Misdemeanor Prosecution (NBER) (ungated) is a new, blockbuster paper by Agan, Doleac and Harvey (ADH). Misdemeanor crimes are lesser crimes than felonies and typically carry a potential jail term of less than one year. Examples of misdemeanors include petty theft/shoplifting, prostitution, public intoxication, simple assault, disorderly conduct, trespass, vandalism, reckless driving, indecent exposure, and various drug crimes such as possession. Eighty percent of all criminal justice cases, some 13 million cases a year, are misdemeanors. ADH look at what happens to subsequent criminal behavior when misdemeanor cases are prosecuted versus non-prosecuted. Of course, the prosecuted differ from the non-prosecuted so we need to find situations where for random reasons comparable people are prosecuted and non-prosecuted. Not surprisingly some Assistant District Attorneys (ADAs) are more lenient than others when it comes to prosecuting misdemeanors. ADH use the random assignment of ADAs to a case to tease out the impact of prosecution–essentially finding two similar individuals one of whom got lucky and was assigned a lenient ADA and the other of whom got unlucky and was assigned a less lenient ADA.
We leverage the as-if random assignment of nonviolent misdemeanor cases to Assistant District Attorneys (ADAs) who decide whether a case should move forward with prosecution in the Suffolk County District Attorney’s Office in Massachusetts.These ADAs vary in the average leniency of their prosecution decisions. We find that,for the marginal defendant, nonprosecution of a nonviolent misdemeanor offense leads to large reductions in the likelihood of a new criminal complaint over the next two years.These local average treatment effects are largest for first-time defendants, suggesting that averting initial entry into the criminal justice system has the greatest benefits.
… We find that the marginal nonprosecuted misdemeanor defendant is 33 percentage points less likely to be issued a new criminal complaint within two years post-arraignment (58% less than the mean for complier” defendants who are prosecuted; p < 0.01). We find that nonprosecution reduces the likelihood of a new misdemeanor complaint by 24 percentage points (60%; p < 0.01), and reduces the likelihood of a new felony complaint by 8 percentage points (47%; not significant). Nonprosecution reduces the number of subsequent criminal complaints by 2.1 complaints (69%; p < .01); the number of subsequent misdemeanor complaints by 1.2 complaints (67%; p < .01), and the number of subsequent felony complaints by 0.7 complaints (75%; p < .05). We see significant reductions in subsequent criminal complaints for violent, disorderly conduct/theft, and motor vehicle offenses.
Did you get that? On a wide variety of margins, prosecution leads to more subsequent criminal behavior. How can this be?
We consider possible causal mechanisms that could be generating our findings. Cases that are not prosecuted by definition are closed on the day of arraignment. By contrast, the average time to disposition for prosecuted nonviolent misdemeanor cases in our sample is 185 days. This time spent in the criminal justice system may disrupt defendants’ work and family lives. Cases that are not prosecuted also by definition do not result in convictions, but 26% of prosecuted nonviolent misdemeanor cases in our sample result in a conviction. Criminal records of misdemeanor convictions may decrease defendants’ labor market prospects and increase their likelihoods of future prosecution and criminal record acquisition, conditional on future arrest. Finally, cases that are not prosecuted are at much lower risk of resulting in a criminal record of the complaint in the statewide criminal records system. We find that nonprosecution reduces the probability that a defendant will receive a criminal record of that nonviolent misdemeanor complaint by 55 percentage points (56%, p < .01). Criminal records of misdemeanor arrests may also damage defendants’ labor market prospects and increase their likelihoods of future prosecution and criminal record acquisition, conditional on future arrest. All three of these mechanisms may be contributing to the large reductions in subsequent criminal justice involvement following nonprosecution.
So should we stop prosecuting misdemeanors? Not necessarily. Even if prosecution increases crime by the prosecuted it can still lower crime overall through deterrence. In fact, since there are more people who are potentially deterred than who are prosecuted, general deterrence can swamp specific deterrence (albeit there are 13 million misdemeanors so that’s quite big). The authors, however, have gone some way towards addressing this objection because they combine their “micro” analysis with a “macro” analysis of a policy experiment.
During her 2018 election campaign, District Attorney Rollins pledged to establish a presumption of nonprosecution for 15 nonviolent misdemeanor offenses…After the inauguration of District Attorney Rollins, nonprosecution rates rose not only for cases involving the nonviolent misdemeanor offenses on the Rollins list, but also for those involving nonviolent misdemeanor offenses not on the Rollins list (and for all nonviolent misdemeanor cases)…. the increases in nonprosecution after the Rollins inauguration led to a 41 percentage point decrease in new criminal complaints for nonviolent misdemeanor cases on the Rollins list (not significant), a 47 percentage point decrease in new criminal complaints for nonviolent misdemeanor cases not on the Rollins list (p < .05), and a 56 percentage point decrease in new criminal complaints for all nonviolent misdemeanor cases (p < .05).
It’s unusual and impressive to see multiple sources of evidence in a single paper. (By the way, this paper is also a great model for learning all the new specification tests and techniques in the “leave-out” literature, exogeneity, relevance, exclusion restriction, monotonicity etc. all very clearly described.)
The policy study is a short-term study so we don’t know what happens if the rule is changed permanently but nevertheless this is good evidence that punishment can be criminogenic. I am uncomfortable, however, with thinking about non-prosecution as the choice variable, even on the margin. Crime should be punished. Becker wasn’t wrong about that. We need to ask more deeply, what is it about prosecution that increases subsequent criminal behavior? Could we do better by speeding up trials (a constitutional right that is often ignored!)–i.e. short, sharp punishment such as community service on the weekend? Is it time to to think about punishments that don’t require time off work? What about more diversion to programs that do not result in a criminal record? More generally, people accused and convicted of crimes ought to find help and acceptance in re-assimilating to civilized society. It’s crazy–not just wrong but counter-productive–that we make it difficult for people with a criminal record to get a job and access various medical and housing benefits.
The authors are too sophisticated to advocate for non-prosecution as a policy but it fits with the “defund the police,” and “end cash bail” movements. I worry, however, that after the tremendous gains of the 1990s we will let the pendulum swing back too far. A lot of what counts as cutting-edge crime policy today is simply the mood affiliation of a group of people who have no recollection of crime in the 1970s and 1980s. The great forgetting. It’s welcome news that we might be on the wrong side of the punishment Laffer curve and so can reduce punishment and crime at the same time. But it’s a huge mistake to think that the low levels of crime in the last two decades are a permanent features of the American landscape. We could lose it all in a mistaken fit of moralistic naivete.