The future of auctions

[Music] Hello and welcome to this episode of the Policy Tracker Spectrum Podcast. My name is Camila Mena and I'm a journalist here at Policy Tracker. Today we're diving deep into spectrum options, joining us is Richard Marston, managing director at Nira Economic Consulting. Richard has worked on more than 100 options since 1999, advising buyers and sellers on format designs and bit strategy in more than 30 countries. Richard, it's great to have you. It's Camila Mena, I'm spread to be here. So you've been involved in a wide range of different options. What have been perhaps some of the most challenging ones you've worked on? Oh gosh, straight into it. Well I've been working on auctions for a long time now. I cut my teeth in the 3G auctions starting in late 1999. So I've now been through effectively three generations of auctions, 3G, 4G and 5G. And there's been a lot of change in that time. We started off with relatively simple simultaneous multiple round auctions, typically with only a few lots. And over the years we've had some very complicated auctions with huge numbers of lots and a big variety of formats, notably including combinatorial auctions. If I would have to pick what have been the most challenging over the years, I would definitely pick out combinatorial multiple round auctions. So I've worked on those both from a design perspective and a bit of perspective and generate challenges. So I was involved in the 2000s in the early development of a format known as the combinatorial clock auction. And developing the rules for that when you're asking bidders to put together bids for combinations and many lots, that was definitely challenging. And then when I left my previous company.econ and joined NERA in 2010, I got very involved on the bidder side in the CCA. And I think what was interesting there is that the original premise of the format was that it was supposed to encourage straightforward bidding. And what evolved that was not the case. There are opportunities as with any complex auction to gain it to your advantage and risks if you fail to anticipate what competitors might do. So a very formative auction for me was the 2012 Swiss auction where the bidding strategy was extremely complex. And we were successful with our clients helping them win the spectrum at reserve price, but in that same auction another bidder had to pay a huge amount for a very similar quantity of spectrum. And it kind of spelled out the difference that good advice can make in these types of situation. It's very interesting. You mentioned CCA for some of our listeners, it might be a bit abstract. So if you could just say what that stands for. Absolutely. So CCA stands for combinatorial clock auction. That was a name invented by Graham Louth who was the head of spectrum markets for off-com in the 2000s. And it's a combinatorial auction formatting which bidders bid for packages of spectrum. So you make many bids for mutually exclusive packages say from one band and another band or together and you can win only one of those packages and the price you pay is based on the opportunity cost of denying spectrum to others. That's already probably sounds very complex to most people. And in your book round by round learnings from the first 35 years of spectrum auctions, you say that initially the case for using auctions to allocate scarce government owned frequencies was advocated by the economist Ronald Kose in 1959. And then 30 years later in 1989, near advised in the world's first spectrum auction in New Zealand, Richard, how did we go from administratively assigning spectrum to more sophisticated auctions that we see today? The first thing to say is when Kose came up with the idea of using auctions for spectrum, it was seen as a rather eccentric idea. He was effectively ahead of his time in suggesting that. And one of the reasons regulators went very interesting then is frankly there wasn't a lot of scarcity in spectrum in 1959. So it was relatively easy to make administrative decisions often, you were just handing out licenses to the people who wanted them because there wasn't competition. And there was no such thing as a mobile phone then. The early licenses were just largely for ball casting. I think that changed in the mid-late 70s early age when mobile and paging were invented. And suddenly there was this new class of demand for commercial use spectrum and there was a limited pool of that spectrum available and potentially multiple parties in each country wanting access to that. First regulator kind of tried to carry on as they had before, which is doling out first come first served or if there's multiple parties applying there's not enough having some kind of comparative selection process. The problem with that is that regulators inherently uncomfortable picking winners in their situations because judging between the applicants is very hard. If they get the decision wrong, they're subject to criticism and that could potentially evolve into legal challenge if they're perceived as not having done it right. So auctions become attractive then because they can solve the problem based on one single objective criteria which is price. And so when my predecessors at Nira had the contract in the late 80s for coming up with a format for selling both broadcast and mobile spectrum in New Zealand they said well why don't you use an auction and the New Zealanders were receptive to that. And that led to the development of the world's first auctions for spectrum which was 1989 1990 in New Zealand. We've seen that over the years spectrum auctions have