Recent Posts

Friday, April 8, 2022

Where Are Supply/Demand Curves?

The reduction of the Fed’s balance sheet is coming, and so we are likely to be involved in debates about how much this will affect bond yields. I am in the camp that the quantities involved do not really matter (which some qualifications). I just want to explain my logic, which involves discussions about “supply and demand” curves for bonds, which vary based on time frame. Although this discussion might appear obvious, my feeling is that there are some hidden assumptions that people make about market price determination.


Since this was a popular area of debate during the QE era, I will just note three areas where central bank buying/selling of bonds will move prices.

  1. Central bank buying/selling is a signal about future rate cuts/hikes. E.g., if the central bank is ramping up bond buying with the policy rate at 0%, it is a signal that they are not thinking about rate hikes any time soon, which should lower forwards. Although this effect exists, there is no reason to expect a stable sensitivity between quantities and yield movements — they can signal just with words, and no bond transactions.

  2. During the Financial Crisis, every major bond investor was at or over credit balance sheet limits. Central bank buying of toxic assets helped reduce those balance sheet strains, and thus would presumably have a major effect on spreads. However, if we do not have such strained conditions, fundamentals (e.g., fair value of a spread equals the default premium plus a liquidity premium) can re-assert themselves.

  3. We can expect a short-term effect and/or a squeeze in a market segment, as will be discussed in this article.

Instantaneous Supply/Demand Curves

If we look at a frozen instant of time, we can get well-defined supply demand curves — at least if we assumed the bond market followed a simplified screen-based trading structure.

Under this simplification, if we take a particular instrument, we have a list of firm bid/offers ordered by price. We will assume the “lot size” is $10 million, and each bid/offer is a whole number of lots. We can then break up each order into individual lots, and so we can label them bid 1, bid 2, … bid n, offer 1, offer 2, offer m. Note that each order book is instrument specific (although if there are spread markets, we then have order books on pairs of instruments).

Let us say that a “price insensitive” seller just dumps $100 million of a bond — their limit sales price is $0. The order will be matched against the first 10 bids (bid 1 - bid 10), leaving bid #11 as the new best bid.

What happens to the “market price” of that particular bond?

  • The best offer price is unchanged — remains at offer #1.

  • The bid price changes from bid #1 to bid #11, and so the new best bid is less than or equal to the original best bid.

The thing to note is that the fact that the seller had a selling price of $0, the market price is still determined by the remaining market participants — the seller just took out the most aggressive bidders. As such, it does not make sense to think of the seller as the “marginal” market participant.

We can look at the order book and map out a well-defined supply/demand curve: based on what your price limit is, you can see how many bond lots you can buy/sell. This is the only point at which we can define a “sensible” supply-demand curve.

Aside: Hydraulic Theory of Prices

This leads into one topic that I see that underlines a lot of popular commentary around markets: prices only change because of buy/sell orders. (One of my old bosses called this theory something like “The Hydraulic Theory of Prices,” but I have never seen the phrase used elsewhere.)

The reason why we have well-defined “supply curve” is that we “froze” the order book to be in its state before the order arrived. In the days before high frequency trading, order books were relatively static, and so it “feels” like how markets work. But this ignores what happens around information releases.

When “market-moving” news arrives, the immediately reaction of everyone is to move their bids/offers in response. In markets with market makers, moving the bid/offer is very easy to do. For markets with an order book, you either need fast algorithms, or you stop trading when the news arrives. (Equity trading has rules about news releases and trading for this reason.) The bids/offers move without any transactions. This flexibility of prices is what allows expectations-based pricing models to work.

Short-Term Supply and Demand

For reasons that I will discuss in the appendix, even if we have order book-based trading, investors generally do not dump all their orders into the queue at the start of trading and call it a day. Instead, they probably have decided upon some strategy (either formal or informal), and will enter orders during the day based on market action and news.

If everyone was a value investor, we could argue that these are “implicit orders” and we could then trace out some “implied order book” with an associated supply-demand curves.

The problem is that investors are not all value investors. The most extreme counter-example would be trend followers that sell if prices fall, and buy if prices rise. In fixed income, the dreaded convexity hedgers in USD markets act this way — which helps turbo-charge moves. In such an environment, the path of prices matters, and so it is hard to define a supply/demand curve like we can for a frozen in time order book.

This “implicit order book” is deeper than what would be seen on a screen at any time, and so it may be hard to gauge the sustainable volume of daily transactions based solely upon a single order book snapshot. Nevertheless, the amount of risk that might be taken in a single day is still small relative to the total risk budget of investors, as discussed next.

Longer Term

Investment firms have risk limits for traders and portfolio managers — and if they do not, they tend to turn into ex-investment firms. If the market is smacked by a wave of large orders, there are limits to how far investors would react.

Those limitations on risk taking can be viewed as tactical — but the tactical risk limits may bear no resemblance to strategic risk limits.

  • It is fairly typical to sub-divide fixed income management to sub-teams by product. If a product class gets cheap, the fixed income portfolio can re-allocate its global risk budget to buy that product.

  • Fixed income issuers will adjust the maturities of their borrowings based on pricing. The standard mandate for treasurers is to “minimise the cost of funding over time” — which implies issuance based on rich/cheap analysis versus their expectations for the short rate (along with spread differentials).

