The effect of district attourneys on criminal justice outcomes

In the United States, elected district attorneys’ offices prosecute over 85% of all felony cases, but we know little about their effect on local criminal justice outcomes. Using a newly-collected dataset of district attorney elections, I show that Republican district attorneys lead to a 18-21% increase in new prison admissions in the two years following their election, while nonwhite district attorneys lead to a 10% decline. In both cases, there are no significant effects on local crime or arrest rates. These results show that the identity of the local district attorney is an important determinant of incarceration rates.

Here is the paper, by Sam Krumholz, on the job market this year from UCSD, that is not his job market paper, here is his full portfolio, public economics and law and economics, to me one of the more interesting candidates this year.


Another paper by a student in an economics department which has nothing to do with economics. Here's a suggestion: shut the graduate division of that department down. They don't have enough work to do and they're not interested in the work they might have.

Maybe if the social scientists in the other social science departments did their jobs properly (and were not afraid to death of mathematics), economists wouldn't need to tread on their turf.

Demographers, geographers (other than David Harvey), and about 1/3 of this nation's sociologists are quite at home with statistical analysis. If the persons in question wish to study social questions, they can enroll in other departments and work with faculty who do make use of quantitative methods.

Art, you old cuck, give it a rest.

But those departments are hamstrung by their cultures... it's not that they can't do the stats, it's that they *won't* do the stats. The stats reveal inconvenient truths.

Isn't incarceration an economic issue? It costs a fortune to run the system. Inmates are rendered economically unproductive. Incarceration impacts on families' economic well-being, future earnings, etc. Without even considering the racial component this constitutes what economists call a natural experiment. If a change in incarceration rate has no effect on crime, what's the real story?

Isn't incarceration an economic issue?

You mean because it implicates resource allocation? Then just about everything is this side of poetry and philosophical speculation is an economic issue.

It costs a fortune to run the system.

Last I checked, allocating general state charges, New York's prison system had a budget of ~ $5.4 bn. That's around 0.5% of state gross domestic product. Other states spend more. New York has tended to invest more in police manpower and less in prison guard manpower. Of course, hiring cops costs money too. I tend to doubt that hiring cops would please the author of the paper.

I don't think street hoodlums are carrying round much human capital. Or is it your contention that mugging people and selling street drugs is a species of entrepreneurial activity?

And, again, have you investigated the deadweight loss from having a robbery rate of 240 per 100,000 rather than 70 per 100,000?

“Inmates are rendered economically unproductive”.

You make sound like an event simply happened to preempt their agency. No, many of them chose harm people, sometimes in the most egregious ways.

Anything having to do with scarcity is economics.


Lawyers who specialize in tournaments.

The government prosecutor is the ultimate bureaucrat — very powerful and unaccountable.
All U.S. prosecutors operate primarily on the highly corrupt practice of plea-bargains; fair jury trials are rare.

Origins of the extreme prosecutorial discretion and power we see in today's America are unclear, but stem from a steady consolidation of bureaucratic police and judicial government powers.

Perhaps the people should take back control of their criminal justice system... by eliminating many/all government prosecutors (elected or appointed)?

Plea bargains go both ways. Many good deals (for the defense) are offered because the prosecutors don't have enough time to look at the case in depth and realize it is provable to a jury.

True enough, though your last "..." sentence fragment does not make logical sense. Read Conrad Black's book on a hilarious but somber take on the American justice system. In a nutshell, for a very small violation of US securities law and alleged destroying of duplicate evidence already submitted to the US government, and contrary to Canadian law, Black, a UK-Canadian citizen, had to spend several tens of millions of dollars in legal fees, bankrupt his publishing empire, and spend time in jail. These fees enriched the Republican former SEC chief Breeden who was the 'private prosecutor' (and fed evidence to the US prosecutor that was later repudiated in court). If the rich can't buy a "get out of jail card" (recall Martha Stewart as well, also railroaded for a technical SEC / perjury violation) what chance do we lesser mortals have? Technically, the government could even nail 1% me for failing to report on time some overseas bank accounts I have. Absurd and with mandatory sentencing you cannot avoid severe criminal penalties once the prosecutor decides to convict you (often for resume enhancing reasons; imagine the feather cap for nailing Ray Lopez). The college admissions scandal (a relatively harmless matter that arguably should have been settled in a civil suit between college and parents) is another case in point. Almost makes you sympathize with those nutters who refuse to answer questions when hailed by police.

