Global missions moves fifty-eight billion dollars a year. It publishes almost nothing about what that buys, and the one statistic everyone quotes turns out to have no study behind it.
Four sources, all public: a 228,223-document corpus of missions and faith-and-work publishing, the full MinistryWatch ratings database covering 1,737 US Christian ministries, IRS Form 990 records via ProPublica, and roughly 640 academic works pulled live from Crossref and OpenAlex. Every number below can be checked.
Nobody here is lying. The problem is stranger than that. A sector built on measurable claims has almost no capacity to measure them — and the holes are not scattered at random. They sit exactly where an answer would cost somebody something.
Follow it upstream and the whole literature drains into three puddles:
| Origin | What it actually is |
|---|---|
| A 2015 endorsement blurb | A pastor's assertion in a testimonial. No source given. The URL now returns 404. |
| A mission agency researcher | Self-labelled "anecdotal evidence" — 2010 interviews with mission workers, not with returnees |
| A 2008 agency publication | "There is a statistic that…" — the statistic is never named |
Everything else is a quotation of a quotation. The citations in between hedge with estimate, guessing, it is said that, informed sources have indicated — and then the hedges fall away and the number walks on alone.
Be fair to the three people at the bottom of this. Each labelled his figure honestly — an estimate, an anecdote, a thing he had heard. The failure is not theirs. It belongs to eighteen years of repetition that sanded the hedges off and handed the sector a fact it had never earned.
So why has nobody checked? Because the number already works. It is grave enough to move donors and clean enough to fit on a slide. A real study might come back at 40 per cent and collapse the programmes built on it. It might come back at 95 and indict everyone. Worst of all it might come back unmeasurable, which takes the number away and puts nothing in its place. The statistic survives because it is useful, and it is useful because nobody has looked.
OpenAlex indexes roughly 250 million scholarly works. Searching titles and abstracts:
One paper. Out of 250 million records, exactly one carries the phrase "missionary retention" in its title or abstract. Scholarship on Mormon missionaries outnumbers scholarship on evangelical missionary attrition by roughly fifty to one.
What Christian work exists is real but tiny, and nearly all of it sits in one journal: clinical-psychological studies of who is suited to the field and what the field does to them.1 The most-cited paper in the entire set has 12 citations. Dwyer, in the next section, has 620 for the same kind of question.
Does a longer assignment produce better or more durable results than a short one? It is the obvious question, and it is not unanswerable. International education has answered it repeatedly. The flagship study has 620 citations and its title gives away the ending:
| Study | Year | Cites |
|---|---|---|
| Dwyer, M. — More Is Better: The Impact of Study Abroad Program Duration2 | 2004 | 620 |
| Engle & Engle — Study Abroad Levels: Toward a Classification of Program Types | 2003 | 313 |
| Engle & Engle — Assessing Language Acquisition and Intercultural Sensitivity Development | 2004 | 285 |
| Medina-López-Portillo — Intercultural Learning Assessment: The Link between Program Duration and… | 2004 | 260 |
| Davidson, D. — Study Abroad: When, How Long, and With What Results? | 2010 | 183 |
Dwyer went back to alumni years later and tested duration against what had lasted. 344 papers match "study abroad" and duration. The instrument is validated, well cited, and would port across with almost no modification.
In eighteen years, nobody has pointed it at a missionary.
The corpus holds 29,803 documents on mobilisation — the work of recruiting and sending people. Search every one of them for what mobilising a single person costs and you get two hits, both false positives about website traffic.
A deeper pass read 2,753 full documents pairing mobilisation language with cost language. Forty-one passages tied a dollar figure to a per-person phrase. Every one came from Stanford Social Innovation Review — a secular journal — about nursing homes, farm programmes, prison diversion and corporate secondments.
Some of this is structural rather than cultural. Form 990 gives an organisation three expense buckets: programme, management, fundraising. Mobilisation is not one of them. Recruiting, assessment, candidate school and pre-field coaching are smeared across two of the three with no way to pull them apart, and the number of workers actually deployed is not a field on the form at all. Neither the numerator nor the denominator exists in public data. Most organisations could not publish the figure if they wanted to.
Every comparable sector publishes the figure anyway:
| Sector | What it publishes | Figure |
|---|---|---|
| Corporate HR | Cost per hire — with an ANSI standard defining what counts3 | $4,129 |
| US Army | Cost per qualified recruit | $13k–24k |
| US Navy | Cost per qualified recruit | $29,700 |
| Peace Corps | All-in cost per Volunteer-year4 | $53,813 |
| Latter-day Saints | Monthly missionary contribution, flat worldwide5 | $500/mo |
| Evangelical missions | Nothing | — |
Two distinctions get collapsed constantly and both matter. The Latter-day Saint figure is what a missionary's family pays to sustain someone already in the field — not what the church spends to raise one, which is unpublished. And the Peace Corps files a public budget justification every year that never breaks out recruitment at all. Strip those away and only two kinds of institution on earth publish a real cost-to-recruit: corporate HR departments and the armed forces.
Meanwhile the figure nobody publishes anywhere is the only one that would settle an argument: cost per person still serving in five years.
Across the same 228,223 documents:
138 to 1. The obvious objection is that the corpus is just thin, and it is not: unreached peoples and church planting fill 20.97 per cent of it, fundraising another 10.56. Against baselines like those, 0.08 per cent on financial transparency is not a sampling accident. It is a subject the field declines to have.
MinistryWatch rates 1,737 US Christian ministries reporting $58.2 billion between them. For 164 it holds no financial data whatsoever. Run those 164 against the IRS Form 990 record and just 16 come back.
The same join throws off a warning as well as a finding. Of the 1,532 organisations appearing in both sources with revenue attached, 85 disagree by more than half. Fiscal-year offsets and parent-subsidiary splits account for most of it. The practical rule: neither source is quotable on its own.
None of this needs a new method. Three studies would close most of the gap and all three are ordinary social science, the kind a competent graduate student runs:
Corpus counts are lexical matches over title and body text, which makes them lower bounds, not censuses. The contrast between clusters is the signal; absolute numbers are not. The 228,223-document corpus is a convenience sample of roughly 100 missions and faith-and-work publishers, not a random sample of the field, and it over-represents organisations that publish online.
"Nobody has run the study" is a claim about what is published. Unpublished or internally-held research would not appear here.
The academic counts come from OpenAlex. An earlier draft of this review used Crossref and badly undercounted the study-abroad literature — Crossref under-indexes education and social-science journals. That error is recorded rather than quietly corrected, on the principle that a review about unchecked numbers should not ship any.
The military cost figures are the weakest rows in this piece: they reach us through aggregators citing 2023 Department of Defense data rather than a primary document, and three sources give three different numbers because "cost per recruit" has no agreed boundary — which is exactly why corporate HR wrote a standard for it.