evolved quite a bit. Can you block us through some of the major design innovations and the main types of auctions? So the New Zealand auctions were very simple they used. They were essentially a series of sequential second price seal bits and second price means that instead of the winning bit of paying the amount to their bid they're paying the opportunity cost of denying the second highest bid. Now just selling one license after another is a little bit problematic because bidders may have want to make choices between the licenses and their preferences may depend on the relative prices. That early format didn't allow that. And then it was another political issue with this second price design which is for some of the licenses some bidders bid very high revealing they had high value but there wasn't any competition or minimal competition and the actual selling price was very low and that didn't look very good from an external perspective because the government was seen as leaving money on the table. So when it came round to the new due auctions in 1994 in the United States they started that and they wanted a solution that solved this problem and that produced what is called the simultaneous multiple round auction invented by Paul Milgram and others which was a real revolution that allowed multiple licenses to be simple. So at the same time in a multiple round process and over the course of those rounds bidders could switch between the different types of licenses so they could treat them as substitutes and they could treat them as licenses they could aggregate. So in the US they were many companies were trying to buy portfolios of regional licenses that they put together to create a coherent basis for launching a network. That format was attracted a lot of competition it was perceived as hugely successful many countries around the world to note and then there was a flurry of countries using that format. In the late 90s for late 2G auctions for example Netherlands Germany etc then many many OECD countries and others in the early 2000s. That early format the SMRA has issues there it can be quite vulnerable to gaming in certain situations it's very vulnerable to demand reduction. So there was always a sort of thought in practitioners well could there be a better format out there and sort of two paths emerged from that the first which I've called a revolutionary path is that combinatorial auctions and the CCA was developed as a combination of a multi round process with package bidding and the idea of that is that it would be more efficient and it would eliminate gaming and then there was a more evolutionary path which is represented by what we call the clock auction and in a clock auction instead of bidding for individual lots as in the SMRA you bid for quantities of lots so say for example in the 900 megahertz band there are typically seven lots and you might want to say three or two of those as opposed to bidding on very individual specific frequencies. So those two paths emerged what was interesting is that many regulators embrace that more revolutionary path starting with off-com developing the CCA that got copied in many countries what was interesting is in the 4G that didn't really work out so well many of the CCA's were produced perceived as producing very high price outcomes somewhat erratic allocation outcomes with some bidders winning a lot more spectrum than others and it was quite hard to justify that and what was probably going on there is that notionally a combinatorial auction is fully efficient because if bidders bid reflect in their values the algorithm just spits out the best possible outcome. So what's probably going on there is the bids going in were not fully efficient and there are a number of reasons for that. One is it's just very complicated bidding in big combinatorial options bidders were probably making errors. Secondly bidders were probably gaming opportunities so they saw ways of placing bids that they knew would raise their competitors cost without raising their own and in certain situations that can become a bit self-fulfilling with rival bidders trying that and that can lead to a kind of ballooning of price and a radical unpredictable outcomes. So if you like that format wasn't delivering in the way it was opposed to and that led to in the 5G era bidders going back towards the clock auction as a replacement for the SMRA. In the 5G era it's been very much the case that the clock auction sort of gained an ascendancy that's been the most popular auction for the last eight years or so. SMRA is still pretty popular but in the modern world where it's quite easy to create generic lots because the equipment is increasingly agile. Typically looks like a quicker format and less exposed to gaming than the SMRA. So if I was making a prediction for what happens next as we approach the 6G era we'll probably see a lot more clock options we'll probably see fewer SMRAs we may not see any CCAs. In your book especially you identify a recent conversions in the prices paid for all the mobile bands in the 600 megahertz to 4 gigahertz range. So we're moving towards 6G as you said in a near of more dynamic spectrum sharing and AI assisted auctions. Are we still going to have auctions or could it evolve to fees or last its renewals for example with the digital network exact? Oh gosh well that's a big question let me break that down into parts. On the first part on prices so that's definitely been a convergence in the prices paid for spectrum across different bands. That said there's still a premium for lower frequency bands and I