that eventually accepts you charges more than the first one you tried, or that your APC was coming out of a budget line that expires in December and the delay leaves you without it. That is real, but it is a calendar problem again, not a direct cost of the rejection.
An uncomfortable finding: rejection is not a signal of bad science
Calcagno et al. (2012) reconstructed the submission history of more than 80,000 articles published in 923 biology journals between 2006 and 2008. Two results worth holding in mind before dramatizing:
- Around 75% of published articles had been submitted first to the journal that eventually published them, ranging from 67% to 87% depending on the subject category. Most of the papers you see in print do not drag a long chain of rejections behind them.
- Articles that did arrive after being rejected elsewhere received, on average, more citations than first-intent papers published in the same journal in the same year.
The honest reading is not "rejections are good". It is that peer review, when it works, improves the work, and that a rejection is not a verdict on the quality of your group. What it unquestionably is, is time. The paper ends up better and appears a year later.
Where it really bites at group scale: the window
Research assessment almost everywhere works in windows. Whatever your national system is called — a six-year research period, a research excellence exercise, a tenure clock, a probationary review — it evaluates a fixed span of time and counts what landed inside it. In Spain, for instance, the current call asks each applicant for five main contributions, with up to two substitutes, for a six-year period (Resolution of 10 December 2025, CNEAI).
That is what makes the delay matter. When five outputs have to fall inside six years, the difference between a paper accepted in December of year six and one accepted in February of year seven is not two months: it is whether it counts at all. Multiply that by the number of people in the group who are inside a window simultaneously — those applying for a research period, those preparing a promotion dossier, those who need to defend a thesis — and the group's editorial calendar and its career calendar turn out to be the same calendar.
What I am not going to do is put a euro figure on a research period that does not come through. That number does not exist in any defensible form: the salary supplement varies by region, by contract type and by year, and the effect on promotions, panels and the right to supervise doctoral students has no market price. Anyone who gives you that figure is inventing it. Months, by contrast, can be counted.
The cost nobody logs: the PI's hours
In almost every group I know there is one person who reads everyone else's drafts, fixes the analyses that do not add up and writes the response to reviewers when things get serious. It is usually the PI, and that work appears on no spreadsheet.
There is, as far as I know, no published figure for how many hours a principal investigator spends reviewing their own group's work. I am not going to estimate it. What has been measured is the distribution of formal peer review for journals, and the pattern is stark: in the biomedical literature, 20% of researchers perform between 69% and 94% of the reviews (Kovanis et al., 2016). Reviewing work does not get shared out; it concentrates on the same people. The same thing happens inside a group, and the bottleneck is usually the person with the least time.
This has a practical consequence that gets lost when rejections are discussed in money terms: each extra iteration of a manuscript does not only cost the first author's time. It mostly costs the time of the one person in the group who cannot be substituted.
Which part of a rejection is avoidable and which is not
This is the question that actually decides whether investing in the problem is worth it. My answer, after a fair number of manuscripts: there are three layers.
The mechanical layer, which is almost entirely avoidable
Nuijten et al. (2016) ran an automated checker over more than 250,000 published p values in eight major psychology journals between 1985 and 2013. Half of the papers using null hypothesis significance testing contained at least one p value inconsistent with its test statistic and degrees of freedom, and one in eight contained a grossly inconsistent value, the kind that may change the statistical conclusion.
That is credibility thrown away, and it is caught by a mechanical check in minutes. The same goes for references that do not exist or are wrong, tables that do not follow APA 7, and results reported without effect sizes or confidence intervals. None of that alone justifies a rejection, but together it builds the impression of carelessness that makes a reviewer look harder.
The methodological layer, avoidable if caught in time
A model that does not answer the question, a sample that will not carry the estimate, a mediation analysis on cross-sectional data presented as if it were causal. This does sink manuscripts, and it is fixable, but almost always before the data are collected. Bought at the design stage it is a conversation; bought with the data already in hand, often the only thing left to do is rewrite the conclusions more cautiously.
The editorial layer, which is not avoidable
Fit with the journal's scope, the editor's priorities, a special issue that filled up, a reviewer having a bad week. This layer exists, it is large, and it cannot be bought off. The only thing you can do about it is choose the journal better and avoid losing four months finding out the paper never fitted.
If you want to know which layer your manuscript is in before you submit, the free AI paper reviewer returns the objections of two simulated reviewers and an estimated editorial decision, and the journal fit checker tells you whether the journal you picked is a reasonable target. Half an hour, and it saves you arguing in the abstract.
How to put a number on this in your own group, without inventing it
You do not need a study. You need a spreadsheet with five columns and the discipline to fill it in for a year:
- Date of first submission for each manuscript.
- Number of journals it passed through before acceptance.
- Date of acceptance.
- Whether the rejection was a desk rejection or came after review.
- The stated reason, in four words.
After twelve months that gives you three figures that are yours and not somebody's paper: how many months an article from your group takes on average from first submission to acceptance, what proportion of your rejections are desk rejections (if it is over a third, your problem is journal choice, not methodology), and which objections repeat. That last column is the one that surprises people: in most groups the same comment shows up in three different manuscripts written by three different people. That is no longer a manuscript problem, it is a shared-criteria problem, and it gets fixed once for everybody.
And what to do with the answer
It depends on what comes out, and this is where honesty has to run in both directions.
If a third of your rejections are desk rejections, you do not need to buy anything: you need half an hour of journal fit before each submission and an internal rule that no manuscript goes out without being read by someone who did not write it. It is free and it moves the needle more than anything else on this page.
If the objections repeat and they are all methodological, you have a capacity problem and it does need resources. The real options — a staff methodologist, training, external consulting, tools — each come with their own drawbacks, and I compare them without varnish in methodological support in a department: what works and what does not.
And if what you need is for the whole group to have pre-submission review without opening a purchase file per manuscript, that is what the annual bundles for groups and departments are for: a fixed amount per year, unlimited reports for the whole team, and the administrative advantage of buying once instead of twelve times. How that purchase actually works — thresholds, budget lines, what your finance office will ask for — is laid out step by step in how to buy statistical support for a research group.
References
- Calcagno, V., Demoinet, E., Gollner, K., Guidi, L., Ruths, D., & de Mazancourt, C. (2012). Flows of research manuscripts among scientific journals reveal hidden submission patterns. Science, 338(6110), 1065-1069. https://doi.org/10.1126/science.1227833
- Huisman, J., & Smits, J. (2017). Duration and quality of the peer review process: The author's perspective. Scientometrics, 113(1), 633-650. https://doi.org/10.1007/s11192-017-2310-5
- Kovanis, M., Porcher, R., Ravaud, P., & Trinquart, L. (2016). The global burden of journal peer review in the biomedical literature: Strong imbalance in the collective enterprise. PLOS ONE, 11(11), e0166387. https://doi.org/10.1371/journal.pone.0166387
- Nuijten, M. B., Hartgerink, C. H. J., van Assen, M. A. L. M., Epskamp, S., & Wicherts, J. M. (2016). The prevalence of statistical reporting errors in psychology (1985-2013). Behavior Research Methods, 48(4), 1205-1226. https://doi.org/10.3758/s13428-015-0664-2
- Resolución de 10 de diciembre de 2025, de la Comisión Nacional Evaluadora de la Actividad Investigadora [Spanish national research assessment criteria]. Boletín Oficial del Estado. https://www.boe.es/diario_boe/txt.php?id=BOE-A-2025-26118