Why does repo exist?

Today's dumb question from the back of the Finance class. (But I would guess some other students might not know the answer either, and some maybe hadn't even thought of the question).

[Update: just to be explicit, I am not asking why lenders want security for loans. I am asking why I don't sell my watch instead of pawning my watch.]

I want to borrow $80 for one month. I have a watch worth $100. I go to the pawnbroker, hand over my watch as security, and borrow $80. I promise to repay the $80 plus interest next month, and the pawnbroker promises to give me back my watch if I do this.

That's like a "repo", which is short for "sale and repurchase agreement". It is as if I had sold my watch for $80, and the pawnbroker had promised to sell it back to me, and I had promised to buy it back from him, for $80 plus agreed-on interest next month. If I borrow $80 on a watch worth $100 there's a 20% "haircut". (The difference is that in a repo I get to keep wearing the watch for the month (I get the coupons on the bond) even though the pawnbroker legally owns it.)

Why don't I just sell my watch instead, then wait till next month before deciding whether to buy it back?

Why do I and the Pawnbroker choose to agree in advance on what we will do next month? Why don't we just wait and see what we will want to do next month? The future is uncertain. We usually wait to get as much information as possible before deciding what to do. We might change our minds when we get new information. If we do make promises about what we will do in the future, there must be some reason that outweighs the benefits of making that decision with better information when next month arrives.

Three possible explanations that come to my mind:

1. Maybe this particular watch has sentimental value, because it used to belong to my grandfather. It's worth $150 to me but only $100 to anyone else. So there's not a competitive market in this particular watch. If I sold it, and then wanted to buy it back, the repurchase market would be a market with bilateral monopoly. The new owner would have monopoly power, and I would have monopsony power. We would haggle over the distribution of the $50 gains to trade. That haggling would be costly, and the outcome would be uncertain. So to avoid those costs and risks, the pawnbroker and I agree on the repurchase price beforehand. By bundling the sale and repurchase together, the price doesn't matter, as long as the haircut is big enough so the pawnbroker has sufficient security.

Or maybe there's a Market for Lemons problem. Any particular watch might be a lemon (have some hidden defect). The owner will have better information than a prospective purchaser on whether the watch is a lemon. If I offer to sell my watch to the pawnbroker, because I need temporarily need cash to buy something else, he doesn't know if i really need cash or if I'm trying to get rid of a lemon. The repo eliminates the Market for Lemons problem (as long as the haircut is big enough). If I'm selling a lemon watch I'm also buying a lemon watch, because I agree to buy back the exact same watch.

Those explanations make perfect sense if I'm pawning my watch. They don't make sense if I'm pawning a Canadian or US government Treasury Bill. Tbills, for a given issuer, maturity, and "run", are fungible. They are all the same.

2. Maybe there are transactions costs of buying and selling watches. There's a spread between the bid and ask price, even though all watches are (by assumption) the same. The market-maker in watches, who quotes bid and ask prices, always puts a spread between bid and ask prices for fear he might make losses when informed traders, who have better and quicker news about things affecting the future demand or supply of watches, decide whether to buy or sell from him.

By pawning the watch, rather than selling it and buying another watch next month, both I and the pawnbroker eliminate the risk that the other is better informed than we are about whether the market price of watches represents a good buying or selling opportunity.

That explanation might conceivably work for Tbills too. Bid-ask spreads are very small, but not zero. But it's not obvious whether it works empirically. Are the transactions costs of a repo lower than on two separate trades?

3. The future price of watches is uncertain. If I know I will want to have a watch again next month, it is as if I have a short position in one future watch. If I sell my watch, I face the risk that the price of watches will be higher next month, when I buy a replacement. If I am risk-averse, I will want to cover my short position by agreeing now on a price at which I will buy a watch next month. I buy a future watch, Cash On Delivery (because I don't have the spare cash now), to cover my short position. Pawning the watch covers my short position, and eliminates the risk.

That explanation too might conceivably work for Tbills. If I have a portfolio full of Tbills, for safe income in my retirement, but I need cash now for a month, I might borrow rather than selling my Tbills. Because if I sold my Tbills there's a risk the price might be higher next month, so I would be able to buy back fewer and my retirement income would be lower. I have short position in safe retirement income that I initially have covered by my ownership of Tbills. When I sell my Tbills I now am net short again. But if I repurchase at the same time I immediately re-cover my short position.

But it's not obvious whether it works empirically. Do the people pawning their Tbills have a future need for those same Tbills for some other purpose? Are they pawning Tbills that have a considerably longer maturity than the loan, and then hanging onto those Tbill after the loan is repaid?

Those three explanations are all I can come up with. Which one of those three explanations (presumably not the first) applies to repos of Tbills? Or are there other explanations I've missed?

Why do I ask? Well partly just out of interest. But also because I think that the answer to the question "why does repo exist?" might matter for monetary policy.

1. Sometimes central banks do repos (and reverse repos, which are exactly the same only with the central bank on the other side of the deal), and sometimes they do Open Market Operations (either sales or purchases of bonds). Whether it matters whether central banks use repos or OMOs, and how it matters, might depend on why repos exist.

2. Some economists have said that there is a shortage of safe assets for repos. And some (I think) have said that central bank OMO purchases worsen that shortage of safe assets for repos. I would understand these questions better if I understood better why repos exist. (If the central bank buys bonds for cash, why do people need the bonds for repos, when they already have the cash?)

Comments

[valuethinker]

Nick

Perhaps you already understand this, but at the risk of stating the obvious.

If I lend money to RBC short term, say via commercial paper (or a demand deposit) and it goes bust tomorrow, then I am unlikely to get much back: Lehmans creditors perhaps 10 cents on the dollar?

By contrast, if I take security over a particular asset from their vault, with a haircut (the degree of overcollateralization) then if they go bust, I have claim to a specific asset and can sell it.

Thus Repos are key to liquidity in interbank markets, because of this high level of asset protection for any short term liquidity loan.

[Matt]

#3

Repo is nothing other than secured borrowing, a loan of cash secured by relatively risk-free securities securities.

If your question is "why do people do so much secured short-term borrowing," the answer I think roughly has to do with the preferences for leverage in the financial industry: returns on equity are higher when there's a cheap way to lever your portfolio.

If your question is "why do people denominate their short-term borrowing as repo" I think the answers begin with worries about the perfection of security interests and the bankrupty treatment of secured lending denominated as lending, though they also shade into things like "Repo 105" where you can for accounting purposes sometimes disguise this lending as something other than lendign.

[Luis Enrique]

was this post inspired by this Felix Salmon blog - if so, well you've already seen his answer, if not, he suggests repo exists because of the "desire on the part of people with money to lend out money but to take no credit risk while doing so".

[Alex Godofsky]

Matt: but if you've got a T-bill, why not "borrow" by selling the T-bill? If buying a T-bill is lending, selling one is borrowing, and "risk-free" borrowing at that. What does a repurchase agreement add to this?

For stocks, etc. it makes sense to me; you are gaining creditworthiness by securing your loan with them. For risk-free assets it seems like an outright sale is best.

Nick: point 3 makes sense for risky assets, if you want to borrow but still want your portfolio to have exposure to that asset (as speculation, as a hedge, or just for the higher expected return). I agree it doesn't make much sense for a risk-free asset.

[Min]

"Maybe this particular watch has sentimental value, because it used to belong to my grandfather. It's worth $150 to me but only $100 to anyone else. So there's not a competitive market in this particular watch. If I sold it, and then wanted to buy it back, the repurchase market would be a market with bilateral monopoly. The new owner would have monopoly power, and I would have monopsony power."

To have monopsony power, don't you also have to be able to afford to buy the watch for more than $100?

[Simon van Norden]

Did you see Tomura's BoC 2012 working paper on this?
[link here pdf NR]

[Simon van Norden]

...or the Monnet and Narajabad paper? [link here pdf NR]

[Nick Rowe]

valuethinker and Luis: Yep. I get that. (I've just updated the post to make it more explicit earlier). Lenders want security for a loan. But why does the borrower borrow against that security? Why not just sell the asset instead of using it as security?

