Hacker News Reader

Comment by dragonwriter | Hacker News Reader

> For example, I studied econometrics, and one regression that is often run is regressing CRIME (a dummy variable for whether someone has committed a crime) on a number of right hand side variables, include BLACK, a dummy variable for being Black. Now it is a well known fact that no matter how many right hand side variables you include, the coefficient on BLACK will always be significant and positive. That is, even after accounting for income, education, etc., Black people commit more crime.

You can't directly measure how much crime is committed by people of different races, so any purported measure of this uses some other proxy, most likely convictions of crime.

Of course, relative rates of conviction for crime may not reflect relative rates of commission of crime. Particularly, they may be distorted by biases in (among other things):

* Crime reporting,

* Prosecution,

* Courts and juries

View on HN

That is an unimportant nitpick, precisely the kind of objection that people are taught to raise in the face of "racist" facts.

I assure you that the academics involved, all certified non-racists (and mostly non-economists if you don't trust them), are aware of this and either consider the bias to be small (in comparison with the effect being measured) or have corrected for it.

No, the fact that the supposed fact cannot actually be supported except by assuming the non-demonstrable non-existence (or irrelevance) of several other kinds of bias is not an "unimportant nitpick", as it means that the "fact" is nothing of the sort, but a conclusion based in non-falsifiable conjecture.

(Note that I haven't said anything about whether or not its "racist" -- that's your deal -- I'm just saying that characterizing it as a fact is unjustifiable.)

No, the fact that the supposed fact cannot actually be supported except by assuming the non-demonstrable non-existence (or irrelevance) of several other kinds of bias...

Is this a parody of logic? Your point is not clear, regardless.

His point is crystal clear: he's saying you can't claim to rigorously study a relationship between race and crime by relying solely on conviction data, because racial prejudice and its legacy (for instance: what neighborhoods different races tend to live in, and whether those neighborhoods are enforcement priorities) has a profound impact on criminal conviction.

But I could also argue that not all crimes are caught in the data, because the law is written wrong. In fact, I could argue the law as it is written is thus criminal. After all, this is a democracy and the people are responsible for the laws. The people are incapable of drafting the laws in a non discriminatory manner (because racial prejudice and its legacy). Or, I could argue the exact opposite. Both are equaly weak/strong. But, this style of debate is incoherent. For it to be coherent, it would require special knowledge. And is otherwise a variant of https://en.wikipedia.org/wiki/Special_pleading

Any empirical work will involve questions of whether you are truly measuring the effect you want to measure.

However I am claiming, based on my knowledge of the field, that researchers have adequately dealt with this, and come to the conclusion that Black people really do commit more crime, after taking into account all observable economic and social effects.

Another replier has already mentioned that you can measure crimes that are less subject to bias in prosecution (like homicide.

If there was an easy way to explain the apparent excess of criminality in Black people by these means, I'm sure I would have read about it in the academic literature.

As all we have is conviction data, we can only see the crimes people care about seeing the criminals punished for.

There are plenty of crimes that go unpunished, and therefore groups can end up under reported, as the crimes they commit are for various reason ignored.

You can measure some crime without convictions. I.e. by ones being reported by crime's victims, or, in the case of homicide, by finding the victims murdered. Some crime like infamous "victimless" crimes can't be found this way, but my opinion is those probably shouldn't exist anyway as crimes.

His point is that being able to measure 'some' crime does not let you make statements about 'how much crime is committed by each race', just like being able to see the first row of trees in a forest does not let you make statements about how much of the forest is a particular kind of tree.

You can make statistical conclusions in some cases. I.e. if you take robberies, and you know that X% of robberies are commonly reported and you know that in Y% of robberies reported robbers were male, and you know that robberies committed by males and females reported with same frequency (note these are hypothetical assumptions, for the sake of example) - you can estimate how many robberies were committed by males, based on certain report statistics - not only on convictions. Of course, this would not be exact - statistics is not meant to be, the whole point is generalizing - but it would show the trends present in the data.

Right, but in this case, one of the points being questioned is whether we can assume that crimes committed by black people and non-black people are reported with the same frequency.

Why won't they be? Do you think somebody robbed by a white person would conceal this because they feel racial affinity to the robber? I have hard time imagining this. But of course real statistics would be helpful here.

They aren't even enforced consistently. Majority-white neighborhoods aren't policed the same way majority-black neighborhoods are.

Err, did you read the thread? We talked about reporting crimes, not policing neighborhoods.

I do not believe you could miss the connection.

This assumes that the reporting rate is the same for robberies committed by males and females. It is entirely possible that e.g. victims are less likely to report robberies commited by women. It's also possible that there is a bias in the reporting because victims tend to assume that robbers are male, even when the robber is masked and it's difficult to tell the gender of the robber.

Reply on news.ycombinator.com