Reg NMS

There is renewed discussion of the possibility of banning payment for order flow in the United States, and the proposed ban appears to be addressing some issues in the wrong way.  Those issues are oddities of "Reg NMS" --- that's "National Market System" --- that convert a dozen or so stock exchanges into a single, abstract "stock market".  And my proposal, ultimately, is that we ban exchanges from charging commissions to liquidity takers.

I want to note that this is ultimately an accounting requirement --- and Reg NMS is why it matters.  If I place an order to an exchange to buy a stock for $64.03, and the lowest price at which anyone is willing to sell is $64.05, the order sits there until someone comes along to sell at $64.03.  If and when they do, the exchange typically charges each of us a commission; I'm really paying $64.031, and they're clearing $64.028.  If the seller were not allowed to pay a commission, you could report the price as $64.031 instead, and charge me a .3 cent commission instead of .2 cents; the same cash changes hands in the same ways, we're just reporting the trade differently.

The problem is that, for many orders subject to certain concessions to the laws of physics (namely the speed at which information can be transmitted), an order to trade "at market" has to go to the exchange with the "best" price.  Because of this, some exchanges actually have a negative commission for liquidity providers --- they might charge me $64.029, and only pay the seller $64.026.  By telling a seller (or, more to the point, the seller's broker) who will only clear $64.026 by selling on my exchange that they can sell there for "$64.03", they can attract a sell order that could perhaps have cleared more (net of commissions) somewhere else.  By requiring that the commission be paid by the trader whose order is resting on the book, you're simply aligning the reported bids and asks with the prices that would actually be obtained by a trader hitting them.

Most of "payment for order flow" similarly constitutes a sleight-of-hand involving careful but economically meaningless distinctions between "prices" and "commissions".

bounded cognition in evolutionary game theory

Suppose you have a set of agents that, for behavioral and strategic reasons, all "cooperate" with each other, but would recognize if one of the other agents started to "defect"; cooperating would be an evolutionarily stable strategy in this context.  If the society gets larger you might expect there to be a point (Dunbar's number, for example) where the agents can't keep track of all of the other agents anymore; suppose, in fact, that we have 1400 agents, each of which is designed to keep track of 140 agents.  As long as all but a couple of agents continue to "cooperate", you're still fine; if the number creeps above 20 or so, then the information required to keep track of who has been cooperating and who has been defecting gets to 140 bits, and so one might suppose that would overwhelm the agents, and there would be a tipping point around there where cooperation would break down.

free will

 At some point I intend to read some philosophical works on free will and determinism, but I want to record a thought I have that may be unoriginal, refuted, or both.

"Free will", as I think about it, I believe entails the ability to take an action that was not predetermined.  ("Causality" on some level requires free will, because I have to have a sensible way of thinking that Y did happen when X did happen but wouldn't have happened if X hadn't happened; I need for the very notion of a counterfactual to make sense.  Part of the reason I'm interested in free will is that I'm interested in causality.)  Perhaps, however, if I live in a world that is ultimately deterministic but in which I in some practical sense can't predict whether X will happen, but can predict that Y will happen if X does, perhaps that gives rise to the perception of "free will" that I require.  If minds of the a similar order of complexity to my own are unable to predict whether I will do X or not, perhaps that amounts to "freedom".

This perhaps fits into the notion of "compatibilism" -- again, I haven't read as much as I eventually should.

rationing

My previous post mentioned power shortages in Texas; to some extent generally, but especially when there are power shortages, retail allocation of electricity in the United States is mostly not done by a market mechanism.  Similarly, the covid vaccine is not being allocated by market mechanisms.  In both cases, more centralized government or quasi-government organizations are making decisions about priorities.

In all cases, it seems that an important consideration for the prioritization schemes should be pump-priming; you have this resource that is bitingly scarce, with a high marginal benefit, and so first and foremost you would prefer not to have to make choices, that is to have as much production as possible.  In both cases we seem to have had failures of this sort; some of the shortage of gas to gas-powered electricity generators appears to have been due to power cuts to gas infrastructure, and while health care workers were given high priority for vaccines, apparently the vaccine makers themselves were not, and have suffered some slowdown because of employees out during the third wave.

