« No, you may not speak to my class | Main | Births Revisited: Births, Population and Per Capita Income »


Feed You can follow this conversation by subscribing to the comment feed for this post.

Why was it a "dumb" question, considering the discussion sparked here? Am I missing some sarcasm?

I really do not understand the argument of "fooling" people by smart/stupid accountants. What exactly is fooling about here? Your views of accounting are very strange. Accounting was created not to fool people but to inform them. When you open financial statements of a company you should know what stands behind any line there. Surely, companies can try to fool and hide information but that is another story.

So you need liquidity and you can borrow it. Your counterparty has excess liquidity but it does not want credit risk of you (and capital requirements if you are a bank).So you make a repo. For one thing this is how monetary policy in Eurozone works - ECB for all practical purposes does reverse repos with banks. In the system with OMO, the central bank does OMO but these are system-wide operations. They do not eliminate individual liquidity imbalances which have to be re-distributed in the most cost efficient manner. Repos simply provide such manner. If tomorrow somebody comes up with a more efficient way and calls it Oper then Oper will be used. Where is fooling here? Of who and by whom?

Additionally, on financial balance sheets not everything can be sold. Most, and I really mean most here, of the balance sheet can NOT be sold. There is a clear reason why it is done this way, everybody (well, in business) understands it and noone considers it fooling. The reason is that those assets are held for net interest income purposes and not for price volatility/trading purposes. That is the main business of most financial institutions. If these requirements were not clear before, the coming IFRS9 standards clearly require institutions to define and describe their business models for *each* of their businesses. These business models are audited and accounting treatment of each of them is approved. The purpose of this is transparency and not fooling.

Finally, if you classify your assets for trading purposes (that is for buying and selling) you have a completely different requirements (esp capital) than if you have those assets until maturity. Again, there is no fooling. It is all about transparency and helping people (investors) understand your balance sheet.

Nick - let me try and explain myself in a clearer manner. Let me ignore rehypothecation for now - just complicates the core concept. I'm defining money as purchasing power, the ability to buy stuff in the market - for which I need access to some form of bank liabilities.

First, let me assume binding reserve constraints a la the 50s - in such a scenario the repo works better than the sale to the bank because one has reserve requirements and the other doesn't (different regs for different CBs - ref the link in my earlier comment).

Now let me throw out reserve constraints - if the bank buys the asset from you and sells it on to another non-bank, it makes no difference. So to create purchasing power ex nihilo, the bank needs to hold onto the asset. It should be obvious why the bank prefers to lend in a near risk-free manner via overnight repo rather than buy risky assets from people. As K said, anything longer than overnight repo is a risky position for a bank (and even overnight repo in times of crisis). So when you say "This looks like the standard loans create deposits story, except that the loan is collateralised." you are right - what the repo market has done is to expand the breadth of collateral with which this works and to make the process near risk-free (thanks to margining, overnight tenor etc). If increase in purchasing power available to non-banks depended on banks buying up risky assets from non-banks we hit some pretty dramatic limits on bank risk appetite pretty quickly. Repo ensures that for all practical purposes (at least in non-crisis times) there are no such limits.

May I ask, Prof. Rowe, what your take is on the shift from a credit-based system to a repo-based system, we are experiencing nowadays?


I agree with what you/K are saying about the superiority of repo over outright purchases, but ideally this superiority is supposed to be fully captured in the haircut. Why would there be an arbitrage for the bond trader in a normal environment with no CB/ sovereign credit shocks?

My other question is - pre-2008, did the explosion in the repo market service a demand for increase in purchasing power, or was it simply servicing a demand to hold bank-like liabilities that pay more interest than actual bank liabilities? Business investment was dropping, actual consumption was being financed by mortgage equity withdrawals (MEWs) and the like, while it was only deposits with money market mutual funds (many of them actually from pension funds!)that were being serviced by the expansion of repo (both quantity and velocity).

The repo market is simply the future market for bonds. By bundling a purchase with a repo you effectively get a forward position. It's very common to immediately repo any transaction, often directly with the counterparty. So you buy and sell the bond/asset today, so that only the back leg to the repo is left.

It's like asking why the future oil market exists. It allows you to go short (sell something you don't actually own) a specific bond. You cannot have short spot position, but you can go short on a forward transaction.

This simplifies cash management (as there is often little relationship between the value of a bond and its risk) and it eliminates FX hedging requirements (as the repo is in the foreign currency which hedges most of the risk). For example you might want to but bunds, but not euros and the repo transaction achieves that for you (you might need to add a bit for the haircut).
It also applies to all kind of assets: equities, fx, etc...

In all the hedge funds I have ever worked, we never actually own the assets. Everything is always repoed out.


“step alone”

But yes after that, agreed


“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.”

Because the net bid-ask structure of repo is symmetrically neutral (sell and buy), and of an outright is asymmetrically preferred (sell or buy)


And I was going to suggest to read “k” closely as well

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

Is the econ blogosphere IN GENERAL asymmetric with respect to mutual liquidity of knowledge seeking between econ specialists and finance specialists?

I suppose that would be natural because econ at the helm of blogging

Pity, because the financial crisis is so important to econ

Repo in finance is a small part of it

But econ could use some term repo at least of finance learning, and with somewhat less avoidance and reflexive rejection of accounting logic to boot

Not to defend the post Keynesians, but their framework is motivated by the desire to repair this asymmetry, I think. And in seeking to repair it, maybe they go a bit overboard at times.


But I’m sure you won’t agree

On transaction costs on repo, I think that a general rule is that transaction cost is a function of the amount of risk transferred in the transaction. Buying or selling a bond transfer a substantial amount of risk between the parties. A repo transaction only transfers a small amount (on an overnight trade almost nothing). The value of a repo almost doesn't change (it only goes up and down with short terms interest rates, but with a very small daily volatility).

Formal future markets exists for bonds obviously, but many traders don't like the fixed terms of the product and prefer to trade OTC to the exact conditions they want.

In standard future markets, it's possible to trade month spreads (e.g. sell nov to buy december). This is functionally identical to a repo transaction. Usually such trades both trade on tighter spreads (usually about half the tick value of teh outright). These contracts are used to keep the term structure of the market in sync and to provide liquidity to teh back months...

varkanut: "Why was it a "dumb" question, considering the discussion sparked here? Am I missing some sarcasm?"

Certainly not sarcasm. Maybe, maybe, just a hint of possible irony. And the discussion it sparked was what I hoped it would spark.

It was a "dumb" question because I didn't know the answer, and I hadn't thought about the question, and I hadn't read the literature. Not my area. But I had read a lot of blogosphere (and some non-blogosphere) discussion of repos, and I can't remember reading one which gave any sort of deep explanation of why repo exists. It was like reading a lot of articles about insurance markets and not one of them saying anything about risk-pooling. If you are talking about increases and decreases in the size of the repo market, and whether it's a good or bad thing, and what policy should do about repo, it really might help the discussion to have some sort of theoretical understanding about why repo exists. How are there gains from trade? Why can't those gains from trade be accomplished some other way? What are the alternatives.

JKH: And you really ought to read this:

Suppose I had asked, in 1957, why corporate debt exists. Why don't companies finance themselves by issuing shares only? All the practical business people, and all the finance guys, and all the accountants, and all the people who were very close to the markets, would have said: "That's a dumb (or very ignorant) question. Some people are more risk-averse than others. The more risk-averse want bonds, and the less risk-averse want shares. So the firm can finance itself more cheaply by issuing both." Then in 1958 two economists, Modigliani and Miller, asked the dumb question, and showed that the standard answer didn't work. The MM theorem was based on a number of assumptions, under which the debt/equity ratio would be indeterminate. So if you wanted to explain debt/equity ratios, you needed to talk about those assumptions, and which ones were false, and how it mattered. And none of those assumptions were at all explicit in the standard explanation of debt/equity ratios.

