A fair amount of my research has been on the economics of vertical contracting and integration, and I sometimes get asked to participate in forums, discussions, amicus briefs, etc. on these issues. I'll be on a panel on vertical mergers at the CRESSE conference in Greece later this month and will also give an advanced short course directed toward practitioners. This post offers a brief summary of a few of the points I will emphasize. (This a long blog post, sorry. Future posts will be shorter :).
1. Recent antitrust complaints, analysis, and court decisions involving vertical mergers in both the U.S. and Europe are a train wreck (my view) for the economics profession, the antitrust community, and consumers. The reasoning in half a dozen recent vertical merger cases is incoherent and inconsistent with economics. This is the "Ugly" in the blog heading.
The biggest cause in my view is the failure of decision-makers to recognize (perhaps due to the failure of economists to adequately explain) the following facts (pardon the jargon in what follows):
(i) linear pricing in a vertical setting is the "same" problem as Cournot’s complements problem, but with sequential rather than simultaneous timing (see, e.g., O'Brien (2008), footnote 19), and
(ii) the main important difference between Cournot’s complements problem and competition between oligopoly producers of substitute products is the sign of the cross partials of demand with respect to price (see, e.g., O'Brien (2018), FTC Hearings on Vertical Mergers, 40:9-17).
Why does this matter? Because the sign of that cross partial is what determines whether the direct effect of the combination is upward or downward pressure on price. In short, the elimination of double marginalization ("EDM") and Cournot complements effects in vertical/complements mergers arise from the same general forces as upward pricing pressure in horizontal mergers, but the effects go in opposite directions.
That is an economic theorist’s way of putting it--my apologies--but the point is so fundamental and has such clear and strong policy implications that it is worth making it in a way that drives home the microfoundations that underlie it. In horizontal settings, a price increase on one product causes positive diversion to the other. Because the value of diverted sales is internalized after the merger but not before, the merger puts upward pressure on price. In complements settings, whether vertical or conglomerate, a price increase on one product causes negative diversion to the other, so the merger puts downward pressure on price, other factors equal. This leads to strong policy implications. Specifically, under linear pricing, mergers that combine substitutes sold by firms with market power create a presumption of harm from the internalization of horizontal pricing externalities, which puts upward pressure on price. However, mergers that combine complements do not create such a presumption because the direct effect of such mergers (setting aside foreclosure incentives for now) is just the opposite--downward pressure on price. This is true whether pricing is sequential, as in a vertical setting, or simultaneous as in Cournot’s complements problem. (There are quantitative but not qualitative differences between these cases.) It follows that it would be dumb to have the same presumption for vertical mergers as the presumption for horizontal mergers, as U.S. agencies seem to be suggesting. This would be inconsistent with longstanding economic theory that originated with Cournot (1838) and remains accepted to this day (e.g., Xavier Vives (1999), Oligopoly Pricing: Old Ideas and New Tools, MIT Press). It would also be inconsistent with the evidence on the effects of vertical integration (see citations below).
There is an investment analog to this point. Suppose that the investments of two firms in a complements setting are complementary, and that investments in a substitutes setting are business stealing. In this case, mergers between complements tend to expand investments, while mergers between substitutes tend to reduce them. Again, it would be dumb to suggest that complements mergers--vertical or conglomerate--should be analyzed under the same presumptions as horizontal mergers. (One can question whether there should be any presumption about investment effects in horizontal mergers. Economic theory and evidence indicates that there shouldn't be. For a brief discussion and references, see O'Brien (2022), Sections on Innovation and Quality Effects) Recent empirical research documents complementary investment benefits form vertical mergers in the smartphone industry. (Chenyu Yang (2020), "Vertical Structure and Innovation: A Study of the SoC and Smartphone Industries," Rand Journal of Economics.). I am not just spouting economics theory---there is empirical evidence.
Vertical mergers can also have important transaction cost benefits. See Dennis Carlton (2022), "Transaction Costs and Competition Policy," International Journal of Industrial Organization. The importance of transactions costs in economics is not as old as Cournot complements effects, but it goes all the back to Coase (1937)!
Of course, firms try to and sometimes do contract around EDM. But if we observe that marginal transfer prices exceed marginal cost, have firms successfully done so? No. Recent research rigorously documents the importance EDM effects from vertical integration in the cable industry. (Greg Crawford, Robin Lee, Michael Whinston, and Ali Yurukoglu (2018), "The Welfare Effects of Vertical Integration in Multichannel Television Markets," Econometrica).
Of course, there can be anticompetitive effects if an integrated firm can raise the costs of a competitor in the upstream or downstream market. There are many settings in which this can occur--this is well known (See Rey and Tirole (2007), "A Primer On Foreclosure," IO Handbook Vol 3.). With different effects pointing in different directions, net effects of a vertical merger depend which effect has the bigger impact on equilibrium effects. Sorting this out requires a real economic model. It cannot be done without a real economic model. But in any event, economic theory and empirical work do not support treating vertical mergers under the same presumptions as horizontal mergers.
