Do longer sentences reduce crime?

I find that Republican prosecutorial offices sentence defendants to longer incarceration spells as compared to their Democratic and Independent counterparts. This increase in incarceration length is driven by longer sentences for both violent and prop- erty offenses, and translates into a persistent increase in incarceration. These sentencing and incarceration enhancements do not lower crime at the county level, indicating that, in terms of public safety, the marginal return to the tough-on-crime stance may be close to zero.

That is from a new AEA paper by Ashna Arora.


Republicans do enjoy finding expensive and ineffective ways to waste taxpayer money. Wars, bailouts, tax cuts for corporations, prisons, walls, etc.

Obama was the first 2-term President in US history to spend all eight years of his presidency at war, and bombed seven different countries.

But thanks for playing!

Make up your mind, Obama was bad for attempting drawdowns on Bush's wars, or he was bad for keeping them going?

Let me guess, it flips with every Tweet.

The author is denying the facts. Crime is down, incarceration lengths are up, duh!

Longer incarceration rates clearly and obviously reduce the crime that incarcerated criminals commit.

And math is hard!

According to all the experts, we have not been at war since the treaties ending WWII because "use of military force" is not war.

On the other hand, if military force is war, the US has been continously at war since at least the 30s, if not before.

Reagan certainly takes the cake with his war with the British Commonwealth, proving the Iron Lady was rather soft.

I don' t know about reducing crime, but longer sentences can be harder to read. ;)

And reading surely anti-correlates with crime...

The shorter the sentence, the more time you have for crime after reading it?

If I remember right, the quickness and certainty of punishment matters a lot more than the punishment length - and you can't make up for the former two simply by making the sentences longer.

This. In particular, I believe there's a good deal of evidence that swift and certain is effective for those on probation. Harder to do swift and certain for those not already in the cj system for obvious reasons.

And if I remember right, people serving prison sentences are not out in the community committing crimes. The crime rate therefore drops by definition.

Tyler and the rest of the left need you to believe that people who serve longer sentences get out and "make up for lost time" by going on a bonus crime-spree due to the longer sentence -- without getting caught, of course -- so that the overall crime rate stays the same.

Your job when you hear this argument is simple: Don't be an idiot.

You are just making the assumption that the model of crime is a set of criminals, which does not change very much, who spend their days committing crimes, so removing them would obviously lower crime.

What the paper gets to is precisely that this model cannot be correct, based on what happens in reality.

There are mepant models where incarceration of criminals does not help. We could be dealing with a baseline demand for crime, and different 'workers' fill said demand. This is what would happen if we were discussing, for instance, crime under prohibition. It's also possible that recidivism is much lower than you expect, therefor making the number of crimes prevented to be far lower than your estimate. Then you add other negative effects, like incarceration itself increasing the chances of committing a crime after you go out, or the disappearance of the prisoner from society turning people increasing the chances of people that were near them to commit crimes.

The best model is the one that predicts reality better, and the traditional tough on crime model just does not match reality. Your job is to look at what happens in practice and see if it matches your model, instead of picking something that feels right to you, despite having no evidence. Arguing with reality is rarely a productive endeavor.

"Arguing with reality is rarely a productive endeavor."

Tell that to any politician from a safe GOP district.

"people serving prison sentences are not out in the community committing crimes. The crime rate therefore drops by definition."
This non-sequitur only makes sense if the supply of criminals is static. Besides for the mechanism that Bob posits, there could also be an offsetting effect of destabilization and financial hardship of families and communities that result from longer incarceration periods, creating new criminals.

"And if I remember right, people serving prison sentences are not out in the community committing crimes. The crime rate therefore drops by definition."


What happens is demand for crime creates a new supply of criminals when existing criminals are removed from the market.

And jail/prison is not the only method of removing crime suppliers from the market: cops killing criminals. Criminals killing criminals. Citizens killing criminals.

Given criminals being killed is "swift justice", the theory that punishment needs to be swift to deter or stop crime gets disproved as well.

