The reasons why the Metropolitan Police decided to curtail its initial investigation of phone-hacking allegations are to be explored both in the public inquiry announced by the Prime Minister and, no doubt, in court cases resulting from the subsequent discovery that hacking was much more widespread than the earlier investigation suggested.

Statistical ineptitude may turn out to be the least important explanation for the failure. But testimony to the Home Affairs Select Committee suggests that, astonishingly, with thousands of telephone numbers to sift through to determine ownership and whether they had been hacked into, the use of (stratified) random sampling by detectives was not used to estimate:  

1. how many per 1,000 of the listed telephone numbers belonged to ‘important people’ – howsoever defined;
2. how many per 1,000 of the listed telephone numbers belonged to persons who had suspected that their telephone had been hacked into – requires that telephone owner be approached;
3. how many per 1,000 of the listed telephone numbers belonged to persons in the ambit of a criminal investigation.

To rule out a 1 per cent hit-rate in answer to any of these questions,  some 300 to 500 telephone numbers would have to have been selected randomly, followed-up on, and ZERO hits of type 1) to 3) discovered.

To rule out a rate of 1 per 1,000 in answer to any of the questions, some 3,000 to 5,000 telephone numbers would have to have been selected randomly, followed-up on, and ZERO hits discovered.

What level of detection did the Metropolitan police aspire to, and how did they guarantee it – before deciding to curtail their initial investigation?

These are matters of statistical science as much as of efficient deployment of police-resources on crime-detection.

Conflict of interest: SMB serves on Home Office’s Surveys, Design and Statistics Subcommittee, but writes in a personal capacity.