Kind of feels like Scott came to his conclusion first and then is now reading the data to support that. But the data doesn’t at all look convincing to me. Clearly there was a spike around May, but the data shows it was starting in the middle of May? Floyd died on May 25, and while protests began the next day in Minneapolis, they really didn’t pick up steam across the country until a few days later. But, for example, the NYC chart shows a clear escalation that starts at the beginning of May.
Like Scott says the Minneapolis Aggravated Assault chart shows it’s obvious but all I see is that before Floyd died, cases were rising dramatically, hit their peak shortly after protests started, slightly dropped soon after and stayed at an elevated pre-Floyd rate the rest of the year.
There is constant up and down in rates of various things. There was an small uptick in mid-May, probably caused by a combination of weather and normal randomness, but instead of being followed by a small downtick like every other small uptick, it went much higher and then plateaued long-term.
I think if the pandemic had started in mid-May, it would be fair to attribute to the pandemic, but the uptick clearly doesn't start in March and so you would need another theory to explain why this was happening in May.
But at least in several of the graphs you provide, the early-to-mid May increases weren’t small.
I would have also assumed the causal effects (especially the ones you suggest) would have been more gradual rather than immediate as the charts suggest, and that there would be a closer correlation between height of the crime wave and height of the protests.
If I were to provide a counter-factual, from what I remember, May is when people finally had enough of lockdowns (perhaps that also explains the size of the BLM protests, people used it to blow off steam), and we were certainly seeing lockdown protests by then as well. I’m personally of the belief that the rise in generally bad behavior over the past few years is attributable to some sort of psychological malaise caused by pandemic/lockdown fears leading to less institutional trust, and you see that starting in May as lockdowns ended. This could even explain some of the bad behavior occurring during the BLM protests.
I looked into this further, and several of these graphs are 7 to 14 day rolling averages, meaning that an increase on Day X will start showing up gradually on day X-7 or X-14.
A nice alternative data source that provides really good precision is CDC wonder provisional mortality data. Here, we can look nationally (smoothing over individual city randomness) at homicides by race by *week*. Since most murders are within race, this is a good proxy for murders by race of offender. The biggest spike in 2020 is for black homicides, and it turns out the huge spike for black murder victims is literally the week following Floyd.
To get a bit jargony, if you run a "structural break" test, it detects the week post Floyd as the most plausible week for a structural break for black people. (And in fact, if you run the same test for car crash deaths, it looks like the structural break happens almost immediately thereafter too. Strongly suggests its about depolicing and BLM).
I actually downloaded the data from CDC Wonder myself and did the analysis myself. Happy to email my code and data to you if there is a way to do that?
CDC Wonder allows you to save urls for specific searches, hopefully these still work:
Er, FWIW, that's usually *not* what people mean by N day rolling (or moving) averages. Usually the window is only backward looking not forward looking. So an increase on day X will get picked up by the 7dma between X and X+6. (Of course, it's possible to center the window elsewhere. No comment on what is being done in said graphs).
I also have no idea about the graph's data source, but you are correct. I'm not aware of anyone who does a rolling average of the week AHEAD of a date. Rarely you will see centered, but when that happens the centering should be explicitly mentioned in the methodology so as to avoid confusion.
You've got it backwards. You can check wikipedia or whatever source you like. Moving averages nearly always take the average of the previous N values, or more rarely, the average of N values in both directions. I've never heard of taking N values forward. So, the Minneapolis Figure 5 is even worse for your argument than you thought and suggests that the protests may actually have occurred at the peak of a crime wave which began to subside at their onset. That doesn't mean your overall hypothesis is wrong, or right, of course.
If you can, please add this to the article for the relevant graphs. This is definitely something worth pointing out right next to those graphs.
Another commenter pointed this out, and I think you did yourself in another recent post, and I think it would be something worth pointing out to 'raise awareness' of generally too.
Murders were declining pretty steadily through 2014, but then BLM emerged at Ferguson in August 2014. Murders started going up where BLM triumphed over local police: in St. Louis, then in Baltimore after Freddie Gray, then in Chicago after the release of the LaQuan MacDonald bad shooting videotape on 11/23/2015.
