The Flawed NRC Report: The Mysterious Case of the Missing Prosecutor, Part 1

(This is Part 4 in my criticism of the recent National Research Council report on incarceration. Here are Part 1,  Part 2, and Part 3.) 

As I argued yesterday, the claim that longer sentences drove prison growth in the 1990s is at worst wrong and at best an extreme oversimplification. Which means that prison growth must be driven by admission-side factors. In fact, I have a paper that points to one factor in particular: the prosecutor. Unfortunately, while the NRC report acknowledges that admissions droves growth in the 2000s, it reaches that conclusion in way that makes it miss the role of the prosecutor. 

This is perhaps the single biggest error in the report’s analysis of the causes of prison growth, so I want to spend a few posts looking at how it makes this mistake, and how failing to identify the central role of the prosecutor poisons efforts to make meaningful policy recommendations more generally.

To its credit, the report does note that admissions drove prison growth in the 2000s. But two key flaws nonetheless persist. First, it gets the dates wrong: admissions drive growth in the 1990s ass well, perhaps even earlier (but the data I have starts in 1994). Second—and this is what I want to focus on the most—its analysis of why admissions grew is simplistic to the point of being unusable.

To figure out why admissions grew, the NRC report disaggregates the path to prison: crimes, arrests per crime, and admissions per arrest. It admits that “admissions per arrest” is a clunky term—we’d really like to filings per arrest and convictions per filing—but that it is a concession to limitations in the data: 

Because national trend data are not readily available for charging and conviction, analysis of imprisonment population dynamics has examined the probability of prison admission given an arrest.

There are two big mistakes with this sentence, tied together by a common thread: charging data is actually quite readily available. First, there actually is an easily-accessible state-level nation-wide database on case filings. And while the on-line version is an annual pdf report, all it took for me to get the data in a computer-friendly Excel sheet was sending a single email.* 

Second, every state Secretary of State (or at least every one I’ve checked with) compiles court processing statistics—usually at the county, not state, level, the importance of which will soon be clear—and makes them available on-line. Sure, the formatting is often inconvenient (pdfs, strangely-designed Excel sheets, etc.), and gathering the data requires going to fifty different websites and downloading multiple annual reports. And states count differently, use different terms, and so on. But “the data require some work to gather and merge” is no excuse for not gathering and mergin them, especially for an authoritative NRC report.

Now the committee is right that conviction data is hard to get. The National Judicial Reporting Program provides data only on those who have been convicted (so you can't detect the conviction rate); the State Court Processing Statistics dataset does provide case-level data on conviction outcomes, but it is a sufficiently wonky dataset that the BJS has issued a special warning against relying on it too much for causal analysis.

But even breaking up “admissions per arrest” into “filings per arrest” and “admissions per filing” is hugely useful. Because here is the problem with using “admissions per arrest”: agreeing that that is the source of growth, if we want to rein in or regulate prison growth, what reforms does that recommend? After all, “admissions per arrest” implicates numerous independent criminal justice bureaucracies. Are prosecutors becoming more aggressive? Are judges using their discretion more aggressively? Are legislators forcing judges to act more aggressively through mandatory minimums? 

Each of these theories yields different reform suggestions—repeal of mandatory minimums if it is legislators, guidelines or other limitations if it is judges, something else enitrely if it is prosecutors. And not only are the reforms different, so too are the jurisdiction: changing legislative outcomes is a state-level action, but regulating prosecutors has to take place at the county level (since in almost every state no state official has any oversight of directly-elected county-level DAs). So knowing it is “admissions per arrest” really doesn’t tell us much.

But adding in felony caseloads tells us a lot. Consider Figure 3B from this paper:

What this figure says is the following: between 1994 and 2008, the probability that an arrest would result in a felony filing** rose from 0.387 to 0.57, an increase of 54%. During the same time, the probability a felony case would result in a prison admission was flat at 0.26. In other words, growth is driven almost entirely by an increase in felony filings.

This solves the “who” question that the report’s analysis can’t answer: growth is driven almost entirely by the prosecutor. Once a felony case is filed, the defendant is no more likely to go to prison in 2008 than in 1994. But the probability that an arrestee finds himself confronting a felony charge soars.

This is a significantly important finding. Almost all analyses of prisons looks at national- or state-level data, and the NRC report is no exception. And this focus, while wrong, makes intuitive sense: prisons are state institutions, so it is logical to compare state outcomes. But while prisons are run by the state, my results indicate that prisoners come from the county.

In other words, to understand prison growth, we must understand the prosecutor. And to understand the prosecutor, we have to turn our attention to the counties. The NRC report simply fails to do this. And in my next several posts, I will examine the very real costs of this failure.

 

 * Strangely, the on-line version only links to the 2010 report. When I first used the data, all past reports were easily available as well; it is unclear why past reports have been taken down or made less accessible. Regardless, all it takes is an email to get all the past data.

** This is just the ratio of state arrests for all index violent and property crimes and non-marijuana drug offenses to total felony case filings in that state in that year.