2. Short cut techniques being used by agencies to analyze vertical mergers---vertical foreclosure arithmetic, vGUPPIs, bargaining models shoe-horned into explaining relative upstream and downstream margins widely asymmetric hurdle rates---do not provide good answers. I suppose this is "Bad" in the post heading. More on this in future papers/posts (but see O'Brien (2022), "Tethering Vertical Merger Analysis," Competition Policy Interational).
3. Somewhere between the "Bad" and the "Ugly" is the following. The Agencies, first in Europe with their Non-horizontal merger regulation, and now in the U.S. with the Vertical Merger Guidelines, treat EDM and Cournot complements benefits as tack-on "efficiencies" that are somehow different than effects from internalizing pricing externalities. "Efficiencies" are given short shrift in antitrust, as the claims are rarely believed. More importantly, treating EDM and Cournot complements effects differently than horizontal price effects is simply wrong as a matter of economics. Just as price effects from horizontal mergers arise from internalizing horizontal pricing externalities, EDM and Cournot complements effects arise from internalizing pricing externalities between complements. These effects come from the same fundamental forces: the internalization of pricing externalities though merger. Indeed, Tirole (1988) calls double marginalization "The Basic Vertical Externality" in an entire section in the Vertical Control chapter of his famous IO book. The general point is that horizontal and vertical/complements effects are "isomorphic" in the sense that they pop out of the same economic framework given different assumptions about the signs of diversion ratios (which are the same as the signs of the cross-partials of demand). (See FTC Hearings (2018) on Vertical Mergers, 142:1-17. I was the source of the question about the isomorphism, which the moderator passed on to the panel. The term isomorphic in this context is just a fancy word for "they're the same thing but with diversion ratios that have different signs." This point is incontrovertible as a matter of economics, but is obviously under-recognized and under-appreciated by decision-makers.)
With apologies to my friends in Europe, I would go as far to say that the bad-to-ugly treatment of vertical/comglomerate effects as tack-on efficiencies traces to the EC's decision to call EDM and Cournot complements effects "efficiencies" rather than price effects in their non-horizontal merger regulation. Unfortunately, the U.S. Vertical Merger Guidelines do the same thing. This has led decision-makers to put these effects in the same box as production cost savings in horizontal mergers, which are rarely considered to be important enough or credible enough to alter enforcement decisions. So now we're in a world where effects from the internalization of pricing externalities (Tirole's "basic vertical externality") through merger are shoved under the rug as tack-on efficiencies. Not good.
4. An argument often made, and one that is an important part of debates over the relevance of EDM, is that vertically-related firms can contract around EDM using nonlinear pricing. If so, there would be no EDM benefit from a vertical merger. I would also put this argument in the bad-to-ugly range. As a general conclusion, it's wrong: the ability to write nonlinear contracts does not automatically eliminate double marginalization. Several factors can prevent the use of nonlinear contracts from eliminating double marginalization:
Incomplete information. If the upstream firm doesn't know the downstream firm's costs, marginal transfer prices typically exceed marginal cost pre-merger. This is screening theory 101. The original cite probably goes to Mirlees (1971), but also Spence (1977), Mussa & Rosen (1978) in the quality context, Maskin & Riley (1984), etc.
Upstream moral hazard. If the upstream firm makes ongoing non-contractible investments, it will require compensation at the margin to be willing to do so, which means that marginal transfer prices will exceed marginal cost. The original rigorous cite is probably Holmstrom (1982)---his moral hazard in teams paper, although I'm sure this point was recognized earlier.
Risk aversion of downstream firms. If downstream firms are more risk averse than the upstream firm, marginal transfer prices are likely to exceed marginal cost to prevent downstream firms from bearing all the demand-side risk. The original rigorous cite is probably Holmstrom (1979), although this was no doubt a folk theorem earlier.
Oligopoly either upstream or downstream. This can generate incentives for an upstream firm to keep marginal transfer prices above marginal cost to soften competition. Even if nonlinear contracts are feasible, an upstream monopolist may want to distort marginal transfer prices to soften downstream competition. Cites include Mathewson & Winter (1984), many papers on vertical restraints, and for cases where there is oligopoly upstream, Ferstman & Judd (1987), Sklivas (1987), Bonanno & Vickers (1988). (Vertical theorists reading this will understand that softening competition in this way requires either that contracts are observable by all downstream firms, or if they are unobservable, that downstream firms do not have passive beliefs.)
4. So far this post has been about the ad and the ugly. What about the "Good?" Here is my take on that. It is good that vertical mergers have in recent years received more attention so that we can have this discussion and straighten some things out. I think having this discussion will cause that to happen, and that's good. It is well-known that vertical mergers can have anticompetitive effects in some circumstances. But let's do the right analysis to figure out when those circumstances occur.
Comments