The Central Role of Prosecutors In Prison Growth

Having dismissed the War on Drugs as first-order cause of prison growth,¹ I now want to provide my explanation for the causes of prison growth. Focusing for now on prison growth during the crime decline that began in the early 1990s--partly due to limitations in available data, partly because the causal story for prison growth during the crime delcine surely differs from that during the crime boom--the story I want to tell is surprisingly simple.

Despite the complexity of criminal justice system and its relationship to crime and other social pressures, since 1994 prison growth appears to be driven primarily by a single factor: prosecutors' decisions to file felony charges.

I will spend the next several posts making my case for this claim, given that it runs so contrary to the Standard Story. In this post I'll just lay out the basic argument, with the support to come in future posts.

Let’s start by first noting where prison growth can occur. There are basically six sources of growth:

1. Increased crime
2. Increased arrests per crime
3. Increased felony filings per arrest
4. Increased convictions per felony filing
5. Increased admissions per conviction
6. Increased time served per admission

Note that each of these factors implicates different actors and different bureaucracies: hyperlocal criminals; city police or county sheriffs; county district attorneys (for the filings); county district attorneys, local defense attorneys, county jurors, county or state judges (for the convictions); district attorneys, state or county judges, and state legislators (for the admissions); and the same—albeit to different degrees and in different ways—for time served, along with state-level parole boards.

If nothing else, this list of actors should eliminate any faith we have in empirical models that use total admissions, total prison populations, admission rate, or incarceration rate as a dependent variable. What does it mean to regress, say, percent black on total prison population? Race surely has different effects on crimes, arrests, prosecutions, convictions, admissions, and time served. So the coefficient is some confusing weighted average of all these effects--but one for which we have no idea what the weights are and thus no idea how to use the result in any meaningful policy sense.²

So, to understand prison growth, we have to understand who is doing how much of the work. Only then can we figure out what causal factors matter where.

What, then, do we see in the data since 1994?

1. Crime has been falling.
2. Arrests per crime have been flat for violent and property crime. Non-marijuana drug arrests³ have risen slightly since the early 2000s, but total property, violent, and non-marijuana arrests dropped by about 10 percent between 1994 and 2010.
3. Filings per arrest have soared.
4. Convictions... well, I have to elide over those here. See the footnote.⁴
5. Admissions per filing (and yes, that's per-filing, not per-conviction) have been flat.
6. Time served has been relatively flat. It really has been. I’m going to back this one up, I promise.

Thus the only real change has been in filings per arrest, and it has been dramatic.

I’ll stop this post here. In the next post, I’ll show clearly why filings have at least been the key source of growth in admissions. The post after that will explain why I strongly believe, contrary to all the Standard-Story explanations I see every day, that time served has been relatively stable.

 

¹As I’ve stressed earlier, and as I will point out a few posts down the line, the War on Drugs may in fact play an important role in prison growth, but it is an indirect one, and one that is substantially more complicated to identify empirically and resolve normatively.

²That’s just one problem. I discuss a host more here. A big problem with using prison population or incarceration rate is that not only is the effect of, say, race averaged over all the stages, it is also averaged over many years: the prison population in 1990 is a function of crimes, arrests, …, admissions, and time served from cohorts admitted in 1990, 1989, 1988, 1987…. So the coefficient on race tells you about its average effect over all these admission cohorts as well. And since the effect of these factors vary over time--do we really think race plays the exact same role in shaping sentences today as it did in 1960?--these coefficients are basically useless.

Sadly, the papers that used this variable generally had even deeper methodological flaws that render their results uninformative even without this conceptual problem. The general shoddiness of empirical work on prison growth really is a black eye for criminology, a self-inflicted injury whose causes I can’t quite figure out.

³Too often we see claims casually linking marijuana arrests to prison populations. It is very hard to go to prison on a marijuana charge, especially a low-level possession charge.

⁴ The court data I use to look at this issue is primarily National Center on State Court’s annual tabulation of total felony filings in state court. This is an aggregate number, not case-level data, and it provides no information on outcomes of cases. As I explain the paper this work comes from, I tried to use a different dataset to look at convictions per filing and admissions per conviction, but the dataset is… a bit wonky, and the results are unclear.