Hacker News new | past | comments | ask | show | jobs | submit login
Why a small handful of counties generates the bulk of US death sentences (plos.org)
88 points by yummypaint 11 days ago | hide | past | favorite | 56 comments

From a section on theory of self-reinforcement

Once an office assembles a staff that has handled a capital trial, it draws upon this capacity to pursue the death penalty in subsequent cases, which further augments the office’s institutional capacity to pursue the death penalty. This self-reinforcing dynamic between capacity and caseload makes it more likely for offices that obtain death sentences to seek the death penalty going forward. Conversely, offices that cease to obtain death penalties (or never obtained death penalties in the first place) may be less likely to reverse course as institutional capacity for death penalty sentencing erodes (or is never developed). … This path dependency may reflect practices of prosecutors who make the charging decisions whether to seek the death penalty, but it may also capture defense lawyering, judges, jurors, and other features of a county that make it more likely to continue to death sentence over time

Wow, Kafka couldn't have said it better himself. Morbid bureaucratic mechanism of horror

Exceptionally well written. I clicked into the source but was disappointed the quotation didn't have a way to read more.

This kind of thing is why I've always argued against the death penalty.

There are people that deserve it but it's very hard to figure out who.

My country has life imprisonment. And we are not woke: you don't get out after 20 years. You get thrown in jail and you will rot in there while society forgets you.

The death penalty is completely unnecessary. Hell some of those in prison after 30 seasons with no future probably wish they were dead.

I argue against life imprisonment for the same reasons.

The Repeat Player vs. The One-Shot Player


(I don't know how good this source is.)

Ideally this dynamic would also mean that defense lawyers and norms would get better over time instead of atrophying.

What is so horrible or kafkaesque about that? Prosecuting death penalty cases is hard and politically contentious. It also not a necessity, as you can secure life imprisonment sentences with much less effort. It makes sense that only those places committed to the practice will continue it.

More people being put to death because the autonomous bureaucracy has gotten better at putting people to death is pretty kafkaesque. It reminds me of the short story with “the machine” that cuts progressively deeper over time.

The story that you're telling leaves out the number one reason why the death penalty is pursued more in some places than others: political will. Simply put, those places that have the resources and political desire to pursue the death penalty do, those that don't, don't. Pro death penalty places this year are more likely to have been pro death penalty places last year. That's all the study concludes.

Does political will really vary that much from one prosecutors office to the next? I would imagine it would have a more even gradient.

Brilliantly put. And moreover, regarding some comments nearby, there was an original commandant in the penal colony who instituted the penalty, and developed the machine. Maybe he knew why?

Over time, the machine has broken down, and all concerned have forgotten what the point of the machine exactly is, but the practice continues, the creaking machine grinding on.

Why isn't it Kafkaesque the other counties are ineffective at capital cases?

The only difference is who you believe morally right.

Sort of? Depends on who is the subject. 'Kafkaesque' doesn't just mean bureaucracy is bad but that it alienates the individual.

Most defendants would probably expediency more alienation if the likelihood of a death penalty depended on which county tried them. Actually, The Trial itself references circumstances very similar to this.

On the other hand, if there was some super depraved criminal that had a chance of walking free due to bureaucracy, then a individual on the prosecution side could be subject to disillusionment and alienation.

So sure either side can be Kafkaesque but since it's more a one vs many i.e the individual facing society, the defendant's situation is more likely to be the loosing end.

1) You make it seem like the county is arbitrary. But you are almost always tried in the place where the crime took place.

2) If you commit a crime in country A, you will be subject to different punishments than if you commit a crime in country B. This isn't really surprising. If you commit a crime in San Francisco, you probably won't be charged even if caught. If you commit a crime in a conservative city, you probably will. Again, not surprising and not absurd. These are different polities and they have different policies.

3) This has almost nothing to do with a self-propelling absurd bureaucracy and almost everything to do with people in some places voting for district attorneys/state officials who want to pursue the death penalty, as well as some DAs offices not wanting to spend the money and suffer the grief of prosecuting a death penalty case, which can require expensive lawyering for multiple decades in many cases.

This thread is so tiresome. People are salivating over this uninsightful study and making sophomoric faux-insightful comments like it's a college literary criticism class.

In almost the entire developed world, there is no death penalty anymore.

No matter which way you fall on the death penalty-morality equation, you have to not be happy that the likelihood that a particular person is executed depends more on a self-reinforcing bureaucracy than on the particulars of the crime itself.

