Seasonal variation in US taxi homicides
Conventional wisdom among US taxicab drivers is that the riskiest time of the year is a few week period leading up to the Christmas holidays.
A review of 211 taxicab driver homicides in the six year period 1998-2003 suggests that risk is indeed elevated in the weeks prior to Christmas.
A surprise in the numbers is that the risk of homicide appears to be substantially higher during two other periods. They are the first six weeks of the year, and the entire month of April.
During the six-year period, the dates on which homicides occurred were not evenly distributed. The first chart shows how many of the 211 homicides occurred in each week of the year. For example, the first line indicates that, over the entire six-year period, a total of seven murders occurred during the first week of the year, i.e. on the dates January 1 through January 7.
Instead of a random distribution, there are three distinct periods in which the number of homicides substantially exceeded the average. Conversely, the chart reveals a long period beginning in May and extending through November in which the number of homicides was generally less than average.
The second chart illustrates the same information in a slightly different way. It shows the number of homicides greater or less than the average for each weekly period. Again, the numbers reflect the cumulative total for each weekly period throughout the entire six-year period. For the sake of clarity, the second chart substitutes month names instead of numbering the weeks of the year.
The data source is the Taxi-L Memoriam List of cab driver murders, which was complied from press sources. That list includes 220 US homicides during the period January 1, 1998 through December 31, 2003. Of the 220 deaths, 211 are included in this review. The other nine were excluded because the precise dates of the crimes are not known. (Of the nine homicides that were excluded, two occurred sometime in January, two in April, two in May, one in July and two in August.)
Does the information that is displayed in these two charts really mean anything? I don't know...I am a taxicab driver, not a statistician. Maybe the apparently elevated incidence of homicide during certain periods of the year is a fluke and will disappear when there are, say twenty years worth of data. My reading, however, is that six years is enough time to smooth out a random distribution, or to reveal a markedly nonrandom distribution.
The experience of the last six years suggests a nonrandom distribution, with increased incidence of homicide in weeks 1-6, weeks 14-17 and weeks 48-50.
This article, by Charles Rathbone, reflects the information available as of January 4, 2004. It is possible that, at some later date, additional homicides will be discovered to have occurred during the six-year period.