  • Asset allocation risk budgets typically dwarf the risk budgets given to the asset class portfolio managers. For example, the bond allocation of a pension fund might be tightly constrained to follow its index/custom duration benchmark, and so bond managers cannot greatly change the DV01 of its portfolio versus benchmark. Conversely, the asset mix team can often move 10% of the portfolio into/out of fixed income.

Since these strategic shifts are necessarily slow, trend-following is not a great strategy. Instead, the considerations would be value, and so the “implicit strategic order book” would resemble that of the instantaneous order book — more buying as prices fall, selling as prices rise.

Since “bonds” compete with “cash” as an asset class, bonds need to be attractive versus the expected returns on cash. (Guess what the rate expectations theory says?)

As a final note, once we take into account this market structure, we can see that “event analysis” is a complete waste of time, and exists solely as a means for academics and central bankers to produce articles that fit their priors.

Central Bank And The Strategic Order Book

If the central bank seeks to influence bond prices with quantities of purchases, they need to overwhelm the strategic order book of all market participants. (If the central bank has a price target — like a yield cap in “yield curve control” — market participants have a price to take into account. They can then trade the targeted bonds like any other instrument that is pegged by the central bank — is the commitment to the peg credible?)

If the duration of the total fixed income market is small enough — or the central bank targets a small segment of the market, like linkers or ultra-longs — they can blow through the implicit order book and plant prices where they want. The usual question is whether the Fed purchases were large enough to do that? I don’t claim to have an answer, but my bias is that they did not. Although they previously took out the most aggressive bids, if the dispersion of views was not that great, the remaining bids may not have been much different than the aggressive ones.

Concluding Remarks

If the Fed decided to dump its holdings all at once, bond prices would collapse. That is not going to happen, and so it is much less clear what the effect of sales will be (other than the admittedly significant signalling effect).

Appendix: A Fossil’s Views on Fixed Income Market Structure

The first thing to note is that I was an analyst, not a trader, and I left finance in 2013. Although I interacted with traders, I was expected to doing important things like produce power points instead of looking over trader’s shoulders to see how they spent their day.

Even in 2013, the Canadian market was dominated by telephone trading, not screen trading. Screen trading was more advanced in the United States, and it is possible that inroads would come into the less liquid market.

I would describe telephone trading as follows. There is a screen based inter-dealer market — where “real money” investors could see pricing, but not trade. These prices are the “market prices” that are used as reference — but these are not all the transactions.

Sell side traders produced pricing runs with indicative prices that salespeople would send to buy side traders (like the ones at my ex-employer). The “indicative” qualifier is important — during a crisis, the sell side traders develop some amazing gastro-intestinal problems and are in the bathroom all day, unable to give firm prices.

Salespeople act as intermediaries between the two traders, since the negotiations are a zero sum game, and can get somewhat heated. (The usual description is that the difference between the buy and sell side is that the buy side yells “F%$# you a$$^*##!” and slams the phone down, while the sell side slams the phone down and then yells “F%$# you a$$^*##!”) (I only heard the hostilities second hand. I was only brought into these calls very rarely when there was some arcane disputes over pricing, and everyone acted polite when sensitive little analysts were in the room.)

Such trading was preferable done when the screen-based trading was quiet — since both sides used those prices as a reference. However, depending on the negotiations, the pricing on the telephone trade can be off the “market” price. Furthermore, only the two counterparties know about the transaction details — unlike public order books. The only way to know about the big transactions is via rumours.

Since I was not a trader, I am not the one to discuss trading techniques. However, it is safe to say that for anything other than a small (by fixed income trading standards) transaction, one needs a certain amount of ambiguity about intentions. The exact deal is only finalised after a period of negotiations. This is different than an (idealised) order book, where investors post binding orders with fixed sizes and prices.

For large transactions in the Canadian market (e.g., implementing an asset mix shift), the negotiations are a multi-day affair. Even though a large real money manager could transact quickly, the dealer counter-party needs to adjust their books to handle a huge slug of risk. For example, they need to source bonds to put into inventory to sell to the investor. If properly done, both sides have an incentive to keep the transaction quiet — they do not want the street front-running the transaction. Although the dealer looks like it has a huge informational advantage, if they move prices too much, the counter-party can walk away — and the dealer is stuck with the “wrong” position.

Old school transparent screen-based systems are not useful for less liquid markets where the on screen depth is not enough to absorb such large transactions. (The wider dispersion of investors in the U.S. dollar markets always allowed more screen-based trading.) Entering a buy/sell order (that will not be immediately filled) is the same thing as selling an out-of-the-money put/call option with an expiry of “how soon can I cancel the order?” for a premium of $0.

I am not an options trading guru, but I think it is safe to say that you do not want to sell options for $0. From this perspective, the rise of high frequency trading/”algos” was a way to recreate the cloud of uncertainty of intentions that telephone trading had. The order books jump around, and so nothing is truly firm. Furthermore, from a management perspective, you replaced the reliance on the fuzzily-defined notion of a trader’s “trading competence” with an allegedly measurable algorithm profitability. It is also a lot easier to hire programmers than experienced traders.

Email subscription: Go to 

(c) Brian Romanchuk 2022

1 comment:

Note: Posts are manually moderated, with a varying delay. Some disappear.

The comment section here is largely dead. My Substack or Twitter are better places to have a conversation.

Given that this is largely a backup way to reach me, I am going to reject posts that annoy me. Please post lengthy essays elsewhere.