MM saw debt/equity ratios as a puzzle. They asked the dumb question. That lead to a much better and deeper understanding of corporate finance.

Pity I'm not Modigliani or Miller. But science starts from seeing the world as a puzzle.

So let me ask this question: would your explanation of repos still work in a world of zero transactions costs? Because you don't mention transactions costs anywhere. If there are no transactions costs, isn't it a bit of a fluke that 4 goods get traded in a single transaction, while in all other transactions only 2 goods get traded? Don't you find that a puzzle, that needs to be explained? There's a whole literature in economics on tied sales, where 3 goods get traded in a single transaction (money and two other goods bundled together, where the seller refuses to unbundle the two goods).

Oh God, I do hope this comment thread doesn't degenerate into another heterodox vs mainstream thingy. Mike Norman's post, and the idiot first comment there, doesn't give me hope.

JKH: Let me put it this way: Yes, certainly armchair economists like me can learn from finance people and practitioners and accountants. And I'm doing that. But you also have to realise that sometimes our "dumb" questions aren't quite so dumb as they look, and that sometimes the "answers" to those dumb questions aren't really answers. Our job is not just to describe the world. And there is more than one way of seeing the world. And some ways of seeing the same world lead to a deeper and more general perspective than others.

I have a question. Are you surprised that the mortgage market exists?

I think there is very little difference between the repo and mortgage market. They are just two forms of secured lending... Obviously the bank doesn't own your house, but it's very close... Especially on non-recourse mortgages.

I really don't see the 3 or 4 goods. A repo is money now vs money later. That's the exchange. The transfer of the good is just a way to cancel/mitigate credit risk...

acarraro: "I have a question. Are you surprised that the mortgage market exists?"

YES! And you ought to be surprised too. And if you are not surprised/puzzled, it's because you aren't thinking deeply and widely enough about why the world is as it is. You are just taking it as given.

For example: suppose I don't have enough money to buy a house. I could rent one instead, and pool my savings with other people's savings to own shares in houses that other people live in. In a world of zero transactions costs that would lead to exactly the same allocation of resources and be equally efficient as my borrowing the money to buy a house.

Why do some people rent, and others own? (This is very much like the Modigliani Theorem). In a world of zero transactions costs everyone would be indifferent. There are transactions costs to renting (the tenants may not take good care of the house, they can't paint the walls the colour they want, etc.) But if there were only transactions costs to renting, and none for mortgages, everyone would own, with a mortgage. So there must be some transactions costs to mortgages as well.

Yes, yes, yes. It certainly is puzzling. A puzzle that can be resolved, but which reappears as puzzling again when you push it deeper.

"I really don't see the 3 or 4 goods. A repo is money now vs money later. That's the exchange. The transfer of the good is just a way to cancel/mitigate credit risk..."

Why don't I do two transactions: I sell my bond to Tom, in exchange for his money now. I then go to Dick, and exchange my promise to deliver money next month for his promise to deliver his bond next month.

ACEMAXX: "Excellent! May I ask, Prof. Rowe, what your take is on the shift from a credit-based system to a repo-based system, we are experiencing nowadays?"

Thanks! I don't know. We cannot explain why repo is increasing if we don't even know why repo exists. We cannot fully understand whether it's a good or bad thing that repo is increasing if we don't know why it is increasing, and why it exists.

(But I thought it had been decreasing again recently, from something i read somewhere?)

Felix Salmon said that repo is dangerous at the system-wide level. He linked to a Keynesian blogger who was quoting Keynes and then saying that repo just gives the illusion of liquidity, because it's liquid at the individual level but not at the aggregate level. I have a hunch they are onto something.

Why don't I sell shares in my house, instead of getting a mortgage? (Why is my house financed by debt plus the occupier's equity?)


Carleton university has an asset: land that was purchased in 1949. It is still on the books at historic cost, totally ignoring the fact that a 2012 dollar is worth a lot less than a 1949 dollar. "A dollar is a dollar is a dollar". If we sold that land at current market value, and leased it back, Carleton's balance sheet would go into an accumulated surplus. That would be a transaction motivated and explained by accounting conventions. If the accountants did inflation-adjusted accounts that motive would be removed.

Acounting takes a large number of numbers and uses conventions to boil them down into a few numbers. Different conventions give you a different bottom line number. If banks did repos under one accounting convention, and didn't do repos under a different accounting convention, then someone is being fooled somewhere, and that's the motive for repos.

Bob Smith @10.01. Very good comment. You might be onto something there.

But I didn't understand this bit: "First, I suppose that there's a farfetched answer,..."

I know you were just throwing it out there, but you never know..

Different rules for different folks. Cost minimization.

If I hold bills and can do fee-generating transactions that you cannot do (or cannot do as efficiently as me), I come to you to get cash to use in those transactions. I make money. You in turn make money lending to me, in the case in question, doing a repo transaction with me.

What's the wedge between the fees I hope to earn and the return you hope to earn? Risk. I take more risk, I get a higher return. Standard stuff. Why don't you simply cut me out and do the transactions that I do? Your rules aren't the same as mine. You do secured lending because that's where you fit in the regulatory scheme of things. (In a less regulated environment, such as 14th century Italy, reputational factors and the risk of loss of life serve a function similar to regulation.)

Why repo instead of purchase and then an unrelated sale (or sale and unrelated purchase)? Because both of us are in the business of repeated, predictable transactions. I always do transactions for fees, you always provide secured credit. If we are always going to do the same thing, then the freedom to make some different decision a month from now is not all that valuable. The reduction in transaction costs is very important when you are making a living by collecting a few basis points, over and over.

Different rules (regulations) for different institutions is only one reason that the lender and the borrower in a repo transaction do the same things over and over. Most firms do the same thing over and over. In finance, though, rules are an important factor. The prospectus creates another limit on activity, much like regulatory differences. Client relationships and reputation are yet another. Point is, if you do the same thing over and over, you don't want freedom to change you mind as much as want to reduce the cost of repeated transactions.

JKH - I think you're replying to Matt in your 5:06 AM comment, not me!

acarraro @0613 AM makes an excellent point with:
"On transaction costs on repo, I think that a general rule is that transaction cost is a function of the amount of risk transferred in the transaction. Buying or selling a bond transfer a substantial amount of risk between the parties. A repo transaction only transfers a small amount (on an overnight trade almost nothing). The value of a repo almost doesn't change (it only goes up and down with short terms interest rates, but with a very small daily volatility)."

Ritwik - I'm not sure I understand your first question. Let us take an example where a pension fund that owns 30y T-bonds wants to repo them out for three months because the pension fund actually wants to keep the economic risk of the duration in the bond. Even if the pension fund was fine with giving up the asset, the bank would charge more simply because it would have to enter into an interest rate swap to get rid of the interest-rate risk on the bond.

On pre-2008, it's hard to say what drove what. I tend to think of the first impulse as the problem of maintaining consumption in the absence of wage growth which necessitated increased leverage. Best way to lever the household while maintaining the appearance of a close-to-risk free financing of this debt is to do this via overnight financing/repo. Financiers pick up the pennies, households get increased consumption and the taxpayer eventually gets run over by the bulldozer when the bill is due.

"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"

That would not earn you any term premium, though, because the separate forward transaction would be done at the implied forward value. The point of repo is that the repurchase is done at the selling price, not at the market price. I still agree that transaction costs are a significant part of the story, though.


Point taken.

(Although I wouldn't be surprised if finance people had ventured into the MM intuition independently around the same time.)

Still, it’s a good comparison.

I’m interested in understanding the 4 goods idea.

If an investment dealer issues an unsecured promissory note instead of repo, does that not involve 4 goods by your meaning? He buys dollars spot, sells note spot; sells dollars forward; buys (redeems) note forward.

Substitute repo collateral for promissory note.

