A hypothetical RDOF do-over: Simulations of the auction results under different rules

I want to like the Rural Digital Opportunity Fund. $16 billion to bring broadband to currently unserved Census blocks had so much potential. But well-documented problems— the fairness of SpaceX’s low penalty weight, aggressive bidding by new entrants, and big claims by fixed wireless providers — have tarnished the outcome. One option, as I’m about to do, is play Whack-a-Mole by trying to write new rules to constrain the results, and hope for a better outcome next time. I worry the problem is more fundamental, with the auction itself. After all, the reverse auction is designed to drive down the government subsidy to provide new broadband. But it isn’t the subsidy money that’s in short supply — it’s the good broadband.

The FCC allocated $16 billion for Phase 1 of the RDOF. By the end, they committed $9.2 billion and covered 99% of the locations that were eligible. But the cost model that underlies that auction predicted that it would cost $26.5 billion to cover all the areas. With broadband money falling out of the sky, what did we get by “saving” $17 billion through this auction format? A whole bunch of headaches, and if you listen to the critics, unnecessary subsidies for satellite systems, and unproven vendors and technology. I’d personally trade the $17 billion for fiber laid to all of these homes, done by union workers employed by a local cooperative.

But lets get to the Whack-a-Mole. I wrote and open sourced code that allows me to simulate the auction, changing certain things about how the auction progressed. It isn’t perfect, but it is close. I’d like to thank Ziggy Rivkin-Fish, whose clear explanation of the auction dynamics for Benton was a huge head start in understanding the complicated rules.

A huge part of the problem is there was too much participation, especially with providers offering (or claiming) gigabit service. No provider could be assigned a block group until the price was sufficiently bid down that the whole auction “cleared” the FCC’s budget of $16 billion. Because of all the gigabit offerings and bidding in nearly every block group, the budget didn’t “clear” until the 13th round of the auction, when providers were bidding for a subsidy 60% of what the FCC’s model implied was needed. No provider received more than 70% of the FCC’s cost model as a subsidy. Most received much less. The auction went to the 19th and last round. The average subsidy was just 35% of the FCC’s cost model.

This is hugely problematic because the only way to correct it is to reduce competition in the auction at the gigabit tier, or reduce the areas that are served. Certainly some of the participants would have liked for LTD Broadband to have been excluded. LTD bid at the gigabit tier, giving themselves the option to provide FTTH or fixed wireless, though after an intensive public campaign they’ve since committed to almost entirely FTTH. As an experiment, I ran a simulation that excluded LTD Broadband entirely.

The impact of excluding LTD Broadband is the total auction commitment rises from $9.2 billion to $10.1 billion. 97% of locations are covered instead of 99%. Space providers increase their winnings to $1 billion. Areas covered by gigabit offerings fall to 81% from 84%.

Another popular target is SpaceX, and to a certain extent the other satellite providers. Beaming internet from space doesn’t have nearly the same marginal cost as running a wire in the ground, and the satellite providers were able to bid late into the auction for low subsidies. SpaceX wound up with a penalty weight of only 20, which gave it an additional advantage.

So the next simulation excludes all the satellite providers. In this scenario, the auction commits $9.15 billion, covering 94% of eligible locations instead of 99% in the real auction. The areas served by gigabit are essentially the same at 84%. The real winners would be providers that are neither gigabit nor space. These providers increase their share of locations from 2% in the real auction to 10% in the hypothetical auction.

There are a couple other auction rules that distort the results. One that Rivkin-Fish pointed out is a gateway for non-gigabit providers into the later rounds. If there is a tie between gigabit offerings in the clearing round (when the budget falls below $16 billion), instead of having just those gigabit bids proceed, all bidders are allowed to bid in the following round. At some point, given the low subsidy in the later rounds, fiber-based gigabit offerings have to drop out, while SpaceX can keep going through the 17th round. By my estimation, $20 million was awarded to non-gigabit offerings in later rounds after 2 or more gigabit providers bid in the clearing round. SpaceX took home 94% of that $20 million.

Another rule with unanticipated consequences is how “minimum scale percentages” drove down prices. The FCC allowed a bidder to say, “winning just one block group in an area doesn’t make sense for us. But if we win X%, then we’re in.” In some cases, the minimum scale percentage wasn’t hit in the 14th round, and as the subsidy level fell, it became even harder to reach the scale percentage. The effect was largely to drive down the price at which SpaceX won these areas, so it’s hard to get too upset about it.

Former Chair Ajit Pai was proud to claim that they structured the auction to be “technologically neutral”. By “technologically neutral” he means if you can claim and convince the FCC your technology can offer gigabit speeds, you can bid at the same tier as fiber-to-the-home. Unfortunately, they got this exactly backwards. Wireless technology should have never been weighted the same as fiber. SpaceX’s weight was too low given their low marginal cost of providing service. The entire weighting system is an exercise in preferencing certain technologies over others.

It’s easy to Monday morning quarterback these results, and I don’t want to diminish what must be tens of thousands of person-hours at the FCC that went into it, but a preferable approach would have been to only allow fiber to the home. Those rules would have served 86% of eligible locations instead of 99%, but the fiber that went into the ground would become valuable for wireless backhaul, satellite base stations, and new economic development. The total cost for the auction shrinks to only $8.8 billion, even though the average subsidy increases by 10% to $1,949 per location. These auction rules are actually a favor to the locations left out: with pending federal subsidies, it’s unclear if RDOF-covered blocks can get more support from coming broadband programs because of GOP fear of “over-building”. These auction rules would have freed them of that bond.

The true outcome of this auction is clear: the imperative to save taxpayer dollars worked against the better connectivity this auction was trying to provide. Sadly, in many areas, we’ll get neither well-spent taxpayer dollars nor good broadband.