Sunday, January 26, 2014

Different goods are differently liquid

An evaporation in liquidity. September 13, 2012, at 12:25:27 to ~12:32 PM, eMini contract. From Nanex.

Steve Roth, who blogs at Asymptosis, recently posted a thoughtful critique of the idea of moneyness over at Cullen Roche's blog. (We've had a series of exchanges before on these questions). Even if my response can't sway Roth it should provide new readers of this blog with a rough overview of where I've been going with the idea of moneyness.

Let's start with definitions. Moneyness is a fancy word for liquidity. In short, it refers to the ease with which we expect to be able to trade something away for another item of value. Our expectations about liquidity are conditioned by an item's historical liquidity and modified by anything that we think could change it in the future, including new market mechanisms that might promote (or demote) that item's liquidity. All valuable goods & assets have varying levels of liquidity, or moneyness. Some will be easier to market when the need arises, others will require more effort.

Roth's first criticism, published on his blog last year, is that the
“moneyness” concept... seems to hinge on a single axis of “liquidity,” when in fact different units of exchange are differently liquid.
He goes on to give more detail at Roche's blog:
Suppose you have $10k in quarters. You can buy all the Snickers bars you want; there's a very liquid exchange market (quarters for snickers bars) out there. (Though need to tromp around to buy $10K worth of Snickers bars does seem to make it less “liquid”…)
But can you buy a car with those quarters? How about treasury bonds? No. Those quarters are completely illiquid relative to cars and treasury bonds.
Now think about treasury bonds. They're completely illiquid relative to both snickers bars and, but extremely liquid relative to fed bank deposits (reserve balances) — if you have 
I don't think I have to stretch this explanation out. Think about fed reserves/deposits — they’re (il)liquid relative to what other goods/assets?
So every financial asset — in fact every real good as well — has multiple liquidities, relative to every other asset/good.
I agree with Roth. Even the most liquid items are only tradeable along a few margins, or routes. The godfather of liquidity, a US dollar chequing deposit, can get you groceries or a car, but can't buy shares in IBM. A deposit at a brokerage can buy you IBM, but it can't get you a bag of groceries or a car. Roth's well turned phrase is worth repeating here, that different goods are "differently liquid", a point that I echo in Long Chains of Monetary Barter.

However, this doesn't mean that we can't arrive at a single combined measure for all of a good's different liquidities. All we need ask an individual is this:
"How much would I have to pay you in order for you to relinquish all rights to trade away your holdings of asset x for one year?" 
What we are extracting here is the individual's reservation price for x's liquidity. In this setup, the individual is allowed to continue to enjoy all the various pecuniary and non-pecuniary returns provided by x during a one year period, save for one return—its liquidity return. We are asking the individual to forgo each and every one of the good's multiple liquidities, or, put differently, the various margins along which x usually trades.

Whatever compensation the individual requires for giving up the right to trade away x along all routes is an indication of the foregone value that they ascribe to each of x's multiple liquidities. By dividing this price by x's total price, we can estimate what proportion of x's overall valuation our individual attributes to the liquidity component.

I think that this meets Roth's criticism, since in effect we are asking an individual to forgo each of an asset's multiple liquidities, all at once. We can go ahead and ask that individual the same question for each asset they own: how much would you have to be paid to forgo the combined multiple liquidities of a? and b? c? In the end, we'll have a list of all the individual's assets, along with the percentage contribution that each asset's liquidity provides to its total value. Having standardized our measurement of liquidity, we can now construct our individual's scale of moneyness for the coming year, ranked from least liquid to most liquid, the most liquid being that good who's liquidity contributes the largest chunk to its total value.

The upshot: the existence of multiple liquidities shouldn't prevent an individual from making private liquidity comparisons across different goods.