Ie, in cities where police fail to charge anyone with crimes against others, as opposed to arresting business people engaged in mutual trade, most murders and injuries are citizen justice, loudly advocated by the NRA. The good guy is defined by the second amendment justice activists, local republican governance. This is, of course, a long American tradition, opposed by liberal do gooders. What could be better than the klan, clearly good guys wearing white robes.

> What happens is demand for crime creates a new supply
> of criminals when existing criminals are removed from
> the market.

Also the quality of criminals is lowered when existing criminals are removed. That is because the market had choosen the best criminal in the post. And after removing him only the second best is available for the duty.

Have they controlled for other factors like the level of crime in the community? I can envisage a scenario in which more violent/crime afflicted communities tend to choose tougher sentencing which would cover up any positive, crime reducing effects of longer sentencing.

It's not a controlled experiment (yikes, the IRB on that one!), but the author looks for a discontinuity in just-barely-elected Republican vs Democrat prosecutors (assuming that the kinds of places that just barely elect a Republican aren't that different from the places that just barely elect a Democrat). Relatively standard "quasi-experimental" technique in economics.

From the full abstract at the link:
> Exploiting quasi-experimental variation generated by close elections, I find that Republican prosecutorial offices sentence defendants to longer incarceration spells as compared to their Democratic and Independent counterparts. This increase in incarceration length is driven by longer sentences for both violent and prop- erty offenses

Thanks. That raises my level of confidence slightly.

The excerpt is equivalent to saying recidivism is irrelevant, which seems quite dubious to me.

I could believe longer sentences are irrelevant to the number of new criminals per unit time, but not to total crime as definitionally a portion of the population inclined to commit crimes have had their opportunity to engage in new crimes significantly decreased while incarcerated.

I don't think that the paper makes that exact mistake -- it tests for total arrests/yr for {violent, property} crime in Republican-prosecutor areas, and so if there's less crime because the criminals are in prison, it would be observed in the lower crimes-committed rate (figure 7), which the paper fails to find.

Now, I think a better criticism of the paper is that its prose conflates "experiment fails to reject 0 effect" (which it does) with "experiment demonstrates 0 effect" (which it does not). The average effect size is not larger than the noise in the author's experiment, but it *is* in the direction of Republican prosecutors (and their longer sentences) reducing crime occurrences. Given that there's an intuitive explanation in that direction (as you point out, at least in locking up criminals preventing some future crimes), a more rigorous treatment would have tried harder to confirm or falsify that claim before holding up "no significant result" as purported disproof of tough-on-crime efficacy.

I won't defend the paper (bc I haven't read it carefully enough). But I can think of one reason why lengthy prison sentences may have no observable effect on crime rates. If a family's main bread winner is locked up for a long period, then someone else will have to take over that responsibility. The next person in line will likely have less job experience (otherwise he or she would have been the main bread winner already), and therefore may resort to criminal activities to make ends meet. If so, then locking people up doesn't reduce the number of criminals; it simply encourages their replacement with a new set of criminals and thus keeps the crime rate and the number of criminal actors, at any given point, relatively stable.

Talk about middle-class insularity. I assure you, very few criminals were functioning as the "main breadwinners" of a family prior to incarceration. I don't have the exact number handy, but something like 92% of New York State prison inmates receive zero visitors during their term, which indicates the strength of their relationships with their families.

It's not middle-class insularity, it's left-wing middle-class insularity. They believe some poor exploited minority gave in to temptation and mugged somebody and hence had to drop out of his PhD program to serve some undeserved jail time because he was falsely arrested by a mean white cop and convicted by a racist patriarchal jury.

I wish these people would go away, but they aren't.

Who believes that? Nobody.

Very unlikely. According to my social workers at the womens' prison, 25% of the children of the inmates are unaware that their mothers are incarcerated. On the male side of things the numbers are likely vastly higher, but it is exceedingly hard to know how many children they actually have. Those who have intact social situations of any sort prior to incarceration are exceedingly rare.

Further, crime doesn't pay. The average drug dealer makes around minimum wage. You can earn somewhat more with property crime, but that tends to require more skill as well. Those who are incarcerated are typically on the lower end of earnings for their various criminal professions. In a nutshell even in the highly rare scenario you suggest, you can replace the lost income pretty easily working for McDonalds.