But then various over-enthusiastic BLM supporters assassinated cops in 2014-2016, helping Trump get elected.
Strikingly, traffic fatalities followed the same trajectory. But it all makes sense if you think of it as how scared are potential bad actors of being pulled over by the cops.
BLM sort of vanished after 2016 as intelligent people like George Soros switched to emphasizing more respectable methods of fighting the New Jim Crow like funding candidates in DA races.
Murders stopped going up in 2017 and dropped an encouraging amount in 2018. At some point in 2019, however, things changed for whatever reason and the post-Ferguson decline stopped and started going back up. This was before both George Floyd and before covid.
The increase before the George Floyd era wasn't enormous, so it could be just random noise. On the other hand, it could be that from mid-2019 the criminal element was once again feeling its oats as during the Ferguson era. Or that could just be hindsight is 20-20.
On the other hand, the black traffic fatality surge is hard to see coming until it suddenly happens after George Floyd's demise.
One possibility is that the kind of people likely to carry and shoot illegal hand guns are a much smaller and more hard core population than the kind likely to drive recklessly if the cops are cowed.
Perhaps the gun criminal element was already beginning to sense a change in the winds regarding enforcement of gun and warrant laws well before George Floyd, while the huge number of average people who'd enjoy driving fast if they didn't have to worry so much about getting pulled over by the cops didn't realize we were entering a new historic era until after the beginning of the celebrated "racial reckoning."
Unfortunately, Scott's tone and reponse to you confirms my feeling that he has no interest in being convinced otherwise. I noted in my comment that this article was a defense of his claim and not an exploratory post. I find it biased, erroneous, and lacking greater context. Sorry not sorry Scott.
There are a lot of comments any time Scott does a political post that sound exactly like something I would have written back in 2016, when I was a dyed-in-the-wool Bernie supporter and quick to gaslight analyses that did not support my left-leaning assumptions.
But when Scott persistently puts in the work and develops these longer posts explaining his reasoning, the burden of proof falls on his cynical detractors to prove him wrong. In this case, the statistics are clear enough (particularly internationally) that I am frankly disappointed to see this kind of obstinance. Scott's "tone" here (as in his post about parties moving away from the center) is based on the weight of the statistical evidence he has found. Your tone, by contrast, seems to demonstrate a lack of interest in accepting this evidence, in favor of nitpicking at the kind of technicalities that will always exist when statistical methods are not perfectly homogenous among jurisdictions. I'd invite you (and others with this view) to kindly consider the overall argument holistically from a neutral point of view, rather than blaming Scott for failing to shift your entrenched priors.
What a passive aggressive comment... and then some directly aggressive. Really. I could say more to the pieces of your comment that have objective merit, but the rest of this leaves me unwanting to carry on the conversation. Toodles
It's this kind of specious hypocritical moralizing that I really dislike and that inspired me to comment previously. If you have something to contribute to the discussion, if you identify something with "objective merit", you're free to say so. If you disagree with the tone or substance of something said, please expand on those specific details. If you don't have a rebuttal or if you were in the wrong, no one will judge you for saying so.
Discussion here should be rigorous and topical whilst allowing for changing minds. It's intellectually virtuous to be able to retreat from a previous position in the face of strong, well-founded pushback. Debate is not a war. Posturing in place of concrete engagement brings the level of discourse down to that of a thread on r/all.
I really don't love being aggressive in any way, shape, or form. I don't believe my previous comment to have been particularly aggressive. But it was fairly blunt. I'm sure we agree on plenty of issues, but on comment quality, I don't believe in compromising. The natural tendency of online discussion (in the absence of self-enforcement) trends towards entropic deterioration.
> If you have something to contribute to the discussion, if you identify something with "objective merit", you're free to say so.
- Agreed
> If you disagree with the tone or substance of something said, please expand on those specific details.
- I did my best to construct and express my thoughts thoroughly. I stand by what I said, especially since no one has answered the questions that I was genuinely asking. You are always welcome to engage with curiosity and learn more.
> If you don't have a rebuttal or if you were in the wrong, no one will judge you for saying so.
