Statistics and statistics, again.

Before Christmas we had the unedifying spectacle of the Home Office publishing unchecked data on knife crime before departmental statisticians and the ONS were aware what was happening.

They were rightly chastised, and Jacqui Smith duly apologised.

Recently crime statistics were published without serious offences being included, it seems because the Home Office realised that there were serious discrepancies in the way that different police forces categorised different offences.

Part of the problem it seems is that in an effort to give a truer picture, - let us give them the benefit of the doubt, government departments almost routinely change definitions and categories. The consequence of changes is that it makes it difficult to compare data over time and to know whether there have been changes in subsets.

(It should be admitted that even in the case of the RPI and CPI, which are carefully calculated and published regularly, the "weightings" employed are adjusted regularly because fashions and patterns of consumption change and items drop out of or come into family budgets rapidly. The same is true even in crimes, especially in the nature of fraud and theft, - 20 years ago credit cards were much less common, mobiles had not arrived and on-line transactions were few.)

With these caveats about all time series published data, it remains true that tinkering will cause confusion and that definitional changes must be communicated clearly to all concerned.