Still trying to figure out why the college thing proceeded.

It's interesting that in the end it will merely jail a handful of B list celebrities and nouveau riche bankers, and kicked off a few coaches of also-ran athletic programs, but largly left the universities untouched. So, there's that data point.

Still looking for a piece of that puzzle.

It wasn’t the bribes. It’s that the wrong people got paid off and the price was too low.

It is not that simple. Brazil has decided to adopt the plea bargain instrument as a way to make courts more efficient, streamline the bureacracy and punish more criminals.

Plea to this DUI with the kid in your car, or we will go after you for felony child abuse.

If anyone drives while intoxicated, and has children in the car, I’m fine with them getting the book thrown at them.

I take it you endorse hate crimes sentence escalators

"Republican district attorneys"..."while nonwhite district attorneys"...

The jaw, she drops. Can we expect a summary deplatforming for this gaffe?

Republican district attorneys see an increase over the prior levels. Democrats do not.

Non-white district attorneys see a decrease over previous levels. White DAs do not.

Not a gaffe, just the piss-poor writing style the dismal science is known for.

Back in the mid-1970s I worked for a committee in my state's legislature. The leadership decided that bills should have economic impact statements. In my committee, I was chosen to prepare the statements. It was the judiciary committee, at a time when mandatory minimum sentences had become the rage among law and order types. To prepare the economic impact statements for mandatory minimum bills, I simply determined the average length of sentence served for the crime, compared it to the mandatory minimum in the bill, then multiplied the difference by the daily prisoner incarceration cost to the state. Sponsors of the bills howled in protest, claiming I did not understand criminals, and that mandatory minimums would greatly reduce crime and hence save the state millions, not cost the state millions as I had projected in the statements. This was resolved by the leadership by exempting mandatory minimum bills from the economic impact statement requirement.

I'm pretty sure that the rate of crime committed against the public by those serving their mandatory-minimum sentences is extremely low.

What's with the U in attourneys? No even wrong. But, gosh, if the intent is to obfuscate, then why not mandataire, advocati, or procuratoribus?

What is also interesting is that the costs of incarceration for felonies is passed on to the state prison system, so the local prosecutor doesn't bear a cost but might get a headline.

One fault of the study, though.

Republican districts are notoriously more criminal, which might account for the higher number of prosecutions.

“Republican districts are notoriously more criminal, which might account for the higher number of prosecutions.”

The opposite of this is true.

Of course, the criminal stats will ultimately depend on where Donald takes up residence.

Just don't commit that crime you were planning in a Red district. The DA will get you.

I don’t get it. The entire United States has a higher criminality rate than the non-United States United States?

Are you mentally retarded or having a stroke? Use your life alert maybe boomer?

Maybe you are not a US citizen because you have trouble with understanding jokes.

Here's an explanation that doesn't assume bad faith: Republicans are more likely to believe people behave rationally; thus, if criminals are treated more harshly (longer sentences, etc.), people are less likely to commit a crime. The problem . . . . drum roll . . . . people aren't rational. That's a problem for economics, which assumes people are. I suspect that committing a crime is a lot like having sex, during which behaving irrationally is not only expected but a plus.

Some weird answers above, but here is the one big bullet point:

In both cases, there are no significant effects on local crime or arrest rates.

We have been told that the high US incarceration rate is necessary to keep crime in check. But, uh, this says no diff? That says we could go to a more humane, shorter sentence, system (as does Europe?) without creating more crime.

And save that incarceration money to boot. Return that budget to education.

No it does not say that.

The operative function on decision to commit crime turns out to be immediate arrest and incarceration.

We also now know reported crime rate is not an independent variable, it’s endogenous to likelihood of resolution of crime rates.

To dumb it down for you, the stats are orthogonal to reality in very weird ways. Local business owners apparently don’t report crime in many areas because they know there’s no response.

As a statistician, always use homicide. It’s the one thing they can’t lie about.

Dude. This report has a dataset, right? The dataset says "no significant effects on local crime," right?

Do you have better data, or is this pure handwaving?

I swear, commentators high or low, they think the primary purpose of the "submit" button is to reject new data.

Retribution is the most important reason to punish criminals.