Luis: yes, that Felix Salmon post did trigger my writing this. But I'm asking a different question, and it's one he doesn't ask.

Matt: "If your question is "why do people do so much secured short-term borrowing," the answer I think roughly has to do with the preferences for leverage in the financial industry: returns on equity are higher when there's a cheap way to lever your portfolio."

I think that's maybe a fourth explanation? In my #3, I pawn my watch rather than selling it because I want to reduce my risk. You are saying I want to increase my risk.

Suppose I sell my watch to buy a bike. I have sold one risky asset and bought a different risky asset.

Suppose I pawn my watch to buy a bike. I now own two risky assets -- the watch and the bike (strictly, in a true repo, I don't own the watch, but I do own the right to buy a watch at a cheap price, and that right is a risky asset.)

[Simon van Norden]

At least some people in the industry [link here NR] seem to like explanation #3, which makes sense to me.

Repo will be attractive if you want to short a particular security (i.e. speculate that its price will fall.) Repo allows me to buy the security now and sell it later at a fixed price. That allows me to turn around and resell the security at a spot price today, hoping to buy it back at a lower spot price in the future, just in time to turn around and resell it to my repo counterparty.

Because repo is often short-term, it is great for short-term speculative short positions. Interest rates are also very competitive and the market is typically liquid for major players.

Sounds like a good way to speculate against indebted european government bonds.

[Nick Rowe]

Min: "To have monopsony power, don't you also have to be able to afford to buy the watch for more than $100?"

Yep, but next month (if all goes as planned) I will be able to afford to pay $150 to buy back my watch.

Alex: good points (I think). We must be careful to define "riskiness" relative to the other risks faced by an individual. If you need to be able to tell the time for the next 10 years, then owning a watch that will last 10 years eliminates that risk, even if watch prices fluctuate. Because you have covered your "real" short position. (I'm using "real" by analogy to "real option theory".)

Simon: Thanks. No I haven't. I will do so.

[Luis Enrique]

Nick,

aha, I see.

Say I own a $100 T-bill, maybe I want to borrow to buy more T-bills. So selling my T-bill to buy a T-bill doesn't get me anywhere. If I borrow against it, I use the money I borrow to buy another T-bill, which I borrow against, use the money to buy another T-bill ... so I end up with highly leveraged exposure to T-bills. Is that right? I may be embarrassing myself in public, never having taken any finance classes.

[Louis S]

Would help to clarify where repo rates are vs. t-bill rates and other sources of funding for the banks.
At the long end of the curve, there will be a clear difference between the yield of the security and the cost of money to finance it in the repo market. So the question "why repo" is not very different from the question "why lend money for term and borrow daily to fund" - and the answer is to capture the term premium.
If the securities used as collateral are of similar maturities to the terms of the repo, this argument doesn't hold as well, unless:
1) the borrowing rate on repo is less than what the mkt makes the gov't pay on t-bills (hard to see why this could be so, except that the haircut makes repo loan a senior claim on the security, and so marginally safer than the security itself (only works in theory if there is perceived to be some risk to the security).
2) The mkt for t-bills is not as liquid as you would imagine, esp for off-the-run stuff. It is easier to just borrow against the value of the portfolio than to dump in onto the market and hope not to move prices against you in the individual securities.

[JW Mason]

This is a very good question. I don't have the answer.

Somewhat OT, but it seems to me that the other question raised by the shift from unsecured lending to repo is, why do we have private banks?

Historically, the reason we have a system of private credit money is that economic actors prefer to hold the liabilities of banks to the liabilities of ultimate borrowers, both because they safer and -- even more -- because they are liquid, they can be transferred to third parties. But if banks themselves now prefer to hold government liabilities to other banks' liabilities, it's not clear what purpose the bank-based payments system is serving.

[Jon]

Regulation. This about is rehypothetication and leverage rules.

We all know the story about how banks create money, but in the real world (futures trading OTC derivatives etc) we don't need money we need to point to an asset to secure our contracts. Repo and rehypothetication is how we create assets to do this without interacting with those monetary policy rules like reserve ratios and the monetary base..

So if you think back to the old story: ten dollars of base is deposited. Then the bank lends 9 dollars to someone else, leaves their promissory note in the vault and gives them the 9 in cash which is then deposited and so forth. Now all the bank customers go to the stock exchange and pledge the value of their accounts as collateral. Yum, we're all making money.

Now someone decides that since banks are insured by the government, this is too risky and bans the use of these accounts as collateral....

This is where repo and rehypothetication come in. In the language of repo, that banking situation is the same as taking a ten percent hair cut and having unlimited rehypothetication--except we don't need banks. And we all still get to claim we have assets when meeting our collateral requirements when really it is all the same asset. (indeed what's happened is that the piece of paper which says we can get the asset back tomorrow is itself the collateral).

The reason there was a financial panic and shortage of tbills is that the haircuts went up--which is e same as raising the reserve ratio:
[link here pdf NR]

The result was a collapse in an important money aggregate and a sudden surge in demand for base money, but the base of the pyramid is the asset like tbills. So when the CB buy tbills to expand their monetary base, they deplete the asset. Why does this matter because the currency of financial markets is not the national currency. So this is more like a devaluation game. Okay, why don't they just switch to using real base money instead of tbills? Because that's illegal, you're now in the realm of banking regulation intended to curtail speculation..

[123]

Yes, the purpose of repo is to increase risk.

"Some economists have said that there is a shortage of safe assets for repos. And some (I think) have said that central bank OMO purchases worsen that shortage of safe assets for repos. "
This is small potatoes. It is like printing money in a denominations not accepted in casinos. A tiny AS shock.

[Nick Rowe]

Simon (and Louis S): I have skimmed the first two papers you linked to. They look really good and interesting. But they raise an empirical puzzle: if the reason for repos is some variant of my #2 (transactions costs of buying and selling bonds quickly) why is it that some of the most liquid assets seem (am I wrong?) to be most widely used in repos? Why do people repo US Tbills, which are some of the most liquid assets that exist? Why isn't the repo market dominated by illiquid assets instead of US Tbills?

Louis S: "1) the borrowing rate on repo is less than what the mkt makes the gov't pay on t-bills (hard to see why this could be so, except that the haircut makes repo loan a senior claim on the security, and so marginally safer than the security itself (only works in theory if there is perceived to be some risk to the security)."

OK, so a repo is doubly safe for the lender. If the borrower goes bust, you can still sell the watch. If the watch goes bust, you can still go after the borrower.

[Matt Rognlie]

Excellent question! I've thought about this too, and a few years ago I did a brief research proposal for a class attempting to suggest some answers. I don't remember all my conclusions too well, but I was similarly puzzled by the "repo a T-bill" phenomenon, and I eventually decided:

(1) As Louis S says, the market for Treasury securities is not as liquid as you would think. There is not a single source of good evidence for this, but you can look at this paper by Griffiths, Lindley, and Winters, who find that the average bid-ask spread for off-the-run one month T-bills never fell below 4 basis points from 1991 to 2001. This doesn't sound like much as a one-off expense, but if your liquidity requirements are changing, it would cost an extraordinary amount of money to simply buy and sell the T-bills. If your average turnover is once a month, you're coughing up 48 basis points per year, which is extremely unfavorable compared to the cost of raising liquidity in the form of repo, which is basically zero.

Other sources offer similar conclusions, though unfortunately all the data seems somewhat old. Fleming and Mizrach (2009) report that the average bid-ask spread for on-the-run 2-year Treasury notes is about 1/128 of 1% of par. This is very low (slightly below one basis point), but even so it's enough to render frequent turnover completely uneconomical. And the bid-ask spread for on-the-run notes is generally far lower than the spread for off-the-run notes, which comprise the vast majority of notes outstanding at any given time.

