3 ms·

Theory and data. Both are hard, especially in the cases I listed. With a good theory of (e.g.) platforms, you could likely identify a few different fundamenta

Theory and data. Both are hard, especially in the cases I listed. With a good theory of (e.g.) platforms, you could likely identify a few different fundamental forces which may push in different directions. Hypothetical example: force X would benefit the platform at the expense of sellers, force Y would benefit the sellers at the expense of the platform. A priori it may not be possible in theory to put a magnitude on those effects: there may be nothing that would say "we can show that X > Y in every case." This is where you need data. (This part is especially hard for platform markets!) You would need data on the platform side AND on the consumer side AND on the supplier side. You write down a model of (generally) consumer demand, supplier response and platform behavior (at least - probably taking a somewhat simplified approach to at least one of those sets of agents). With your estimates, you would try to answer the "Is X > Y?" question posed above. You can also simulate the impacts of counterfactual policies, among other things.

lisper 4y ago
Thanks!
nybble41 4y ago
> This is where you need data. More specifically, this is where you need data which does not and cannot exist. Specifically, an objective scale which would allow you to compare the cost or benefit of policy to the platform and the sellers. There is no such objective scale, because value is subjective and not comparable between different economic actors, or even for the same actor at different times. Any model which assigns numeric "estimates" for these costs and benefits so they can be compared is substituting the authors' own subjective valuations in place of those of those who will be affected by the policy.
- huitzitziltzin 4y ago
  > More specifically, this is where you need data which does not and cannot exist. This is wrong. All important work in my field in the past 25 years has been empirical. > Any model which assigns numeric "estimates" for these costs and benefits so they can be compared is substituting the authors' own subjective valuations in place of those of those who will be affected by the policy. Very, very wrong. You’ll never get away with picking your own parameter values in modern empirical IO. If you don’t know the field that’s fine. I probably don’t know your field either. But don’t make these overconfident and totally wrong claims about what’s done in a field you don’t know anything about.

lisper 4y ago

Thanks!

nybble41 4y ago

> This is where you need data. More specifically, this is where you need data which does not and cannot exist. Specifically, an objective scale which would allow you to compare the cost or benefit of policy to the platform and the sellers. There is no such objective scale, because value is subjective and not comparable between different economic actors, or even for the same actor at different times. Any model which assigns numeric "estimates" for these costs and benefits so they can be compared is substituting the authors' own subjective valuations in place of those of those who will be affected by the policy.

huitzitziltzin 4y ago
> More specifically, this is where you need data which does not and cannot exist. This is wrong. All important work in my field in the past 25 years has been empirical. > Any model which assigns numeric "estimates" for these costs and benefits so they can be compared is substituting the authors' own subjective valuations in place of those of those who will be affected by the policy. Very, very wrong. You’ll never get away with picking your own parameter values in modern empirical IO. If you don’t know the field that’s fine. I probably don’t know your field either. But don’t make these overconfident and totally wrong claims about what’s done in a field you don’t know anything about.