Thread by @conlon_chris

So I was waiting for @nberpubs to tweet this out last week (which happened Friday pm). Anyway, here is my paper with @msinkinson and Matt Backus. This is the #commonwnership thread for RTE Cereal. 1/11
chrisconlon.github.io/site/bcs_cerea…

The reason we chose cereal (besides the obvious: that's what IO economists do) is that there is a lot of variation in the degree of common ownership across firms and over time. 2/11

For example Kellogg's is 20% owned by a family foundation and is basically indifferent to competitor's profits (weight < 0.2), and should be a strong competitor (turns out it is a high price, high margin firm). Common ownership can be highly asymmetric. (GIS weights KEL ~0.5) 3/

Quaker Oats (part of Pepsi) has lots of small investors and according to our other paper this means they should love competitor profits (turns out they have low prices and margins) 3/11

Post gets bought, sold, spun off, IPO'd a bunch of times and so it's common ownership varies over time. This provides a lot of variation (not necessarily exogenous) in profit weights that we can try to line up with costs/prices. 4/11

Most of lit regresses price on (ownership augmented) concentration indices called MHHI. We get a robust and significant 3% price reduction for each 1000 point MHHI increase, implying that common ownership REDUCES prices. Not time to call #FTC yet, but off to a strong start. 5/11

There are several reasons not to run these kinds of regressions. One is that we take all of that nice cross firm variation and flatten it into a single market-level measure. The other is that when products are differentiated you can get spurious results (in the Appdx) 6/11

When we run our semiparametric test which accounts for product differentiation, we can strongly reject Common Ownership at 95% confidence level (these are t-stats). See the details here: 7/11

https://twitter.com/conlon_chris/status/1354069359535284225?s=20

Finally, we stick an "internalization" parameter which nests own profit maximization and common ownership. (Personally I dislike this approach - but the audience always wants to see it). Anyway, we find approximately zero internalization of CO effects with a SE of around 13%. 8/

We can reject 20% internalization of common ownership at 90%. With very small amounts of internalization the implied markups become indistinguishable from one another. 9/11

We should point out that FAILING TO REJECT SMALL DEGREES OF INTERNALIZATION IS NOT THE SAME THING AS EVIDENCE IN FAVOR. In fact it is the opposite. In short things could not have gone worse for common ownership in our data 10/11

What does this mean for #commonownership? It means it isn't showing up in the prices of cereal. It doesn't mean CO is wrong or impossible. It does mean we need to examine more markets, and most importantly that we keep collecting necessary ownership data 11/11

The results regarding common ownership are here (separate thread)

https://twitter.com/conlon_chris/status/1354079020061757440?s=20