In an idealized market, the gas systems would have outbid other buyers for power, and vaccine producers, being paid marginal benefit for additional vaccine production, would outbid most buyers to allocate those vaccines to their own employees.  In a (politically and otherwise) realistic market, I don't know whether that would have worked out.  In any case, it seems like a useful point for future designers of rationing schemes to keep in mind; if a scarce good is extremely valuable, then using some of it to produce more of it is probably a good idea.

energy prices

 A week or two ago, Texas experienced some unusual weather which, by means of a couple different mechanisms, reduced the supply of electricity while increasing its demand.  The wholesale price of electricity shot up to $9 per kWh, where it is capped, and many people had their power shut off altogether as they were sacrificed to preserve electricity for priority customers (e.g. hospitals).  Other people kept receiving power; some of them have retail providers who pass along the wholesale price of the power, and many of them are now seeing very large power bills, some over $10,000.

There are a few things that I've been thinking about related to this.

• If you were to draw an actual "marginal benefit" curve, based on some plausible measure of customers' own valuation of marginal units of power, it doesn't seem to me like it would be as inelastic as these price dynamics imply. My impression is that a lot of the people who were shut off would have been unwilling to pay $10,000 to instead have power through the week, and especially that, given the choice between paying $10,000 to use that much power or $5,000 to have half as much, they would have opted for the latter. Perhaps I'm simply wrong here; suppose I'm not.
  • How well were the customers paying variable rates able to track and control their use of electricity? Presumably they have some kind of smart meter that at least knows when they used power; are their meters able to communicate the price in real time, maybe shut off appliances or change thermostats? (Did these people take steps to lower thermostats themselves?)
  • What is it about the way the market is structured that caused the price to get well ahead of marginal benefit (assuming, again, that that's what happened)? Perhaps if more consumers had been on variable-rate plans the system would have been more robust. Any power that was used that was not worth the price to the customers was presumably used either in ignorance of the price or under an arrangement in which the person deciding to use the power was insulated from the cost.
• The supply side, on the other hand, does seem really to have been quite inelastic; the marginal cost, I'm guessing, was well under $1 per kWh up to a very high percentage of the amount actually being supplied, with a sharp upturn at that point. Classical Marshallian welfare analysis suggests that, in such a context, "producer" surplus will be very high. In a rational expectations model with free entry, sellers would expect these kinds of episodes and the surplus would be used to cover fixed costs and/or induce entry. In the real world, it does look like a nice reward for the producers that were able to keep producing power, perhaps because they had prepared better for a situation like this, or perhaps because they were lucky. (Usually there's some of each.) I don't know whether it's likely to induce local improvements to robustness or not.
• Some of the drop in supply seems to have been that gas-fired plants weren't able to get natural gas. Few if any retail gas customers seem to have lost gas, however; it was apparently prioritized first to households, with power plants lower in priority. I have heard the word "obviously" attached to this decision, perhaps because cutting off gas to households that use gas during a cold snap would mean those houses would lose heat, yet what did happen is that many houses that use electricity for heat had their heat cut off during a cold snap. I don't know that the physics of the gas pipes would allow households to have some throttled quantity of gas, but it seems likely that some alternation — rolling gas-outs designed to provide gas to households when it's likely they're dropping below some temperature like 45 or 55 degrees, but provide some gas to the plants producing the power to heat other houses — would have been better on the whole.

Runoff voting with runoff

Voting for a single winner between two candidates is straightforward; if the candidates and voters are all to be symmetric, then asking each voter which candidate to prefer and electing the candidate with more votes is relatively[1] problem-free and is the obvious way to determine a winner.  With multiple candidates things get tricky; in particular, our usual plurality system results in "vote splitting", where the result depends as much on which candidates decide to run as it does on who the voters prefer; for example, perhaps Alice gets 40% of the vote, Bob gets 39%, and Carol gets 21%, and it may be the case that more voters preferred Bob to Alice, but some of them voted for Carol instead. 

