I understand the point of this post and the original one. Further, I agree wholeheartedly that innumeracy drives bad policy (housing in SF anyone?) and mathematical models aren't used enough. In a nutshell, I would love for mathematically-able people, like Chris, to be in charge of policy everywhere. All that said, this whole exercise seemed a bit futile and kind of pointless.
Given enough parameters and assumptions about it, any model can be made to produce one or another result. (I know, I know I should shut up and just code to make that point, but bear with me) In this sort of policy debate, the harder part is actually finding the assumptions that are consistent with reality[1] and sorting out the moral and ethics questions behind whatever policy.
In this case, it is far less interesting to find that a hypothetical implementation of basic income or basic job would cost the government N trillion dollars per decade than it is to answer associated the moral question: Given that the GDP of the US is M trillion dollars, and that the Ginni coefficient is g, and that these candidate implementations would cost N trillion dollars, should the government redistribute wealth in some capacity as to guarantee a basic income? a basic job? Tinkering with the models doesn't really help there. All it does is change N. That helps when trying to answer the question, of course, but not too much.
[1]: Another example where this is the hardest step: It is easy to "explain" why there are so many more male STEM professors by noting that the distribution of <your favorite math standardized exam> has a std deviation slightly higher for males than females and assuming that a professor needs to be > 4std dev above the mean to succeed. The trickier part is measuring the distributions accurately and testing whether professors really need to be > 4std dev above the mean to succeed. fwiw, I believe that it is likely that the distributions are good enough and one does need to be that smart to be a prof, but testing those assumptions is the difficult and interesting part, not doing some numerical integrals of a gaussian.
1 comment
[ 2.7 ms ] story [ 9.2 ms ] threadGiven enough parameters and assumptions about it, any model can be made to produce one or another result. (I know, I know I should shut up and just code to make that point, but bear with me) In this sort of policy debate, the harder part is actually finding the assumptions that are consistent with reality[1] and sorting out the moral and ethics questions behind whatever policy.
In this case, it is far less interesting to find that a hypothetical implementation of basic income or basic job would cost the government N trillion dollars per decade than it is to answer associated the moral question: Given that the GDP of the US is M trillion dollars, and that the Ginni coefficient is g, and that these candidate implementations would cost N trillion dollars, should the government redistribute wealth in some capacity as to guarantee a basic income? a basic job? Tinkering with the models doesn't really help there. All it does is change N. That helps when trying to answer the question, of course, but not too much.
[1]: Another example where this is the hardest step: It is easy to "explain" why there are so many more male STEM professors by noting that the distribution of <your favorite math standardized exam> has a std deviation slightly higher for males than females and assuming that a professor needs to be > 4std dev above the mean to succeed. The trickier part is measuring the distributions accurately and testing whether professors really need to be > 4std dev above the mean to succeed. fwiw, I believe that it is likely that the distributions are good enough and one does need to be that smart to be a prof, but testing those assumptions is the difficult and interesting part, not doing some numerical integrals of a gaussian.