No, it depends on the local politics of the death penalty. Some states have abolished it, some have not. District Attorneys often run on pro or anti death penalty platforms.

You are doing the opposite of anthropomorphizing. You are attributing to the faceless bureaucracy what is really just the result of different political preferences.

> The only difference is who you believe morally right.

This is also a theme frequently explored by Kafka. Especially in the machine short story I mentioned. People who believe they’re “right” in sentencing a person to death will come up with all sorts of ways to justify their amoral actions and Kafka created several characters who do.

The point is a random outcome in one case may perpetuate in a way that creates the commitment, and a random loss early on may prevent that commitment.

The Penal Colony. That story is etched into my brain. Ugh

>This model shows that the cumulative number of death sentences previously imposed in the same county is a strong predictor of the number imposed in a given year.

Sorry to be dismissive, but this is obvious. People who ate a lot of steak last year are likely to eat a lot of steak this year. People who did X a lot last year are more likely to do X a lot this year than people who did X less last year.

Some district attorneys' offices are more interested in pursuing the death penalty than others. How often to pursue the death penalty is a political question. Liberal district attorneys are much less interested in it than conservative district attorneys. And because the last 40 years of case law have made it very difficult to correctly prosecute a death penalty case without getting overturned for some reason on appeal, it is a significant financial investment for a prosecutor's office to pursue the death penalty. And the expense is not just the prosecution but the decades of appeals and other legal actions that will follow. So it makes sense that death penalty prosecutions are centered in places committed to pursuing the death penalty.

Obvious in hindsight?

Myself, if I'd been asked to guess the distribution of death penalties an hour ago before I'd read this article, would have cited all kinds of things as plausible factors and gone in completely the wrong direction.

We're a bunch of programmers, we like to talk here about our biased mental models of things. Give us data, let us fine tune our models and make them more realistic for those future discussions.

Real data, even when it confirms something 'obvious', is still good, right?

No, obvious in foresight too. If you asked me which places are likely to give the most death penalties this year, I'd just give you the list from last year.

I know that something being dressed up as SCIENCE makes us want to think it's profound. But most of the time, research findings are steaming piles of worthless crap that people pretend are insightful because their careers depend on it.

On the other hand, people who bought a house last year are far less likely than the general population to buy a house this year.

The non-obvious part is that the death penalty follows the first model and not the second. You could reason from first principles that the death penalty's intended effect is to prevent crime by executing a criminal, which is a long-term action like buying a house and not a short-term one like buying a steak. Or you could reason that its intended effect is to deter crime by presenting a deeply undesirable outcome for criminals. The surprise is that it's neither, and the counties so often have more people to execute.

You could also reason that killing people is good for getting certain kinds of votes, and generating lots of wasteful and stupid appeals work is good for outside counsel. (Aka donors)

> On the other hand, people who bought a house last year are far less likely than the general population to buy a house this year.

Is this even true? This is a non-obvious claim where the exact opposite phenomenon is allegedly true of other seemingly one-time purchases (e.g. cars, household appliances).

I sort of wonder if this is the data behind Amazon’s bizarre tendency to do things like advertise the same GPU to people who just bought one: I’ve always assumed that it’s just bad ML, but I’ve had this nagging thought that repeat purchases are somehow correlated (if only because seeing an ad for the thing you just bought might remind you that you need to purchase a related accessory in a way that seeing the accessory wouldn’t)

Yes this comes up a lot. If something like 2% of people ever buy a GPU, but 5% of GPU buyers will buy another one soon, then you spend a lot less overall and also get a better conversion rate advertising only to GPU buyers even if most of them think you're silly for showing them the ad.

It also preys on FOMO (show a better deal, they might cancel and re-order from you), brings you top of mind if they need a refund/replacement in short order, and does work for general brand awareness for next time, or conversations with a friend (since the ad knows you are now a GPU buying type of person). "Yeah I got XYZ GPU but I saw ABC GPU is cheaper and probably what you need." I probably missed a few things, but briefly it wouldn't be done if there wasn't a reason with the number of metrics they have these days.

I think it's even less likely to be true in geographical terms; counties where people bought more houses last year are probably more likely to be the places where people buy more houses this year. Even if the individuals who did buy houses are no longer in the market.

> You could reason from first principles that the death penalty's intended effect is to prevent crime by executing a criminal, which is a long-term action like buying a house and not a short-term one like buying a steak. Or you could reason that its intended effect is to deter crime by presenting a deeply undesirable outcome for criminals. The surprise is that it's neither

This doesn’t say anything about the intended effect.