(2) But this leaves the question of why transactions costs are so high for Treasury securities compared to repo. How can daily repo transactions cost almost nothing, while buying and selling a similarly sized portfolio of Treasuries every day would be completely ruinous? I think the answer is that the bid-ask spread on Treasury securities, like pretty much all spreads in financial markets, arises from asymmetric information. You might think that asymmetric information is pretty minor issue when it comes to a quasi-guaranteed, relatively short-term nominal asset like a T-bill or 2 year T-note... and it is, but it's not quite minor enough to eliminate the last couple basis points of spread. (The spread for a 1 month T-bill seems particularly baffling, and I'm sure it has to do with details of market microstructure far too arcane for any of us to have a clue. But after some thought it's possible to imagine some reasons for asymmetric information; key rates like LIBOR or the effective federal funds rate fluctuate somewhat from day to day based on the short-term liquidity needs of financial institutions (with the federal funds rate often going haywire on the Wednesday end of the 2-week reserve settlement period), and the clever person on the other side of the trade could be taking advantage of you. This adverse selection equilibrium is self-perpetuating, since with cheap alternatives like repo there is no particular reason for banks to be moving T-bills in massive volume every day, raising the chance that your trading partner is out to screw you.

The advantage of repo is that it is doubly, or arguably triply secured: there's the value of the security itself, the haircut, and then the creditworthiness of the repo borrower. For someone to lose money on repo, the value of the security needs to fall by more than the haircut and the repo borrower needs to fail, all within the term of the agreement. The nice thing about repo-ing Treasuries is that, barring a fiscal crisis, their prices are negatively correlated with the probability of borrower failure. A world where banks are failing is a world where there is a flight to liquidity, and anticipated short rates fall. This makes repo pretty much as close to riskless as you can imagine, with truly negligible frictions from asymmetric information.

The picture isn't so pretty when we have other, sketchier securities being used as repo collateral. There, asymmetric information might still be a problem for repo. But it is evidently even more of a problem for the securities themselves (much worse bid-ask spreads), meaning that repo is far preferable to buying and selling the securities for liquidity.

[Nick Rowe]

Matt Rognlie: Hey, it's great to see you are still alive! I miss your blogging! (I will actually read your actual comment later!)

[Josh]

Nick,

Perhaps it has something to do with mismatch between receipts and payments. For example, suppose that I am a firm. I need to make some type of purchase today (say because of payroll or a contract), but I will not receive the income receipts necessary to make that purchase until tomorrow. I hold little or no cash in my portfolio because of its zero nominal rate of return. One thing that I could do is sell my Treasuries today, make my purchase, and use the income received tomorrow to purchase Treasuries. On the other hand, I could pledge the Treasuries that I have as collateral and borrow money from another firm overnight -- i.e. a repo. I will choose the latter option when the transaction costs associated with selling and buying Treasuries exceeds the overnight interest payment on bonds (which I have to believe would be almost always).

[Nick Rowe]

Matt Rognlie: Very good comment. But it raises another question: if I would be buying and selling Tbills frequently enough that the transactions costs matter, why don't banks take over and replace the repo market? I sell my Tbills to a bank, and the bank holds the Tbill as an asset and my deposit as a liability, and pays me interest. That way I can get a 0% haircut.

Are the transactions costs of banks greater than the transactions costs of repos? Or is it regulation?

Now I need to re-read Jon's comment.

[Ritwik]

Money demand, Nick. Repos exist to meet the money demand. Repos are used to expand the liabilities of the shadow banking system. There is a huge demand for banking (and hence 'safe')liabilities that also bear interest.

Repos expand the money supply/ velocity of money without the central bank needing to. Pledged collateral is the best deal that portfolio demanders of money can get.

Without repos, the developed world would have faced the deflationary pressure of this increased money demand long ago.

[Phil Koop]

"Suppose I pawn my watch to buy a bike. I now own two risky assets"

Yes! That is one of the most important reasons for "borrowing" using repo: leverage. Suppose you spot an anomaly in the bond market: if the bond is priced too low, it is usually only by a handful of basis points - not worth getting out of bed for on a cash basis. But if you can lever up 20:1, that's a different story. This is possible because haircuts for government bonds are small. And of course, if the bond is priced too high, you want to short it and then a cash transaction is impossible. A more liquid security is better-suited for this application than a less liquid one because it has a lower haircut and smaller bid-ask: anomalies are easier to spot and to trade.

The plain old-fashioned "positive carry" - borrowing short and lending long in a rising yield curve environment - mentioned by some commentators is also an important motivation.

[Nick Rowe]

Ritwik: OK, but why not regular banks? Sell your Tbills to the bank, and the bank credits your account and you make payment by cheque?

[Phil Koop]

"why don't banks take over and replace the repo market"

Now you are talking from the lender's perspective again. For corporate depositors in the US, the repo market is the bank deposit market. Corporate deposits are not insured and therefore such depositors demand collateral. Despite this, the haircuts demanded by corporate depositors vary according to the perceived creditworthiness of the repo counterparty; repo is not in practice "informationally-insensitive." That is because corporations do not actually want to have to deal with the legal and market costs of counterparty bankruptcy.

[Phil Koop]

Here is a related example - loosely a "repo" - where borrowing a security is much more efficient than an outright sale and repurchase.

Suppose that a small Canadian hedge fund wants to put on a US equity position; that requires USD cash. Of course, the fund could buy the USD outright with CAD to fund the position, but then it would be exposed to FX risk when the position is liquidated. This could be hedged with an FX forward (or other instrument) but the hedge would be expensive, particularly in small volume and for an unknown notional amount. It is much cheaper to borrow the USD and post CAD as collateral.

[Phil Koop]

There is also an opportunity cost motivation. A bank that makes a market in bonds necessarily holds inventory, and this consumes appreciable funding capacity. The return earned by this inventory is generally not particularly attractive compared to the cost of funding. However, since bond transactions do not settle same-day, it is not necessary to have actual possession of a bond when it is sold. But the market maker would like to know the P&L of a transaction with certainty. So it can buy a bond, repo it out to recover the funding, and unwind the repo when it sells the bond.

[dwb]

a lot of repo activity is used for hedging (related to #3). If a corporation plans to issue debt, then while you cannot short bonds, you can enter into a reverse repo which has the same P/L profile as the underlying treasury.

Also, there are transaction costs - but there can also be tax and accounting related issues with selling a position and recognizing a gain or loss (and then re-entering it within 30 days).

banks not only don't want to shut the market down, they make money as intermediaries.

someone may have mentioned it, but another issue is that many bonds settle ("T+3") - three days from now whereas via FEDWIRE you can settle treasuries overnight. So theoretically you can have money that you only need to invest for say 2 days thereupon you need to *guarantee* that there is cash in your account when the other bonds settle (or vie versa: you may have sold bonds but the deal has not settled, so you can lend them out for a few days).

there are various types of repo: there is general collateral (which can be just about any high-grade liquid corporate bond or asset backed), but sometimes specific bonds trade "special" meaning there is specific demand to borrow a particular bond/note (hence you make more if you lend it out). bonds can be scarce for many reasons (its been parked away in a portfolio that does not want to sell for tax reasons, someone is short it, or there is a lot a corporate debt being issued whose reference price is that particular bond (think a large corporation issuing debt at the on-the-run ten year treasury + 100 bps).

dig out Fabozzi's handbook of fixed income (a tome) i am pretty sure it has many sections on repo.

[MP]

As others have said, I think the primary motivation for the borrower is leverage. I've never done a repo on T-Bills, but I have on much riskier assets. And the reason, every time, was to get leverage in a relatively painless way.

[Matt Rognlie]

Nick, thanks. I am very much alive, though following a few comments I need to get back to work. : )

Matt Rognlie: Very good comment. But it raises another question: if I would be buying and selling Tbills frequently enough that the transactions costs matter, why don't banks take over and replace the repo market? I sell my Tbills to a bank, and the bank holds the Tbill as an asset and my deposit as a liability, and pays me interest. That way I can get a 0% haircut.

Are the transactions costs of banks greater than the transactions costs of repos? Or is it regulation?