So is the 4 goods idea a particular issue for repo?

"Why does a repo have lower transactions cost then the separate sale and purchase?"

You would be price-taking (paying the bid/ask) on both legs of these separate transactions.

I don't think it's true that with no transaction cost people would be indifferent between owning and renting. You would still get a change in wealth as house prices go up and down... Obviously it's simply a transfer between people, but there is a transfer.

There is currently no market where I can exchange my promise of money for his promised of delivering a bond. Actually there is, but only for a few specific bonds. As I said above in exchange traded markets you get the 3 transaction (spot, forward and spread/repo). But this requires 3 prices instead of two. There is an arbitrage between the 3 prices, so the market usually gives you only two prices and you can calculate the third. In some market spot and forward are the quoted prices in others spot and repo...

A repo market is just a different way to quote a forward market, like quoting bond as a price or a yield. There is no economic difference.

Surely we don't need to explain why we want a future/forward market? To me that's just asking why we want market in general and that's a bit too basic...

I really don't see how repos are dangerous frankly. They are just secured loans. We had secured loans for ages. It's probably mostly about technology. It's easier to keep track of the assets than it was in the past so the actual transfer of ownership is the best way to mitigate credit risk. I am not sure what you mean about repo being liquid. The ability to repo is very variable. The credit risk on repo is small as long as the collateral is very good or you have a high haircut. If collateral is worthless, than you get full credit risk on counterparty (and at a bad time, since you know they just lost money of the asset they pledge to you)...

Phil: "That would not earn you any term premium, though, because the separate forward transaction would be done at the implied forward value. The point of repo is that the repurchase is done at the selling price, not at the market price."

You lost me there. Could you explain that bit more slowly please.

JKH: sorry for being a bit antsy. That Mike Norman post and comment pissed me off a bit, especially after all the angry responses to my first year text post from permanently indignant people with reading comprehension problems.

"If an investment dealer issues an unsecured promissory note instead of repo, does that not involve 4 goods by your meaning? He buys dollars spot, sells note spot; sells dollars forward; buys (redeems) note forward."

Dunno. I don't think so. But it's the right question to ask. My head may become clearer later.

"You lost me there. Could you explain that bit more slowly please."

I think I misunderstood you - you meant to suggest a series of 1-day separate sales & repurchases, not a term forward (the point being that the repo market *is* the forward bond market.) Sorry.

Yes, Nick.

I did say "maybe they go a bit overboard at times".

(Where "bit" is rather flexible in magnitude measurement terms.)

Mind you, he did say "operational answers" in that post.

Just slouching in the direction of wondering about optimization of 2 types of input here.

Maybe it is a bit heretical, but I thought of MM as clever on paper, useless in practice. Does it actually help understand the word? I am not convinced frankly...

There is a certain tendency here to look for the one true explanation of repo, which in my view does not exist. The market has various types of agent and must be structured to satisfy all of them simultaneously. So you have:

1) Large depositors who want a secure demand deposit.
2) Market makers who want to finance inventory.
3) Long investors who want to pick up a bit of yield by lending their securities (sec lending is not economically equivalent to repo - you have to pay to lend money! - but the mechanics are the same and it is part of the same market.)
4) Arbitragers who want to take advantage of anomalies in relative prices.
5) Speculators who want to take an unhedged leveraged or short position.

The market is a jigsaw puzzle and every piece must fit. For example, the depositors who are the ultimate source of funding in the system want good collateral but don't want to be exposed to the market risk of that collateral; their requirements would not be answered by separate sale and repurchases with different counterparties.


I was referring to K's argument (perhaps this wasn't your point)that a repo transaction converts a 30 day t-bill into an overnight loan. Ideally, the superior return and risk on the 30-day t-bill ought to have been captured in the haircut. So why should you do the repo? Is it simply that the credit risk is being converted into roll-over risk, with roll-over risk somehow being mispriced vs. credit risk?

(Is this 'mis-pricing' of roll-over risk the key fact that explains the existence of any bank? But I digress.)

I understand your example, where the repo exists to create an avenue for pure funding, without any associated risks. In that sense, the repo market simply exists to minimize transaction costs by combining the two transactions into one. It's the classic Fischer Black split of each bond being seen as a risk free funding + interest rate swap + credit default swap, the repo removing the need to sell the IRS + CDS together with the pure funding. I don't think this was K's point.

I don't quite agree with your characterization of the drive behind repo markets. My understanding is that households that needed to finance consumption in the face of stagnant wages were able to do so against housing equity. There is no sui generis need to convert this leverage into risk-free financing, as the lending is collateralized/ not being extended at riskfree rates anyway, and household borrowing cycles are far too long(infrequent)to explain the explosive growth of an overnight market.

Ultimately, there has to be some demand for an 'overnight' lending, i.e. demand deposits. If it makes 2% return on the side, even better. Stash 80% funds in this, gamble with the other 20%. Isn't that what most funds were doing/ are doing these days anyway?


"Are the transactions costs of a repo lower than on two separate trades?"

It's definitely not a dumb question, but the underlying assumption seems to be that you can use repo to "replicate" a position in the underlying collateral by simply repeating the repo trade, and that repo is therefore a substitute for outright ownership. This is not the case.

First, a series of overnight short rate loans, is economically not at all the same risk as a term fixed rate loan. Repo is *overwhelmingly* overnight with long term assets as collateral. So the kind of repo that is actually used in the market *cannot and is not* used as a substitute for holding economic risk in the underlying asset.

On the contrary:

Lets say you already have invested your entire wealth in various assets, but you would really like to add some 30-year treasuries to the portfolio for whatever reason. So you borrow a some money and buy some long bonds. Now you *could* get a 30 year fixed rate loan and use it to buy the bond, but then you'd be stupid since the loan and bond would be a wash economically (in reality you'd lose big time since your borrowing costs are higher than the US govt). You don't want the loan itself to be a market-risky position, so you just want to roll it at the short rate. That way the loan will never change in value. All it does is enable the position in the 30-year bond.

Your next question is how will you get the loan? The answer, of course, is that you'll use whatever collateral allows you to get it the cheapest. Maybe there's strong demand in the repo market for a 10-year bond in your portfolio (short sellers really want to sell *that* bond) today. So you use *that* bond as collateral for your repo loan which you get at a really low rate. Maybe tomorrow it will be some other bond, and you'll use that. The point is that there is *no* relationship between any one asset in your portfolio and your financing method.

So repoing an asset is used as a way to *increase*, not decrease your exposure to that asset or some other assets you want to own. This seems like the exact opposite of your hypothesis that repo is used as a substitute for selling the asset; it's used as a way to buy more.

On the other (reverse repo) side of the repo trade there are variety of closely related motivations for entering into the trade. The common principle is that someone who is lending overnight against collateral achieves relatively *less* market risk. A bank who has received net reserves during the day needs to unload those reserves and faces the choice between buying risk assets from a bank in a deficit position *or* just lending that bank the reserves using those risk assets as collateral. So the GC loan is a way for them to reduce their risk and allowing the deficit bank to increase theirs. The special bond repo trade is similar, though not as simply explained by standard portfolio theory. People want to borrow *particular* bonds because they want to be short bonds. They may be credit traders, market makers in bonds or derivatives, inflationistas, gold bugs, Republican members of congress, whatever. But shorting bonds is just another version of "wanting less market exposure." So the repo trade is *always* a case of the borrower (collateral lender) wanting to get more market exposure and the lender (collateral borrower) wanting to get less.

I can't think of *any* case in which a market participant would borrow an asset in order to gain exposure to it. On the contrary.

I guess the question from the back of the class was quite interesting.
Hope I can contribute to the debate, even with null knowledge of financial economics. I apologize in advance for any confusion in the equations which is more than likely.

