Tuesday, 26 April 2011

Bob Cox, WTF??!!


Bob Cox: Publisher of the Winnipeg Free Press,
in front of machines designed (in part) using
theories and principles derived from evil
statistics.

Image Source: The Uniter
 Bob Cox is the publisher of the Winnipeg Free Press. On the paper’s website, we can find an article attributed to him on statistics and public policy. While it’s pretty old (Feb. 2, 2011), the sheer stupidity it represents is timeless and thus has to be addressed. The article is named “Focus on Crime, not statistics” (How the hell do you do that without knowing anything about the nature and cause of crime, something only statistical analysis can determine?).


Lies, damn lies and statistics.

The old line about how easy it is to manipulate statistics came to mind again this week as yet another critic of Statistics Canada took aim at the agency's data on crime rates.

Statistics Canada has been reporting falling rates of overall crime and violent crime for some time, a stance that is at odds with the federal Conservative government's efforts to toughen criminal laws and build new prisons.

There are exceptions to the overall trend. For example, Manitoba's violent crime rate increased by 10 per cent in 2009, primarily due to a 25 per cent increase in robberies. But Statscan data show national rates of crime reported to police dropping.
... The Macdonald-Laurier Institute added to the Conservative argument with a report that says crime is increasing and Statistics Canada's methods are suspect. Among the criticisms: the agency's reporting on crime rates does not account for unreported crime, even as other Statistics Canada data show people are increasingly not reporting crimes such as break-ins.
... It is easy to criticize crime statistics
... A prime example of this happened last Christmas in Winnipeg. Take a look at the statistics and they seem to show drunk driving increased dramatically. What really happened is that police targeted locations, such as bars, that are more likely to produce impaired drivers, so they laid more charges.
Police themselves said it is unlikely that the prevalence of drunk driving changed in a single year. In fact, the new enforcement strategy could end up reducing impaired driving in the longer term by acting as a more effective deterrent.
... In all areas of crime, we would do better if we focused on discussion on how to curb illegal activities, rather than whether the statistics are accurate.”


So, because Cox has miraculously found out that statistics aren’t infallible and need proper context, we should chuck them out altogether? And, for that matter, why is the report of one Rightwing Think Tank the end all be all for the non-partisan and world renown Statistics Canada Agency? The report, had Cox bothered to dig a little deeper, ignores the effect of including “threats” as violent crime,showcases total number of crimes rather than per capita rate, and is full of other holes.

There’s serious stupidity in Cox’s reasoning, a sort of “throw out the baby with the bathwater” mentality. Because statistics aren’t infallible, we should focus on the qualitative nature of crime and the best mechanisms to fight it. What are the best ways to measure these? Statistics is, because everything else tends to be either horribly anecdotal or based on ungrounded intuitions.

Most statisticians know statistics can be misleading if there are various sampling biases, which is why the science works to minimize these. Because Bob Cox is either just learning about these, too lazy to learn more details, or too intellectually rigid to learn anything new does not mean this generation should throw out some of the best, most effective public policy tools around.

And, finally, it would be useful to keep this in mind:

It is easy to lie with statistics, but easier to lie without them.
.
(Quote usually attributed to Frederick Mosteller)

No comments:

Post a Comment