Isn't this essentially what repo is, except that in your example it's restricted to being between banks? Assuming that bank A and bank B reverse this transaction at the end of some predetermined period (which might just be a day, especially if the bank selling the Tbill and then repurchasing it thinks of this as a deposit), it's repo. If there is no commitment to reverse by the end of a predetermined period, and instead (for instance) bank A is just selling the T-bill to bank B for a deposit in bank B, then you have all the usual adverse selection problems of trading, which imply a bid-ask spread?

[Matt Rognlie]

By the way, I didn't mean to dismiss the other function of repo, which is basically to serve as a convenient form of collateralized borrowing. (Many, many others have mentioned this.) This may well be the main purpose of repo, and there is a definite (though not precise) analogy to other forms of collateralized lending. You borrow to buy a house rather than pay rent because there are economic reasons that make it efficient for you to own it. (Alignment of ownership and occupancy mitigates agency problems, etc.)

Similarly, you borrow to own an asset (or borrow against that asset rather than selling it when you need money) because there is something about that asset that jibes with the overall design of your portfolio. Maybe you're using it as a hedge, maybe you have good reasons to speculate on it, etc. It makes more sense for you to own that asset than for your creditor to own it (where "ownership" is meant in the sense of being the ultimate claimant to its returns; of course, part of the definition of repo is that formal ownership passes temporarily to your creditor, but this isn't the kind of ownership that usually matters for financial assets). My guess is that repo is just a convenient, regulation-friendly way to arrange a securitized lending transaction that has (from the perspective of a creditor) many of the features of a bank deposit.

What puzzled Nick and I is the use of securities with extremely stable values (e.g. short-term Treasuries) as collateral in repo transactions, because aside from maybe a few hypertechnical considerations, there isn't much reason to incur a debt in order to buy an asset with almost identical characteristics. This is where all the stuff about repo being more liquid than T-bills comes in.

This reasoning doesn't seem like it should be very relevant right now for banks, since most of them are sitting on massive piles of excess reserves, which from a bank's perspective are the most liquid asset of all. (Pretty much every transaction, if you look closely enough, is ultimately settled via Fedwire; even the alternative clearinghouses/payment networks are just mechanisms to more efficiently net out payments before eventually settling whatever remains with Fedwire.) There is an interesting implication here: if the "collateralized deposit" rationale for repo is now all that matters, we should see short-term Treasuries being used as collateral far less than before.

[Sergei]

Hm, you think too theoretically about the "problem". Repos exist because there are normally balance sheet and/or accounting constraints which do not allow for outright selling of assets. If you take liquidity book of a bank it is 99.9% booked as HtM and not 0% as you hypothesize. HtM means hold to maturity and this accounting treatment allows banks to avoid mark-to-market volatility of their holdings in the p&l statement (also on sales you do not want to show p&l). However it is still a liquidity book and therefore should be able to provide liquidity. This is where repo comes in because it is a secured lending from those who is long liquidity to those who is short liquidity, it does not require balance sheet sales of assets, can be used to liquidity management operations and allows to eliminate counterparty credit risk because lending is secured and assets are therefore ring-fenced.

[rsj]

Nick, are you asking:

1. Why do short term collateralized loans exist? Why are not all loans either long term loans or uncollateralized loans?
I think it shouldn't be hard to imagine the answer.

2. Given that there is a market for short term collateralized loans, why don't banks monopolize this market and squeeze out, for example, private sector non-financial corporate lenders? What competitive advantage would banks have to do this?

[rsj]

Or perhaps the real question is

3) Given that there is a market for short term collateralized loans, why would the loans be structured so that failure to pay allows the counterparty to "keep" the collateral (because failure to pay is equivalent to failure to buy back the collateral) as opposed to requiring the lender to undergo a foreclosure proceeding?

IIRC, muslim banking is similar to a repo, in that the lender buys the house and the borrower has a contract to gradually buy it from them for a certain price at a later date.

[Phil Koop]

@rsj

I took Nick's question to be, "why would anyone secure a loan of money with collateral that is similar to money?" The point of my CAD/USD example is that sometimes it is convenient to collateralize a loan with something that literally is money!

There are other examples where money is used as collateral, but the asset being collateralized is not liquid. I presume that Nick would not be puzzled by that - he would just invert the terms as say that the asset is the "true" collateral.

[rsj]

"why would anyone secure a loan of money with collateral that is similar to money?"

Ahh.

I would say that Treasury Bills are similar to money because they can be repo-ed so easily. Or less extreme, the ease with which they can be repo-ed is a function of how "money"-like they are.

[K]

Nick,

First, I think you underestimate what constitutes risk to a bond trader. While a 3-month t-bill might look like a riskless instrument to you, to the trader on a bank's short term funding desk, it represents an arbitrage between her funding costs and the yield on the t-bill. *All* her profits come from trading the spread between short term instruments. While you see a yield of 5%+/-0.25%, depending on Fed actions over the next couple of meetings, the traders p&l and and entire job is just the +/-0.25%. So to the people who are trading t-bills, there is nothing risk free about it. As you move out the term structure, it should be clear that there is nothing riskless about bonds for almost anyone.

Second, when a trader in a bank, broker/dealer or hedge fund (or other leveraged player) buys a bond, she will finance that purchase in whatever way is cheapest. A straightforward way is just to borrow at the institutions internal funding cost, typically libor based. Typically that is not the cheapest, though. If another trader wants to short a bill, bond or whatever, she has to reverse repo it and then sell it. But when you are short selling e.g. a 5-year bond, you really want to sell the most liquid 5-year around for the simple reason that you need to buy *that* same issue back when you want to close your short position. Particular issues can get "parked" in large portfolios and be near impossible to find, in which case a short seller could find it extremely expensive to close out their position. For this reason, short sellers strongly prefer the most liquid, "on-the-run" bond issues, and *therefore* the repo rates on those securities are significantly lower than the risk free (fed funds) rate. So if you buy such a liquid treasury bond, you don't want to fund your purchase at the general risk free rate; you want to *repo* it to someone who wants to short sell it and who therefore wants to lend you money against it at a very low rate.

I think that explains a lot of the repo market. There's also a "general collateral" (GC) market which is a repo market where the collateral is determined buy the borrower *after* the deal is made. So it's not useful for getting particular bonds to short sell, and for that reason the rate is more of a pure risk free interest rate. This is the market the Bank of Canada manipulates in OMO repos.

GC is a very important market for clearing interbank balances. I.e. if I write you a check, then your bank takes on a liability to you and my bank no longer has a liability to me. So my bank has to compensate your bank with a payment of some reserves. Seeing as my bank doesn't have excess reserves (in fact, since we're Canadian, my bank doesn't have reserves at all), it borrows those reserves from your bank (who doesn't want to hold excess reserves anyways). But, since your bank doesn't want to have credit exposure to my bank, it extends the loan against GC.

[Nick Rowe]

rsj: there's Phil's and Matt Rognlie's answer, as to what the puzzle is. But you can also think of it more abstractly: a repo is two transactions bundled into one. Which is puzzling. Why would we bundle the two transactions into one? And one of those transactions is a forward transaction.

I can understand why I might sell a watch. It's because I want the cash more than the watch, and the buyer wants the watch more than the cash. Gains from trade.

I can understand why I would buy a watch next month. It's because I will want a watch more than the cash, and the seller will want the cash more than the watch. Gains from trade again.

1. But why don't I wait until next month before buying the watch? Why do the repurchase now?

2. Why bundle the two transactions together? Why do I sell a watch and buy a future watch from the same person at the same time in one double-transaction?

If we are talking about real assets like watches, the puzzle is easily resolved. Transactions costs, and hedging against the uncertainty of future prices. Can those same things also resolve the puzzle when we are talking about short term bonds?

