The inefficiency of large, infrequent transactions

In a conversation with Matt Levine, Tyler Cowen asks:

COWEN: Like you, I’m mostly an efficient markets guy, but when I look at initial public offerings I’m very baffled because investment banks take such a huge cut.
If you needed to argue, “Well, they need the cut because they talk up the security, and in the absence of their efforts, no one would be interested, and it’s worth it,” maybe that argument works. But it seems somewhat to stand in tension with an efficient markets hypothesis, which suggests the thing will find its appropriate level without any particular investor having to talk it up.
Furthermore, attempts to get around the current mainstream system of IPOs have not always been successful. Auctions have been tried. Israel has tried other methods. Spotify is giving it a go. We’ll get further data, but they’re not obviously doing better.
How do you reconcile IPOs in their current form continuing, the investment banks taking such a huge cut, and some version of efficient markets hypothesis actually making sense? Do you see what I’m asking?

I wasn't at Amazon for its IPO, but I did work on a huge convertible debt offering there. At the time it was the largest ever done. After the transaction had cleared, I received a box at work from Morgan Stanley, who'd been our banker on the deal. Inside was a Tumi carry-on bag, personalized with Amazon and Morgan Stanley logos. Everyone on our side who worked on the deal received one. My guess was the bag retailed for $500 or so without the personalization.

It's always an ambivalent feeling, receiving such a pricey gift from a business partner after a deal. I've received similar customized luggage from law firms after their assistance on something. It feels as if you bought yourself something you didn't want. I generally believe in efficient markets, but I believe even more strongly that law firms and bankers don't do anything out of affectionate generosity. Call it the efficient business transaction hypothesis, in which you always pay for every bit of service rendered. There's a reason you don't receive a Tumi bag from your dry cleaner after laundering a big batch of shirts.

So, as with Cowen, I've long puzzled at the exorbitant take on the part of bankers in deals, regardless of the volume of work done.

Levine answers:

LEVINE: How do I reconcile? One version of efficient markets is that, in the absence of news, the price yesterday is going to be the price today, or whatever. There’s some sort of continuity of prices.
The IPO is a huge discontinuity. You don’t have a price and then you have a price. If your notion of efficient markets is a straight line of the price not moving very much or of the price instantly incorporating information, it’s not unintuitive that you’d have a big squiggle at the start, that you wouldn’t really know what the price is for three days, and then you would.
I don’t think it’s unusual that the first trade of a stock would not be the price that it settles to in a week, but then, the second week would be pretty close to the first week.

A good answer, but the word "discontinuity" hints at another aspect of this type of transaction which hints at why an IPO may not exhibit aspects of an efficient market.

I long wondered the same of the real estate market, another space in which real estate agents have been able to maintain their 6% rake despite the advent of the internet. It's a bit of a puzzle considering how many other markets the internet has turned so efficient that almost all the profits went out of them.

There's a class of transactions which retain abnormally large margins which share several qualities.

  • They are rare, or infrequent transactions for one side, usually the buyer. Many people may only conduct one of these in their lifetime.
  • They are really large relative to most transactions in a person's lifetime, anomalously so.
  • They don't have well-known price anchors, often because the service is good is unique, making exact comps squishy.
  • They are transactions which have some symbolic or mystical value which can justify overspending because how do you put a price on magic?
  • They are time sensitive.

Let's take those in order. The low frequency of such transactions weighs in favor of the seller, not the buyer, because the buyer has no experience. IPOs, housing purchases, weddings. The experience asymmetry is really information asymmetry. One should always be skeptical of playing any game where your opponent has to explain all the rules to you. The three card monte and other cons always begin with the con man explaining the game to you, no?

The anomalous size of the transaction works against you in several ways. One, when so much money is involved, one tends to experience inflation on all fronts. If you're buying a condo that costs a million dollars, or, in the case of San Francisco, a lot more than that, what's another couple hundred dollars in fees here, a few thousand there? In the scheme of things, 6% to your agent doesn't even cross the mind after you've been beaten out on several dream condos in a row by a suitcase of cash from a Chinese buyer who never even showed up. Relative to the transaction size, it's nothing. Human cognitive deficiencies on this front are well-studied.

The lack of price anchors doesn't help. What should a wedding dress cost? You can ask your friends, but, to crib from the Marines, "This is my wedding dress. There are many like it, but this one is mine. It is my life."