What I think is much more plausible is that the longer a parental figure is incarcerated the more their children will adopt antisocial beliefs and attitudes. We know that children without stable paternal influences are on average more likely to become incarcerated themselves. It is not too much of a stretch to assume that increasing lengths of incarceration result in more children lacking parental influence during puberty and falling in with gangs at greater rates. If the father of a 7-year-old boy is incarcerated for 3 years, he gets out when his son is 10 and likely has a non-zero impact on his son staying out of gangs. If he gets 5 years, his son is 12 and likely has to navigate his first serious interactions with gangs without paternal support. 7 - years means that we are looking at the prime gang recruitment window being entirely during a fatherless period.

>he gets out when his son is 10 and likely has a non-zero impact on his son staying out of gangs.

The gangbanger father, you mean? Yeah, I'm sure he's an exemplary Dad. Assuming he has even met the kid, his non-zero impact is likely to be "Junior, get out of school and into this gang. You'll be making bank and getting respect at 12 years old. And even if you screw up as bad as I did, you only have to do three years.... with your friends!"

The data suggests that it does. Crime is a young mans game and shockingly gang bangers tend to outgrow their gang affiliations. The majority of gang members are typically under 20. Say he became a father at 17 (very common among incarcerated individuals), he is arrested (again) at 24. He serves say 5 years after that. By thirty most people don't like hanging around stasis obsessed teenagers and leave the gang. Keeping him to 32 will undoubtedly prevent him from committing crimes, but I would have the lead parental role for his son be a washed up 27-year-old has been than a 19-year-old in the peak age of violence and anti-social behavior.

Remember I am not saying that dad is good, just less bad than his replacement(s).

Letting out the other 60-80% of offenders who don't have children (and this increases the younger age bracket the crimes go), so that the 20-40% can putatively offer a slightly improved role model to their sons, which they will not reliably do at all, seems largely a bonkers consideration unless longer sentences have totally zero value.

The implication of the study is not that longer sentence length is irrelevant to creating new criminals, but that longer sentence length is an impetus to creating new criminals. This would offset recidivism (whatever amount exists) and keep the level of crime the same as a jurisdiction with shorter sentences.

Signaling. That's the point of longer sentences. It signals that Republicans are tough on crime. The Wall. Signaling. It signals that Trump is tough on illegal immigration. Nuclear weapons, aircraft carriers and battleships, bombers and fighter aircraft, enough weaponry to destroy the world and everyone in it. Signaling. It signals that the politicians who vote for them are tough on our enemies, real and imagined. Signaling. Caplan shouldn't be bashful when it comes to signaling. Higher education may well be mostly about signaling, but it's harmless signaling. The kind of signaling I have identified isn't. Come on Caplan, tell us more about dangerous forms of signaling.

That is so insightful and courageous (and relevant), to ferret out political malfeasance from 1948.

The wall is signalling? Which is easier, climbing over a wall or just walking across a border?

See if you can answer the question without engaging one of your heuristics.

Hmmm.... So the real question of interest to me here is how much, if at all, this contributes to the US's overall incarceration and for what crimes it's used for. Finding some marginal effect that Republicans give slightly higher sentencing for some serious crimes, but it's not really a very large effect, is not so important for that.

So I'll skim read:

Abstract: I find that Republican prosecutorial offices sentence defendants to longer incarceration spells as compared to their Democratic and Independent counterparts. This increase in incarceration length is driven by longer sentences for both violent and property offenses, and translates into a persistent increase in incarceration.

This paper shows that Republican prosecutorial offices lead to longer incarceration spells as compared to their Democratic and Independent counterparts. These effects are identified using
quasi - experimental variation generated by close elections of chief prosecutors at the county level over the period 1980 - 2014.

The data indicate that these effects are driven by the lower use of
alternative sentences such as restitution
, and translate into a persistent increase in incarceration.

This rise in incarceration does not lower crime at the county level, as arrests across a variety of categories remain unchanged.

We observe large, significant increases in sentence length for violent crimes such as rape and armed robbery, as well as property offenses like burglary, auto theft, grand larceny and receiving stolen property. There does not exist a consistent pattern of effects on sentences for drug offenses.