Actually, though my head is still not clear on this, there's a parallel here to barter vs monetary exchange. In a barter exchange I do one transaction: I swap an apple for a banana. In a monetary economy I need to do two transactions to do the same thing. First I sell my apple for money; then I sell my money for a banana. We normally explain monetary exchange by explaining why the sum of the transactions costs of those two transactions is actually less than the transactions cost of the one barter trade. Because of coincidence of wants, etc.

[Nick Rowe]

Sergei: "Hm, you think too theoretically about the "problem"."

Nah, the trouble is that other people aren't thinking theoretically (abstractly) enough! So they can't see the puzzle! ;-)

So your theory is that repo exists in order to fool the accountants? That's actually a theory that has some merit in some cases. Like that Greek swap thingy? But it can't explain all or even most repos, can it? The accountants ought to have figured it out by now. And when I pawn my watch I don't care what my accountant says about it.

[rsj]

Nick @5:09,

Why do people borrow? It is two transactions -- first borrow, then repay.

Let's assume that we are not puzzled by why people borrow.

Then there is a need to borrow against collateral because risk is reduced.

Without including risk, there is no reason to require collateral. But even if you borrow against collateral, what happens to your collateral when your borrower goes bankrupt? You have to wait in line with all the other creditors. What happens if your borrower refuses to pay? You have to sue them to get the collateral. Banks are finding out that it is not easy to part households from their homes, even though those homes are collateral. It can take a year of litigation.

With a repo, the borrower gives (or rather sells) you the collateral for less than its market value, and he buys it back later on. He is repaying a loan, but because you own the collateral outright, there is no need to charge a premium for the risk of litigation or bankruptcy. If the borrower runs away, you keep the collateral without needing to sue them for it.

[Sergei]

Nick, sorry, it is a puzzle for you and other people because you are more interested in abstract theory than in real world. So what I said is not a theory. It is real world. And it is not about fooling accountants because accountants created these rules because accountants are interested in the real world. You can theoretically out-think yourself but the point of repos is not to opportunistically trade assets but to reduce unnecessary volatility of p&l statements. And volatility of p&l statements is BAD. There is no theory about it, it is just for all practical purposes BAD. If you sell an asset, you have to book a p&l gain on it. Full point. But repos are not about trading, they are about short term liquidity management. Liquidity has a price and this price is different from prices and their changes of any other asset in the world be it financial or real.

[Ashwin]

Nick - you asked "Are they pawning Tbills that have a considerably longer maturity than the loan, and then hanging onto those Tbill after the loan is repaid?" They are - the overwhelming majority of repo transactions in the private market have a maturity between overnight and three months and the assets involved have a longer maturity. The case when both the repo and the asset expire on the same date is a much rarer product known as repo-to-maturity which is traded for entirely different reasons (usually accounting-driven).

Obviously this links in with the transaction cost argument as well - if I have a 5-year T-bond and keep buying and selling it on an overnight basis, it can get expensive quickly. But this is also a matter of flexibility - I can decide to roll my overnight repo today, not roll it tomorrow if I have some extraordinary requirements, again roll it day after and so on.

If you want the gory details and all the other reasons why the repo market exists, try Moorad Choudhry's book 'The Repo Handbook' - chapters 5,6,7 and 13 should be sufficient to put you to sleep. Worth noting that from the lender's perspective, it is more than a secured loan in that legal title actually passes over to the lender. So it is in fact better than a secured loan - as a lender, the asset belongs to you and is in your possession and can be liquidated by you. Repo of riskier assets can also carry not only an initial haircut but also a variation margining process where changes in MtM are exchanged on a regular basis. So again not just a secured loan by any means. In fact the legal docs go out of their way to ensure that the first leg is treated as a genuine sale so as to avoid getting caught up in bankruptcy proceedings.

The implications for monetary policy partly arise from the fact that they subvert reserve requirements even in normal times when we don't have excess reserves and interest on excess reserves etc - quoting from page 22 here :
"The ECB for instance applies a zero reserve requirement to repos, regardless of counterparty, whereas the Bank of England only excludes repo transactions with other banks (equating them with interbank loans). Since the United States reserve requirements impact only on transactions accounts, repo liabilities are not included."

To take an example, govt issues T-bonds, I buy them and repo them with bank for cash -> increase in money supply.
To take a more typical pre-2008 example, party A tranches some MBS into a large AAA super-senior tranche, I buy this tranche and repo them with bank for cash -> increase in money supply.
Neither case has any impact on reserves even if they were binding. Also illustrates why the textbook credit-money distinction is increasingly redundant.

The corollary for the current situation is that when the CB buys assets that are already repoable with negligible haircut, it has no impact on money supply. As a holder of the T-bond, I could have converted it to money anyway if I had so desired.

On the whole reduction in safe assets via QE argument, the logic depends upon the bonds supporting a larger quantity of repo-money via the process of rehypothecation and this ladder being taken away via QE. Absolutely possible but I'm not convinced.

[Determinant]

GC is a very important market for clearing interbank balances. I.e. if I write you a check, then your bank takes on a liability to you and my bank no longer has a liability to me. So my bank has to compensate your bank with a payment of some reserves. Seeing as my bank doesn't have excess reserves (in fact, since we're Canadian, my bank doesn't have reserves at all), it borrows those reserves from your bank (who doesn't want to hold excess reserves anyways). But, since your bank doesn't want to have credit exposure to my bank, it extends the loan against GC.

That's not the way the Canadian Payments Association works, AIUI. Cheques generally are cleared in ACSS by netting transactions at the end of the day; the balance gets made up at 10AM the next day through overnight loans. In the far more important and far larger Large Value Transfer System, payments are secured through either Bank of Canada balances (Tier1 1) or against a pledged pool of assets from the banks involved (Tier 2). Still, transactions are netted and settled at the end of the day, not instantaneously. LVTS is a deferred net settlement system, not a real-time gross settlement system.

The US and Canada use different payment models due to the fact that there are much fewer players and the players are larger in proportion to the economy in Canada.

[Bob Smith]

Nick, I don't think the issue is to so much fool accountants, rather its to engage in transactions which are consistent with a particular accounting characterization. The accountants know about repos, but they don't care, because the economic reality is consistent with the accounting characterization, notwithstanding the fact that, legally, they're a sale and repurchase. You might not care about the accounting characterization of pawning your watch, but if had to preparing financial statements for your investors, you might. It also might matter if the tax results follow the accounting treatment (I believe, but stand to be corrected, that the US characterizes repos as secured loans for tax purposes).

Another explanation that was given to me(by someone who is knowledgeable about the sector - infinitely more so than I am) was that from the "lender's" perspective, a repo is perceived as being safer than a collateralized loan. In your pawnbroker example, he may have a security interest in your watch, he may have possession of the watch, but the watch isn't his. If you go bankrupt and don't repay your loan, he will probably be able to acquire title to the watch and sell it, but its potentially messy and it may take time (I don't pretend to be a finance lawyer, but intuitively that makes sense - that's why creditors, even secured creditors, spend a lot of money in bankrupcty proceedings). In contrast, if you repo your watch and if you don't repurchase it from the pawnbroker, hey, legally it's already his. If you go bankrupt, and get a stay against your creditors, not his problem, he's not a creditor (unless the value of the collateral has declined, but then only to the extent of the shortfall), he's the proud purchaser of a watch, he can sell it off the minute you fail to fulfill your obligation (in fact, he can sell it off before that time if he wants). The additional security for the creditor may be marginal, but all else being equal it might mean that you can carve a basis point or three off your cost of capital. In that case, you can see why a market for repos would emerge.

[Anonymous Quant]

The forward market for tbills is not as liquid as the spot market or the repo market. Gains from trade on the spot and in the repo market.

[Bob Smith]

"3) Given that there is a market for short term collateralized loans, why would the loans be structured so that failure to pay allows the counterparty to "keep" the collateral (because failure to pay is equivalent to failure to buy back the collateral) as opposed to requiring the lender to undergo a foreclosure proceeding?"