One increasingly popular way of addressing this is with instant runoff voting; voters in Maine (and, going forward, Alaska) rank their options, indicating that one candidate is the voter's first choice, another the second choice, and so on.  When the votes are tallied, each vote is counted for the candidate ranked first, but if no candidate gets a majority, the candidate with the fewest votes is eliminated, and ballots are recounted, being credited to the highest ranked remaining candidate.  Older runoff systems in the United States ask voters to come back to vote in subsequent rounds with fewer candidates, but the instant runoff with ranked ballots can do this automatically without calling the voters back to the polls because the voters have implicitly left instructions on how they wish to vote in the runoff.

One problem with this procedure is that it can frequently eliminate popular second choices.  For example, it may be that the 40% who voted for Alice prefer Carol to Bob, and the 39% who voted for Bob prefer Carol to Alice; Carol is the top choice of fewer voters than the others, but is nobody's least favorite candidate, and in particular she would win the election if either of the other candidates dropped out.  A majority of the voters prefer Carol to Alice, and a majority prefer Carol to Bob, but Carol is dropped from the runoff, and one of candidates who she would have beat will win instead.

This is not too hard to fix with a relatively small change to the procedure: when deciding which candidate to eliminate, we look at both of the two candidates with the fewest votes, and eliminate whichever is ranked lower on more ballots.  In this case, we see that Bob and Carol have the fewest first-place votes, so one of them will be eliminated; because 61% of voters prefer Carol to Bob, we eliminate Bob.  Alice and Carol are in the runoff, where Carol wins.

Maine has published the ballots from the House of Representatives election for 2018 in its second district, and, while I have trouble figuring out exactly how the state interpreted some of the ballots, I can use them for an example.  By my count, Poliquin was the top choice of 134,358 voters, Golden of 132,145, Bond of 16,650, and Hoar of 6,996.  In Maine, this was enough to eliminate Hoar; in my system, we first compare him to Bond.  Of the ballots expressing a preference between them, there were 43,131 more ballots that had Bond ranked above Hoar than vice versa; it is only after we observe this that we eliminate Hoar.  This leaves three candidates, and we redistribute the 6,996 votes for Hoar; Poliquin now has 135,275 votes, Golden has 133,381, and Bond has 19,313.  Because Golden beats Bond by 96,458 votes when only those two candidates are considered, we eliminate Bond.  Less than 4,000 of her votes transfer to Poliquin, though; in the final round, he has 139,238 votes, while Golden has 142,664.  Golden is therefore elected.

Golden	132,145	133,381	142,664
Poliquin	134,358	135,275	139,238
Bond	16,650	19,313
Hoar	6,996

---

[1] I'm going to assume away exact ties, as I so often do.

selling volatility

It occasionally happens that, for some exogenous reason, some market participant or group of participants sells a lot of options on a stock, and it is then noted that this means the sell-side of Wall Street is often involuntarily buying them, which typically means that those sell-side shops will go out and delta-hedge; since these move one-for-one with the stock when the stock price is high, and don't move when the stock price is low, that entails buying the stock as it rises (nearish the strike price), and selling as it falls, thereby stabilizing the price.  

All of this is well and good, but I'm inclined to skip some steps; the buy-side agents are selling volatility, and that should generally push down the price of volatility.  In most markets, selling a thing pushes down the price, and when ordinary supply and demand work reasonably well, reequilibration at the lower price will entail moving to a situation in which the marginal cost of the thing is lower than it was at the beginning, and the detailed mechanics don't really matter that much.  "Volatility" may seem abstract, but it works the same way.

An important special case here is the mortgage bond market; homeowners there have (and to some extent use) an option on interest rates, and when those become more likely to be exercised, that tends to destabilize long-term interest rates.  Again, you can walk through a discussion of interest-rate investors trying to match shrinking durations, and eventually you find dropping interest rates being pushed lower and rising ones pushed higher, but if you're just interested in the dynamics of interest rates themselves, you can skip all these steps; a bunch of people who have no idea what duration, convexity, or volatility mean are unwittingly long volatility, and that pushes up prices.

