Matched cohort, 1,280 young people, administrative records

Criminology research poster example

A criminology research poster example has to show how the comparison group was built before it shows the gap, because selection is the first objection any reader raises. This A0 portrait board reports a fictional matched cohort study of youth diversion and twelve month reoffending, 640 diverted young people against 640 court processed, with four charts, a five entry methods protocol and conclusion bullets that carry each rate with its count.

Create a poster with OneCraftA0 Portrait, printed at 841 by 1189 millimetres

The whole board

The poster at full size, exactly as it prints. Every number, citation and caption on it was written for this example, so the layout is being judged on real content.

Youth diversion and reoffending at twelve months: 1,280 matched young people

A. Sultana ¹, K. Marlowe ¹ ², J. Petrakis ²

1 School of Criminology, Barrowfield University · 2 Youth Justice Data Unit, Barrowfield University

Figure 1. Proven reoffending within twelve months: Div diverted, Court processed through the childrens court.

Source: matched administrative records, three financial years.

Figure 2. Reoffending by prior record and pathway, the four matched subgroups.

Source: matched administrative records, 1,280 young people.

Figure 3. Cumulative reoffending in the diverted group by month of follow up.

Source: 640 diverted young people, proven offences only.

ABSTRACT

Police in this jurisdiction may divert a young person to a community programme instead of laying a charge, and the decision sits with the individual officer. We asked whether diversion is followed by less reoffending than a court appearance for young people who look alike on paper. Using administrative records for 1,912 eligible young people aged 12 to 17, we propensity matched 640 diverted to 640 court processed cases on age, prior contacts, offence category, postcode disadvantage and time of year.

INTRODUCTION

Diversion is cheap, fast and popular with police, and the case for it usually rests on the cost saving rather than on outcomes. The outcome argument is harder to make because the young people who get diverted are not the young people who get charged: officers divert the ones they judge least likely to come back, which builds the answer into the sample. Matching does not solve that, but it narrows it to the factors an officer can actually see in the record, and it lets the size of any remaining bias be argued about openly rather than assumed away.

REFERENCES

1.Eldridge, T. and Marlowe, K. (2025) Matching on the visible record in youth justice research. Journal of Criminological Methods, 14(2), 177 to 201.

2.Nakamura, Y. (2024) Diversion programmes and proven reoffending: a review of cohort designs. Youth Justice Review, 31(1), 8 to 34.

3.Petrakis, J. (2026) Officer discretion at the diversion threshold. Barrowfield Criminology Working Paper 2026-01.

METHODS

Design. Retrospective cohort built from administrative records, propensity matched one to one, with twelve months of follow up from the diversion decision or the court date.

Population. All 1,912 young people aged 12 to 17 recorded as eligible for diversion across three financial years in one jurisdiction.

Matching. Nearest neighbour on age, prior contacts, offence category, postcode disadvantage decile and time of year, matched without replacement.

Outcome. Any further proven offence within 365 days, taken from the same administrative records rather than from self report or from police contacts.

Analysis. Risk difference with bootstrap confidence intervals over 2,000 resamples, with matched pairs resampled together so the pairing is preserved.

Figure 4. Matched pairs by prior record: 384 first offence pairs and 256 with a prior offence.

Source: matched administrative records, 640 pairs.

RESULTS

The 12 point gap holds in both prior offence strata and does not depend on any single year of the cohort. Pairs matched within the same postcode decile show the same difference as the full set. Half of the reoffending in the diverted group had happened by month six. Of the 1,912 eligible, 632 found no close match on the five variables and were left out, leaving 640 pairs, 1,280 young people. Selection on what an officer can see has been matched away; selection on what an officer can sense has not.

DISCUSSION

Twelve points is large enough to matter and small enough to be eaten by selection if officers can see something the record cannot. The subgroup pattern argues against that reading, because a hidden judgement about risk should shrink the gap among young people with a record rather than widen it. Two designs would narrow the selection problem further. The first is a discontinuity: officers in this jurisdiction follow a scoring tool with a hard cut off, and young people either side of the cut off are close to randomly assigned. The second is a trial of assignment by roster rather than by officer judgement in the two districts where both pathways are equally resourced.

19% reoffended after diversion, 122 of 640

31% reoffended after court, 198 of 640

Gap 12 points, 95% CI 7.1 to 16.9

Present in both first offence and prior offence pairs

Block by block

What each block on the board is for, in the order a reader walks it.