And it would better to burn the money than to spend it on our public educational system. De-fund socialism!

Using a newly-collected dataset of district attorney elections, I show that Republican district attorneys lead to a 18-21% increase in new prison admissions in the two years following their election

This, I assume, means instances of Dem->Rep switching, not that every reelection of Reps (Rep->Rep) leads to a 20% increase.

That suggests that Dem->Rep switching probably driven by underserved demand for higher sentencing? Reps probably get elected based on 'tougher sentencing' but this is not discretionary, rather driven by their electors.

You can get an idea of what gets this chap out of bed in the morning by looking at his blog


If America doesn't act now, São Tomé may become a new Syria. If we lose São Tomé to sectarian strife, all bests are off.

I mean, all bets are off.

Tulsi is on it!

Good point. Representative Gabbard is tough on terror. But can we wait until she is elected?

Elected attorneys are not only among the most powerful officials in the country, they're also at the head of the queue for further electoral advancement. State attorneys general are in the best position to run for governor because they can produce their own headlines, which they do. District attorneys not only commonly move into the attorney general role, they are familiar enough with voters to successfully run for other positions as well. They must appear tough on crime to be elected, unless their constituency is made up of minorities.

"Using a newly-collected dataset of district attorney elections, I show that Republican district attorneys lead to a 18-21% increase in new prison admissions in the two years following their election..."

Note that the comparison is with respect to districts in which a Republican DA replaces a Democrat DA (irrespective of race/ethnicity).

This conclusion doesn't surprise me considering that, in general, I would expect a Republican candidate to promise tougher prosecution and sentencing. In fact, if anything, because of the methodology used, I supect the results are underestimated. The author uses the two year period after the election. However, it is important to consider that 1) there is usually a delay between the election and the time one takes office; and 2) there is also a delay (on average more than 121 days between arrest and sentencing). Thus, a significant part of the incarcerations after the two-year period following the *election* represent the policies of the prior DA. This is probably also evidence that Republican DA's are adhering to their campaign promises.

"...while nonwhite district attorneys lead to a 10% decline."

This is the comparison in districts in which a non-white DA replaces a white DA (irrespective of party affiliation). The author goes on to note that the main difference is with respect to drug cases.

I would be curious to know any differences between conviction/acquittal rates. Are Republican and/or white prosecutors more effective litigators? Does this account for some of the increased incarceration rate?

"In both cases, there are no significant effects on local crime or arrest rates".

This is a much more problematic conclusion. Due to the delay referenced above between an election and sentencing, one would not expect any additional deterrant effect for at least one year. Plus, is the change in incarceration rates "significant" to begin with? What is the definition of "significant"? If a 20 percent increase in incarceration rates leads to a 5 percent decrease in crime, I would consider that significant. Most importantly, any deterrant effect as a result of tougher prosecution and sentencing probably would not show to any significant extent until after that two-year period.

Re last point re effects: I think they lagged election and sentencing by two years. As for significance, it is probably a statistical significance test.

" I think they lagged election and sentencing by two years."

No---he clearly states he took the two years after the election. See this from the paper if it was not already clear from the abstract:

"First and most importantly, in my primary analysis the dependent variables are the changes in the primary outcome variables in the
two years following the election relative to the year prior to the election. I use a two year window because the typical election term is four years and differencing requires data from the year prior to the election."

Isn't taking it two years after the election lagging the result?

Please explain to me why it isn't.

Let me explain it to you in elementary English:

The author used the period from the election date and ending two years after that. Thus, the results reflect not only the period of time that the newly-elected DA had not even been sworn into office and some period after that in which he effectively had little control over the trial results.

Your comment also suffered from a second problem. You stated: "I think they lagged election and sentencing by two years". You used the conjunction "and". As I noted above, the average time from trial to sentence would be at least 120 days. Thus, the author would have included sentences handed down for trials controlled by the newly-elected DA's predecessor(s). Thus, as I indicated, it is likely that the average sentences under a Democrat to Republican change would be even greater than he measured. For the reasons I stated, I don't think the deterrent effect, if any, would have been reflected in such a short (and over-lapping) time frame.

In answer to Bill @ 47. To be even more clear, the author did not use the period two years after sentencing. The relevant time period used was always two years from the election date, if we are to take the author's own description seriously.

Comments for this post are closed