- Agreed
> Discussion here should be rigorous and topical whilst allowing for changing minds. It's intellectually virtuous to be able to retreat from a previous position in the face of strong, well-founded pushback. Debate is not a war. Posturing in place of concrete engagement brings the level of discourse down to that of a thread on r/all.
- Agreed 100%
>I really don't love being aggressive in any way, shape, or form. I don't believe my previous comment to have been particularly aggressive. But it was fairly blunt.
- Generally speaking words are a clumsy vehicle for the human experience, stepping on toes is easy to do without realizing it and I recognize it happens and try to call it out without holding it against people. I'm having issues with the site so I can't revisit your prior comment. The thing I always try to center on is the behavior or actions and not the person. e.g. saying "specious hypocritical moralizing" rather than calling me a "specious hypocritical moralizer". e.g. saying "that was a dumb comment" instead of "you are dumb".
> I'm sure we agree on plenty of issues, but on comment quality, I don't believe in compromising.
- I'm not sure what minimum bar you are expecting but as far as I know there are no requirements here, only guidelines on civility.
> The natural tendency of online discussion (in the absence of self-enforcement) trends towards entropic deterioration.
- 100% I have left many communities because of this. There is a lot about how this community operates that I appreciate, including the thoughtful and thoroughness of Scott and commenters. That being said, I don't believe in a minimum bar to entry... especially in a case where the other party was amassed a large amount of data.
In this case, I am accepting the data Scott presented and questioning the broader context in which it is presented, as well as what Scott actually accomplished here. I don't see a problem with that kind of meta commentary and I do find it constructive, at least to anyone interested in the scoping, definition and conceptualization of problems. In other words, Scott defined an initial scope and I questioned it. With more time and interest, I could amass more data and present a counterclaim to Scott but I am not trying to do that. I am just trying to ask a rigorous question about his methodology and purpose. Again, nobody seems to have answered these questions for me and that's fine. I remain curious and life moves on.
Kind of feels like Scott came to his conclusion first and then is now reading the data to support that. But the data doesn’t at all look convincing to me. Clearly there was a spike around May, but the data shows it was starting in the middle of May? Floyd died on May 25, and while protests began the next day in Minneapolis, they really didn’t pick up steam across the country until a few days later. But, for example, the NYC chart shows a clear escalation that starts at the beginning of May.
Like Scott says the Minneapolis Aggravated Assault chart shows it’s obvious but all I see is that before Floyd died, cases were rising dramatically, hit their peak shortly after protests started, slightly dropped soon after and stayed at an elevated pre-Floyd rate the rest of the year.
There is constant up and down in rates of various things. There was an small uptick in mid-May, probably caused by a combination of weather and normal randomness, but instead of being followed by a small downtick like every other small uptick, it went much higher and then plateaued long-term.
I think if the pandemic had started in mid-May, it would be fair to attribute to the pandemic, but the uptick clearly doesn't start in March and so you would need another theory to explain why this was happening in May.
But at least in several of the graphs you provide, the early-to-mid May increases weren’t small.
I would have also assumed the causal effects (especially the ones you suggest) would have been more gradual rather than immediate as the charts suggest, and that there would be a closer correlation between height of the crime wave and height of the protests.
If I were to provide a counter-factual, from what I remember, May is when people finally had enough of lockdowns (perhaps that also explains the size of the BLM protests, people used it to blow off steam), and we were certainly seeing lockdown protests by then as well. I’m personally of the belief that the rise in generally bad behavior over the past few years is attributable to some sort of psychological malaise caused by pandemic/lockdown fears leading to less institutional trust, and you see that starting in May as lockdowns ended. This could even explain some of the bad behavior occurring during the BLM protests.
I looked into this further, and several of these graphs are 7 to 14 day rolling averages, meaning that an increase on Day X will start showing up gradually on day X-7 or X-14.
A nice alternative data source that provides really good precision is CDC wonder provisional mortality data. Here, we can look nationally (smoothing over individual city randomness) at homicides by race by *week*. Since most murders are within race, this is a good proxy for murders by race of offender. The biggest spike in 2020 is for black homicides, and it turns out the huge spike for black murder victims is literally the week following Floyd.
