Read on Twitter
Let me jump in this swashbuckling fight between @matthewstoller & @conlon_chris
A THREAD for academic IO (= Industrial Organization) economists interested in competition policy
1. *Trends* & *aggregates* are important. They are much used in other sectors of the economy
1/ https://twitter.com/matthewstoller/status/1337489895414239232
A THREAD for academic IO (= Industrial Organization) economists interested in competition policy
1. *Trends* & *aggregates* are important. They are much used in other sectors of the economy
1/ https://twitter.com/matthewstoller/status/1337489895414239232
2. An array of indicators *is* useful for policy makers that need to take decisions in *finite time*
3. IO has often refused (with the exception of John Sutton and few others; disclosure: I was John’s student) for the past 30+ years to say there’s anything useful in those
2/
3. IO has often refused (with the exception of John Sutton and few others; disclosure: I was John’s student) for the past 30+ years to say there’s anything useful in those
2/
4. IO is not very tolerant to other fields. And it has become self-referential (look at our declining citations, please)
5. We look down at macro/finance/etc. requiring on impossible standards of identification to believe *any* empirical result they may come up with
3/
5. We look down at macro/finance/etc. requiring on impossible standards of identification to believe *any* empirical result they may come up with
3/
6. True, De Loecker @jan_eeckhout et al rely on strong assumptions (but have advantages too). But hold on, also BLP does. And BLP's underlying assumptions are just as unreasonable (though different) than those used by the macro etc literature
4/
4/
7. Specifically, why should a static model of horizontal product differentiation, in narrowly defined markets, under a specific assumption about error terms, be better than other models? There is so much that is just not incorporated in BLP
5/
5/
8. Back to concentration/margins. There is *no way* one can produce meaningful trends which span decades many countries & the whole economy using BLP. Acknowledge that
9. At best we can do a few industries in the US & compare two points. Interesting, but what do we learn?
6/
9. At best we can do a few industries in the US & compare two points. Interesting, but what do we learn?
6/
10. Ask ourselves: has BLP been useful for *policy*? Has it been used in policy? Yes or no?
11. My answer is no, since we have insisted on a single set of tools (demand estimation) to address a broad & important policy question. And it’s not much used either: think about it
7/
11. My answer is no, since we have insisted on a single set of tools (demand estimation) to address a broad & important policy question. And it’s not much used either: think about it
7/
12. Personal view, informed by a bit of experience in policy: this attitude de facto means we are never moving the bar & it’s exactly what corporations with economic and political power need to preserve the status quo
13. Way forward. Different approaches are complementary
8/
13. Way forward. Different approaches are complementary
8/
15. IO guys please do play to dig into mechanisms (which is important!), but also please welcome other contributions
16. Use brainpower to "tool up" the methods such that cross-sectoral analysis is reliable as possible (instead of looking only at what we can *measure*)
FIN/CIAO
16. Use brainpower to "tool up" the methods such that cross-sectoral analysis is reliable as possible (instead of looking only at what we can *measure*)
FIN/CIAO