1) The profit that the pawn shop will receive is S(t)*r (where I abstract from the haircut) s the spot and r the agreed net rate. Now, if you would sell your watch and buy a future(assuming homogeneous watches)the abstract counter party of this trade would receive F(t)-S(t+1). So doing a repo the pawn shop does not bear any risk(assuming no default), while doing the trade in the market the abstract counterparty is bearing the risk.

For risk averse(and profit maximazing) pawn shops F(t)-E(S(t+1))>S(t)*r and let d denote the positive number that makes the inequality an equality(similar to a risk premium), F(t)-E(S(t+1))=S(t)*r + d. The expected value of your portfolio with repo is E(S(t+1))-S(t)*(r) and the expected value of your portfolio with the market trade is S(t)+E(S(t+1))-F(t), thus (ii)Valuerepo-Valuemarket = F(t)-S(t)(1+r), then substituting(i) in to (ii) Valuerepo-Valuemarket= d + E(S(t+1))-S(t). So a repo is convenient as long as the
pseudo risk premium ('d') can compensate the expected drop in value of your asset (of course for appreciation is always convenient). I find this quite intuitive because considering that you always end up with the watch in the market process you are able to re-buy it for a lower price, whilst in the pawn shop you cannot internalize in the deal this drop. And the reverse with appreciation if it is expected the future price will be more expensive (internalizing the expectation) whilst the repo deal remains unchanged to your advantage.

2) Depending on the collateral required for a future if it is more liquid (and you precisely seek liquidity) or bigger than the haircut you will prefer to do a repo.

Congratulations on this blog, one of the best to learn economics. I hope you find this ideas useful!

PS: Shouldn't we know all this things that you ask as economists before knowing dynamic programming or Matlab?


"a repo transaction converts a 30 day t-bill into an overnight loan."

I'm not sure exactly what you want to say, but I definitely wouldn't put it that way (I hope I didn't!). If you use a t-bill to get a repo loan, then *economically* you have a t-bill *and* a repo loan. You *still* have the t-bill and are exposed to the full economic consequences of ownership (even if technically it's not yours for the next 24 hours). So you haven't exchanged it for overnight risk.

The point of using the t-bill to obtain a loan has *nothing* to do with how you feel about the t-bill. You do it because it happens to be the cheapest way of obtaining your desired financing (leverage).


To respond to your earlier question:

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.

Bond dealer inventories are trivial compared with the amount that is bought and sold, borrowed and lent, in the bond markets, by dealers, hedge funds, treasury areas, ETFs, etc They all have access to the trading and lendin markets. Bonds cycle multiple times through the clearing system on a given day. There are trillions in face value of daily trades trades for billions in bond issuance and a trivial amount of dealer "inventory". Dealers typically do not maintain an inventory of bonds at all. They just try to capture a modest spread to fulfill customer orders, and hedge customer trades in real time. If a given bond has a small issuance, it may be harder to borrow, all else equal, and the repo rate specific to that bond may reflect these supply conditions. The "special" repo rate will eventuall be low enough to induce some porfolio owner of the bond to lend it out. During the Clinton- era US surplus, when bond issuance was smaller, you often had to lend money at zero in exchange for borowing scarce 10- year notes. I have no insight at all into the question of the sufficiency of "safe bonds" in the context of the current financial crisis.

Nick: "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?"

Purchases and sales come with brokerage fees and one is subject to the bid-ask spread.

Treasury repos -- at the very least -- eliminate the bid-ask spread. Often times the haircut is near zero. Even the haircuts on structured debt products were near zero prior to the financial crisis. See Gorton and Metrick: http://research.stlouisfed.org/publications/review/10/11/Gorton.pdf


“The point of using the t-bill to obtain a loan has *nothing* to do with how you feel about the t-bill. You do it because it happens to be the cheapest way of obtaining your desired financing (leverage).”

Athough the fact that you’re financing it at all (at the margin) does say something about how you feel about the bill.

From there, it’s a choice as to the best financing method.

And in fact repo may be the only alternative that’s cheap enough to cause you to hold onto the bill in the first place.

Unless you're otherwise constrained to hold onto the bill.

acarro: MM *by itself* is indeed useless for any practical purpose, and I would much prefer an experienced practitioner over MM. MM is like a purely negative result. But MM revealed that the practitioners' understanding must rest on some implicit assumptions, that perhaps were hidden to the practitioners themselves. Maybe, by thinking about MM, and trying to figure out what those implicit assumptions are, practitioners could improve their understanding, and adjust in different circumstances, depending on cases where those assumptions were or were not true. Or maybe, they would find that their implicit assumptions were in fact always true, so they didn't have to change what they thought, and the only benefit is that they have the satisfaction of knowing that their understanding is now built on a deeper foundation.

My head hurts. I mean literally this time. Probably because I was crawling under the MX6 last night trying to replace the oil pressure sending unit. I may take a break.

Nick: If banks did repos under one accounting convention, and didn't do repos under a different accounting convention, then someone is being fooled somewhere, and that's the motive for repos

What is your understanding of why people have conventions? Looks like you call it fooling and therefore harmful.

Yes, there are different accounting rules. Some assets are held for mark-to-market purposes in trading book (say gold), some assets are held for income purposes in banking book (say loans, bonds) and some assets are just assets and simply used/consumed (amortized) in the process (say land/buildings). Accounting rules require you to define your business model for each of your business lines. Large number of numbers as per your definition? The definitions of business models allow accountants to cut this large number down to just a few numbers. But what is the problem with it?

Is it that all you are puzzled about is why everybody does NOT use market values in accounting process? Because if it was the case then most likely there would be no need and market for repos. But then your argument is completely different from the whole discussion above.

Is it all you are puzzled about?

Re the disc. with Ashwin:

Repo's allow a party to engage in maturity transformation; selling bonds for cash does not. The purpose of the repo mkt during the 2001-2007 shadow bank growth was not to finance trading positions (that earn money through market making), but to finance maturity transformation of AAA or AA rated (quasi riskless) securities.

As Carolyn Sissoko has written (Synthetic Assets blog), a system more reliant on unsecured lending is less coupled, better able to weed out lemons, and therefore more robust. The chart below gives an indication of how far some E. banks have gone in terms of encumbering their assets: in an unsecured lending market they arguably would have failed long ago. Now, if they fail the value of their collateral will also tank. This is the inherent "tight coupling" of a collateralized banking/shadow banking system.


I think that you hit the idea a while back with transaction costs and risk shifting. As Bob Smith and others have noted, there a variety of "transaction costs" including tax, accounting, and bankruptcy-related issues.

Yes, a repo is equivalent to a set of transactions (but now, sell forward, etc). Please can and do sell bonds forward. However the point you are still missing I think is that each bond is idiosyncratic. there are reasons one might need a *very specific bond* to hedge some risk. The demand for a very specific bond may be greater than supply, hence the need to create essentially a derivative transaction (a repo).

Just to clarify a point which K has also made which may not be obvious to all - the initial "seller" in a repo transaction retains all the economic benefits and risks of the underlying instrument. A repo is a "sale" only in the legal sense so as to allow the "buyer" to liquidate the asset on default and avoid having to go through bankruptcy proceedings. In the economic sense, it is secured lending (which is obviously cheaper than simple secured lending due to the legal benefits of being treated as a sale).

So obviously this is not equivalent to a sale today and a purchase tomorrow or vice versa if both are made at then-current market prices because in this case you are not economically exposed to the asset as a seller in the interim 24 hours. If I decide the market prices in advance, then it is just a secured loan with some legal benefits.

And here's a link to the book by Moorad Choudhry

^^ "(but now, sell forward, etc). Please can can do sell forward"

ugh: should be "(buy now, sell forward, etc). People can amd do sell forward"


To respond to your earlier question, "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."

