The Volatility of Monthly Crime - District of Columbia Crime Policy

advertisement
HIGHLIGHTS
■■ Month-to-month changes in crime are normal.
■■ Even during decreasing crime trends—as with
homicide over the past decade—many months will
show increases (followed by other months with
decreases).
■■ Trends can only be identified using data compiled over
time; single months compared to the prior month or
the prior year are insufficient to identify a trend.
Brief
March 2011 No. 10
The Volatility of Monthly Crime
BY Akiva Liberman and Meagan Cahill
The views expressed are those of the authors and should not be attributed to the District of Columbia Crime Policy Institute, the
Urban Institute, the Brookings Institution, their trustees, or their funders.1
I
n recent years, the overall trends in crime
in the District have been down. Yet, news
accounts often report surges in crime.
Monthly burglary, robbery, and homicide
are volatile and can go up even during a long-term
downward trend. Does an increase in crime in a
given month—compared with last month or last
year—suggest that crime is on the increase?
The key to reconciling short-term changes in
crime with longer-term trends is to understand
the volatility of crime. In this brief, we explore
this issue by examining monthly changes in crime
during the past decade against the background of
longer trends.
Figure 1. Homicide
35
Homicide
30
Yearly Avg.
25
20
15
10
5
0
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Homicide
Figure 1 shows the number of homicides in the District of Columbia over the past decade, which declined
considerably. But the numbers swing wildly from month to month. From January 20012 through December 2010,
82 percent of months showed a swing in homicides—in one direction or the other—of at least 10 percent from the
previous month. In most months, homicide changed by at least 25 percent (see table 1).
Table 1. Homicide
Of course, increases in homicide
generate the most concern and news
coverage. Homicide increased by at
least 10 percent in over a third of
months (39 percent), and increased
by at least 25 percent in 27 percent
of months.3 Therefore, news
BRIEF: The Volatility of Monthly Crime
Table 2. Robbery and Burglary: Change from the Preceding Month
reports about homicide could have accurately reported
large increases in monthly homicide more than a quarter
of the time. Yet, such reports could be deeply misleading
against the large overall decline in homicide.
With an average of 16 homicides per month over this
period, just a few additional homicides generate large
percentage changes.
Robbery and Burglary
One key reason for the volatility of homicide is its rarity.
Increases in robbery or burglary also prompt
considerable concern and media attention. (Robbery is
the taking of property from a person by force or the
threat of force, while burglary is the taking of property
from a residence.) At the yearly level, robbery was fairly
stable over the past decade, while burglary declined
somewhat (figure 2).
Figure 2. Robbery and Burglary
500
400
300
200
Robbery
100
Yearly Avg.
0
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
600
Burglary
500
Yearly Avg.
400
Both robbery and burglary are around 20 times as
common as homicide, with average monthly reports of
340 robberies and 349 burglaries. As a result, they are
considerably more stable than homicide. But they still
show considerable volatility (table 2). For both crimes,
changes of at least 10 percent occurred in about half
of months, and changes of at least 25 percent in about
10 percent of months. Increases of at least 10 percent
occurred in about 30 percent of months. Increases of
at least 25 percent occurred in robbery in 3 percent of
months, and in burglary in 7 percent of months.
Same Month, Last Year
300
200
100
0
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Is the apparent volatility in month-to-month crime an
artifact of a yearly cycle? Figure 2 seems to show yearly
cycles of increases and decreases in robbery and burglary.
Are months more stable when compared to the same
month in the previous year? Surprisingly, no. Comparing
each month’s robbery or burglary with the same
month in the previous year finds as much instability as
comparisons to the previous month. Details are as shown
Page 2 of 3
BRIEF: The Volatility of Monthly Crime
Table 3. Change from Same Month, Previous Year
in table 3.4
How can these results be reconciled with the yearly cycles
visually apparent in the graphs? The answer is that the
monthly timing of that cycle varies across years. For
example, in 2000, 2001, and 2002, robbery peaked around
the New Year, while in 2005, 2007, and 2008, robbery
peaked months earlier.
Conclusion
Why is monthly crime so volatile? Largely, because each
crime is relatively rare. A slight change in the timing of
offending—which can be prompted by something as
innocuous as good weather—can lead crime to rise one
month and fall the next. As a general rule, the rarer a crime,
the more it fluctuates month to month. Therefore, homicide
is much more volatile than robbery or burglary. Similarly,
monthly crime numbers in smaller areas, such as Wards or
Police Districts, are much more volatile than for the District
as a whole.
Another reason is that crime is part of a dynamic interplay
between offenders on the one side, and law enforcement
and alert citizens on the other. Spikes in particular crimes are
met with increased vigilance and law enforcement efforts to
interrupt and prevent long-term increases. When successful,
these responses to crime surges prevent them from turning
into longer-term trends.
When is a monthly surge a leading indicator of a longerterm crime increase? Only occasionally. Usually, short-term
changes do not become long-term trends. Unfortunately, we
simply cannot tell in advance.
Data used in this report were provided to DCPI by the Research and Analysis Bureau of the Metropolitan Police Department and
were independently processed and analyzed by the authors.
1­
In this brief, we examine monthly changes starting in January 2001. We compare each month to the prior month, and also to the
same month in the previous year (see table 3). We graph months from one year earlier, starting in January 2000.
2
3
Decreases in homicide, by at least 10 and 25 percent, are the volatility number minus the number of increases.
4
This is also true for homicide, which does not show as apparent a yearly cycle.
Page 3 of 3
Download