would expect that to remain because lower frequency spectrum is inherently more scarce than other bands. So those trends in spectrum prices have very little to do with the choice of auctions and whether you're using auctions or not. They're just reflecting the fact that there's a lot more supply of spectrum and the technology has evolved so that higher frequency spectrum is a better substitute for lower but never completely so I would expect those premiums to remain. In terms of what happens next well right now there isn't really a good alternative to using auctions for selling new spectrum bands. There may be instances when governments effectively negotiate with a relatively stable industry and say well we'll hand out certain new spectrum to you in say return for coverage commitments or other quality of service rules. So we have seen that for example that happened in New Zealand a couple of years ago. So there may be a bit more of that but I would expect to see auctions for new spectrum to continue to be the norm through to the mid 2030s at least. It may be a different story for renewals. The US has long had a policy of automatic renewal of mobile spectrum rights if they're being used according to some simple basic coverage rollout obligations. I think there's increased interest around the world in whether spectrum licenses should have longer durations or be renewed automatically. That's a trend that's gaining especially with the digital markets act and the EU so we may see fewer auctions for expiring licenses and a greater trend for those to be directly renewed. And in more specific details how do you think that the DNA is going to have an impact? Well I think the DNA pushes market forces a bit further up and it's trying to favour the role of the market. It's always well remembering that auctions are just a means to an end. They're getting spectrum out of the hands of the regulator and into the hands of the industry and the most important arena where competition takes place is not an auction. It's downstream. So the DNA is effectively attempting to promote that downstream competition. There's a series of reforms that reflect spectrum licenses and allocation which are simply designed to give the market a bit greater role in getting that spectrum into the hands of the users in a consistent manner. A big innovation is potentially having longer licenses and maybe even more indefinite licenses. That's really good from an operator perspective because it gives them planning certainty. I don't think there's a revolutionary change for the digital markets act but it will probably incrementally encourage more license extension. I see and so you're saying that auctions will probably be the main way in which spectrum will be auctioned. Will it be the same multi-band auctions that we have today or could it move to what off-com is doing with high density areas? Well so it depends on the type of spectrum banks. So for mobile the next two big bands on the horizon that is probably six gigahertz and 600 megahhertz will probably auction for those will probably look more similar to the current wave of auctions. I think off-com has raised the possibility of some kind of high density division at six gigahertz. We'll see where that goes. That obviously worked very well for millimeter wave spectrum where there's very clear dividing line between the use case for exclusive use in cities and in rural areas. It's not so clear that with six gigahertz which will probably heading to be something or more of a mainstream mobile bat. So you know the case for national licenses is much stronger there than a millimeter wave. I think one of the things to remember is that more and more spectrum is being allocated for mobile use and as that supply goes up operators have greater choice. Now obviously the amount of spectrum they need to run a modern network has gone up hugely as well but to be fair to regulators many have done a pretty good job over the last decade in bringing spectrum to a market ahead of when it's really needed for capacity and I think that's really helped in bringing down spectrum prices over the last decade and regulators as I should say in Europe are doing are looking ahead at six gigahertz needed there and if that's coming in the early 2030s that may well be ahead of demand so I wouldn't expect a big rise in spectrum prices. There might be a shift away from the convergence that we've seen in recent years. In recent years we've had as I said this convergence between lower high frequencies but looking forward low frequencies are always inherently a bit more scarce so it may be that the premium for those solidifies but it much lower levels than the senior decade ago. In terms of looking ahead you talked about the use of spectrum for mobile do you think there's going to be satellite auctions and would they be similar or would it be different from a normal mobile auction? Yeah that's a really interesting question so satellites obviously a very exciting area right now. It's worth drawing the distinction between the MSS bands and the higher frequency bands that are starting and the like are using for their fixed broadband services. If we start with the MSS bands there are basically low frequency the nature of the spectrum means that they're often better used for exclusive use by satellites and in that sense they look a little bit like mobile bands. I think there's a very strong case for auctioning access to MSS spectrum in fact we've done that on behalf of Saudi Arabia near a auction a few years ago. The European Commission went well embrace an auction for the