To get a bit jargony, if you run a "structural break" test, it detects the week post Floyd as the most plausible week for a structural break for black people. (And in fact, if you run the same test for car crash deaths, it looks like the structural break happens almost immediately thereafter too. Strongly suggests its about depolicing and BLM).
Can you link the specific analysis you're using?
(I'm aware of Steve Sailer's version, but I'm specifically trying NOT to cite Steve Sailer here)
I actually downloaded the data from CDC Wonder myself and did the analysis myself. Happy to email my code and data to you if there is a way to do that?
CDC Wonder allows you to save urls for specific searches, hopefully these still work:
Black Homicides:
https://wonder.cdc.gov/controller/saved/D176/D293F081
Black Motor Vehicle Deaths:
https://wonder.cdc.gov/controller/saved/D176/D293F820
Why not cite Steve Sailer?
Why not? Fear of being canceled?
Er, FWIW, that's usually *not* what people mean by N day rolling (or moving) averages. Usually the window is only backward looking not forward looking. So an increase on day X will get picked up by the 7dma between X and X+6. (Of course, it's possible to center the window elsewhere. No comment on what is being done in said graphs).
I also have no idea about the graph's data source, but you are correct. I'm not aware of anyone who does a rolling average of the week AHEAD of a date. Rarely you will see centered, but when that happens the centering should be explicitly mentioned in the methodology so as to avoid confusion.
Centered is pretty common in historical data analysis precisely because it retains correct timing of events.
You've got it backwards. You can check wikipedia or whatever source you like. Moving averages nearly always take the average of the previous N values, or more rarely, the average of N values in both directions. I've never heard of taking N values forward. So, the Minneapolis Figure 5 is even worse for your argument than you thought and suggests that the protests may actually have occurred at the peak of a crime wave which began to subside at their onset. That doesn't mean your overall hypothesis is wrong, or right, of course.
If you can, please add this to the article for the relevant graphs. This is definitely something worth pointing out right next to those graphs.
Another commenter pointed this out, and I think you did yourself in another recent post, and I think it would be something worth pointing out to 'raise awareness' of generally too.
Murders were declining pretty steadily through 2014, but then BLM emerged at Ferguson in August 2014. Murders started going up where BLM triumphed over local police: in St. Louis, then in Baltimore after Freddie Gray, then in Chicago after the release of the LaQuan MacDonald bad shooting videotape on 11/23/2015.
But then various over-enthusiastic BLM supporters assassinated cops in 2014-2016, helping Trump get elected.
Strikingly, traffic fatalities followed the same trajectory. But it all makes sense if you think of it as how scared are potential bad actors of being pulled over by the cops.
BLM sort of vanished after 2016 as intelligent people like George Soros switched to emphasizing more respectable methods of fighting the New Jim Crow like funding candidates in DA races.
Murders stopped going up in 2017 and dropped an encouraging amount in 2018. At some point in 2019, however, things changed for whatever reason and the post-Ferguson decline stopped and started going back up. This was before both George Floyd and before covid.
The increase before the George Floyd era wasn't enormous, so it could be just random noise. On the other hand, it could be that from mid-2019 the criminal element was once again feeling its oats as during the Ferguson era. Or that could just be hindsight is 20-20.
On the other hand, the black traffic fatality surge is hard to see coming until it suddenly happens after George Floyd's demise.
One possibility is that the kind of people likely to carry and shoot illegal hand guns are a much smaller and more hard core population than the kind likely to drive recklessly if the cops are cowed.
Perhaps the gun criminal element was already beginning to sense a change in the winds regarding enforcement of gun and warrant laws well before George Floyd, while the huge number of average people who'd enjoy driving fast if they didn't have to worry so much about getting pulled over by the cops didn't realize we were entering a new historic era until after the beginning of the celebrated "racial reckoning."
Unfortunately, Scott's tone and reponse to you confirms my feeling that he has no interest in being convinced otherwise. I noted in my comment that this article was a defense of his claim and not an exploratory post. I find it biased, erroneous, and lacking greater context. Sorry not sorry Scott.
There are a lot of comments any time Scott does a political post that sound exactly like something I would have written back in 2016, when I was a dyed-in-the-wool Bernie supporter and quick to gaslight analyses that did not support my left-leaning assumptions.