I'm posting this at the moment because Matt Levine is writing about greenshoes, which are options granted to underwriters of IPOs, for the purpose of enabling the underwriter to stabilize the price of the IPO.  The details of the mechanics are a bit different from what you might think based solely on that sentence, but, as I keep saying, it's not that sensitive to mechanics; if the company thrusts options into the financial system, that has the effect of pushing down volatility.  Some people don't like greenshoes, and a problem with them that Levine argues is probably hypothetical is noted in his newsletter.  It has made sense to me for a long time to let locked-up shareholders sell at some minimum price before the lockup period ends, e.g. you IPO at $40 and tell certain classes of shareholders that they can't sell below $60 in the first six months, rather than tell them they can't sell at all.[1]  I wonder now whether it makes sense to distribute warrants with the initial share allocation.[2] [3]

It's hard to remember one's previous state of ignorance, especially when it's been so long, but I believe that 20 or 25 years ago I assumed an IPO worked more or less like a direct listing with a capital raise; until much more recently I didn't realize that those weren't even allowed.  (I became aware of that when discussion of changing that heated up, and it looks like in 2021 it will finally be permitted to do what I assumed in the 1990s was the way a company went public.)  I mention this in substantial part by way of warning that the hard parts of finance are legal and customary, where, as is true of most people, by "hard parts" I mean "the parts I don't know and/or understand so well."

---

[1] In a certain kind of theory, this should make no difference; forward-looking investors on the day of the IPO know these shares will be available to the supply side of the market in six months, and should price that in.  A lot of my practical concerns with ideas I'm putting forward in this post are related to the extent to which this fails in practice, and the related fact that the amount of genuine "float" in the early days of a new issue is quite small.

[2] In this case the size of the early float of shares endorses my proposal, insofar as the reasons the float is so small are likely not to apply to these warrants.

[3] One of the complaints about the greenshoe is that companies don't like not knowing how many shares they're selling; as I've argued in some previous posts, a lot of companies should be offering more of a bona fide, upward-sloping supply curve anyway.  If the warrants execute, it's likely the investors think the company can make good use of their additional dollars, while if they don't, they don't, and the company should have some idea what it would best do with the additional investment should it be forthcoming.  It is perhaps worth noting here that if demand is volatile, an elastic supply curve is a way of absorbing some of that volatility into quantity rather than price; if you don't like talk of "selling volatility", perhaps the fact that a warrant exercise increases the quantity supplied is more intuitive.

index minus

In a certain kind of ideal market, securities lending fees should be zero; at any positive price, there would be effectively infinite supply.  In actual practice, a lot of shareholders, for institutional or other reasons, don't lend out their shares.  Those that do lend out shares tend to be price-inelastic, at least above some de minimus price; they'll lend out all of the shares they have available.  Retail borrowers typically don't directly pay a lending fee; they pay it indirectly in various ways, and those ways tend not to respond to the level of the actual fee; if the actual fee is higher than what the broker is getting from the customer, they are simply unable to borrow.  Retail demand, thus, is also very inelastic for most possible prices, but with a dramatic elastic region around a particular price that depends on things like nominal interest rates.

Index investing has grown a lot in the last generation (after growing a lot in the generation before that); a lot of people simply own the S&P 500 in some literal or practical sense.  The saver owns the same fraction of every company in the index.  Some retail investors have, for a while, adopted an "index plus" strategy of putting most of their money in a passive index fund but putting a little bit in stocks that they select.  Increasingly, people are interested in something like an "index minus" strategy; they are largely agnostic about most stocks, except that they think they should own "stocks", but there are a few stocks that they think are overpriced, or think are bad for ESG-related reasons.  My impression is that, at least at the retail level, they mostly end up just buying an ESG fund, or buying an S&P fund and perhaps selling a couple of stocks short, and that that's about the best they can do.

I'm trying to work out in my head whether there's a sensible market for an index minus brokerage-type account; the broker here would guarantee that a customer could in some sense "sell short", but only in the amount that that customer owned the stock.  To the extent that different customers "short" different stocks, part of the custodial assets of the company could be in a Vanguard S&P fund or do something with futures.  To the extent they don't, the company might have to do some of the work of a fund itself,  perhaps including some securities lending, and would have some of the same economies of scale that most mutual funds do.  I think this ends up looking a bit like a subsidy to customers selling hard-to-borrow securities, but the limit on the size of the sale keeps that manageable, and it may be an attractive enough idea to enough people looking to put their money in stocks, many of whom would only sell cheap-to-borrow stocks anyway, that there's a business opportunity to be had here anyway.