MSS spectrum with the the rights for which you're exploring next year. It's much more challenging to see how an auction might be run for the higher frequencies satellite bands which are currently based on shared use but there is a potential problem there. The KU and KAA bands are getting more and more congested because of the huge increase in the number of satellites going up and we may not be that far away from a point where they become so congested that some kind of pricing mechanism is needed to encourage some users to migrate to higher frequency bands which remain relatively empty. That's probably not going to look like your current spectrum auction because what your need is a price signal to persuade bidders to migrate. It's very different from an exclusive use environment where you're simply saying here's a frequency I'm allocating it to one user. In this case we're looking at spectrum that can accommodate many users but we need a mechanism to try and relieve the congestion. So I think that's going to be a super interesting area. We may see exchanges for access to that kind of spectrum eventually emerge and that might eventually open up whole new ways of looking at how you allocate spectrum. For now, mobile is definitely focused on exclusive use access, but it's not impossible to imagine that as the amount of spectrum from mobile grows that will eventually move to some kind of more dynamic platform where instead of mobile operators always having their own exclusive use spectrum, perhaps a larger amount of spectrum is shared and it's priced on a more dynamic temporal basis. So what could be the pricing mechanisms? Well, the pricing mechanism might be some kind of forward auction saying, do you want capacity in three months ahead and then a late auction, do you need extra capacity a day ahead and you have different bands, different amounts of frequencies in those different auctions and much more escaped for spectrum to move around across operators according to their particular needs at any time and indeed there might be other use cases competing for the spectrum, but we are more than a decade away, perhaps longer from that kind of mechanism, but that is the kind of mechanism already exists in electricity markets. Electricity is obviously a very fungible product. Megahertz are much less fungible because of the limitations of the technology that's employed in megahertz, but as the technology becomes more flexible, it's at least possible to imagine a future where there's much more flexibility for operators to change what they're using from day to day. And slightly changing the focus. We know that spectrum is a scarce resource and its distribution really shapes entire markets. How do you ensure that auctions promote competition instead of entrenching it? Well, okay, so I think it's a really good idea to distinguish the outset always between competition in an auction and competition in a downstream market. So the most important goal of any auction should be to further competition in downstream markets. And I think if you look back, there are many examples in the past of regulators probably forgetting that and focusing too much on generating competition in the auction, which can produce high revenues, but may end up with allocation results that reduce the number of downstream operators or impose very heavy price burdens on them, which limit their ability to invest in downstream markets. So I would always say that most important thing is competition downstream. And then when you're thinking about that, it's often the case now that many countries only have three mobile operators. We're probably already at a case where can't really afford to lose one of the mobile operators. And that puts serious constraints on how regulators can run auctions. You need spectrum caps. You may need to do things such as two stage auctions where you have a first stage that guarantees a critical mass spectrum tool, at least three operators and a second stage in which they can compete over the residual spectrum. The details of how you design the auction become very important. And it may be in many cases, and we've already seen this, that regulators have to accept a loss of competition in the auction in order to guarantee an outcome that will support their downstream competition objectives. I see. And a significant difference from auctions that run, for example, in the commercial sector is that maximising the revenue is often not a primary objective. Would you say that there are ideal levels of competition and how does revenue play a role in the auction? Well, you know, it's up to each regulator and potentially the government what, how to define the objectives for particular auction. And if you're reasonably wealthy country, you probably don't need to be focusing on revenues from a spectral auction. You probably should be focusing on the efficiency of allocation and promotion of downstream competition. That said, it's a scarce state owned resource and it's reasonable to expect a certain return for taxpayers from for allocating this to private operators. The best approach generally is for governments to do some assessment beforehand about what is an acceptable level of revenues. It's not so hard these days to take a conservative estimate of the market price, for example, using benchmarking. You can use that to set a level of reserve price that should guarantee an acceptable outcome. Then it doesn't matter so much if the auction is not competitive because that's kind of baked in. I think where things have gone wrong is when governments have set reserve prices very