But when Scott persistently puts in the work and develops these longer posts explaining his reasoning, the burden of proof falls on his cynical detractors to prove him wrong. In this case, the statistics are clear enough (particularly internationally) that I am frankly disappointed to see this kind of obstinance. Scott's "tone" here (as in his post about parties moving away from the center) is based on the weight of the statistical evidence he has found. Your tone, by contrast, seems to demonstrate a lack of interest in accepting this evidence, in favor of nitpicking at the kind of technicalities that will always exist when statistical methods are not perfectly homogenous among jurisdictions. I'd invite you (and others with this view) to kindly consider the overall argument holistically from a neutral point of view, rather than blaming Scott for failing to shift your entrenched priors.
What a passive aggressive comment... and then some directly aggressive. Really. I could say more to the pieces of your comment that have objective merit, but the rest of this leaves me unwanting to carry on the conversation. Toodles
It's this kind of specious hypocritical moralizing that I really dislike and that inspired me to comment previously. If you have something to contribute to the discussion, if you identify something with "objective merit", you're free to say so. If you disagree with the tone or substance of something said, please expand on those specific details. If you don't have a rebuttal or if you were in the wrong, no one will judge you for saying so.
Discussion here should be rigorous and topical whilst allowing for changing minds. It's intellectually virtuous to be able to retreat from a previous position in the face of strong, well-founded pushback. Debate is not a war. Posturing in place of concrete engagement brings the level of discourse down to that of a thread on r/all.
I really don't love being aggressive in any way, shape, or form. I don't believe my previous comment to have been particularly aggressive. But it was fairly blunt. I'm sure we agree on plenty of issues, but on comment quality, I don't believe in compromising. The natural tendency of online discussion (in the absence of self-enforcement) trends towards entropic deterioration.
> If you have something to contribute to the discussion, if you identify something with "objective merit", you're free to say so.
- Agreed
> If you disagree with the tone or substance of something said, please expand on those specific details.
- I did my best to construct and express my thoughts thoroughly. I stand by what I said, especially since no one has answered the questions that I was genuinely asking. You are always welcome to engage with curiosity and learn more.
> If you don't have a rebuttal or if you were in the wrong, no one will judge you for saying so.
- Agreed
> Discussion here should be rigorous and topical whilst allowing for changing minds. It's intellectually virtuous to be able to retreat from a previous position in the face of strong, well-founded pushback. Debate is not a war. Posturing in place of concrete engagement brings the level of discourse down to that of a thread on r/all.
- Agreed 100%
>I really don't love being aggressive in any way, shape, or form. I don't believe my previous comment to have been particularly aggressive. But it was fairly blunt.
- Generally speaking words are a clumsy vehicle for the human experience, stepping on toes is easy to do without realizing it and I recognize it happens and try to call it out without holding it against people. I'm having issues with the site so I can't revisit your prior comment. The thing I always try to center on is the behavior or actions and not the person. e.g. saying "specious hypocritical moralizing" rather than calling me a "specious hypocritical moralizer". e.g. saying "that was a dumb comment" instead of "you are dumb".
> I'm sure we agree on plenty of issues, but on comment quality, I don't believe in compromising.
- I'm not sure what minimum bar you are expecting but as far as I know there are no requirements here, only guidelines on civility.
> The natural tendency of online discussion (in the absence of self-enforcement) trends towards entropic deterioration.
- 100% I have left many communities because of this. There is a lot about how this community operates that I appreciate, including the thoughtful and thoroughness of Scott and commenters. That being said, I don't believe in a minimum bar to entry... especially in a case where the other party was amassed a large amount of data.
In this case, I am accepting the data Scott presented and questioning the broader context in which it is presented, as well as what Scott actually accomplished here. I don't see a problem with that kind of meta commentary and I do find it constructive, at least to anyone interested in the scoping, definition and conceptualization of problems. In other words, Scott defined an initial scope and I questioned it. With more time and interest, I could amass more data and present a counterclaim to Scott but I am not trying to do that. I am just trying to ask a rigorous question about his methodology and purpose. Again, nobody seems to have answered these questions for me and that's fine. I remain curious and life moves on.