Buying cars is also a rare transaction, but it's unlikely you're buying a car model that's never been purchased before. As soon as the internet revealed dealer pricing, it was easy to just call a string of dealers around town and offer to pay $500 over that, nothing more.

The special nature of such transactions supports a "je ne sais quoi" premium. Are you really going to skimp on your IPO, or pinch pennies when it comes to your wedding? I know brides who would sooner cut dinner from the wedding festivities than skimp on their wedding dress. If the event has symbolic value, then almost any price can be justified to oneself. Different cultures support different such mystical price premiums. In the United States, people will spend to no end on their dogs, for example. It's the same for babies or children in many cultures. Are you really going to try to save a hundred dollars or so on a stroller which will transport your precious child for several of their formative years, or not pay the $25,000 tuition for your potentially genius of a child to attend the coveted private school every one is trying to get into?

The time-bounded nature of such transactions just prevents you, as the buyer, from using a long time horizon for leverage. Sellers, of course, will use the time bound as a cudgel in every way possible. If you don't nibble they have many more transactions to come, but you as the buyer have no such luck. The wedding invites have been sent, your investors are eagerly awaiting your IPO, you need to secure housing before your current lease runs out.

The internet has, in many ways, ushered in an age in which old axioms of supply and demand economics aren't always applicable, or which make edge cases commonplace. Infinite supply, zero marginal transaction costs (especially in production and distribution), these are suddenly prevalent. It's almost nostalgic, therefore, to ponder the economics of infrequent transactions, or of markets where supply is fixed (the limited edition sneaker market, for example, or the Bay Area housing market).

The high margins prevalent in large, infrequent transactions suggest that businesses might create some moats simply by replicating some of the qualities of such transactions, even if the underlying businesses don't consist of those. What is subscription pricing but a way of grouping many, frequent transactions into many infrequent transactions where it's difficult to determine the unit cost of all the components of the bundle? I suspect half the reason for the phenomenon in which tech oscillates between bundling and unbundling is to find the efficient size of the bundle that the consumer will bear. We won't all have dozens of Patreon accounts that we support in the future, but neither will we pay $300 a month for some mega video streaming bundle that includes everything under the sun.

Every transaction is a moment for a buyer to reconsider whether it's all worthwhile. At some point, I expect many newspaper will band together to offer some bundled subscriptions. Magazines already have tried this but the very concept of a magazine, and the typical frequency of issues (monthly) is difficult to construct a high value bundle around as compared to newspapers.

I also suspect that the ride-sharing services like Uber and Lyft will continue to experiment with some sort of monthly fee with unlimited rides within reason (perhaps with mileage overage payments). Uber has already experimented with offering pre-payments for a bundle of fixed price rides for a month, and I expect that's just the first step towards some sort of lock-in car-as-a-service subscription pricing. It's exhausting to duke it out with competitors one ride at a time, and it trains buyers to shop for the best price on each ride, something that's easier than ever given that both services' apps allow you to enter your destination to see the price ahead of booking a car.

Let's suppose that consumers do possess some limit to the number of transactions they're willing to put up, some finite cognitive load for payments. If so, then any seller must consider the number of other subscriptions in the marketplace when understanding how well their own bundle will do. And in that case, we should start to see more bundles of bundles. In the tweet above I suggest a few, there are likely many others.

The problem with creating bundles of bundles is divvying up the credit. It's unlikely all components of the bundle contribute equally, so the highest value component is typically hesitant to join in. However, in the digital age, assigning credit is much simpler. So, just as ESPN earns much higher carriage fees in the cable bundle than, say, Comedy Central, can be done with exact usage figures in real time.

The funniest example of the efficiency of bundling (well, it is to me since I don't have kids) is the division of childcare. When I'm around parents, it's somewhat striking how much of a constant low-level tension arises from splitting up such duties. Whose turn is it to change a diaper, or to go hold a crying baby? With the exception of the wealthy who can afford full-time childcare help, such tasks are among the most frequently occurring transactions in adult life. Each is a moment of negotiation, and thus an opportunity for friction.

Thus, parents come up with long-term divisions that reduce the number of such transactions. Sometimes it's an agreement that the dad handles all dishes and diaper changes while the mom does all the feeding and crying. Parents also will come up with bundles which they trade, to reduce the overall number of negotiations. The husband get to go on a guys golf trip for a week, but later that year, the mother gets a girls trip to Palm Springs.

Sometimes efficiency is just household equilibrium, and you can't put a price on that type of harmony.