Which leads into Roth's next criticism, that estimates of liquidity differ across individuals. This presumably (Roth doesn't go into detail) hampers the effort to strip out a single measure of moneyness:
the liquidity of many assets depends on who you are. If you're a bank, your treasury bill is more liquid than if you're an individual, cause the bank can trade it for reserves and the individual can't. 
Again, I agree with Roth. Viewed from the eyes of a drug dealer, a chequing deposit is surely much less liquid than cash, while from the eyes of a typical nine-to-fiver, the opposite would be the case.

However, the charge of subjectivity shouldn't preclude us from extracting a market price for liquidity. After all, markets provide prices for diverse consumption goods like wheelchairs, which though an integral part of the life of an eighty-year old, from the perspective of a healthy twenty year old might be worthless. Milk is off-limits for the large portion of the population that is lactose intolerant while being popular with the rest—but markets are still capable of spitting out one price for milk. A particular good's liquidity, like a wheelchair or a carton of milk, is a consumption good the utility of which varies from individual to individual, yet in a competitive market these varying preferences should nevertheless interact together to create one market clearing price for these goods.

What follows is basic microeconomics. We can construct an individual's demand curve for x's liquidity by querying how much he or she would be willing to pay for those services at various prices. If we do this for all individuals and all assets, we can construct market demand curves for each asset's liquidity. Given a set of supply curves (supply is a totally different post), we can then submit this data to a Walrasian calculator to determine the market prices for these liquidities. These prices can be used to calculate the contribution made by liquidity to each asset's total market price, and from there we can proceed to construct the market's scale of moneyness, the asset with the most moneyness being that asset whose price is made up of the largest liquidity contribution.

The upshot is that the many differing personal scales of moneyness that Roth draws attention to can be reconciled by a market-wide moneyness scale. I hope that adequately answers Roth's points. One issue worth mentioning here is that we rarely get an opportunity to see living-breathing liquidity prices. As Nick Edmonds, who blogs here (and who should be on your reading list), points out:
I'm not sure that it is possible to extract out a market clearing price specifically for liquidity services, because it's assets that are traded not services. Each asset comes with a bundle of features yielding utility or disutility, not just the liquidity aspect.
Put differently, the difference between liquidity and a wheelchair is that liquidity doesn't stand alone as its own good but rather coexists as a service attached to an already produced good. Decomposing that service and its respective price from the rest is tricky.

Let me point out that fixed income markets do often provide accurate decompositions of the market price for liquidity. For instance, assume that FDIC-insured banks are offering chequing accounts yielding 0% and 1-year fixed term deposits yielding 2%. If we were to ask the market: "How much would I have to pay you in order for you to relinquish all rights to trade away your holdings of chequing accounts for one year?" ... the answer is 2%. So 2/100ths of the value of each chequing dollar is comprised of a 1-year liquidity return.

I've also spent some time trying to isolate the price of liquidity in equity markets, this post provides some detail.

But my best answer to Edmond's point is that this is a case of missing markets. We really don't have accurate prices for moneyness yet. One of the goals of this blog is to think about what these markets would look like, how you'd build them, and what they'd be useful for.


  1. I think Roth hits the nail on the head insofar as why it is so hard to "guesstimate" the size of liquidity premia for different assets because their benefits have so many sources and go in so many directions. In the simplest example, thinking of assets generally as a "consumption storage technology", many people prefer assets with high liquidity both for their ability to move them closer to current consumption ("consuming your capital" when you're laid off unexpectedly) and also move them further away from current consumption (selling your treasuries bills to buy stocks after a market collapse). A liquid asset is simultaneously dry powder, like a perpetual option on more volatile assets, or a consumption good (there is the long chain of barter for any non-pure-money assets, though, but I'm not entirely sure how important that is when thinking about this. As long as bid-ask spreads and commisions are de minimis and volume is large). There's a beautiful Borges passage from his story the Zahir that captures this concept perfectly which I'll copy below if you bear with me.

