Bigger requests get refused more - except in policing, where they don't
Counting how many numbered sub-questions and question marks a request's documentation contains gives a rough, honest proxy for complexity - not a perfect one, but a real structural signal. Across every release we can measure this on, more complex requests are refused more often and take longer to answer. Policing is the exception, and an interesting one.
The overall pattern
| Complexity | Refusal / exemption rate | Median response |
|---|---|---|
| No lists or questions | 16.1% | 33 working days |
| 1-2 sub-parts | 21.1% | 35 working days |
| 3-5 sub-parts | 22.3% | 39 working days |
| 6+ sub-parts | 30.8% | 59 working days |
Both measures climb in a straight line. A six-part request is refused roughly twice as often as a plain one-line question, and takes nearly twice as long to answer. That's not a surprising shape in the abstract - more moving parts, more chances for one part to hit an exemption, more work to review before responding - but it's the kind of thing that's easy to assume and rarely actually measured.
Where policing breaks the pattern
Splitting the same buckets by niche, policing & justice tracks the overall shape up to a point, then reverses:
| Complexity | Policing & justice refusal rate |
|---|---|
| No lists or questions | 17.3% |
| 1-2 sub-parts | 19.1% |
| 3-5 sub-parts | 20.6% |
| 6+ sub-parts | 15.9% |
Everywhere else in this dataset, the most complex requests are refused the most. In policing, the most complex requests are refused about as often as the simplest ones - lower than the 3-5 range that precedes it. One plausible read, consistent with what we found in our piece on why slow doesn't simply mean refused: a genuinely large, multi-part policing request (the kind asking for outcomes broken down by year, offence type, and demographic group at once) tends to get treated as a serious data-compilation job worth doing properly, rather than triggering a reflexive exemption. Smaller-to-medium requests may be the ones more likely to get a quick administrative "no" instead of a full disclosure effort.
That's an interpretation, not something we've independently confirmed - but the reversal itself is a solid, well-powered result: over 1,600 six-plus-part policing releases sit behind that 15.9%, not a handful of edge cases.
Reading this responsibly
Complexity here is a structural proxy - numbered list items plus question marks, counted across the whole release text - not a judgement of how hard a request actually was to answer, and it isn't reliably isolated to just the original request wording versus the rest of the document. "Refusal or exemption" is the same text-presence signal used throughout this site: a document mentioning an exemption isn't necessarily one where it was the operative reason for withholding.