Is the murder rate down or is medicine more competent?
Abram Katz January 07, 2003
New Haven’s apparently diminishing murder rate is certainly good news.
So are statistics showing other crimes going down.
The numbers are so reassuring that many may not want to examine where the percentages and minus signs came from.
Crime fighters and civic leaders properly take pride in the decline of murder, but do we really know what the numbers mean?
Epidemiologists and sociologists have wondered recently what accounts for the national downturn in murder and mayhem.
There’s no obvious answer. Millions of firearms remain in circulation. The economy does not promise widespread prosperity.
People are not getting kinder and gentler. If they were, crime would drop dramatically.
Is beating someone to a pulp "kinder" than shooting him in the head?
A suddenly law-abiding, altruistic society would not shift from killing to merely maiming. Evil would disappear.
Researchers at Harvard and the University of Massachusetts published a chilling hypothesis in the British Medical Journal recently.
The United States is as violent as ever, they reasoned. What’s improved is the skill of emergency department doctors and nurses.
New surgical and medical techniques were developed to treat battlefield casualties in Vietnam.
These life-saving techniques are now employed in emergency rooms, the authors suggest.
Ten years ago, a grievous gunshot wound to the gut was fatal. Now doctors save the victim. "Murder" becomes "aggravated assault" or a lesser crime.
To an official who only keeps an eye on murders, the trend seems to suggest a more peaceful republic.
Not that society would be better off with more deaths than assaults.
But if the murder rate is decreasing because trauma care is improving, eventually medicine might drive law enforcement.
Suppose doctors advance to the point that all but the most severe wounds can be treated successfully and with few lingering complications.
A large-caliber gunshot wound to the back is no worse than a twisted ankle. Murder becomes a misdemeanor.
Not that we need to wait for a science fiction future. Sensible gun control right now could make a much healthier society.
Meanwhile, when we scan annual crime statistics, we need to remember certain statistical principles.
One of them is the "regression line," which encompasses several related ideas.
A regression line, as the name suggests, is a line drawn through a dots plotted on a graph. The dots could represent year and number of murders, for example.
Since the numbers of murders do not follow a set function, the points are a smear. The line attempts to show how the dots are related.
If the number of murders varies a lot from year to year, drawing a meaningful regression line is difficult. A trend is not clear.
Likewise, a small sample size can skew the line in strange ways. Imagine years with 3, 2, and 1 murders respectively. Though they are not significantly different, a 3,2,1 line looks like a sharp drop and a 1,2,3 line appears to be a crime wave.
Statistics is by no means elementary. Discerning real effects from coincidence can be difficult.
Crime statistics are easier, in one sense, because in absolute terms lower is always better, assuming that reporting is reliable.
Doctors comparing procedures to see which saves more lives is another matter.
That’s why these studies are statistically rigorous, and analyzed with standard deviations, confidence intervals, means, medians, variance, and other tools.
Why aren’t crime statistics analyzed in the same way?
Perhaps because one less person killed is good, whether or not it’s "significant."
But a declining murder with a 95 percent confidence level is more comforting than a declining murder rate unexamined.
©New Haven Register 2003
