No, I'm not looking at "Floyd effect" vs lockdown, etc., and I haven't spent a lot of time on homicide vs other causes because it's smaller than ones like drug ODs.
Thanks! I have the data by state, so was planning to dig into that as well, but wanted to get this high level stuff off my laptop and into the interwebs 😉 I'm not trying to steal your material, I swear! 😂
Ah, that wasn't my point -- a lot of different people have been looking at this.
I wasn't trying to look at homicide specifically much because suicide & drug ODs (and motor vehicle accident deaths) are so much more significant... and ALSO took a huge jump during the pandemic.
We should chat separately, because I think I could give you pointers of what to look at... I've been watching this stuff even before the pandemic.
Interesting. But. One of the problems for us "Covid truth seekers" is that, whatever indicator you choose, there will always be counter-example countries (even if you restrict to "the West"). There has not been much of an increase in homicides in Germany, for example - but then, homicides are much rarer around here (the USA has ten times more homicides but only four times more people).
I have some sort of idea in my head that I need to figure out how to articulate better. Basically, all of these things (drug overdoses, homicides, other health issues) can be affected at the margins. Imagine some normal distribution showing people's willingness to commit homicide (from "No, never" on left to "I've killed someone" on the right). The US & Germany would be centered at different points (the US being further to right). Now something happens (say lockdowns), and people are the distributions shift right... this shift may push a lot of people in US into the "willing to murder" category, while in Germany the mass of the distribution is still to the left of "willing to murder" but nearer than before.
My day job involves credit portfolio models of CreditMetrics type. Default of an obligor is triggered if an abstract, normally distributed variable (usually called "asset value") exceeds a certain threshold. The position of the threshold corresponds to the obligor's credit rating (better rating <-> higher threshold). In order to model dependent default of different obligors, the individual normal distributions are connected by global variables (industry, or country, say). Dependent on a realisation of the global variables (a "scenario"), the normal distributions of the obligors' asset values are transformed, which makes default more or less likely.
Just think of individuals instead of obligors, committing murder instead of default, personal psychology instead of default threshold, and different global factors influencing individuals differently (a "USA factor", a "Germany factor", a "poverty factor", an "availability of weapons" factor, a "lockdown factor", ...).
If you can point me to any references on this type of modeling, I would appreciate it... just basically using my intuition to get my head around this conceptually, but would be happy to learn more formally.
There are several books on credit portfolio modelling (or credit risk modelling); usually quite mathematical in nature (e.g., Bluhm/Overbeck/Wagner). Here's a link to the credit risk chapters of my own book (Risk Model Validation, 3rd ed):
Not to show off, but simply because we wrote the book with a more general audience in mind.
When I give a talk about the material, I usually also apply the toy model and parameterization procedures to a very different problem (what is the probability about the next pope being from Italy?), just to underscore that one can quite easily enter absurd territory when applying models.
This old introductory paper by P. Schönbucher might also be helpful:
Thanks, will take a look. Although 99% of what I post here will involve simple statistics, linear regression, etc., I don't mind going into some deeper math.
I've done some analysis re: homicides here:
https://marypatcampbell.substack.com/p/the-geography-of-homicide-states
That's by state.
No, I'm not looking at "Floyd effect" vs lockdown, etc., and I haven't spent a lot of time on homicide vs other causes because it's smaller than ones like drug ODs.
more here:
https://marypatcampbell.substack.com/p/video-us-mortality-trends-2020-2022-9e9
Thanks! I have the data by state, so was planning to dig into that as well, but wanted to get this high level stuff off my laptop and into the interwebs 😉 I'm not trying to steal your material, I swear! 😂
Ah, that wasn't my point -- a lot of different people have been looking at this.
I wasn't trying to look at homicide specifically much because suicide & drug ODs (and motor vehicle accident deaths) are so much more significant... and ALSO took a huge jump during the pandemic.
We should chat separately, because I think I could give you pointers of what to look at... I've been watching this stuff even before the pandemic.
🙏 I was just joking .... as far as I'm concerned we are part of the same "Army of Davids" to use a Glenn Reynolds (Instapundit) term.
Would be happy to connect further as we clearly have similar interests here.
Have you seen the group I'm working with:
https://www.insurancecollaborationtosavelives.org/
that's more for the physiological causes of death, for which there can be less political/more medical interventions -- but gotta start somewhere.
I would like to find something re: the fentanyl/suicide stuff, if I could -- I do Movember each year, but that's at a remove.
Interesting. But. One of the problems for us "Covid truth seekers" is that, whatever indicator you choose, there will always be counter-example countries (even if you restrict to "the West"). There has not been much of an increase in homicides in Germany, for example - but then, homicides are much rarer around here (the USA has ten times more homicides but only four times more people).
Right.
I have some sort of idea in my head that I need to figure out how to articulate better. Basically, all of these things (drug overdoses, homicides, other health issues) can be affected at the margins. Imagine some normal distribution showing people's willingness to commit homicide (from "No, never" on left to "I've killed someone" on the right). The US & Germany would be centered at different points (the US being further to right). Now something happens (say lockdowns), and people are the distributions shift right... this shift may push a lot of people in US into the "willing to murder" category, while in Germany the mass of the distribution is still to the left of "willing to murder" but nearer than before.
Does this make any sense or am I babbling?
Oh yes, I can relate to that way of modelling...
My day job involves credit portfolio models of CreditMetrics type. Default of an obligor is triggered if an abstract, normally distributed variable (usually called "asset value") exceeds a certain threshold. The position of the threshold corresponds to the obligor's credit rating (better rating <-> higher threshold). In order to model dependent default of different obligors, the individual normal distributions are connected by global variables (industry, or country, say). Dependent on a realisation of the global variables (a "scenario"), the normal distributions of the obligors' asset values are transformed, which makes default more or less likely.
Just think of individuals instead of obligors, committing murder instead of default, personal psychology instead of default threshold, and different global factors influencing individuals differently (a "USA factor", a "Germany factor", a "poverty factor", an "availability of weapons" factor, a "lockdown factor", ...).
If you can point me to any references on this type of modeling, I would appreciate it... just basically using my intuition to get my head around this conceptually, but would be happy to learn more formally.
There are several books on credit portfolio modelling (or credit risk modelling); usually quite mathematical in nature (e.g., Bluhm/Overbeck/Wagner). Here's a link to the credit risk chapters of my own book (Risk Model Validation, 3rd ed):
https://www.file-upload.net/download-15182285/RMV3-CreditRisk.pdf.html
Not to show off, but simply because we wrote the book with a more general audience in mind.
When I give a talk about the material, I usually also apply the toy model and parameterization procedures to a very different problem (what is the probability about the next pope being from Italy?), just to underscore that one can quite easily enter absurd territory when applying models.
This old introductory paper by P. Schönbucher might also be helpful:
https://www.econstor.eu/bitstream/10419/78427/1/bgse16_2001.pdf
Thanks, will take a look. Although 99% of what I post here will involve simple statistics, linear regression, etc., I don't mind going into some deeper math.