Following up on my last point, I think (though I don't pretend to be a bankruptcy expert either) the answer is that, the loan agreement would be subject to the applicable bankruptcy law, and the possibility of a stay of proceedings against creditors, in the event that the borrower declares bankruptcy. So long as the collateral is the property of the borrower, it would be subject (or at least potentially subject to that stay). With a repo, the creditor doesn't have to worry about the bankruptcy of the borrower, because the collateral is his property, not the borrower's.

[Nick Rowe]

Ashwin: very useful comment.

rsj: "Let's assume that we are not puzzled by why people borrow."

If I own a real asset, like a farm, it is no puzzle why I should borrow rather than selling my farm. High transactions costs of selling and buying, and I'm the best manager of that asset. But if I own a financial asset like a bond, it's more of a puzzle. Why should I be a borrower and a lender at the same time? A puzzle, not a contradiction. We are exploring the possible answers to the puzzle. Transactions costs, and different risks, and accountants, and government regulations.

[Nick Rowe]

Thanks for all the good comments everyone. I'm still mulling them over. There seems to be more than one possible explanation, and not mutually exclusive. I'm thinking of how to categorise them, and their implications for monetary policy.

At the moment, I'm still thinking in terms of two broad categories: transactions cost explanations (my #2); and risk-shifting explanations (my #3, only broader). I see the first as more "money", and the second as more "finance".

[Phil Koop]

"So they can't see the puzzle! ;-)"

Yes, I must confess that's true: I can't see the puzzle. Phil the philistine once again.

Fixed terms of redemption are necessary condition of the leverage argument. When I take a position expecting to "buy low and sell high", I can never be sure about that selling high part. But at least I can be sure of my buying price - so long as there is a repurchase agreement. If I had to buy my bond back at market price, then I could never profit from my position.

[wh10]

Great post and great comments. But Nick, I don't understand your resistance to the idea of repos allowing firms to minimize earnings volatility. Do you doubt that firms like to minimize their earnings volatility?

[jt]

It seems the only reason for a Tsy repo market to exist (i.e. that resolves Nick's puzzles) is to speculate (long/short) on future interest rates.

[Nick Rowe]

Anonymous Qunat's comment leads me to the following thought:

Are there really two separate theories of repo: my #2 transactions costs; my #3 risk shifting?

Suppose there were zero transactions costs. Would we still do repos to shift risk? I don't think we would. Because instead of doing a repo, I could just as easily do two separate transactions: sell my Tbill to Tom; buy a Tbill forward from Dick. It would be a total concidence (the same coincidence of wants that is invoked to explain money vs barter) if Tom and dick were the same person, i.e. if the person who wanted to buy my Tbill also wanted to sell me a Tbill forward.

So the #3 risk shifting argument also requires transactions costs. Or else some reason why that coincidence of wants isn't really a coincidence.

[Nick Rowe]

wh10: Thanks! Definitely some of the best comments on any WCI post (definitely not the best post, except insofar as it lead to the best comments). I expect the "accounting" theory might explain *some* repos. But the comments here from the people who know the market better give explanations that have nothing to do with accounting. And the accounting theory does seem to rest on fooling people. And if an activity were so widely known you would have thought that people would have gotten wise to it by now.

[Bob Smith]

"And the accounting theory does seem to rest on fooling people". People (investors, perhaps?), maybe, but not accountants.

[Nick Rowe]

Possible opening lines for my next post on this subject?:

"The central puzzle of monetary economics is why people use two transactions when one would serve the same purpose. (Why do I sell apples for money, then turn right around and sell that money to buy bananas? Why don't I just swap my apples for bananas?) The central puzzle of repos is why people use one transaction when two would serve the same purpose. (Why don't I sell my TBill to one person, then buy a Tbill forward from another person?). The answer to both questions must be "transactions costs". But why are two transactions cheaper than one in the first case, and one transaction cheaper than two in the second?"

[Bob Smith]

"But why are two transactions cheaper than one in the first case, and one transaction cheaper than two in the second?"

Just to be clear, in each case there are two transactions, in the apple and banana world there are two transactions with three people, whereas in the repo world there are two transactions with two people.

Surely the distinction is that in your apples and bananas world there are two goods and money, while in your (hypothetical) repo world there's really only one good and money. You're not repo-ing your t-bills to buy bananas, you're repo-ing your t-bills today to buy t-bills tommorow (and have cash on hand in the meantime), and the lender is buying t-bills today to sell them tommorow. In the repo world there is always a double co-incidence of want, because there's only one good and the difference between the buyers and sellers is simply one of time preference. Sure, you could sell the t-bills to one person and repurchase them from another, but why bother, when your friendly neighbourhood banker has developed a a package deal.

[AlephBlog]

Going out on a limb here because this is not my area of highest expertise.

Repos are done because they are two transactions that get treated like one. Sell a long dated asset, with an agreement to repurchase, and it gets treated as a short term loan.

That makes a lot of difference for how the repo transaction is treated for: GAAP and Regulatory accounting, regulatory capital, Asset-Liability testing, and bankruptcy. Since the repo gets treated as a short collateralized loan:

1) It gets favorable capital treatment
2) It gets treated as a cash-like borrowing for interest-rate testing purposes, rather than a short-dated financing of a long-dated asset. Trouble is, in a crisis, with mark-to-market and falling collateral values, there is both liquidity and credit risk.
3) It doesn't get caught in bankruptcy.
4) It's cheaper than borrowing directly to hold onto the long dated asset.
5) the asset gets to stay on the balance sheet, with a corresponding entry on the liability side for repurchase transactions.
6) you have the flexibility to end the repo financing if the terms are no longer desirable, or if you want to sell the asset outright.

My main problems are the capital treatment of repos, and asset-liability treatment for financial institutions. Even safe assets being repoed are subject to low-frequency, high-severity events akin to a run on the bank, where no one wants collateral, everyone wants cash, and no one wants counterparty risk. It would be nice for the financial regulators to make the banks do a stress test to makes sure banks have enough capital, such that if the repo haircut went up by a factor of 5-10x during a general credit/liquidity crisis, the bank survives.

[Bob Smith]

Aleph: "3) It doesn't get caught in bankruptcy"

Just to be clear, that a repo doesn't get caught in bankrupty isn't a function ofthe fact that it's treated as a short term collateralized loan. That's a function of the fact that it ISN'T a short-term collateralized loan, notwithstanding that it's treated as such.

[JKH]

It’s about trading, mostly, and financing trading inventory.

And it’s about risk management, which means hedging costs as you go along. You know the original cost of the bonds, and the repo cost, so you know what your accrued cost will be at your chosen point in the future (repo maturity). You pick the future point that you want to free up your flexibility (often just overnight) to make a decision to sell outright rather than repo again.

“The future is uncertain. We usually wait to get as much information as possible before deciding what to do.”

I’ve seen that before here. That’s wrong. That’s not how risk management works. You don’t wait. You make term financing decisions now. If the world worked as you describe, there’s be no bonds or no term structure on anything in finance. Everybody would “wait” and never decide, while all interest rates clocked in continuous time.

[Bob Smith]

Nick,

To follow-up on my point, let's imagine and apple and banana world with no cash. So that's equivalent to our repo world with t-bills and cash. And let's suppose that this is a world where the banana holders want apples today, but not tommorow, while the apple holders want apples tommorow, but not today. In this case, wouldn't you expect to see reciprical transactions between apple holders and banana holders? Heck, even in the absence of any meaningful transaction cost, that's the world that will arise since one agreement will always be easier to enter into than two.

[Bob Smith]

JKH: "You know the original cost of the bonds, and the repo cost, so you know what your accrued cost will be at your chosen point in the future (repo maturity). You pick the future point that you want to free up your flexibility (often just overnight) to make a decision to sell outright rather than repo again."

Just to play devil's advocate here, surely that would also be true if you sold the bonds to one person and simultaneously entered into a forward purchase agreement for bonds with another person (I concede the point that, in practice, that isn't likely to be feasible). I think Nick's point is that economically (and legally) that isn't really much different from a repo transction, except that it's with two counter-parties rather than one. But maybe the answer is that, sure, but isn'tthe counter-party to the forward agreement going to want to hedge his position? He could do that by entering into a forward with the person to whom you sold the bonds (or some other bond holder). But if he wants to do that, why not just hedge his position with you in the first instance by offering you a repo transaction.

