> If someone got a job three months in, and becomes no longer eligible, then they will suddenly and discontinuously lose their benefits three months in, which is uncorrelated with changes in their underlying circumstances
Why is losing benefits due to getting a job considered "uncorrelated with changes in underlying circumstances" -- getting a job is not random (is it?), and it does change underlying circumstances.
You can compare their spending the month before they lost access to food stamps and the month after. Since they lose access 3 months after getting a job, you wouldn't expect them having gotten a job to cause a major difference in their spending to suddenly occur when they lose access.
Shouldn't economics, at least in a democracy, really be about the search for good policy, good policy defined as any policy that tends to increase the general welfare of this and future generations?
As for data analysis and the search for causal relationships, it would be a lot more scientific if error bars were included in the data being analyzed, given the well-known inaccuracy of many economic observations. Physicists include error bars in the measurements they report. Why don't economists?
Error bars, visualizing the range of their uncertainty. I think we find in economics, as in many other fields attempting to measure human behavior, what I call the fallacy of misplaced empiricism. This is the assumption that you can measure what in fact cannot be measured with the precision required to draw much of a conclusion. That's why econometrics failed. Am I wrong about this? Read Morgenstern's book on the uncertainty of economic observations.
High quality neighborhoods here appears to mean average family income, which means parents with higher than average intelligence and other heritable personality traits that contribute to success. Maybe army families who choose (or can afford) to live in these neighborhoods share those traits, in which case "neighborhood quality" is not what is driving the correlation. Parental quality is the confounding variable, which itself cannot be measured with much precision (average IQ or SAT scores alone theoretically excepted, except such scores are rarely publicly available).
Maybe I am missing something here? Tell me if I am, but make it simple.
PS I see that study you link tpis reporting correlations to two decimal places. Quote: "Assignment to a county with a 10 percentage point higher college graduate share raises college attendance by 0.09 percentage points per year of exposure during preschool ages, but by 0.19 percentage points per year during high school."
> If someone got a job three months in, and becomes no longer eligible, then they will suddenly and discontinuously lose their benefits three months in, which is uncorrelated with changes in their underlying circumstances
Why is losing benefits due to getting a job considered "uncorrelated with changes in underlying circumstances" -- getting a job is not random (is it?), and it does change underlying circumstances.
You can compare their spending the month before they lost access to food stamps and the month after. Since they lose access 3 months after getting a job, you wouldn't expect them having gotten a job to cause a major difference in their spending to suddenly occur when they lose access.
Nicely explained… thx
I’d say making conditional predictions
Shouldn't economics, at least in a democracy, really be about the search for good policy, good policy defined as any policy that tends to increase the general welfare of this and future generations?
As for data analysis and the search for causal relationships, it would be a lot more scientific if error bars were included in the data being analyzed, given the well-known inaccuracy of many economic observations. Physicists include error bars in the measurements they report. Why don't economists?
Is your contention that economists do not include standard errors?
Error bars, visualizing the range of their uncertainty. I think we find in economics, as in many other fields attempting to measure human behavior, what I call the fallacy of misplaced empiricism. This is the assumption that you can measure what in fact cannot be measured with the precision required to draw much of a conclusion. That's why econometrics failed. Am I wrong about this? Read Morgenstern's book on the uncertainty of economic observations.
"But that's all we have," is no defense.
There is also some interesting quasi-experimental evidence on the impact of neighborhood quality on children's outcomes:
https://www.nber.org/digest/202410/neighborhood-quality-and-childrens-outcomes-insights-military-families
High quality neighborhoods here appears to mean average family income, which means parents with higher than average intelligence and other heritable personality traits that contribute to success. Maybe army families who choose (or can afford) to live in these neighborhoods share those traits, in which case "neighborhood quality" is not what is driving the correlation. Parental quality is the confounding variable, which itself cannot be measured with much precision (average IQ or SAT scores alone theoretically excepted, except such scores are rarely publicly available).
Maybe I am missing something here? Tell me if I am, but make it simple.
PS I see that study you link tpis reporting correlations to two decimal places. Quote: "Assignment to a county with a 10 percentage point higher college graduate share raises college attendance by 0.09 percentage points per year of exposure during preschool ages, but by 0.19 percentage points per year during high school."