It says a lot about the actual dynamics, but there is no necessary relationship between those two things.

It not only doesn't say anything about the intended effect, it can't as it is designed, and the intention is different across the people involved - juries, victims, judges, and voters.

I would go even farther: I am highly skeptical of the paper’s analysis. The analysis seems to assume that the rate of death penalties is entirely predicted by a handful of time-invariant parameters plus the prior death penalty rate. Under such a model, if the parameters chosen do not adequately predict the rate, and the rate is roughly constant over time, then of course the prior rate predicts the rate. This seems tautological. The fact that a model of this type does not support the author’s theory of self-reinforcement — the model is equally consistent with the death penalties being predicted by the second letter of the name of the county or the “racial threat” mod 0.1 or just about anything else.

(I have not rigorously verified my objection, but I’m moderately confident I’m right.)

You're right the authors have not proven causality (they did find correlation though). They don't claim that either, they say it's consistent. Causality is typically difficult to prove.

However, to verify your objection you need to first give the parameter that they have not used, and second come up with a good causality why your parameter explains the variation better (hint second letter of the county name is not it).

I beg to differ. The authors did not find a parameter other than death penalty rate that predicts death penalty rate. They claim that this implies that the high death penalty rates happen because the death penalty rate was high. This authors are _wrong_: this conclusion does not follow from their analysis; their analysis is unsound.

A demonstration of unsoundness does not require a proof that the conclusion is wrong; unsoundness means that the argument is wrong. The onus is in the authors to present a sound argument, not on the reader of the paper to demonstrate that the conclusion is incorrect.

(As a different example, Andrew Wiles’ first proof of Fermat’s Last Theorem was unsound. This didn’t mean that the theorem wasn’t true; it meant that Wiles failed to prove it. Similarly, in my opinion, the authors of this paper have not provided sound evidence in favor of their claim. This doesn’t mean I disbelieve their conclusion.)

”A death sentencing system based on racial dynamics and then amplified by self-referential dynamics is inconsistent with equal protection of the law, but this describes the United States system well.”

"A death sentencing system based on racial dynamics and then amplified by self-referential dynamics is inconsistent with equal protection of the law, but this describes the United States system well."

This reads strangely to me. It's like saying:

"Our research found that 2+2=4. In combination with racial dynamics, this is inconsistent with equal protection of the law."

I get that they are saying it amplifies, but they injected it in a weird way that makes their research sound more important to a broader conclusion than it is.

They could have left out the racial dynamics parts and the sentence would have been just as true. So not like your objection. A justice system where the chances of being given the death penalty is highly dependent on what county you are prosecuted in is inconsistent with equal protection of the law.

It's fine. You elect your local DA, mayor, and police chief. You also vote for your governor, state legislators, state judges (in most states), and state attorney general (in some). They get to make important decisions, like how strongly to enforce the law and what punishments to seek. Otherwise, it wouldn't be worth voting for them. In San Francisco, most crime is basically tolerated and punishment for most offenders is not sought. In more conservative parts of the country, that is not the case. If you don't want to be executed, don't commit murder in places where the death penalty is legal and exercised.

Local variance in how strictly the law is applied is expected. I don't really see that as violating equal protection.

There is no death penalty in most counties and in the ones that have it there are many counties with very small populations and a small handful with mega cities like Dallas Texas in them.

I assume you mean to imply that population alone explains the title of the post? My guess is the argument goes:

(1) Crime scales with population.

(2) Death sentences scale with crime.

(3) Therefore, death sentences should scale with population.

(4) The U.S.'s geographic distribution results in very dense clustering such that only a handful of counties have a lot of population.

(5) Therefore, we would expect to see only a handful of counties account for the majority of death sentences.

However, this line from the abstract seems to make (2) unlikely:

> The number of death sentences in a given county in a given year is better predicted by that county’s previous experience in imposing death than by the number of homicides.

Thus, their argument is that something much deeper is going on than just population-level trends.

I think the argument is more that there are only a handful of states that use the death penalty with any regularity, so most death sentences should be expected in the major population centers for those states.

The best predictor of death sentences is going to be whether or not you are in Texas.

I mean, probably, but the article is explicitly only looking at states with the death penalty:

> In the analyses below, we include only states allowing the death penalty in the year of analysis. Because we focus on the geographical variability in the use of the death penalty, we exclude the US military (which has sentenced 15 individuals to death since reestablishment in 1984, but carried out no executions) and the federal government (which has issued 79 death sentences since reinstatement in 1988, and carried out three executions).

