Wednesday, June 20, 2018

Bringing a gun to a knife fight (spectrum edition)

As I pointed out in “Satellite spectrum efficiency” the satellite industry can’t win a spectrum auction fight with cellular since the cellular industry generates more $/Hz. This obviously generalizes to any number of industries competing in a license auction; the industry that generates the most $/Hz will always win. (The question of how overwhelming the win is, as a function of differences between industry $/Hz distributions, is left as an exercise.) So what?

One implication is that market mechanisms to divide up an allocation between industries, where all compete for the same licenses, will yield an unequal division of the spoils – and quite probably a monoculture, as the less financially efficient industry is squeezed out. It’s unlikely that this will be the social welfare optimum, since non-market values don’t necessarily correlate with market values.

(One might predict that satellite could disappear as an industry if it was required to compete on an equal basis with cellular for inputs like spectrum allocations. However, just national interest considerations will probably keep it alive, since a key client - the militiary - cares more about resilience, and thus supplier diversity, than cost. As Galen Pospisil pointed out to me recently, the U.S. DoD accounts for a large fraction of U.S. satellite business revenues. According to a 2017 study by consulting firm Research and Markets, the U.S. government "is the single largest purchaser of commercial satellite services in the world, and accounted for $1.34 billion in 2015." )

The problem is exacerbated by regulatory processes. Allocations and auctions require service rules, and these rules are promulgated by regulators. The richest industry (i.e. the one with the highest $/Hz) will be able to afford more lobbyists than anyone else, and the resulting service rules (even if ostensibly industry neutral) will favor the richest industry. This will make the licenses being auctioned more valuable to the richest industry and less valuable to others – making it even more likely that it will take most or all of the licenses in the auction.

Once one industry dominates an allocation, regulatory capture means that any subsequent changes to service rules will favor that industry’s business model even more. Since there is a great deal of inertia in companies’ loyalty to a particular business model, it is unlikely that a band will shift to another use, even if it make economic sense in the abstract. One will end up with a semi-permanent monoculture. For an example of the inertia in an allocation, one need only look at the broadcast TV allocations, and how long it has taken to repurpose the UHF bands to cellular use (and that only partially). It is striking that none of the broadcasters became cellular operators, but either sold out or stayed put.

Policy implications are in the eye of the beholder. I conclude that one can’t rely on the market to divide up frequencies between different industries, uses or applications (used interchangeably here), since I don't believe financial metrics correlate perfectly with the total social value of a particular industry, let alone the value of industry diversity. As Gabor Molnar pointed out when we discussed this, we’re back to beauty contests. There may be a way to combine public interest preference with market mechanisms, but that will necessitate a thumb on the scale.

(Gabor also observed that this takes us back to questions about facilities-based competition. Apparently there are results in the economics literature indicating that facilities-based competition increases social welfare. That leads to the question of how one quantifies diversity in commercial ecosystems. We agreed that monoculture is bad, but characterizing diversity requires defining markets, a notoriously contentious question.)

Another implication is to auction bands only when necessary, and not to put everything on the market at once. (The dominant industry of the moment will always argue strenuously that it is essential to put everything on the market NOW, and that not doing so will doom the nation to darkness.) This mitigates the problem of locking in all available bands to the currently dominant use. If licensees were dynamic operators that easily and frequently changed their business as technology evolved, this wouldn’t be needed, but they’re clearly not. While some of the inertia is due to the inflexibility of licenses granted to date, companies in general – not just those in the spectrum game – are usually wedded to their business-of-origin, and seldom change their spots. (I can only think of Nokia, the exception that proves the rule; you know, from paper pulp to rubber and cables to telecoms.)

No comments: