Did the Body-Worn Cameras Really Cut Complaints by 30%?
Memo: Since we issued body-worn cameras (BWCs) to all patrol officers in January, public complaints against police in our district fell 30%, from 100 in the previous year to 70. Therefore BWCs caused the fall. I recommend expanding BWCs to all districts and citing this as proof in the annual report. Other facts: a customer-service training programme began in March; two officers with the highest complaint counts transferred out in February; the complaints hotline changed in April to require in-person statements; citywide complaints fell 12% over the same period; 90% of officers wore BWCs but only 60% activated them; no routine footage review was introduced. The BWC rollout cost HK$1.8 million in the district. You are the staff officer asked to review the memo before the annual report is finalised in two weeks.
1. Identify the main logical flaw in the memo and explain two alternative explanations for the 30% fall.
Read the worked answer
Main flaw: post hoc ergo propter hoc - because the fall followed the BWC rollout, the memo assumes the rollout caused it. A 30% fall is a correlation, not a causal estimate.
Alternative 1: the customer-service training began in March and could have changed officer behaviour; the fall may partly reflect training. Alternative 2: the complaints measure changed. Requiring in-person statements in April would suppress or filter complaints, so a fall in recorded complaints may not be a fall in dissatisfaction or misconduct. A third alternative is selection: the two highest-complaint officers transferred out, which mechanically lowers the district's complaint count. Citywide complaints also fell 12%, so some of the district's fall is a general trend, not a BWC effect.
A simple benchmark: if the district had followed the citywide 12% fall, it would have had about 88 complaints (100 × 0.88), not 70; the extra 18-point gap is interesting but still not proof. Also, with only 60% activation, many patrol encounters were not recorded, which weakens any claim that 'BWCs' did the work.
2. What evidence would make the causal claim more credible? Describe a practical comparison the district could use next year.
Read the worked answer
A credible design needs a counterfactual: what would complaints have been without BWCs? Practical options: (1) a difference-in-differences comparison with a similar district that did not get BWCs at the same time, using pre-rollout trends; (2) a phased or randomised rollout by team or shift, so early and later teams can be compared; (3) record activation rates and analyse complaints by activation status, while recognising that officers may activate selectively; (4) separate complaint types (rudeness, excessive force, procedure) and substantiated from unsubstantiated complaints; (5) control for the training programme, staff transfers and reporting changes; (6) check whether the fall began before January.
The district should publish the method, not just the headline. If a comparison district also falls 25%, the BWC effect is likely small. If the district falls much more than the comparison district after adjusting for activation, the case is stronger but still not definitive.
3. Rewrite the causal claim as a defensible one-sentence statement for the annual report.
Read the worked answer
Defensible version: 'Public complaints against police in the district fell by 30% in the year body-worn cameras were introduced, from 100 to 70; because a customer-service programme, staff transfers and a change in complaint reporting occurred in the same period, the reduction cannot be attributed to body-worn cameras alone and will be evaluated with a comparison district next year.'
This keeps the number, removes the causal assertion, names the confounders, and commits to a test. A shorter version: 'The 30% fall in complaints is consistent with a possible benefit of body-worn cameras, but it is not proof of cause because other changes coincided with the rollout.' The rewrite should not hide the fall, but it should not overclaim it either.