Bond dealers keep trivial inventories compared to the size of debt issuance, and the size of debt issuance is trivial compared to the size of daily bond purchases and sales, borrowing and lending. The most traded bonds cycle hundreds of times through the clearing system in a single day. Big players have netting agreements that allow them to deliver/ receive only the net position for a given bond on a given day. Many players, including hedge funds, insurance companies, government agencies, corporate treasury areas, ETFs and public funds have direct or indirect access to the repo market, and can be a source of securities for lending. In fact, for some of the afore mentioned classes, repo is a cheap source of finance for their inventories.

If a particular bond has a small issuance size, then all else equal it will be harder to borrow, and the bond- specific repo rate will reflect that scarcity. The short positions in a bond (whether resulting from bond sales or uncovered bond lending) need to buy or borrow the collateral from those who own or have already borrowed it. So the shorts offer to lend money against the collateral at a low rate. In the US, the FED will induce shorts unable to cover in the repo market to buy back the bonds, in order to maintain an orderly market. A "failure to deliver" is economically like a repo at a zero rate, but brings the added odium of regulatory displeasure. The impact will be lower bond yields for the scarce bonds, all else equal, since their purchase can be financed, for some period of time, until the squeeze unwinds, for free or near free. No different than q stock squeeze.

I have no insight at all on the "scarcity of safe bonds" in the context of the current financial crisis. I am just discussing conventional repo mechanics.


"And in fact repo may be the only alternative that's cheap enough to cause you to hold onto the bill in the first place."

Agreed. Unless you're a leveraged player, you'd think you'd never own that issue. Of course, the weird thing is that real money investors buy on-the-run bonds that trade special all the time. The only justifiable reason is that they are huge and they need liquidity and "specialness" tends to coincide with liquidity for the reason that the shorts *really* value liquidity too.

But it is crazy to buy a special bond and not repo it, which is why lots of large pension funds have the ability to use repo: i.e. they repo the bond and use the proceeds to buy higher yielding short term paper. (And hopefully they are smart about it...)

Ashwin: Exactly.

I think that you hit the idea a while back with transaction costs and risk shifting. As Bob Smith and others have noted, there a variety of "transaction costs" including tax, accounting, and bankruptcy-related issues.

Yes the bankruptcy issue... In the US repos are the only kind of lending not subject to a stay upon bankrupcy. As such repos are the most senior of all possible loans. It was Volcker who got this codified as law after a bankruptcy judge treated some repos as loans instead of sales. So the law says they are sales...

This is outside the relm of govt bonds, but maybe it will add something to the discussion about why a repo market exists:

I've been in the business of purchasing super risky bonds. As a condition to making these purchases I often demanded leverage from the seller (if a bank). Repo was typically the compromise between the ideal borrowing I wanted (long term, no mark to market, cheap) and what the bank wanted (no lending against the asset at all). It was not clearly an attractive risk proposition for the bank; repo was mainly provided to entice asset buyers and shed assets.

Ashwin: "So obviously this is not equivalent to a sale today and a purchase tomorrow or vice versa if both are made at then-current market prices"

True, but it would be equivalent to a sale today and purchase tommorrow at the current market price through a forward purchase agreement entered into today - as you point out, legally, that is what a repo is (albeit with one counterparty, rather than two). In that example, the seller retains the economic exposure to the underlying bond. I wonder how the accountants would record that transaction on the books of the seller?


"repo was mainly provided to entice asset buyers and shed assets"

I think this is *always* the core function of a repo trade: to enable the lender (of cash) to reduce market exposure, and the borrower to increase it.

You'll like this Nick...

Think of a pure exchange economy. Party B wants to obtain a certain risk asset, and party A wants to get rid of it. But there's no double coincidence of wants, because B doesn't have anything of equal value that A wants in exchange, so they can't complete the trade. The only way they can complete the trade is for A to lend B the value of the asset. The loan exactly offsets the purchase, so no medium of exchange is required (just a unit of account). But since A doesn't trust B, A must hold the asset as collateral. Every day they renew the loan at the new, current rate of interest. If, one day, A decides to stop lending to B, B must immediately sell the asset back to A at the spot price on that day. So B has the full economic risk of the asset as long as the arrangement continues. A, on the other hand, has replaced the asset with savings at the risk free short rate.

Looking over the comments, I mostly see convoluted forms of, "Yeah, two and three are basically right."

Makes sense.

K, Ashwin

I think I made lots of implicit mental model assumptions, so let me try to clarify.

I understand the bit about repo simply being a collateralised loan, except even cheaper than a collateralised loan due to the legal treatment as a sale. Consider the example of t-bills for a moment. Now why does it matter whether it is the bank or the non-bank holding onto the bill - if the bank wanted its cash back tomorrow, it could easily liquidate the t-bill tomorrow. Presumably the bank really cares about the one day price risk of a t-bill. Which it has priced in by offering the repo at a cheaper rate than the corresponding secured loan. It is operating at its indifference curve, so to speak.

So the repo transaction has achieved exactly one thing - it has kept the one-day price risk of a t-bill with the non-bank rather than the bank. I was finding it difficult to imagine why different entities would evaluate about a 'money' asset like t-bill so differently, so I guess I made some implicit assumptions when putting forth my point about roll-over risk.

Say that the 30 day t-bill would have traded at 0.55% had there been no repos and the bank would have been forced to buy the t-bill. Now that we have repos, the economic risk is better allocated and the non-bank is happy to hold it only at 0.5%. Why is it accepting the 5 basis points on t-bills lesser than the bank? Presumably because it is assuming that it will be able to roll over its repo tomorrow. If the investors (banks and non-banks) were doing pure 'credit' risk pricing, t-bills would have been 0.55%. Now that the investor (non-bank)is doing a roll-over risk pricing of t-bills, she is willing to accept just 0.5%. This is the 'mis-pricing' that i was talking about. There is cheaper funding everywhere, yes, but only because credit risk assessment has been converted to a roll-over risk assessment. The 30 day loan has been converted to 30 one day loans. Note that the non-bank does not really benefit in the net. Presumably, whatever it has gained through the cheaper rate of repo borrowing, it has already lost through the lower yield on the t-bond.

The other way to posit things would be to say that the non-bank genuinely has greater utility from/ need for a t-bill, between today and tomorrow, than a bank. Hence the repo market has helped the market achieve the social optimum. That seems a bit hard to digest.

Note that things remain the same even with longer bonds/ ABS. Assume for a moment that the ABS is actually being priced 'right', so that a AAA tranche used as collateral is actually AAA. Again, the only risk that is being borne is the one-day price risk. If the AAA model is working fine, the only reason it matters whether it is the bank or the non-bank holding on the price risk is if the non-bank really derives more utility from holding the AAA asset than the bank for a day, each day. And it has already paid for that utility by accepting a lower return on the AAA asset, than if there was a counterfactual pure credit risk pricing of the asset.

One can see why securitization of the final loan would bring down interest rates through a better allocation of risk to whoever wants to hold it. But once the underlying asset has been securitized (or is already a security, like a T-bond), the repo market lowers rates all around only because repo-borrowers, whoever they are - banks or non banks - are systematically saying that daily roll-over risk is cheaper than the daily credit/price risk. The 'correct allocation' of economic risk (to the bank rather than the non-bank)over the 30 days or the 30 years does not really matter, because whether held to maturity or sold in between, the credit/price risk of the security is the same for the bank and non-bank. If the non-bank is more willing to hold this economic risk than the bank, it has also accepted a lower yield for doing so.

And that leaves the final question of - if the utility function of banks and non-banks really is different, with banks not wanting the one day price risk, why is that the utility function of the banks? Perhaps because they also have one-day liabilities - money market deposits.

Which is why I said that the market for repo arises from some combination of mispricing of roll-over risk and a greatly increased money demand. Securitization arises from the need for cheaper credit/ leverage achieved via a 'better' allocation of risk. But not repo of the securities themselves. That is pure portfolio money supply being created to match portfolio money demand.