And with this in mind, they still found that death sentences didn't correlate well with the homicide rate:

> The table also lists the rate of death sentences per 100 homicides and the rate of homicides per 100,000 population. If there was a direct link among these variables, we would expect some consistency here. But we see very little. In fact, the correlations are surprisingly low; in fact, the rate of death sentences per 100 homicides and the rate of homicides per 100,000 population correlate at -0.12. The counties with the highest raw numbers of death sentences listed in the table include not a single county that ranks in the top 100 with regards to death sentences per 100 homicides.

First, even the death penalty rate isn't approximately equal across the sample, so it's not just population. And second, they are saying that the death penalty rate doesn't even correlate well with the crime rate. And yet there are still a few outlier counties that have an abnormally high death penalty rate! That is what is unintuitive. I mean, sure, it makes some sense that some counties are more likely to issue the death penalty than others, even where it's legal, but it's a huge skew, and it's very different from saying "Texas has a lot of people."

I would probably qualify (3) with 'in counties that choose to implement it' which would make your model (which I agree with) fit the observations in the abstract.

I wouldn't think most counties decide to implement the death penalty based on number or types of crime (aside from 100 year outliers) as 'life in prison without possibility of parole' is also satisfactory in most cases. I would wager that the first execution in most counties occur after a specific election rather than a specific crime.

As they say, any map of social effects in the US degrades into the population density map of the US, as they say.

Normalizing by population density might help, but only in a limited way, because huge concentrations of people create different effects than pretty sparsely populated areas.

Sentences are imposed by judges, so if a judge in a given county has a hard-on for death sentences, you'll see more of them in that county.

Not just that but some judges are elected and research (I know, no cites) shows that elected judges get tougher in election years.

Now you're making me wonder if they controlled for elected judges.

The death penalty requires a jury to vote for it (usually unanimously) in most if not all US states.

All states nowadays, since Ring v. Arizona requires it.

It would be interesting to see a breakdown of whether male vs female judges have more or less death penalty sentences.

"The number of death sentences in a given county in a given year is better predicted by that county’s previous experience in imposing death than by the number of homicides."

That's not exactly a surprising result. A lot of people (probably including me) just won't use the death penalty unless there's a serial killer who videotaped his crimes and swore to break out of jail. The ones who will use the death penalty in more cases are not uniformly distributed.

Only ~21 states currently employ the death penalty. Within each state, a very small number of counties include high-crime areas.

The paper’s conclusions follow from this.

tl;dr "This model shows that the cumulative number of death sentences previously imposed in the same county is a strong predictor of the number imposed in a given year. Results raise troubling substantive implications: The number of death sentences in a given county in a given year is better predicted by that county’s previous experience in imposing death than by the number of homicides."

How many executions in 2020?


How many people murdered due to the Reign of Terror 2020?


Wow, your trendy pet activism of the bailed out rich killed more people than bin Laden. How about a study on that?

You also mention Jim Crow. May we look at the slavery you commit today, far worse than any other time in human history. Where do you think all your drugs, and sex workers, and cheap clothes and cheap electronics and child soldiers come from?

I think this study is orthogonal to whether or not the death penalty is a good thing.

There are a lot of things that are self-reinforcing. Take for examples, unions. Whether or not you think unions are a good thing or bad thing, likely you will be more successful unionizing a company that is located where there already are a lot of unions. For example, one of the big disadvantages of the Alabama Amazon union organizers was that they were in a part of the country where there was a strong anti-union sentiment in general.

To get to something more related to law enforcement, likely certain areas of the country will have more success in convicting police officers of murders.

For example, the Chauvin trial, instead of being held in Hennepin County, was held in a small rural county in norther Minnesota would most probably been harder to convince the jury to convict on all three counts and probably would have had a higher chance of a hung jury(which is why the defense asked to moved the venue).

Yes different parts of the country, and even the state will have likely different outcomes. This will be true for a lot of things. This observation is orthogonal to whether what you want is right or wrong.

After all, no one would argue that Chauvin’s conviction should be thrown out since a different jury in a different part of the country might have reached a different conclusion?

I fully expect the defense to argue on appeal that venue was improper. People have had criminal convictions overturned for improper venue before. Minneapolis suffered an extreme level of political turmoil due to the George Floyd death, and local public officials there spent the last year talking about basically nothing else.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact