Wednesday, August 04, 2010

Truth, Lies and Un-Statistics

According to Stockwell Day, SOME statistics are useful and reliable for informing public policy; those would be the unreported ones. When asked why his government is planning on investing billions of dollars in new prisons during a recession when crime rates have been dropping for decades, Day responded that he is concerned about unreported crimes. Now to be fair, Statistics Canada does collect self reported data on crime victimization - on the soon to be voluntary Stats Can Long Form Survey. It is also true that some 34% of crimes are unreported - a number that has also been pretty consistent for decades. BTW - I wonder if they asked anyone if they were victims of homicide?

While the reporters present had some probing questions on what statistics the Treasury Board President was relying on, my question is a little more basic. Let's assume that unreported crimes (the majority of which the reason given for not reporting was that it was not important enough) are a cause for alarm. Why are we building prisons?

Generally unreported crimes are also un-investigated, un-prosecuted, un-tried, un-convicted and un-sentenced. So maybe we should spend un-billions to build un-prisons for these un-convicts instead.

3 comments:

Anonymous said...

Check out this analysis on the real reason for the new jails: it depends on the definition of "crime:"

http://blog.paulmckeever.ca/2010/08/04/shedding-light-on-day-unreported-crimes-code-for-cannabis-offences/

Hugh MacIntyre said...

Stat Can does not collect data on victimization in the census. They do it on a voluntary survey.

CoteGauche said...

Thanks for the clarification Hugh.

You are correct, the data come from the StatsCan General Social Survey on Crime and Victimization.

However the fact that this is a voluntary survey, sent to 24,000 households, that only asks about 8 specific areas of crime, makes reliance on such data to rationalize massive spending on a new crime bill is even more tenuous. Voluntary surveys naturally have a selection bias and statscan has already said that Day misinterpreted the data.