    "I reflected that there is nothing less material than money, since any coin is, in all truth, a panoply of possible futures. 'Money is abstract', I said over and over, 'money is future time'. It can be an evening just outside the city, or a Brahms melody, or maps, or chess, or coffee, or the words of Epictetus, which teach contempt of gold; it is a Proteus more changeable than the Proteus of the isle of the Pharos. It is unforeseeable time, Bergsonian time, not the hard, solid time of Islam or the Porch. Adherents of determinism deny that there is any event in the world that is 'possible', i.e., that 'might' occur; a coin symbolizes our free will."

    He's describing a pure liquidity value asset, he might as well have been talking about Bitcoin! To get more to the point though, I agree with you that I don't see why a market price would resolve the issue. I especially don't see why the subjectivity of the liquidity value has anything to do with it. It sounds like he's merely saying that the market demand curve has a non-zero slope, in which case all I can think to say is "big whoop? who cares?"

    Edmond's point is the more damning one. How would you securitize a service like liquidity? and who would the buyer be? The first question is pretty much solved already for assets traded on an exchange, simply contract not to trade in them for a given period of time. The second question, not as much. If I bend my mind a little bit I can imagine brokers, short sellers and market makers being interested in the product, but imagining these markets is difficult, and simply guessing at the magnitude of the value is hard, though thinking about what attributes of an asset push its moneyness value up or down is manageable. More volatile assets probably have a low liquidity premium, as the only reason you hold volatile assets in the first place is if you expect to own them for a while anyway as a long term investment. Bid-ask, volume. The quantity of items something is exchangeable for. Companies like Square tip the balance against the liquidity value of currency in your pocket in favor of the liquidity value of your checking account. What other things might affect the value of an assets liquidity?

    Also, a thought, many studies use option prices to study the average value of services thrown off by securities during the life of the option, such as voting rights. Could something like this work for liquidity? or would we just be comparing the liquidity of options to stocks? And sorry for the long post.

    1. Good comment.

      On modeling liquidity as an option, there was a long post at Nick Rowe's place a few year's back on the subject that has some gems;

      "How would you securitize a service like liquidity?"

      It would have to be synthetically. The owner of the stock deposits it at a clearinghouse for x years, who in turn will lend it on. The clearinghouse guarantees the owner all dividend payments, votes, and the return of the stock after x, plus extra yield to compensate for foregone liquidity.

      Sellers of liquidity would be value investors in for the long term, who wouldn't mind forgoing liquidity for a few years in return for some extra yield. The buyers, as you mentioned, would be short sellers and market makers looking to have stable inventory.

    2. JP, do you know if there is long-term private market for securities lending? I'm curious whether hedge funds can secure long-term borrow, e.g. via Berkshire (in the same way they sold 20-year puts). I've never heard of it. BTW if you look at Interactive Brokers they maintain a public list of their borrow rates per security, but I'm not sure those are liquidity services per se because they can be bought in at any time.

    3. Good question. As far as I know, there is no long-term private market for securities lending. Any idea why? I'd imagine there would be demand for this. A Buffet-style investor would love to lend long term and earn higher yield, a short seller would love to borrow long term to avoid being bought in and short squeezed. I wonder if the deep-seeded dislike of short selling is at fault for these missing markets. Thoughts?

      You're right that borrow rates aren't currently a great indicator of liquidity services, since these are immediately callable... the lender doesn't forgo any liquidity at all by lending out stock.

    4. "Good question. As far as I know, there is no long-term private market for securities lending. Any idea why? I'd imagine there would be demand for this. A Buffet-style investor would love to lend long term and earn higher yield"

      I suspect there is no natural market for long-term borrow. (1) Large asset managers hold stock for ~1.5 years on average. (2) Single stock risks are high and are managed as part of a portfolio (i.e. asset allocation). People don't even seem interested in single-stock futures (see e.g. (3) If the asset manager uses some sort of portfolio insurance, then by definition they can't give up liquidity, so in this case maybe the cost of liquidity is equivalent to an option.

