[Unsurprising 'News'] When Race Tips the Scales in Plea Bargaining

The fact that it's just left up to chance and that there's no way to verify one way or the other the quality of representation either group got basically destroys the results because of the uncontrolled variable. It's worse that it's a variable that we know has a large influence on outcomes......but that's completely dismissed by those who want to push a narrative.

 

If it's random assignment and the sample size is sufficiently large, the quality of representation is supposed to be the same. If they aren't as effective when representing black people, what is the explanation?

 

The random assignment just means that there's no bias that goes into what attorney is assigned to what offender, it doesn't suggest that the quality will be the same. Also, a state like Wisconsin, which is where the samples come from, has a larger than normal rural population and a much smaller than normal black population and both of those factors can play into quality of representation and overall outcomes. Nearly all of your sample when it comes to black offenders will come from a few densely populated areas (90% of the black population in Wisconsin live in just 6 counties Milwaukee, Dane, Racine, Kenosha, Rock, and Waukesha) where public defenders will have a higher case load than those operating in less populous areas which can have a negative affect on outcomes. The overall white sample will have these kinds of factors diluted by offenders from small towns, suburbs, and rural areas. Another factor is that in Wisconsin the black population is significantly younger than the overall population of the state (28 compared to 38), that's another factor that can skew results.

I mean, we can talk for days about just how many factors can affect outcomes, and that's why trying to isolate just one factor is so difficult and potentially impossible.

 

Ok, good points.

 

A PDF for the paper can be found here:

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3036726

From the paper:

"The analyses presented later in the article restrict the sample to cases filed and adjudicated in Dane County. Focusing on a single county is advantageous to the extent that it helps maintain a number of factors relating to law enforcement and the criminal justice system (such as district attorney office and judges) constant."

"As discussed earlier, there are several factors that affect judges’ sentencing decisions. For example, the type and severity of the crime, the number of concurrent convictions, and the criminal history of the defendant all play a role on sentencing determinations. Other demographic characteristics of defendants, such as age and gender, also appear to play a role in sentencing outcomes. One concern is that black defendants are more likely than white defendants to exhibit those characteristics that correlate with longer sentences.

To provide a more rigorous analysis of the role of race in sentencing decisions we can estimate the following specification:
...
The outcome of interest, Senti, is the length of the sentence associated with case i. Blacki is an indicator variable equal to 1 if the defendant’s race was coded as African-American, while Otheri is an indicator variable equal to 1 if the defendant was coded as a race other than African-American or Caucasian. The vector X contains an additional set of defendant controls, including (i) the age of the defendant, 123 (ii) the defendant’s gender, and (iii) the defendant’s prior criminal history.
...
The results for this baseline specification is presented in column (1) of Table 2. The coefficient on the indicator variable Black measures the difference in the length of sentences received by black and white defendants after controlling for the other variables included in the regression. Black defendants receive on average sentences that are 1.23 months longer than white defendants, a difference that is statistically significant and represents 32.6 percent of the average sentence in this sample of cases (3.77 months) "

So, the analysis did control for county and other factors like age, gender, etc.

 

Ah, I read that but I must have missed the part about it only being Dane county. That does eliminate some variables, and I acknowledge that they certainly tried to control for many other factors as well, I'm just skeptical about how certain factors can be adequately controlled for.

 

Back when I was a DA, we would usually make offers without knowing the race of the defendant, especially in misdemeanor cases. The race of the defendant is generally mentioned only in a caption in the front of the police report, and most people skip ahead to the narrative where the observed facts are laid out by the reporting officer. So, in formulating an offer one would have read that the suspect was observed (for example) taking three items from store shelves, hiding them in his waistband, and then exiting the store without paying. There could be a note that he was captured on the store's surveillance system and apprehended by the Loss Prevention Officer upon leaving the store. One would then write a summary like this in the front of the file on a case information sheet. Then one checks the rap sheet for criminal history and note that on the file if any exists. Based on the facts and criminal history, one notes an offer in the file and moves on to the next one in the stack. Most times, I wouldn't know what the defendant looks like until I saw him in court for the first time, and there would already be an offer noted in the file by that point.