So the effect seems to be driven by Dem prosecutors using "alternative strategies" for sentencing "violent and property" crimes.

Republican DA is associated with higher imprisonment and lower alternative strategies like fines, probation, restitution for sentencing. It's not in "If sentenced, then higher time", rather than "It sentenced then >p jail".

But are Democratic DAs really using jail time sentences less overall for very serious crimes? It seems like it's unlikely to me.

The data is clearer than the qualitative analysis. See Figure 1 -- R/D prosecutors hand down nearly the same number of sentences during their term. And then see Figure 2 -- D's hand down nearly twice as many 1-12mo sentences, with R's clearly handing down more 3-5yr, 5-10yr, and >10yr sentences.

This seems to support D's being more likely to hand down less time plus some "alternative strategy", not just no time at all.

Probably, but Dems are not really going to be handing down fewer >10year sentences by substituting 1-12mo sentences plus alternative strategy (fines or whatever), so the substitution effect must be mainly between 1-12 plus alternative strategy and 3-5 year surely?

Well, Fig 2 suggests that R's hand down ~30% more >10yr sentences than D's, so those are being substituted into something...

(Feels pretty natural to guess that they're downshifting 11yr sentences into 9yr sentences, 9's into 8's, and 5.5's into 4.5's, etc. But the paper apparently doesn't do any kind of identification to test that.)

Given that the paper's arguing for a counterintuitive finding, I'm disappointed that it conflates "we found a right-direction effect but could not exclude 0" (which it does) with "we demonstrate 0 effect" (which it never quite says, but frames things to suggest".

Actually, I'm not disappointed with the author, since this style of analysis is sadly common in the social sciences, but I am disappointed with Tyler for choosing to (misleadingly) highlight a paper that claims too much in this way, when the paper's true answer to the blog post title is more like "dunno" than "no".

And to save Prof. Cowen the need, there are other interesting papers at the link, relating to the economics of crime.

I thought prosecutors recommended a sentence, and the judge “handed down” or actually set the sentence. Presumably the set of judges doesn’t change when a new DA is elected.

Or is it the case that the judges are rubber stamping esentially all of the plea agreements made which are essentially all of the cases?

We see that while Republican District Attorneys lead to large increases in sentence length in the period 1980-2004, these effects are absent in the period 2005-15. This indicates that the judicial branch may be capable of blocking, and in fact, entirely offsetting the influence of political preferences of prosecutorial offices.

I only saw a table breaking down the results on the length of incarceration pre and post 2004, but not on subsequent crime in those periods. Regardless, there's no need to worry about Republican prosecutors anymore because our black-robed masters have "fixed" the "problem."

Thank goodness various judges have put perv Larry Nassar away for the sum total of 360 years. That will convince other pervs not fool around with little girls.

Longer sentences decrease readability.

One might also ask whether absence of sentencing increases crime. A first estimate is that every repeat offense is preventable. That would certainly be the case for capital crimes like murder and rape.

Fig 1 of the paper suggests that R and D prosecutors give an equal number of sentences; it's just that the Ds' sentences are shorter on average.

Not to say that your question isn't an interesting one to ask, just that this paper's quasi-experiment can't investigate it.

I fear the growth of a new subfield--identity economics.

I assume that if there is no punishment whatsoever, that crime will increase. Perhaps I am wrong, but go with it. For crime rates to not change, given that more people are in prison, the proportion of criminals in a population (including the prison) must go *up* at some point. But that can't be right at the extreme, either- if you kill off all criminals, then there would be a reduction in crime as well.
So what is clearly going on is noise in the middle.

This can't be literally true -- increasing the sentence for jaywalking to 50 years in prison won't reduce the incidence of jaywalking?

Highly unbelievable. Do they offer a hypothesis of how crime rates don’t go down when the criminals are in jail?

The answer is that "it's complicated". Incarcerating someone has a myriad of effects: it removes a known criminal from the streets for a period of time, it diminishes already tenuous paternal bonds, it normalizes incarceration as part of the "life script" for related populations, it increases the incarcerated's exposure to gangs and "hardened criminals", it familiarizes the incarcerated with the prison system, it reduces supply of criminal labor and may increase demand by the illegal economy, it provides a cautionary tale to others, it provides law enforcement with additional points of leverage ...