[Nick Rowe]

Bob: "Just to be clear, in each case there are two transactions, in the apple and banana world there are two transactions with three people, whereas in the repo world there are two transactions with two people."

I disagree.

Suppose I go to Tom, and swap a basket of my apples plus carrots, for a basket of his bananas plus dates. That's one transaction (contract), with 2 people and 4 goods.

Now suppose I go to Tom, and swap a basket of my current TBill plus my $84 next month, for a basket of his $80 today plus his Tbill next month. That's one transaction (contract), with 2 people and 4 goods. (I'm adopting the Arrow-Debreu "dated goods" approach, in which an apple to be delivered next month is not the same good as an apple to be delivered today).

"In the repo world there is always a double co-incidence of want, because there's only one good and the difference between the buyers and sellers is simply one of time preference."

The fact that trade is in dated versions of the same two goods, and that one of those goods is the medium of exchange, must indeed be part of the answer. But even then it's not obvious there will be a coincidence of wants if there are differences in time-preference. Yes, it would be obvious if there were only two time periods; just as a coincidence of wants would be obvious if there were only two goods -- apples and bananas. (In a barter economy with n goods there are (n-1)n/2 markets; in a monetary exchange economy there are (n-1) markets; and when n=2 you get the same number of markets (one) in both.)

And *somehow* (but I'm not yet sure how) this money/barter stuff must be tied in with the idea of shadow banking.

Oh, I really do find this stuff fascinating! I can't help it!

[Tim Worstall]

Hmm, slightly unconvinced. Not by hte argument, but by the examle.

Pawning something includes an option value: I can decide not to redeem my pawn.

Repo doesn't.

To really stretch things, a bit like the difference between a future and an option. "Must" and "May".

[JKH]

Bob Smith,

Supply and demand.

The counterparty to the repo has the reverse repo.

There's all sorts of institutional demand for outlets through which to invest cash on a short term basis. It's part of institutional liquidity management. Reverses are very liquid because they can be contracted to as short maturity as desired - overnight is most common. And reverse repos pay a contracted interest rate for a contracted term, so interest rate risk is taken out of the equation.

A lot of it is about the most efficient way of managing liquidity risk and interest rate risk together.

The repo borrower just has to come up with the collateral to satisfy the credit risk - but in a way that's secondary to the main purpose from either a supply or demand perspective.

The repo borrower is motivated to finance inventory.

The reverse repo lender is motivated to deploy liquidity.

Both are interested in hedging interest rate risk for those two different purposes.

[Nick Rowe]

Sorry I'm not responding to all comments. It doesn't mean they aren't good comments (they are). It just means my brain can't keep up.

[Nick Rowe]

Tim: fair point. I suppose it doesn't matter much if the haircut is big enough. (And I don't think you are stretching things at all when you talk about options. The pawn ticket is an option, not an obligation, to repurchase my watch.)

[Edeast]

I thought i showed there are n^2 markets if you allow, people to trade everything for everything. 2 dogs plus some corn for a sheep. Etc. What you are saying is accurate if you only exchange single goods with each other.

[jt]

Slightly off topic: ... is there any difference between these two sequence of events for a Tsy repo:
(1) repo shadow banking: i.e. simple collateralized loan where no vertical money is created, and
(2) leveraged purchase of a Tsy financed by a bank, then repo`d as in (1), i.e. vertical money creation
This maybe similar to a previous comment on rehypothecation, but I wonder whether it also has a bearing on Nick`s comment on the relation of (Tsy) repo`s to monetary policy.

[Edeast]

Or 2^n. I'm out of here.

[Nick Rowe]

Edeast: "What you are saying is accurate if you only exchange single goods with each other."

Correct. I had forgotten. But your n^2 can't be quite right either. E.g. if n=1 there should be 0 markets, and if n=2 there should be 1. Maybe (1/2)(n-1)^2 ??? (My bad math strikes again).

[Nick Rowe]

jt: what is "vertical money"?

[Nick Rowe]

JKH: "That’s wrong. That’s not how risk management works. You don’t wait."

Agreed. That was (roughly) my explanation #3. I buy my watch forward because I want to insure against the risk that watch prices will rise.

[jt]

Sorry, I meant horizontal ... bank money.

[Edeast]

Just wanted it to be the powerset (2^n) to make my proof of uncomputability accurate. But what you are saying makes sense, so dunno.

[Nick Rowe]

Alephblog: and it's certainly not my area of expertise either ;-) (but I'm much farther ahead now than I was yesterday).

Your explanation for repo seems based on regulations. I always have a methodological problem with theories based on regulation. I'm never quite sure how to state the counterfactual. "If instead of the existing set of regulations, the regulations were X, then we wouldn't see repos." But how would we define X in such a way that the theory was both true and interesting? Sometimes it's obvious, and sometimes it isn't.

[Nick Rowe]

Edeast: once we allow baskets (weighted averages) of goods to count as goods, I think there is an infinite number of goods. There is an infinite number of weighted averages of apples and bananas (like price indices).

jt: the terminology with which I am familiar is: outside money (central bank money); inside money (commercial bank money, that is redeemable on demand at a fixed exchange rate for outside money).

[jt]

Horizontal = inside money. Cheers.

[Nick Rowe]

Ashwin: "To take an example, govt issues T-bonds, I buy them and repo them with bank for cash -> increase in money supply.
To take a more typical pre-2008 example, party A tranches some MBS into a large AAA super-senior tranche, I buy this tranche and repo them with bank for cash -> increase in money supply."

What does "cash" mean in this context? Does it mean "demand deposit at the bank"? If it does, then is it any different from my simply *selling* the bond to the bank, with no repurchase agreement? (Unless the repo has no required reserves and the simple sale does.) This looks like the standard loans create deposits story, except that the loan is collateralised.

[Bob Smith]

Nick,

First, I suppose that there's a farfetched answer, it may well that there isn't a double coincidence of wants in the Repo market, but we don't observe the unfulfilled wants. How would we know if Repo "borrowers" enter into "synthetic" repo transactions by selling their t-bills to one party and entering into a forward with another? Maybe the repo market is just the subset of a broader group of transactions where there is a double co-incidence of wants.

More seriously, is the difference that one party (the "lender") doesn't have a time preference? Or, at least, is willing to enter into repos over a range of time periods. The double coincidence of wants, in that case, isn't all that remarkable, given that one party (the "lender") is willing to enter into transactions over a range of time periods. In that sense, we really do live in an environment with two periods, now, and any time in the future. That surely isn't that remarkable a suggestion - we see the same thing in the loan market, where lenders are prepared to enter into loans over a range of terms.

The other point is that the lender doesn't really want the t-bill, so does "double" coincidence of wants really matter? As long as it would hold it's value and is reasonably liquid, the t-bill could be anything - indeed, in practice, I don't think the "lenderL much cares about the characteristics of the particular t-bill so long as it's worth at least 102% (or whatever) of the repurchase price. The T-bill is just a means of ensuring cash tomorrow. In that sense, the repo is really a cash-today vs. cash-tomorrow transaction, indistinguishable from a secured loan (subject, to possible application of bankruptcy law).

In that sense, it isn't that different from the pawnbroker. No one wonders about the remarkable double coincidence of wants which causes pawnbrokers to be willing to acquire watches and boomboxes and what have you in the event that you fail to repay your loan. The pawnbroker doesn't want his collateral, he just wants the cash he can sell them for. Obviously, the market for his collateral isn't that liquid and is riskier, which is reflected in his interest rate, but it's conceptually the same.

[Nick Rowe]

Ashwin: "The corollary for the current situation is that when the CB buys assets that are already repoable with negligible haircut, it has no impact on money supply. As a holder of the T-bond, I could have converted it to money anyway if I had so desired."

