« Rotations | Main | Great to be back and greetings from Washington! »

Thursday, May 02, 2013

What is a Drug Crime? Part II

In my last post on drug offenses, I noted that the standard definition of a “drug incarceration”—someone convicted of a drug possession or drug distribution offense—may provide too narrow a view of how the war on drugs is driving prison populations. How, for example, should we count the person whose life is upended by a minor drug arrest and ultimately turns to crime as a result: isn’t his incarceration for, say, robbery in some ways tied to the war on drugs? Or how about drug-market related violence that arises in no small part because there is no legal market nor legal ways to handle business disputes?

It isn’t possible to address all these here. For example, it is almost impossible to determine which violent crimes were drug-market related from official statistics, except perhaps for homicide.1 But two datasets gathered by the BJS provide at least some indirect evidence on the extent to which prior drug arrests and convictions may be responsible for subsequent non-drug incarcerations.

To foreshadow my conclusions: both suggest that the effect of prior drug arrests is weak. The evidence is by no means clear-cut or dispositive, but neither is it wholly ambiguous.

First consider the results from the State Court Processing Statistics. The SCPS was a survey of felony trial cases conducted biennially between 1988 and 2006. The cases were chosen to provide a representative picture of criminal trial case outcomes in the 75 largest counties in the United States, which account for about half the crime in the country.2 The survey followed cases from initial filing through disposition or one year, whichever came first, and thus recorded outcomes in plea cases as well as trial, avoid the tip-of-the-iceberg problem. Among the questions asked in the surveyed cases is whether the defendant was ever arrested or convicted of a felony or misdemeanor in the past. Unfortunately, the survey did not gather any information on what those offenses were, so the measure is quite crude.

Looking at the 2004 survey, there were approximately 57,500 trials in the 75 most populous counties.3 Of these 57,500 cases, about 59% had at least one prior felony arrest, and another 10% had no prior felony arrest but at least one prior misdemeanor arrest. Looking at convictions, 44% had at least one prior felony conviction, and another 15% had no prior felony convictions but at least one misdemeanor conviction. (Similar patterns hold when I look separately at those facing charges for violent, property, drug, or “other” crimes.)

So 60% have had some sort of conviction, and 70% some sort of arrest. But, reframing, this means that 40% of felony defendants were facing their first conviction of any sort, and that 30% had never even been arrested—and for these defendants, it is hard to see how to attribute their current state to the war on drugs.

Furthermore, for the 60% with prior convictions and 70% with prior arrests, at least some fraction of these cases involve defendants with no drug-related priors. It’s hard to say more than that. One could try to make an educated guess, like “it seems likely that perhaps no more than half the cases in the large counties involve defendants with prior drug convictions,” but that is nothing more than a guess.

To get around the coarseness of the SCPS, I decided to look also at the 2004 wave of the Survey of Inmates in State Correctional Facilities. The SISCF provides incredibly detailed questioning about inmates’ current and past crimes, demographics, etc., etc.: it contains well over 2000 variables. For prior criminal history alone, it asks about the top five charges from the inmate’s first arrest, as well as the top five conviction offenses for each of the last ten (!) times he was incarcerated. The 2004 survey sampled 14,499 inmates, weighted to represent the more than 1.2M people in prison that year.

To crudely estimate the effect of the war on drugs, I created a variable that recorded whether the inmate had reported a drug offense for any of the five convictions for his prior ten incarcerations. The results? About 10% of all inmates have some sort of drug prior incarceration (perhaps in conjunction with other, more serious offenses), and for about 14% their first arrest was drug-related (again, perhaps in conjuction with other, more serious crimes).

When it comes to marijuana, the great boogey-man of the War on Drugs story, with everyone from Ernest Drucker to Adam Gopnik pointing to the arrest of marijuana users as a major driver of prison growth, the numbers are even smaller: only 2% report a prior conviction for marijuana, and only 4% reported a marijuana charge among the charges at their first arrest.

Now it is important to note that the SISCF doesn’t provide information on prior felony convictions which do not result in incarceration, which is a major omission. A felony conviction, even absent incarceration, can have all sorts of collateral consequences that may push convicts toward a life of crime. So these estimates are lower bounds on the number of felony convictions due to drug charges. But they are very low floors.

There is another limitation to these results: they look at a stock, not a flow—at total population, not admissions. This is a particularly tricky and important issue, one that has snared other people writing on this topic, and so I want to address more fully in a future post. But the short point is that if we are concerned that any sort of short contact with the prison system can have destabilizing effects throughout the lifecourse, then the SISCF data will underestimate the number of such contacts and thus potentially understate the importance of prior drug convictions.


1The FBI’s Supplementary Homicide Reports do provide information on the motive for the killing, including “narcotics law” and “gangland killing.”

2There are 3,141 counties in the United States, so half of all crime takes place in less than 2.5% of all counties—although these counties also contain about one-third of the nation’s population, so crime isn’t as “overconcentrated” as the 2.5% number suggests. It’s easy to forget how much of the US is sparsely populated, which is why these red/blue maps are so much more useful than the ones based on geographic borders.

3Officially, the SCPS observed only 15,761 cases that year. But the data provides weights to calculate population-level estimates. It is thus important to remember that the numbers here are imprecise estimates, not (theoretically) precise census counts.

Posted by John Pfaff on May 2, 2013 at 10:56 AM | Permalink


TrackBack URL for this entry:

Listed below are links to weblogs that reference What is a Drug Crime? Part II:


I was unclear in my post, and I've corrected it a bit to fix that. The SCPS started following a case when the case was filed, so it looks at cases that are resolved both by plea and by trial, and so it avoids that 99%-of-cases-settle problem, since it includes both types of cases.

Posted by: John Pfaff | May 2, 2013 11:24:06 AM

Another limitation—how is it valid to extrapolate from extremely rare cases that go to trial to ordinary cases, which do not? Given collateral consequences, etc., doesn't someone without a criminal history have a greater incentive to go to trial? In general, should we expect the 90-99+% of cases resolved without trial be represented by the ones that do? No.

Posted by: j | May 2, 2013 11:14:31 AM

The comments to this entry are closed.