low, well below is sort of minimally acceptable outcome and then sometimes operators are able to take advantage of that outcome looks bad. Then the other extreme, the auction is gerrymandered in some way to create incentive for operators to try and knock each other out or fear that they'll be knocked out. Then you can get very high prices because the price is not really reflecting just the spectrum but the operating value of being in the market. Then that can have very negative implications downstream. I think the challenge is to avoid those extremes and find that sensible middle ground. You mentioned downstream competition. Can you explain that for the listeners there are a bit less familiar with the spectrum auctions? By downstream competition I mean the competition between mobile operators to cell phone mobile contracts and that is where the real benefit to consumers in society come from having good high quality mobile services and low competitive prices for mobile services. Ultimately the regulated most important objective must be to promote that and not so much the outcome of any particular auction. You mentioned auctions going wrong. I imagine that could be for example when bidders perceive that the press is too high or too low or the key bidders want too much or too little spectrum. What happens in practice when auctions go wrong? Well I can give you some examples of auctions that have probably gone wrong with various degrees of effect. If we go back to 2012-2013 the Netherlands and the Austrians both held CCAs and there were only three operators there was quite an obvious way of sharing it but the format incentivized them to bid strongly rather than share the spectrum and the problem was they probably started bidding against each other based on their enterprise value not just the intrinsic value of spectrum. The result of that was very high price outcomes and it got very close to a big operator being knocked out of the market. That didn't happen but they ended up spending a lot of money and then they ended up in litigation against the regulator. They didn't win the regulator won but in hindsight that was not a situation that the regulator would have wanted and I think that showed up the difficulty of that format in very high stakes big auctions. If we fast forward to Norway in 2013 they came up with a what's called a first price seal bid so they basically said okay just tell us what you're willing to pay for these licenses as a package and the highest bidders will win and the others get eliminated and the problem with the first price seal bid is you're not really bidding based on your own value you're bidding to try and out bid the competitors just enough so you have to shake your value down. What happened in that auction was one of the M&O's Tally 2 actually got eliminated and having failed to win critical 4G spectrum they didn't see a path forward to continuing the market so they exited. Now they were replaced by a new entrant ice which has been reasonably successful but it's taken time for them to build up in the market. That kind of auction is very high risk because you're exposing the market to very unexpected results and it's not clear that that outcome is efficient at all and so it can be pretty disruptive. If we give you a sort of a more sort of farcical but lower stakes version the Portia Geese in 2021 ran a traditional SMRA for a very big 5G auction with dozens and dozens of lots and they had very low bid increments written into the auction rules. That auction ended up lasting almost 9 months over 1,727 rounds. That's obviously not a sensible way to run an auction. Now it delayed the launch of 5G arguably in Portugal because the auction took 9 months other than that the outcome was probably fine in the end but it does show the importance of getting sensible auction rules and really thinking about the implications of the decisions that you're making and that's really one of the reasons why the clock is overtaking the SMRA because the SMRA has these issues with it for very large auctions it can take just too long and that's one of the reasons I suspect that the clock will continue to grow in use relative to the SMRA. And you've talked to about sensible auction rules before we wrap up, what's one piece of advice you'd give policymakers designing future spectrum auctions? Well, I would pay attention to the full range of things that matter. So it's not just about the auction format of the auction rules. They're important, but you also need to ensure that you get the reserve price right and you have appropriate qualification criteria for the bidders and that you are thinking about the downstream market and what range of outcomes are reasonable or not. So you don't need to rely wholly on the market. It's best probably to design an auction in which you are allowing the market to determine the outcome within an acceptable range. Putting a lot of effort in the upfront into understanding what that acceptable range is in terms of who the bidders are, what prices are acceptable and what allocation outcomes are acceptable from a downstream competition perspective. If you get those details right, almost any auction format is probably going to work and then you can focus on picking the auction format that's going to deliver in an acceptable amount of time. So that would be my advice. Richard, thank you for joining us today and sharing your expertise. If you've enjoyed it, feel free to subscribe. The Policy Trackers Spectrum Podcast is available on all major platforms. Thanks for listening and see you next time. Thanks everyone. [Music]