(apologies, but skipped the comments)

Nick, I think for financial assets you are right about points 2 and 3. In gold markets, for instance, you'd rather lend or swap (ie. repo) your gold than sell it (upon the anticipation of buying it back at some future point) because you might fear that, come time to buy the gold back, the future price could be much higher, or that the gold market could be illiquid and you might not be able to buy.

Incidentally, you can also sell your gold and buy a futures contract. Selling spot and buying a futures contract is financially equivalent to swapping (repoing) gold - in both transactions you'll lock in a guaranteed price and will avoid the risk of illiquidity. So your question: why repo? is similar to the question: should I sell some asset and simultaneously buy a futures contract or sell it and take the risk of buying back at spot at some future point in time.

So, the purpose of a repo is to buy liquidity when one holds an illiquid asset. In return for a fee and the possibility of losing the asset one can obtain cash (or some other more liquid asset).

Ok, let me have one final go at clarifying what I really want to say.

There are 4 types of loan transactions

1)Uncollaterlaized loan - Has counterparty risk. The lender's valuation tree is 0 or payout. Hence, full payout is a high number, say 105 for 100 lent out.

2) Collateralized loan - No counterparty risk, only price risk. Lender's tree is payout or asset. Hence the price risk of the asset is borne by the lender. But the lender still has the asset, so full payout is a lower number, say 103 for 100 lent out.

3) Sale today, repurchase tomorrow at tomorrow's price - This is a collateralized loan + a buy-back guarantee. So full payout is an even lower number, say, 102 for 100 lent out.

4) Repo (sale today, re-purchase tomorrow at a fixed price) - Collateralized loan + buy-back guarantee + price guarantee. So the lender's valuation is payout or fixed price asset = simply payout. This is a truly risk-free loan. So full payout is the lowest of the four, say 101 for 100 lent out.

Now Nick is modelling the repo as 3, and wondering if the buy-back guarantee which is the difference between 2 and 3 is really economically beneficial to the transacting parties. He is also wondering about the existence of 2 itself, as a corollary issue.

K/Ashwin are showing that a repo is 4, not 2 or 3 (and also showing that 2 and 3 are themselves different). Which is fine.

My question is - what is really the difference between 2 and 4 which reduces the interest charged from 3% to 1%? Ashwin says that a bank does not want to hold on to a 30 year bond, and so would have to enter a rate swap to get what it really wants. It does both transactions with the same party, and hence charges only 1% not 3%.

Fine. But the bank does not have to hold on to the 30 year bond for 30 years. It does not really face the interest rate risk. The repo loan has only been extended overnight. Purchasing power has been created for only one day. Had there been no repo (and so the asset was the bank's rather than the non-bank's), the bank would have simply liquidated the position tomorrow, bringing purchasing power back to where it was (this is equivalent to a repo not being rolled over). The only risk it bears is the risk of singe-day price movements.

So my questions (and answers) are:

1) What is it that makes the bank so risk averse that it does not even want to hold on to a AAA asset, and only wants to lend for one day at a time? My answer is - demand for money market deposits.

2) Given that the repo transaction enables the one day price risk to remain where it is supposedly most efficiently held (with the non-bank rather than the bank), this will be factored into the pricing of the AAA being used for collateral. So, AAA assets are already trading at a discount to where they would have traded in a world without repo where both banks and non-banks had to hold them. The non-bank thus has a lower rate on loan achieved only through a lower yield on its asset. This part of its net benefit washes out. So any net benefit that arises to it is only because it is borrowing successively at a one-day horizon. But presumably it would have to keep borrowing every day, i.e. the system is underpricing roll-over risk.

The bank accepts peanuts due to its risk aversion. The non-bank net benefits by the under-pricing roll over risk. The borrowers at the end of the AAA asset benefit, but their main benefit was already achieved when their borrowing was securitized. Its successive repo-ing is not particularly necessary for them.

K: "I think this is *always* the core function of a repo trade: to enable the lender (of cash) to reduce market exposure, and the borrower to increase it."

I won't quibble too much with this, because it's very often true (like in my original example), but I'll quibble a little. You can repo assets completely unrelated to the lender. Lender A provides repo to buyer B who bought assets from seller C. It happens because A hopes to do business with B in the future, and that "business" may not involve buying/selling assets at all.


I think you are complicating things, and there is some stuff I don't agree with...

"2) Collateralized loan - No counterparty risk, only price risk"

Don't agree. The collateral is only in case the counterparty defaults. So there is *default contingent* price risk. If the borrower is risk free (e.g. the CB) then the loan is at the risk free rate and the presence of collateral is irrelevant.

"3) Sale today, repurchase tomorrow at tomorrow's price - This is a collateralized loan + a buy-back guarantee."

Not the way I see it. A decision to buy something tomorrow is of no economic consequence. You can change your mind any time between now and then. I don't see any loan here, or any guarantee. This is just a sale today plus, as always, decisions to be made in the future. Everything else is irrelevant.

"4) Repo (sale today, re-purchase tomorrow at a fixed price)"

Agreed, but...

" - Collateralized loan +buy-back guarantee + price guarantee."

Now I'm confused. There is no *plus* "buy-back guarantee + price guarantee." A repo is economically *exactly* equivalent to just the collateralized loan. The only difference is that under bankruptcy the priority of claims on collateral can be complicated under some circumstances, in some jurisdictions. But forget that. For our purposes, just assume that if the borrower defaults, the lender keeps the collateral, and makes a claim in bankruptcy of an amount equal to the difference between the loan amount and collateral value at the moment of default. That's how it's *supposed* to work, and in that case there is *no* difference between the 2) collateralized loan and 4) repo.

If you have a repo with a very large haircut, or equivalently, a loan with a very large amount of excess collateral, the loan and repo rates will be exactly equal to the risk free rate. So in summary, 2) and 4) are the same, and identical to 1) if the borrower is risk free. 3) is just a sale, which is nothing like the other three, which are loans.

[I'm ignoring the case where the lender chooses some special collateral. The present discussion is for the case of general, arbitrary collateral with no convenience yield serving only as security against the loan]

"what is really the difference between 2 and 4 which reduces the
interest charged from 3% to 1%?"

Ignoring said legal difficulties of actually taking collateral in some cases, none at all. The rates are 1) identical 2) equal to the risk free rate if there is enough good collateral.

"But the bank does not have to hold on to the 30 year bond for 30 years. It
does not really face the interest rate risk. The repo loan has only been extended overnight."

I don't follow. If you own a 30-year bond, you have a lot of rate risk. It's a very volatile instrument, whose value changes (in real or nominal terms) as much as most stocks. Repoing it has *zero* impact on your risk profile. As I and Ashwin have said above, repo (or borrowing against collateral) does *not* transfer the economic risk of the collateral. Repo is not about transferring collateral risk. It is about lending, and *possibly* enabling the parties to make *other* trades that do transfer the risk of the collateral assets. See MP's very illustrative comment above.

In your last few paragraphs, you seem to be considering the case of repo as part of enabling a bank to unload a troubled asset (MP's example). That's one interesting theme, but by no means the principle volume of repo. The vast majority consists of trillions of dollars daily of overnight risk-free loans from one financial institution to another. If you want to answer the question of *why*, in general, we have repo, you need to consider all of it.

Good Lord: what a lot of comments! I am late to the party and I hope I don't repeat something that's been said already - I have just been able to skim through the bunch.

Nick: I think the pawnshop analogy may mislead. The difference, as you note, is that the borrower in a repo still in effect owns the bond - or gets the returns from it, at any rate. It would be as if I pawned my guitar but still were able to play it whenever I wanted. Then why would I not sell it? Because I want a guitar to play and I don't want to tie up all my funds in the investment. I take 500 dollars of my own and buy a guitar. Then I sell it to you for 400 with an agreement to repurchase it tomorrow for, say, 405. I take the 400 dollars and do whatever I like with it. Tomorrow I see if you'll rollover; if not, I find someone else to buy it under repo, repaying you out of the proceeds. I play whenever I want and I value daily "guitar services" at north of $5/day.