About the only time the appearance of the defendant is important is when the case turns on the identity of the perpetrator. Then you look at tape/photos, etc. to see if the defendant is the same person as the suspect.

 

Interesting, why do you think blacks get worse deals then for similar circumstances as white men? But misdemeanor don't plea bargain do they?

 

Could be many reasons. Not all DAs operate the same way, so it is entirely possible that there are racist actors as well. There could also be variables that are not being accounted for. This sample could be unrepresentative of nationwide trends. Social science is very messy. It isn't like laboratory conditions. A controlled experiment would be to give the same DA two identical fact patterns and simply swap out the defendant and see if that changed the outcome. You can't do that in the real world, because there are no two identical fact patterns. You will see some charges with very standardized offers though. For example, a first time DUI in my county with a BAC of between .10 and .14 without significant poor driving conduct is going to be offered essentially the same deal regardless of the identity of the defendant (2 days jail, fines/fees of about $2,000, 3 month DUI class, 3 years probation, six month license suspension). Other charges have a wider variety of conduct involved, and thus a wider array of plea offers.

Absolutely they do. There isn't nearly enough time and resources to have a trial on every misdemeanor violation. The plea bargain can be as simple as the DA making an offer at arraignment to a defendant who doesn't even have a lawyer yet, or more involved negotiations with defense counsel. Sometimes a defendant will enter an open plea to the judge, in which case there is neither trial nor plea bargain, but that is less common (at least in my jurisdiction).

 

Someone doesn't understand random assignment.

Randomly assigning people to a treatment (or control) means you don't have to control for a wide variety of variables which could otherwise confound a study. The people who were randomly assigned an attorney have gone through that randomization process (as have the attorneys), which means that there isn't a systematic, uncontrolled variable affecting your data because you randomized specifically to wash out all of those things.

This is the very basis of controlled experiments, which are the gold standard, but is a VERY strong control in this study.

The variable is accounted for because its assignment is random.

I never claimed it was.

Yes, you can, because attorney quality was randomized.

They're called "facts," not "narrative."

Given that it appears that you don't understand the difference between a conspiracy theory and a quantitative academic study which was peer-reviewed and published in a respected journal, you'll have to expect that your inappropriate, frantic laughter just won't affect me in any meaningful way.

 

Said by someone who doesn't understand that random assignment is the single best possible way to control for the variable.

It is called "omitted variable bias," which is DRAMATICALLY worse than any possible bias that might result from random assignment.

The only one here who believes a "narrative" over facts is you.

"Narrative" - A story, wheither fictional or factual, which can be told independent of the facts behind it.

"Evidence" - Data or analysis which could be used to support or undermine a narrative.

So far, the OP brought up evidence and then you told a narrative. I mentioned the details of the evidence and you made up some **** and told another narrative. Then, we went back and forth, with me continuing to mention the evidence followed by you telling a narrative.

The only one here who has failed to provide any evidence at all while only telling a narrative is you.

 

That same attorney will be assigned (presumably) to many cases within that jurisdiction. That's part of what makes random assignment of the attorneys such a powerful control in this case. Unless you can present compelling evidence (not just your bullsh!t narrative) that some attorneys were assigned on a basis other than random, then the random assignment is as good as gold on that variable.

Controlled for when you look at the county specific analysis.

Also controlled for by the county level analysis.

County.

Level.

Analysis.

Explicitly controlled for in the study.

Only "impossible" if one ignores the facts in favor of the narrative they desperately want to believe instead.

 

The mechanism is certainly interesting here. Maybe in addition to any bias experienced (consciously or subconsciously) by the DA and/or judge, perhaps police officers tend to write more scathing reports when they apprehend blacks as opposed to whites. That study could be done by textual analysis, could be interesting.