But if there were no repos, I could also have sold it to convert it into money, by selling it. But I didn't, until the central bank did something to persuade me to sell it to the central bank and convert it into money.

[Nick Rowe]

Josh: " I will choose the latter option [repo NR] when the transaction costs associated with selling and buying Treasuries exceeds the overnight interest payment on bonds (which I have to believe would be almost always)."

Yep, but aren't there transactions costs of doing a repo too? Why does a repo have lower transactions cost then the separate sale and purchase?

[JKH]

"To take an example, govt issues T-bonds, I buy them and repo them with bank for cash -> increase in money supply."

That's not correct.

That transaction alone only uses bank reserves.

It doesn't increase the money supply - i.e. doesn't increase bank deposit liabilities.

[Michael]

Conside a bond dealer. They have inventory. This inventory is over a variety of terms, and exposes them to interest rate risk. This risk is large compared to the expected profit in the retail bond business. If they structure their repos to match the durations, they will hedge that risk via the repo 'shorts'.

[Jon]

Note: I didn't make it clear when I was pecking my answer away... but I'm convinced that the CB purchasing tbills is a problem. I think currency/bank reserves can substitute for the asset creation done through repos--its just more costly and not preferred.

I'm open to a discussion about CBs negative interacting with the repo market... but mostly in my first comment this morning I was mostly trying to explain the point of view that I think motivated Nick's inquiry--I don't advertise that as my own fast opinion.

[Nick Rowe]

Michael: So bond dealers "borrow" their inventory of bonds? OK, that makes sense. But the stock of bonds in dealers' inventory must be a very small percentage of the total stock of bonds, no? And if this were the only source of the repo market, we presumably couldn't talk about a 'shortage of safe assets for repos". Because presumably if the total stock of bonds shrank, dealers' desired inventories would shrink too, roughly in proportion.

JKH: I think that when Ashwin said "cash" he didn't literally mean currency or reserves.

[Nick Rowe]

Jon: "I think currency/bank reserves can substitute for the asset creation done through repos--its just more costly and not preferred."

And that is because they are treated differently by regulation? Would it be too crude an interpretation of your view to say that the whole repo market is just a way to avoid regulations on banks?

[Michael]

Sorry if this is redundant; I did not read the whole thread through. Traders generally "repo out" a bond as a source of finance for the bond itself. I buy bonds and simulataneously borrow the money for the purchase by means of repo. I get a "good" rate since government securities ("govvys") are good collateral. T-bill collateral is both high credit quality and short dated, so there is small default risk and limited price volatility. Therfore, the "haircut" will be small and I can borrow almost the entire amount needed to fund the purchase. Further, certain "bonds", at least in the days before massive government deficits deficits and enormous issuance sizes, were systematically scarce. For example, traders shorted the US 10 year note as a hedge to MBS inventory, corporate bond inventory, and many other products. So these shorts always needed to borrow the govvy collateral and it was always in short supply. As incentive to owners/ potential lenders of these "special" (to use the jargon) bonds, the shorts would offer to lend money at a cut rate. So if I wanted to buy the systematically scarce bonds, I could borrow the money at very low rates, sometimes zero (even back when money market rates were above 6%. That alone would e strong incentive to use the bonds as collateral for funding. If one needed or wanted to short those bonds, one always had to count on the pain of the "special" borrowing rate. Speculating on the ebb and flow of specialness, through term repo trades divorced from outright purchases and sales, was a big business. The repo business has a richly deserved reputation as a rough corner of the bond business.

[Jon]

"And *somehow* (but I'm not yet sure how) this money/barter stuff must be tied in with the idea of shadow banking."

Yes. http://www.imf.org/external/pubs/ft/wp/2010/wp10172.pdf

[wh10]

JKH- I don't follow your reasoning. If someone trades a Tbond with a bank, the bank gains an asset an a deposit liability. That increases the money supply. If the person withdraws the cash, the deposit turns into currency, which is still an increase in the money supply.

[wh10]

Also- I was going to say, is this not something that could be answered by asking smart, informed market participants? Some might have commented on this thread already...

[Michael]

Gosh, a better answer to the question: Generally, traders sell bonds they do not own. They may sell them for speculation, hedging , or spread trading (buy bond x, sell bond y). They must deliver the bonds they sell short the next day at least by US convention), and so must borrow them back. That is why they do (reverse) repo. Old saying that works if the sayer is from Tennessee: "He who sells what isn't his'n buys it back or goes to prison." But he can borrow it for a while :)

[JKH]

wh,

"govt issues T-bonds, I buy them"

[JKH]

Nick,

"govt issues T-bonds, I buy them"

that alone uses reserves; that's how bonds are settled

[wh10]

Ah right, I skipped the first step.

[Nick Rowe]

wh10: "Also- I was going to say, is this not something that could be answered by asking smart, informed market participants? Some might have commented on this thread already..."

yes and no. Take insurance for example. Why do I bet that my house will burn down? Because at the market price, it's a good deal for me. Why does the insurance company bet that my house will not burn down? Because at the market price it's a good deal for the insurance company. That's all the market participants need to know. But it doesn't really answer the question of why the market exists -- why it makes sense for two people to take opposite sides of the same bet. Under what conditions will the same deal be a good deal for both? You can't just ask one side of the market. Are they both risk-loving? Do they have different beliefs about the probability of my house burning down? These are all conceivable answers to why the market exists. But a better answer is risk-pooling.

[wh10]

But if we assume the govt sells bonds to spend, and they do, then repoing with a bank would be a net add to the money supply.

[K]

There are a lot of comments here, but if you want to get at the motivations of the main agents for participating in the markets, I'd reread the comments by Michael, JKH, Ashwin, and myself above (we're all basically saying the same thing). The main issue that was missed in your post is that bonds are *not* risk-free instruments, and for whatever reason (different future income/consumption streams, different preferences and expectations) some people want to be short and some people want to be long. Also, repo is almost 100% overnight.

The classical portfolio theory view is simply that as well as choosing some appropriate asset mix, some agents who are relatively less risk averse will increase their total portfolio size *beyond* their total wealth by borrowing a the risk free short rate from agents who are relatively more risk averse and therefore choose to hold some of their wealth as risk free short rate loans. The loans are risk free by virtue of the fact that they are at the short rate and collateralized by a liquid portfolio of assets.

[rsj]

K,

" The loans are risk free by virtue of the fact that they are at the short rate and collateralized by a liquid portfolio of assets."

If the central bank was not ready to discount bills, what do you think the repo market would look like? I think there is still risk in being short an overnight liability and long a 90 day liability; there is an institutional key to why the repo market is less risky.

[Too Much Fed]

Didn't lehman brothers use repo to "hide" assets off balance sheet?

[K]

rsj: " I think there is still risk in being short an overnight liability and long a 90 day liability"

Absolutely. But the the risk is almost entirely in the 90 day liability. But if we just look at GC type trades where the motivation is *purely* lending, then let's just look at the risk in the repo trade. (In a GC trade the bond borrower doesn't generally sell the collateral - they just hold it to give it back tomorrow). The only risk then is that the collateral loses more value than the haircut (and the counterparty defaults) which is extremely unlikely if the collateral is treasuries, the loan is overnight, and the haircut is adequate. So *I* don't see why the repo market would depend on government support.

The main risk, I guess, is liquidity in the treasury market. If the market breaks down like post 9/11 then overnight risk can suddenly turn into multi-day risk just as the volatility of the collateral is spiking. So I guess government could have a role in providing emergency liquidity in the bond market but it's not obvious why the same function can't be provided by the private sector.

[Jon]

If the central bank was not ready to discount bills, what do you think the repo market would look like? I think there is still risk in being short an overnight liability and long a 90 day liability; there is an institutional key to why the repo market is less risky.

There was a large repo market for asset-backed paper, none of which could be discounted by the CB under then prevailing rules. In the repo market, there are haircuts depending on risk--just like the pawn broker will lend you less than your watch is worth. So the question remains: why lend when you can sell.

For an answer, see my previous posts...