      I don't ever recall Buffet writing covered calls for example (re:BRK annual report). I think he would view selling liquidity in the same camp as it is predominantly investment-driven rather than "insurance-driven" (like selling the market puts). The puts he sold required little collateral commitment so in general he is very reluctant to sell liquidity, free options and *encumber* assets.

      "I wonder if the deep-seeded dislike of short selling is at fault for these missing markets. Thoughts?"

      Dislike may not be the right word, maybe distrust that there is an information disadvantage. There is so much media attention on frauds that no asset manager wants to be caught holding one.

    5. "I suspect there is no natural market for long-term borrow. (1) Large asset managers hold stock for ~1.5 years on average."

      I suppose one could make the argument that that if the average holding period is so short, managers expect to be able to get in & out of a market very quickly, in which case market makers need a ready supply of inventory to quote tight two-markets. Long-term borrowing would secure them a trustworthy supply.

  2. I'm still thinking through this, but thought I'd ask one question now.

    Your hypothetical question on the value of relinquishing trading rights is an interesting one, but when I think about it in practice, I don't always get liquidity as the thing that is being priced. Take my house, for example. Right now, I might be very settled and not wanting to move. So the value of the right to sell may not be worth much to me. On the other hand, I might be anticipating a change of job with a move to a different area, in which case the ability to sell may be highly valuable to me. This may then mean my house ranks differently on your scale of moneyness, but is that really due to any difference in liquidity?

    btw, thank you for the reference to my blog. I find your posts on this topic very interesting, so it's nice to know you think I have something to contribute.

    1. "...but is that really due to any difference in liquidity?"

      It's not due to any physical difference in liquidity but rather a difference in the subjective value attributed to liquidity.

      We probably need to make a distinction between the physical characteristics that make up a good's liquidity and the value people place on that liquidity. Even if the physical characteristics remain constant, the value can change if an individual's situation is altered. This is exactly what we want to measure -- the subjective value that people attach to liquidity. We're not so much interested in the objective usefulness of diamonds relative to water, but the marginal value attached to these goods.

    2. OK, but what does it means then for to say that, for me, asset a is more liquid than asset b? In my example, my house might have a higher liquidity value than Treasuries, but I don't think we would say that it is actually more liquid, even if just from my perspective. Or maybe we don't actually need to know the answer to that.

    3. That depends. What do you mean by the word "actually"? Are you referring to the "actual" underlying properties and mechanisms that make some item liquid or not? Maybe my comment to Squarely Rooted below is useful. I appreciate your line of questioning, you're making me think hard about these issues.

    4. Nick,

      I think of it this way. Let's say you can only own one good (either the treasuries or a house). You're hungry and need to eat very soon or you die. Which of these two goods can you dispose of immediately at a subjectively lower sacrifice in order to obtain food? That one is to you more liquid. I suspect that typically that would be the treasury rather than the house.

      If you only own a house, and need to eat, you're basically screwed, because you can't sell it at a good price quickly enough (and there are other costs like appraisals, notaries, taxes, ...).

    5. I think that's why I'm not entirely convinced by JPK's test. I think it tells us something useful, but I've a feeling that it's picking up things other than liquidity as well. But I find it hard to put my finger on why.

      I thought about this a bit in terms of the Nick Rowe type idea that the market for any good (or asset) is in fact two markets - a market for the good and a market for money. So selling my house is also buying money. It seems to me that we're interested in the buying money bit, but want to exclude thinks that might specifically drive selling the house. But, I haven't really been able to construct a good alternative test that draws this out.

  3. "Moneyness," methinks, is in practice a nexus of a few different factors; one of those factors, I'd wager, is second-order liquidity, ie, how much liquidity can you achieve with either what you have or one trade away? For example, take Roth's forty-thousand quarters. While it's true that, under most circumstances, you can't buy a car with $10,000 in quarters, its trivially easy to exchange those quarters for liquid media - paper cash, deposits - that are "car-liquid." On the other hand, the Snickers bars are a lot harder to trade for something you can use to buy a car.