I think you are not properly closing the loop in the monetary system. Lets say that C has a bank, D. As B's deposits move from A to D bank, A must transfer reserves to D to compensate them for assuming that liability. Now A is short reserves and D has excess reserves. In order to flatten their reserve positions (without changing their risk profiles), they do a GC repo trade sending the reserves back to A. Now A just has two offsetting repo trades (which only differ by the type of collateral and the rate). So A is not at the end of the repo "chain," because A borrowed the money from D who borrows the money from C (the deposit), who *was* unloading market risk.

Any time somebody buys something from someone else, there is an instantaneous chain of (hopefully) risk-free loans that propagate through the banking system from the seller to the buyer. If the buyer is purchasing a capital asset from the seller, that produces an offsetting increase in the risk-free short rate asset by the seller and an equal decrease in that asset by the buyer. That chain is effected through the financial system via repo loans. This is the sense in which I'm saying that a repo always involves a lender trying to decrease market exposure and a borrower trying to increase it. It's that the ultimate lender is *always* the seller and the ultimate borrower is *always* the buyer.

Maybe that's all trivially obvious and pedantic. I find it somehow enlightening.

I'm way behind in the comments here.

But I wanted to take another stab at the (original) question, which was "why borrow against collateral if you can just sell the collateral".

Without disagreeing with any of the offered explanations -- transaction costs are higher, accountants don't want to mark the asset to market -- I want to try to make a simpler explanation which, I believe, generalizes the others.

There is heterogeneity of beliefs as well as financing needs. Everyone who is long an asset thinks it is worth more (to them) than the market price, and everyone who is short an asset thinks it is worth less than the market price. The market price is the average of these. Here, "worth more" also includes non-price properties in the sense that the owner of the asset values the duration of the asset more than the market, or they value the risk profile more than the market.

This divergence of believes is the fundamental (e.g., non-technological) source of bid/ask spreads, as well not wanting to mark to the market.

If you are long collateral, you do not want to sell it right now for what you believe is less than its value to you because of a short term funding need. You only want to sell it when the market price is such that the market values the security more than you do.

But you may have a short term liquidity need for cash in the present. It is better for you to keep the undervalued collateral and use it as assurance against your default rather than sell it.


Yes. And the fact that you use a particular asset as collateral against a loan is completely unrelated to whether or not you'd sell that particular asset if you weren't able to get the loan. If you were going to sell some asset to get cash, you'd rank all your assets by how much you think they are undervalued, and sell the least undervalued asset, or something like that. Totally unrelated to which of your assets might be most suitable for a collateralized loan.

There is simply *no* relationship between the desire to sell a particular asset and the choice of using that asset as loan collateral. Unrelated decisions, nothing to do with each other, shouldn't ever be discussed in the same context.

If I need cash I first have a choice to make: sell something or get a loan. If I decide to sell something, maybe the vintage Ferrari is a better idea than the house. If I decide to get a loan, I'll probably use the house, and not the Ferrari, as collateral. The fact that I used the house as collateral is in no way or shape evidence that I don't want the house or would consider selling it.


Agreed. If you were forced to sell an asset to cover the loan repayment, or if you could not get a loan (say the market was not working) then you would sell the least undervalued asset first. Because that is idiosyncratic to you, there is no reason to believe that the least undervalued asset is the most suitable for collateralizing the loan.

In fact, if anything, you would expect a negative correlation, because the best collateral is stable in value whereas there is likely to be greater disagreement about assets more volatile in value.

And here again, undervalued refers to your own peculiar set of plans and needs, and not necessarily because there is a disagreement about payouts (although there may well be a disagreement).

Edmund: "Looking over the comments, I mostly see convoluted forms of, "Yeah, two and three are basically right." Makes sense."

That's the sense I get too. Except "convoluted" isn't really the right word. What they are mostly doing is giving more specific and realistic examples of my #2 and #3, which I only sketched in the most stylised way. They are putting some real-world flesh on my #2 and #3.

And you could read rsj's 4.59 as saying "Hey Nick, your #1 wasn't so daft after all, especially if you get to wear your watch even while it's in pawn (which you do in a repo, because you get the coupons on the bond). The very fact that you own your watch probably means you value it more highly than anyone else, for whatever reason (it doesn't have to be sentimental value)".

Which makes me wonder if my #1, #2, and #3 really are separate and distinct. There are really three questions: transactions costs of buying and selling watches; my valuing a watch more than the market; my valuing this particular watch more than the market.

What I am still mulling over is how those three questions are related, in the explanation of repos.

BTW, Arnold Kling gives basically(?) the same answer as Michael above. It's the bond dealers.


I think Andy's answer is quite misleading. Yes, auto dealers finance their inventory. But compared to the financing of consumer owned vehicles, dealer financing is very small. It's the same in the securities financing world. Some people want to have more assets (cars, bonds) than they can afford and some people want to hold less. The people who want to hold less provide loans to the people who want to hold more, who then take their borrowed money, buy assets with it, and use those assets as collateral against the loan. Intermediaries like car and bond dealers and their *minuscule* inventories are irrelevant in the big picture.


Ok, got it. A repo is just a collateralized loan. It increases the borrower's balance sheet, as well as the lender's. An outright sale does neither. So the 'why borrow against collateral' question reduces simply to 'why borrow', and the answer is just as trivial.

I think the dealer example is interesting because it helps clarify something about repo that 'borrowing against collateral' perhaps doesn't. The collateral does not necessarily precede the borrowing. The borrowing is needed to purchase the securities, which then serve as collateral themselves.

Secondly, while dealer inventory many be miniscule compared to the stock held by investors, dealer inventory may be turned over several times in a week, while the extra securities purchased by investors in a week may be trivial. Which is to say that from the perspective of financing, dealers may be a bigger chunk of the repo demand than the investors.

In this sense, dealer financing through repo is a sort of 'real bills' discounting - a kind of working capital financing. The term of the repo should match the average holding time of each security on the dealer's balance sheet. Investor financing through repos is liquidity transformation - they should have ideally taken out a loan/bond that's roughly in line with the term structure of the assets that they propose to hold (like how households get 30 year mortgages). The fact that investors are able to engage in maturity transformation beyond having the mandate to do so explains the explosion of repo beyond its 'socially optimum' levels, and is a consequence of the demand for money market deposits from the supply (of repo) side. Yeah, I'm sticking with the money demand story. ;)


When you say "money demand" you mean desire to hold non-interest bearing money, rather than t-bills because you might need to buy something?


I mean the desire to hold money market deposits, which are more 'money' than t-bills and yet pay interest in excess of bank accounts (and close to t-bills). Zoltan Pozsar/ Manmohan Singh have documented how a big part of this demand comes from asset managers who actually have the mandate to invest into long term securities (for cash management as well as principal protection purposes), thus inverting the entire logic of maturity transformation. I am suggesting that this money demand is the 'cause' of the desire of banks to repo-lend.


Your alternative explanations include the following wording:

# 1 “the outcome would be uncertain”

# 2 “eliminate the risk”

# 3 “the future price of watches is uncertain”

These are all variation on risk management.

BTW, the same thinking applies to the JP Morgan “London Whale”, in which everybody was asking - why didn’t JPM simply exit the position instead of trying to re-hedge it? In both cases, it’s a choice between selling the position or financing it in some sense. And you finance it if you think your expected forward position will be superior from a risk management perspective.