Title band and three authors: The title gives the comparison and the sample size: youth diversion and reoffending at twelve months across 1,280 matched young people. Three authors carry generated superscripts over two affiliations, a school of criminology and the youth justice data unit that holds the records, so a reader can see where the data came from before reading a figure.
Four charts across the top: Four numbered figures take the upper half of the board. Figure 1 sets proven reoffending after diversion against a court appearance as two bars. Figure 2 splits reoffending into the four matched subgroups, first offence and prior offence by pathway. Figure 3 traces cumulative reoffending in the diverted group at three, six, nine and twelve months. Figure 4 shows the make up of the matched set, 384 first offence pairs and 256 with a prior offence. Each caption carries its own source line.
Abstract and introduction: The abstract states the policy question, the design, the age range, the 1,912 eligible young people and the five matching variables. The introduction explains why the comparison is hard: officers divert the young people they judge least likely to return, which builds the answer into the sample, and matching narrows that problem to what the record shows rather than removing it.
Methods as a labelled protocol: Five entries: a retrospective matched cohort from administrative records, all 1,912 eligible young people aged 12 to 17 across three financial years, one to one nearest neighbour matching without replacement on five variables, any proven offence within 365 days as the outcome, and risk differences with bootstrap intervals over 2,000 resamples of matched pairs.
Results and discussion: The results block says the 12 point gap holds in both prior record groups, in each year of the cohort and within postcode deciles, and that half of the diverted group's reoffending had happened by month six. It also carries the exclusion step: 632 of the 1,912 eligible found no close match, leaving 640 pairs. The discussion weighs selection against the subgroup pattern and proposes two stronger designs, a discontinuity at the scoring tool's cut off and assignment by roster in two districts.
Conclusion bullets with counts: Four bullets close the board: 19 percent reoffended after diversion, 122 of 640; 31 percent after court, 198 of 640; a gap of 12 points with a 95 percent interval of 7.1 to 16.9; and the gap present in both first offence and prior offence pairs. The counts divide to the printed rates, so a reader can check them standing at the board.
Where the exclusions sit: There is no participant flow block on this layout, so the results text carries the step in one sentence: of the 1,912 eligible, 632 found no close match on the five variables, leaving 640 pairs or 1,280 young people. 1,912 minus 632 is 1,280, so a reader can check it standing at the board. A layout with a flow block would show the same step as boxes.
How to adapt this board: Retype the title, authors and abstract, replace the five method entries, and put your headline rates in the conclusion bullets with their counts. In the current version of this layout only Figure 4 is a live chart with its own data table; Figures 1 to 3 are image frames, so export those three charts from your analysis software and upload them as images. Keep Figure 4 or its equivalent, because showing the make up of the matched set is what lets a reader judge the subgroup figure.

What makes this board work

Every rate carries its count

The conclusion bullets give 19 percent as 122 of 640 and 31 percent as 198 of 640, and the gap with its interval. A percentage with no denominator is the most common fault on a real criminology board and the first thing a reviewer challenges.

The subgroup figure answers the obvious objection

If officers divert young people for reasons the record cannot see, the gap should shrink among those with a prior offence. Figure 2 and the results text show it does not, which is the board's argument against hidden selection, offered as an argument rather than proof.

The limit is stated where it belongs

The results block ends with the sentence the whole study turns on: selection on what an officer can see has been matched away, selection on what an officer can sense has not. Putting it beside the result, not in a footnote, keeps the claim honest.

Questions people ask

What goes on a criminology research poster?

The policy question, the population and how the comparison group was built, the outcome and its follow up window, the headline difference with an interval, a flow that shows who was excluded, and at least one subgroup split. Administrative record studies stand or fall on the exclusions, so show them.

How do I show a matched design on a poster?

Name the matching variables in methods, show the size of the matched set, and show the result inside the strata the matching was built on. This board does all three: five variables in the matching entry, 384 and 256 pairs in Figure 4, and reoffending by prior record in Figure 2. If your layout has a flow block, add the exclusions there too.

How many figures fit on an A0 portrait board?

Four works when they sit together in one band, as on this board: pathway bars, the four subgroups, cumulative reoffending by month and the matched set. Each chart then needs a short caption and few categories. Beyond four, each figure shrinks below the size where its labels can be read from a metre away.

What text size should I use for a poster?

Body text on this board sits near 24 pixels on a 1656 by 2340 canvas, about 12 millimetres printed at A0, which reads at a metre. Section headings are roughly double. If a block will not fit at that size, cut words rather than shrinking the type.

Should conclusion bullets repeat the key numbers?

Yes, when the layout has no stat blocks. On this board the conclusion bullets are the only place the headline rates appear with their counts, 122 of 640 and 198 of 640, so a reader who skips the charts still gets the finding. Keep each bullet to one number and its denominator.

How do I adapt this board to my own study?

Retype the title, authors and abstract, replace the five method entries, and write your rates with their counts in the conclusion bullets. Each figure holds its own data table, and the chart type can be switched among ten kinds. Keep one figure for the make up of the matched set, because it lets a reader judge the subgroup split.

Build your own in about a minute

The button below opens the generator with this use case already described. Change the wording to match your own, generate, then edit anything you like.

Make one like this

Sources

Written and checked by the OneCraft team. Last checked 29 September 2026.