<< The Consequences of Unseriousness, Part III | Main | Where's Al Sharpton When We Need Him? >>

Increase in Crimes the FBI Does Not Count

| 1 Comment
Last month, you may recall, the FBI announced its final Crime in the United States report for 2012, showing an increase in violent crimes of 0.7% and a decline in property crimes of 0.9%.  (Press release here; prior post here.)  Today, the Bureau of Justice Statistics, a different branch of USDoJ, announced the results of its National Crime Victimization Survey (NCVS), showing violent crime increased 17.7% and property crime increased 15.0%. 

The press release is here.  The full report is here.  Percentage changes are in Table 9 on page 9.

What gives?  How can these two measures of "crime rates" produce such dramatically different results?  The devil is in the details.
These different measures, by two different agencies, have different definitions and different methodologies.  Each has blind spots, seen by the other.

The FBI's Uniform Crime Reports (UCR) are the basis for Crime in the United States.  This is full count, not a sample, of crimes known to the police.  For the violent crime figure, only assaults that are classified as aggravated are counted.  "Forcible rape" is counted, not the broader category of "sexual assault."

The NCVS calls people on the phone, like public opinion polls, or visits them in person and asks people over 12 about crimes committed against them personally.  Simple assaults and sexual assaults other than rape are tallied and go into the violent crime category.  Crimes never reported to the police, and therefore invisible to the UCR, are counted.  The NCVS does not include homicide, as only living, direct victims are surveyed.  It does not include crimes against children under 12.* 

"Given these differences," says BJS (p. 9), "the two measures of crime should be considered to complement each other and provide a more comprehensive picture of crime in the United States."

Overall, the UCR leans more heavily toward more serious crimes, while the NCVS is more weighted to the less serious crimes.  Crimes not reported to police will be, on the whole, less severe than the ones that are reported.  Homicides, not included in NCVS, are the most severe of all.  The inclusions of simple assault and sexual assaults other than rape also give NCVS a lower center of gravity.

Most crime indexes, including UCR and NCVS, simply count crimes and do not weight them for severity.  Such unweighted indexes tend to be dominated by the least severe crimes included in them, because crimes generally number in inverse relation to their severity.  Inclusion of less serious crimes in an index therefore changes that index substantially.

So what does it all mean?

Why is there a skew toward less severe crimes in 2012, with a big jump on the low end of the scale and little change higher up the criminal ladder?  One possibility is a general softening of criminal justice policy nationwide.  The self-designated "smart on crime" people are proclaiming nationwide, based on flimsy evidence, that we have been locking up lots and lots of people we don't "need" to lock up.  They are finding receptive ears in financially strapped governments, even among people who ought to know better.  Efforts to drive down the incarcerated population start with those convicted of lesser offenses.  That means more persons who have committed such offenses in the past are on the street instead of in jail.  It also means that those on the street disposed to such offenses have less to fear if they commit them and are caught. 

This also means victims of those offenses have less reason to report and more reason not to.  The perpetrator of a lesser offense is now more likely to dance in and out of the revolving door of criminal justice, and may come looking for the reporting victim next.

These are, of course, hypotheses.  But they are entirely plausible.  If anyone has alternative hypotheses, fire away in the comments.


* There are additional differences.  Because it is a sample and not a full census, NCVS numbers have a margin of error (or confidence interval) to allow for sampling error, just as polls do.  The sample size for 2012 is N=162,940.  That is large, but still only 1% of all victimizations are sampled.  The UCR is intended to be a full count and does not have the random sampling error issue, although there are problems with missing data.  The NCVS does not produce state or local data, as the sample size would be too small.  The UCR does.

1 Comment


One thing that strikes me is that the year-to-year crime differences found by the UCR are small (less than one percent), while those found by the NCVS are much larger (at least 15 times as large for each category). It's hard to believe that changes that big are an illusion. Something is going on here, and it's not good.

Overall, I agree with your hypothesis that, since the states following this "Smart on Crime" baloney are making at least some effort to release the less dangerous criminals, the increase in crime would show up in the less dangerous categories.

But crime is still crime, and an elderly couple whose life savings have just been wiped out by a "low level, non-violent" swindler have been devastated nonetheless.

As I said in my earlier entry, when you prematurely release criminals, you get more crime. And none of us should be fooled into thinking that future releases will not include people who are plenty dangerous, however one defines that word.

Leave a comment

Monthly Archives