We don't even know which direction each of these effects run. In a highly noisy dataset that is drawn for the idiosyncracies of elections I highly doubt we can do anything other than pitch a just so story. Maybe the dominant factor is the children of inmates and the timing for when the inmates are incarcerated. Maybe the dominant factor is the ability of the DA to offer more credible threats in the future. I am just not seeing things here that are worth actually basing policy changes on.

I don’t think it’s that complicated. Lots of factors, but one factor almost certainly dominates. Quarantined criminals can’t commit crimes. (The kind of people who commit violent crimes are crap fathers anyway, so you want weak paternal bonds with them.)

While I applaud the move to legalize things that shouldn’t be illegal (drugs and prostitution and whatnot) I’m very skeptical of what seems to be a recent trend among libertarian to champion softer sentences for real crimes. If you’ve ever held a loaded gun to a person’s head in order to take their stuff you should be in a cage until you die.

A couple of questions:

1. If recidivism rates are that high, how does changing initial prison length effect crime rates? After all, if we are sure that criminals will reoffend an initial sentence of say 5 years instead of 7 merely means that the offense occurs two years earlier. With the way sentences increase with repeat offenses, very few prisoners will commit fewer offenses over their criminal careers.

2. How do we account for crimes committed against the quarantined? After all assault is legally a crime inside of a prison, as are rape, theft, and all manner of other things. I am willing to buy that it is less terrible to have these crimes relegated to those who have committed such offenses ... I just don't see how the aggregate number of crimes drop just because we place all the criminals on an island.

3. Let's grant that ex-cons are terrible fathers, okay, who replaces them? Teenagers are going to seek role models and paternal influences. The most likely replacement for an ex-con is going to be a gang member. Me, I would prefer that the "parenting" is done by a late 20s/30s washed up has-been from crime world rather than a statistically much more violent late teenager without as strong of natural connection to the child. Ex-con parents, in my experience, as less bad than their typical replacements.

4. Why are drugs less terrible than holding a gun to another's head? Alcohol and marijuana, for instance, result in massive numbers of innocent people dying due to intoxicated driving. Likewise, a huge number of criminals who "hold loaded guns to heads" do so because they are using psychoactive compounds that we know lower inhibitions against things like assault and murder. At a population level both armed robbery and drug use have a statistical chance of leading to dead innocents. Right now drugs (including alcohol) kill far more innocent people than armed robbers.

Re; drugs, murder is worse than suicide, in the case of alcohol we rightly sentence the dangerous drivers and the hit and run perps not generally people using alcohol.

No we don't.

Per survey data drunk drivers get arrested after 80 episodes of drunk driving. Seventeen million Americans admit to driving drunk annually, yet only around 1-2 million are arrested annually. Of those who actually have their license revoked for DWI/DUI/etc., half will continue to drive without it. And on it goes.

If you kill someone with a gun accidentally (e.g. mistake the other guy for a deer), in many states we give you a harsher sentence than if you kill them with a car while drunk.

We barely make a dent in the criminal alcohol use population, we give them far weaker sentences when we do arrest them, and they are not locked up near as often nor near as long. Straight up alcohol is far more likely to be contributory to your death (as a non-drinker) than armed robbery is. Yet we punish the latter more reliably and more harshly.

If we treated other crimes with similar risks of innocent death as we treat drunk driving, we would empty the prisons.

1. That's why a person who points a gun at a someone's head to take their purse shouldn't get 7 years instead of 5. He should get 25.

2. We should do our best to prevent those crimes in prison. On balance, though, I'd rather they happen to criminals than non-criminals.

3. I don't think urban teens are likely to choose their washed-up, has-been fathers as role models. However, if we immediately put those violent late teenagers in prison for long periods of time, they're not there to be bad role models. Releasing violent criminals into a community is bad for that community.

4. I'd favor equally harsh sentences for drivers who kill people while intoxicated.

1. Ahh, but that is not what this study examined. Marginal changes in prison sentences may well be dominated by second order effects while gross changes in sentence length would be quite different.

