3 ms·

> The most novel feature of my return to education calculations was that I tried to count everything that matters. I took the countless papers that start with

by cafxx 5y ago

> The most novel feature of my return to education calculations was that I tried to count everything that matters. I took the countless papers that start with the standard return estimates and tweak them with one novel complication. Then I merged all the tweaks that seemed convincing to me to get final policy-relevant numbers. (emphasis mine, highlighting in the author's own words enough wiggle room to potentially undermine for the purposes of general policy whatever number comes out of said estimates; if you pick and choose the conditions, you can get any arbitrary value as an estimate; I'm not arguing his estimate is not valid, it very likely is - but only under the author's choice of values)

parksy 5y ago
I had a similar reaction. I haven't heard of the author or their modelling prior, so haven't had much time to process the data, but I thought I'd bite, and took a look at some of the spreadsheets. The bit I was unclear about is where the predictive modelling takes place. I can see the raw numbers and how they've been composed across the main variables like education spending, education level, workforce participation and income, social benefit from workforce participation, and cost of crime. But I'm not seeing how the model itself is predictive. Where would I look to see the effect of dropping education spending by 50%. How does that impact crime rates and costs? What about the generational effects of deeply-connected criminal families and their children vs having more opportunities to participate and contribute to innovation etc. Does the model assume a linear relationship, i.e. n times the number of criminals = n times the crime rate = n times crime costs, or does it account for network effects such that n times the criminals have n^n crime rate, and crime costs are n^n - and is there a point where criminal justice processes break down and crime breaks out? Anyway save from writing a novel, it seems like the author did some homework to pull the various numbers together but I think some high level diagram of their model, inputs, calculations and outputs would be a better way to capture interest. Analysing something like this is a job, relying on the kindness of friends and internet strangers seems like a funny thing to get caught up on when the author themselves acknowledged the time and energy they had to spend pulling all of this together. I'm maybe going out on a limb but the lack of interest probably doesn't indicate that no one cares about the numbers, nor must it indicate people would rather use emotional intuition, just that there's probably better and more complete models out there used by social scientists and economists that follow tried and tested conventions that take many more factors into account, not my wheelhouse though, if no such standard models exist I'd be very surprised.

by cafxx 5y ago

parksy 5y ago

I had a similar reaction. I haven't heard of the author or their modelling prior, so haven't had much time to process the data, but I thought I'd bite, and took a look at some of the spreadsheets. The bit I was unclear about is where the predictive modelling takes place. I can see the raw numbers and how they've been composed across the main variables like education spending, education level, workforce participation and income, social benefit from workforce participation, and cost of crime. But I'm not seeing how the model itself is predictive. Where would I look to see the effect of dropping education spending by 50%. How does that impact crime rates and costs? What about the generational effects of deeply-connected criminal families and their children vs having more opportunities to participate and contribute to innovation etc. Does the model assume a linear relationship, i.e. n times the number of criminals = n times the crime rate = n times crime costs, or does it account for network effects such that n times the criminals have n^n crime rate, and crime costs are n^n - and is there a point where criminal justice processes break down and crime breaks out? Anyway save from writing a novel, it seems like the author did some homework to pull the various numbers together but I think some high level diagram of their model, inputs, calculations and outputs would be a better way to capture interest. Analysing something like this is a job, relying on the kindness of friends and internet strangers seems like a funny thing to get caught up on when the author themselves acknowledged the time and energy they had to spend pulling all of this together. I'm maybe going out on a limb but the lack of interest probably doesn't indicate that no one cares about the numbers, nor must it indicate people would rather use emotional intuition, just that there's probably better and more complete models out there used by social scientists and economists that follow tried and tested conventions that take many more factors into account, not my wheelhouse though, if no such standard models exist I'd be very surprised.