redding.dev
	mobility-sequence Mobility: Introduction This is the first in a sequence of posts on income mobility in the US. It tours various factors that people commonly believe impact a child's economic success in adulthood and checks the academic literature to see how strong these effects empirically are. However, before we get into that, I would be failing my duty as a writer, if I didn't note a couple things at the outset. Variance First, there is a crucial difference between "X doesn't explain much variance in income" and "X doesn't have any significant effect on income". For instance, suppose you were to learn that the college you went to only explained 5% of the variance in income. That would mean the difference between attending a 25th percentile college and a 75th percentile college would roughly a 25% increase in income. More generally, variance explained follows $r^2$ and effect sizes follow $r$, so that means a relatively small source of income inequality (0.05) can have a relatively large effect size (0.22) Conversely, maybe IQ correlates with income at r~0.4; that would mean IQ only explains about 16% of the variance in income. These interpretations are two sides of the same coin. One ($r$) answers the question "if I changed X by a bit, how much would Y change?". The other ($r^2$) answers the question "if I completely eliminated this source of inequality, how much would income variance be reduced by?". Outside The Box After this post, I will focusing almost exclusively on income mobility within the society of the US and other developed nations. There are enormous alternate aspects that I will be completely ignoring. Probably the most obvious two are: where and when you were born. But that sentence hides other determiners of income, where nearby Everett branches could have someone's outcomes ending quite differently. This could be due to individual-level luck or due to "structural factors". For instance, we happen to live in a society where tech companies use algorithm interviews to evaluate candidates. We used to live in a society where this was done using brain teasers, and, before that, we lived in a society where it was done with more conventional interviews. We could also imagine a future where most tech companies used other evaluation methods (IQ tests, fake projects, probationary periods, etc). It should be obvious that all of these evaluation methods correlate with job performance somewhat, but none correlate perfectly, and it is almost certainly the case that most people hired by (e.g.) top tech companies are greater outliers at algorithm-question-answering than software-engineering. I'm making absolutely no claims about who deserves what, but factors like how tech companies interview candidates is a kind of "luck" that isn't captured by statistical models and is often overlooked - at least by the lucky ones. Finally, income mobility is a relatively narrow (myopic) outcome metric. On a shallow level, it ignores other aspects of class such as educational attainment and occupational sorting. On a deeper level, it ignores the bulk of the rich experiences that make up human life. For the rest of this sequence of posts, I will be almost entirely ignoring all the above. Mea culpa. Satisfaction Not Guaranteed I am not an economist or sociologist. I have no advanced degree. Many of the questions I grapple with here are hard or impossible to answer. I do my best express confidence only when I feel it is warranted.

mobility-sequence

Mobility: Introduction

This is the first in a sequence of posts on income mobility in the US. It tours various factors that people commonly believe impact a child's economic success in adulthood and checks the academic literature to see how strong these effects empirically are.

However, before we get into that, I would be failing my duty as a writer, if I didn't note a couple things at the outset.

Variance

First, there is a crucial difference between "X doesn't explain much variance in income" and "X doesn't have any significant effect on income".

For instance, suppose you were to learn that the college you went to only explained 5% of the variance in income. That would mean the difference between attending a 25th percentile college and a 75th percentile college would roughly a 25% increase in income.

More generally, variance explained follows $r^2$ and effect sizes follow $r$, so that means a relatively small source of income inequality (0.05) can have a relatively large effect size (0.22)

Conversely, maybe IQ correlates with income at r~0.4; that would mean IQ only explains about 16% of the variance in income.

These interpretations are two sides of the same coin. One ($r$) answers the question "if I changed X by a bit, how much would Y change?". The other ($r^2$) answers the question "if I completely eliminated this source of inequality, how much would income variance be reduced by?".

Outside The Box

After this post, I will focusing almost exclusively on income mobility within the society of the US and other developed nations. There are enormous alternate aspects that I will be completely ignoring.

Probably the most obvious two are: where and when you were born.

But that sentence hides other determiners of income, where nearby Everett branches could have someone's outcomes ending quite differently. This could be due to individual-level luck or due to "structural factors".

For instance, we happen to live in a society where tech companies use algorithm interviews to evaluate candidates. We used to live in a society where this was done using brain teasers, and, before that, we lived in a society where it was done with more conventional interviews. We could also imagine a future where most tech companies used other evaluation methods (IQ tests, fake projects, probationary periods, etc).

It should be obvious that all of these evaluation methods correlate with job performance somewhat, but none correlate perfectly, and it is almost certainly the case that most people hired by (e.g.) top tech companies are greater outliers at algorithm-question-answering than software-engineering.

I'm making absolutely no claims about who deserves what, but factors like how tech companies interview candidates is a kind of "luck" that isn't captured by statistical models and is often overlooked - at least by the lucky ones.

Finally, income mobility is a relatively narrow (myopic) outcome metric. On a shallow level, it ignores other aspects of class such as educational attainment and occupational sorting. On a deeper level, it ignores the bulk of the rich experiences that make up human life.

For the rest of this sequence of posts, I will be almost entirely ignoring all the above.

Mea culpa.

Satisfaction Not Guaranteed

I am not an economist or sociologist. I have no advanced degree. Many of the questions I grapple with here are hard or impossible to answer. I do my best express confidence only when I feel it is warranted.