    Here's a better example. At one point I considered investing in a local small-business. The funds I would have used for that were currently invested in an index fund of stocks traded on large public exchanges. One of the reasons I didn't end up investing was my concern about liquidity; ie, too much of my net wealth would have been relatively illiquid when it was in the form of equity in a local small business, as opposed to in the form of equity of a larger publically-traded business. In this case it was not just about first-order liquidity but also second-order liquidity - ie, it was much easier to get to deposits from public stock than private stock.

    Or riffing off those great stories about detergent as currency - - laundry detergent is liquid relative to cash and drugs, but not "car-liquid" - but cash can be "car-liquid." On the other hand, other products that might have superficial similarities to detergent, like air fresheners, dryer sheets, dish soaps, etc, are much harder to get to "car-liquid" in just one trade.

    Perhaps a measure of moneyness is to arbitrarily select a comparator, like insured deposits or simply cash, and create an index of how costly/difficult/complex/time-consuming it is to trade from X to that comparator; that would be a rough approximator of at least one aspect of "moneyness."

    1. Hi Squarely Rooted,

      "Perhaps a measure of moneyness is to arbitrarily select a comparator, like insured deposits or simply cash, and create an index of how costly/difficult/complex/time-consuming it is to trade from X to that comparator; that would be a rough approximator of at least one aspect of "moneyness.""

      My thoughts on this are similar to what I responded to Nick. Do we want to try and measure the physical characteristics of a good's moneyness, e.g. time required to trade relative to some comparator, perhaps by comparing bid ask spreads to see which are the narrowest? Or do we want to find out how individuals value that moneyness? That would entail trying to determine the prices people put on liquidity. I still like the second route. When economists try to understand apples, they don't go about their task by measuring their sweetness or colour. They're only interested in supply, demand, and the price of apples. It should be the same with liquidity.

      Re: first order liquidity and second order liquidity. The way I've set up the hypothetical question, all your orders of liquidity would be foregone at once, in the same that all of Roth's multiple liquidities would be void for a year. This all-or-nothing setup removes any nuances to liquidity, making it clear to the individual what foregone services they must be compensated for.

  4. Regarding quarters, snickers bars, cars and bonds – the cost of liquidity can be reduced by taking advantage of a facilitating transitive process rather than a direct one, can’t it? Using quarters to buy a car is probably a lot easier if you take the quarters to your bank first.

    A small point on the 2 per cent term deposit:

    Not all of that two percent differential is the cost of liquidity – because it’s a fixed rate, some of it is a premium for hedging interest rate risk.

    Another point – risk managers assigning “value at risk” calculations to market trading positions incorporate a “time to close” assumption for liquidity risk (i.e. time to liquidate the position), which affects the VAR calculation in proportion to the square root of time as I recall. This is a calculation methodology that can be applied across heterogeneous types of traded instruments and even generally non-traded instruments. It’s been a while since I was involved, but somebody working in that area currently may be able to shed some light on this.

    1. I noticed you were a major participant in the link I mentioned in my comment to John Hawkins. Do you remember that conversation?

      "...the cost of liquidity can be reduced by taking advantage of a facilitating transitive process rather than a direct one, can’t it?"

      Or, put differently, the liquidity return on quarters, snickers bars, cars and bonds can be increased if we introduce one more margin along which they can be exchanged--the bank deposit margin.

      Your other points are well-taken.

    2. That was five years ago - I remembered it when I saw the toothbrush option. And another commenter there reminded me of my participation in a Steve Randy Waldman post a year before that.

      Net result after 6 years – I still have a very poor intuitive feel for convenience yield.


