A lot of guff is written about crime statistics in the UK, the recorded crime figures, and the British Crime Survey (BCS). Here are some facts which should dispel a lot of the myths: 1: Police recorded crime statstics. Every year since 1998, there has been a rise in overall recorded crime. Every year, this is blamed on "changes in reporting". It is true that changes in reporting have occurred, but in 1998 and 2002. In 1998, more offences were included, and in 2002, the National Crime Recording Standard was introduced. Whilst these changes will account for the rises in 1998 and 2002, they cannot account for the rises in-between, and the rises since, no matter how many times they repeat this. A change in reporting can affect the absolute value, but will not change the differential (i.e. the rate of change) after the year in which the change takes place. If that were true, then crime could in reality remain constant, but the amount of recorded crime would tend to infinity, which is of course impossible. When looking at the stats, therefore, it is not meaningful to look at the change from 1997 to 1998, and likewise from 2001 to 2002, since there is not consistency in reporting between each of these pairs of years. However, every other pair of values can be meaningfully compared, since no change happened within them. Thus, the increase in the last year cannot be blamed on changes in reporting which happened in 1998 and 2002. The total recorded crime stats look like this: 2. The British Crime Survey. This is the government's preferred measure, principally because it says what they want people to hear - that crime is down. The only thing that's really certain about the BCS is that if it had shown consistent rises since 1997 and the recorded crime had seen a fall, the Govt would be saying that the BCS is just a survey, and is therefore unreliable. The reality is that the BCS is just a survey, and is unreliable. The media never go into details, but people are left with the assumption that the BCS is a huge survey of people in Britain. This is not the case, however: The 2000 BCS had a sample size of 19,411 people aged 16 and above, plus a further ethnic booster sample of 3,874 (source). That is 0.042% of the British population, and therefore cannot be considered as representative on an epidemiological basis: it's about 1 in 23,800, i.e. roughly 1 person per mid-sized market town completed the survey. Also, the "ethnic booster sample" means that AT LEAST 16% of the survey respondants were ethnic minority (since if the initial sample is representative, 8% of those selected will also be ethnic minority (2001 census)). If we assume that the initial sample is ethnically representative, 1,533 of those asked in this initial sample will also have been minority ethnic. Thus the total proportion of the survey population who are minority ethnic is likely to be 23.2%, i.e. ethnic minorities are over-represented by almost 200% Representative? I think not. Of course, the figures gained from the survey are then multiplied up to the whole population of the UK. Other problems with surveys are also serious: we don't know whether the questions asked have been the same from year to year, so they could have been engineered to produce a fall in crime (survey bias). Since only 1 in 23,800 completed it, the chances of completing it twice are slim, and the chances of knowing at least one other person who has completed it before are also slim, so there's no basis for comparison. The survey results for total crime look like this: A few more points (with some repetition): When the Labour party talks about rises in crime under the Tories (like in Cpl Ripper's e-flyer) and falls under Labour, they are talking about the BCS data, i.e. data based on interviewing 0.042% of the population with minority ethnics over-represented by 200%, and have possibly been subject to survey bias. The recorded crime figures show exactly the opposire (see above). If the BCS showed rises & the recorded crime showed falls, Neue Arbeit would be calling the BCS what it is - an unreliable, unrepresentative survey of too few people.