2. Fair enough

3. Experience with them as patients leads me to disagree, kids will choose their dads even when he robs them. Incarcerating violent late teenagers in prison requires meeting burdens of proof that are difficult to meet. When the victims are unwilling to press charges or cooperate with the police (which is very likely if they are also gang members) you have to wait until the violent teen does something dumb enough to get incarcerated.

4. But that is not the question. DWI is doing something illegal with high odds of killing an innocent person. Armed robbery is doing something illegal with high odds of killing an innocent person. Vehicular homocide is not the question, DWI (DUI) is. If we are throwing the book at people who infringe upon the rights of others, by far the most generally harmful category of crime (incidence x likelihood of death) is DWI.

Do the data show or suggest whether prospective criminals elect to commit property crimes and/or violent crimes in counties with Democratic prosecutors to avoid committing crimes in counties with Republican prosecutors?

If prospective criminals are not alert to the prosecutorial zeal native to one county or another, why not? Don't prospective criminals read the criminology journals and the sociological surveys?

By the by: how many state and/or Federal prisons' libraries subscribe to and carry academic journals devoted to criminology and the sociology of crime?


Crime is so hyper local that putting a police car on a block with robberies decreases robberies overall instead of shifting over two blocks.

Incarceration is a great example of how most politicians are incapable of marginal thinking. Incarceration was responsible for a significant part of the huge reduction in crime in the 90s but then politicians kept pushing more and more incarceration even after evidence showed that the marginal impact of more incarceration was now very limited. Today, our incarceration rate is almost certainly higher than what would pass a marginal cost/marginal benefit analysis.

This is a silly study. What would the mechanism be? Are people under the impression that criminals are following the elections of local prosecutors? And then, having observed that a republican had been elected, shying away from committing crimes until a democrat retakes the office?

Stiffer sentences obviously reduce criminality. We can debate how much and where the break points are, but this study isn't a good look at that.

What this study shows is that potential criminals aren't aware of changes in expected sentencing that are driven by the political wins and losses of local prosecutors. That's not a surprise.

Are people unaware that "tough on drunk driving" sentencing initiatives have had a major effect?

I don't like this study, because it is upsetting!

I find it credible that sentencing has little effect on crime rates. I doubt that most burglars, car thieves, and muggers are capable of rational risk assessment; crime actually does not pay. Secondly, I doubt that the risk of getting caught is high enough to make even prudent street crime perps think twice.

It doesn’t “pay” in terms of income, let alone a secure steady income, but it “pays” in other areas? If it means involvement in a gang and the attendant lifestyle, it has provided status (of a sort), camaraderie, connections, protection and probably some job security and income. Don’t forget excitement, if only of a non cerebral kind, and belonging.

We're tiptoeing around the classic Fox Butterfield Moment -- wondering why the prisons are so full when crime is down

Tyler making sure that he has continued employment at Bloomberg and continued invites to posh parties.

Who cares if drivel like this gets cited in policy? At least the hors d’oeuvres will be delightful at the parties!

Given the bright flashing factual error in the quoted excerpt (prosecutors don't sentence defendants), I don't have a lot of confidence in this paper.

Basically, if they are trying to explore the effect of sentencing, they should correlate to legislatures and judges. If they are trying to explore prosecutorial behavior, they should look at charging, not sentencing. And if they don't understand who does what in the criminal justice system, they shouldn't write about it.

Do a couple of variables tell the whole story? Fake news?

So unless the recidivism rate for these criminals is zero then while locked up they aren’t committing crimes.

Does he control for Democratic areas simply don’t charge crimes? In California we made anything under $950 a misdemeanor and LE are reluctant to even attempt to catch these guys....massive amounts of car break in’s in SF. More shoplifting and more bike stealing

But if they’re not reported as crimes....

And on another note, there are no prosecutions for election fraud in blue cities!

If longer sentences reduced crime, you'd have only used one period in that paragraph.

David Roodman came to the same conclusion: "I estimate, that at typical policy margins in the United States today, decarceration has zero net impact on crime outside of prison. That estimate is uncertain, but at least as much evidence suggests that decarceration reduces crime as increases it."

Comments for this post are closed