  5. Just found this blog, very good stuff! I have been tinkering with a kind of paleo-monetary model ( in which I argue (similar to what you mention at the end) that the nominal interest rate is really the price of liquidity services of money. Of course, in practice this is usually bound up with other things, often some amount of risk premium. In the cases where the risk premium is negligible (like treasury debt) you are only measuring the difference in liquidity services between money and that debt. Since that debt is highly liquid, this underestimates the true value of the liquidity of money relative to other less-liquid goods. To have a perfect measurement you would need a market for loans where each borrower was assured of being paid back but was unable to sell the loan at any time before it matured. Still, a lot of prices are difficult to measure, obviously this doesn't mean that they don't exist.

    1. Yep, I pretty much agree with all of that. Glad you like the blog. The best way to get a perfect measure would be via a structure like a futures market. No one worries about counterparty risk when they buy or sell futures because a central clearinghouse like the CME stands in between all transactions. The contracts would be for fixed 1m, 3m, 1 year, 3 year terms etc., thereby getting you a clean measure for liquidity services over each time frame.

  6. Regarding lending of securities: I'm aware of the process of shorting a stock though I've never done it (except through options). If those loans are all callable at any time, why is it that everyone who owns a stock does not put it up for loan at all times (which would most likely drive the rates on them down to zero)? I imagine the answer to this question will reveal some type of liquidity preference but I don't know, someone enlighten me.

    1. Not every institution necessarily wants to lend out stock to earn extra revenue since the loan involves counterparty risk, and they may not feel that the return outbalances the potential for a large loss. To control for this risk, stock loans are usually collateralized with cash. Because of this, a stock loan doesn't reveal much about liquidity preference. Not only can a lender call back a stock whenever they want (so they don't forego liquidity), but they also receive as collateral an even more liquid instrument than the stock they've given up. Securities lending is a fascinating business.

    2. Yeah that's a good point (they receive an even more liquid instrument than what they give up). So I agree it says little (probably nothing) about liquidity preference.

  7. I'd suggest that liquidity is tradability, and tradability needs differing views of the future, a flexible price, and more than one bidder. If any of these are missing, then "liquidity" disappears.

    The problem with quarters is that their price is fixed at the bank. Silver dimes now carry a premium, and I am sure that you can buy a car with quarters -- provided that you discount $0.20 cents on the quarter in car terms.

    So there are no bad tradable assets, just bad prices.

    Bank reserves versus insurance company maturities versus household income utility are simply heterogenous views -- these differing utilities create trades.

    As long as the "second best" trade is close in price to the proximate trade, then it will get done. But as long as a bank is willing to pay you $0.25 on the quarter, you'll never accept a $0.20 discount for a car purchase.

  8. I missed this! Thanks so much for responding, and apologies for not reciprocating. I sort of gave up on keeping up with my RSS feeds a while ago, am relying too much on Twitter... After writing this I'll spend some time trying to catch up with your posts.

    Not sure I have much add here, but a few thoughts:

    "If we do this for all individuals and all assets, we can construct market demand curves for each asset's liquidity."

    "we can then submit this data to a Walrasian calculator"

    Some deep theoretical problems with this (starting with the practical problem that a Walrasian calculator doesn't exist). You can't just sum individual demand curves into a market demand curve. Steve Keen's gone after this at length, and there's lots of backing literature supporting his point.

    I don't know how important that is, though. Depends what one wants to do with the measure(s) one derives.

    Your idea of the market in foregone liquidity is a good one. It's hard to think of existing markets in uncollateralized foregone liquidity that could be used to back out liquidity premia.

    I suppose one measure that would be interesting to derive is overall market liquidity preference. Seems that IS/LMers, at least, could put such an empirical measure to good use, FW that's worth...

    You've probably addressed that. Time to do some reading and catch up...

    Given my rants about what "money" is, would you consider changing your blog name to Liquidityness? ;-)


    1. Liquidityness, lol. Dunno, it sounds like laundry detergent.

      The Walrasian thing isn't important at all. I just used it as a shortcut in order to keep the post from being too long. The key here is really all the foregone different liquidities.