With reference to K’s general explanation, you cannot disengage the holding of the asset from the decision to repo it in order to raise cash for purposes that you claim to be independent of that asset in particular. This is because, if you don’t repo the asset, the financing for that asset must be traced to another source. Balance sheets must balance. So then it’s a matter of comparing those two marginal financing costs. There’s no free lunch in terms of just pretending to choose an asset from somewhere in the portfolio to repo for purposes of raising cash that you claim to be quite separate from financing the asset itself. The choices in term of the effective marginal funding cost for that particular asset must be part of the equation. Otherwise, you delusionally think that you're raising cash by repoing a “freed up” asset - when the alternative/existing financing cost for that asset might itself be more expensive than the repo opportunity. It’s a case of false “mental accounting” or false compartmentalized accounting.

BTW Arnold Kling’s explanation is a simpleton second hand version of a general operational description, based on his reading of a 30 year old textbook on the money market (I bought it 30 years ago). Does the fact that it comes from an economist make it extra noteworthy? (an economist with zero exposure to the actual market, obviously, given his reliance on a textbook resource). Nothing wrong with his observation, but since an economist said it, you pay new attention to that operational perspective?

JKH: Some of those risks refer to transactions costs and non-fungible assets, and others don't. If you are a monetary economist, like me, (and if you are looking at possible relationships between repo and the demand and supply of money) that distinction is absolutely crucial. Because in a world of zero transactions costs we wouldn't use money. And the ability of money to lower transactions costs depends a lot on money being fungible. So distinctions which might appear irrelevant to individual participants in repo markets will be very relevant to me (and vice versa). Famous "fallacies of composition" tell us that the individual's perspective doesn't always add up at the aggregate level. For example, each individual can always sell/repo his bonds, so thinks his bonds are very liquid. But it's not possible for all individuals to sell/repo their bonds at the same time.

Arnold Kling has extra credibility. He understands money/macro. Plus he worked for Fanny (or was it Freddy?). He's got both perspectives (or should have).

(BTW, did you read my (second) response in comments to Tom Hickey, and his good reply?)

Yes, Nick. Good exchange there.

In terms of the economist/philosopher fusion, maybe you’ve covered this, but here’s a follow up question:

When you ask the question “Why do repos exist?” can you answer that effectively without also answering the related question “how would they not exist?”

Meaning – how would the world look if they did not exist?

Would there be an alternative form of collaterized financing using the same asset?

Would such collateralized financing instead disappear altogether?

If it disappeared, how would holders of Treasury bonds finance those assets?

And how would the Fed operate?


So it becomes a bigger question, doesn’t it?

An almost impossible question?

But maybe you’ve covered all that because I haven’t read all comments closely.

But is the question framed as precisely as it should be?

My reading of it roughly is that your question implies a specific counterfactual by comparison – which is the open risk sale and buy back rather than the contracted sale and buy. And that would be a narrower question.

But maybe I’ve missed something bigger.


Maybe I missed this as well, but did you compare barter repo with barter open risk sale and buy back?

The risk analysis should be the same as with a monetary economy.

Don't bid ask spreads also exist in barter transactions?

And I'm not sure you can separate the issue of transaction costs from that of risk.

JKH: Good comment.

"So it becomes a bigger question, doesn’t it?" YES!

"But is the question framed as precisely as it should be?"

No. My question wasn't framed as precisely as we would want it to be. But a large part of what we are really doing when we ask a question like "Why does repo exist?" is trying to explore what those counterfactuals might be.

Why do I (and the pawnbroker) repo my watch for one month? As an alternative I could:

1. Not do any exchange at all.
2. Do an unsecured one month loan.
3. Sell my watch
4. Sell my watch and buy it back one month later.
5. Sell my watch and buy another watch one month later.
6. Sell my watch, and buy another watch forward from a different person (not the pawnbroker)
7. Do something else...

And part of asking the question is to try to figure out what all the (relevant and interesting) alternatives are.

In other words, you don't really know what the question is when you ask it, and the only way to figure it out is to ask it, play with a possible answer, and then see if that answer covers all the possible counterfactuals.

(Plus, an answer that works well if I'm repoing my watch might not work at all if I'm repoing a Tbill (and maybe vice versa), so we are trying to figure out the range of applicability of a possible answer.)

And this is why the comment thread has been both so interesting and at the same time so "all over the map". It *has to be* all over the map, just because people have very different counterfactuals (implicitly) in mind. And some people, with one counterfactual in mind, and hence one answer in mind, may think another answer makes no sense at all.

At one extreme, someone who thinks the counterfactual is "do nothing", is going to answer: "Because I need the cash, duh!"

And close to that, a person who thinks the counterfactual is "do an unsecured loan" is going to answer: "Because the lender needs security, duh!". (But even in that case, it's not quite so obvious, because sure the lender wants as much security as possible, but the borrower doesn't, otherwise all loans would have a "pound of flesh" clause attached.)


That makes sense, as a process of exploration and development of question and answer together.

So do you have a “most universal” “answer in progress” to your question at this point in time?

And one of the reasons I'm talking about pawning a watch rather than repoing a Tbill, (even though the analogy is not as good as I want it to be, because I don't get to wear the watch when it's at the pawnshop, and I have the option rather than the obligation to buy it back) is to try to push the operations people outside of their comfort zone, where everything is familiar to them, and they take it all for granted, and see if we can come up with a *general* theory of repo, as well as explore the similarities and dissimilarities between different types of exchanges. Because at one level a repo is just an exchange, and like all exchanges, the explanation to why it happens is that both parties (think they) benefit. But how can both parties benefit at the same time? Comparative advantage, different preferences, etc., there's a whole slew of economics devoted to trying to see a multitude of different exchanges as being fundamentally the same.

JKH: "That makes sense, as a process of exploration and development of question and answer together."

Very good way of putting it. Which is why it looks like a confused mess. Which it is, in one sense.

"So do you have a “most universal” “answer in progress” to your question at this point in time?"

I keep thinking I do, starting to write it up, then realising I don't. Or I haven't got the answer in its simplest and most general form. I think some sort of "transactions cost" will be a necessary part of any answer.

Maybe some day there will be a 4.00am post on the subject!


I don't disagree with what you are saying (as discussed above), but I think it's somewhat beside the point. But let me rephrase my point then, with better attention to details of "special" repo.

Assume that the portfolio is made up assets, some of which can be repoed below the risk free rate. When you buy such "special" assets, you do so with the intention of repoing them since otherwise you are getting arbitraged. In fact, you borrow against each such asset until the marginal cost of additional borrowing is equal to the risk free rate. And when you evaluate the risk/return characteristics of each asset, and decide on portfolio construction, it is in the context that the asset is financed by that level of borrowing. In that context, *additional* borrowing is at the risk free rate, generally by borrowing against non-special assets. (Or possibly above the risk free rate as you get to very small haircuts or try to repo illiquid and opaque assets). Similarly, the decision to deleverage is achieved *not* by reducing repo on special bonds, but by parking part of your assets at the risk free rate (e.g. in a GC repo).

So if, for special bonds (and stocks), you define "asset" as a combination of the security and special repo, and define the "asset" amount as just the haircut, then it is true that you can segment the portfolio construction, and leverage decisions.

And yes, that's a slight fudge, and yes there are even more technical details we can consider, but I don't think any of it casts any particular light on the fundamental reason for the existence of repo, i.e. to construct a risk-free loan from relatively risk-averse to relatively risk-seeking investors to enable each one to tune her risk asset exposure to levels above or below her total wealth (and even to negative amounts of risk asset exposure). If somebody wants to decrease their asset holdings, then (barring a wealth transfer) it's *impossible* to achieve that without:

1) somebody else increasing their asset holdings, *and*
2) the first person providing a loan to the second person.

It's a no-brainer that the loan should be collateralized, given the extra asset holdings. That's why we have repo. It's the same role that is always provided by the banking system, but just for securities.

JKH: "Maybe I missed this as well, but did you compare barter repo with barter open risk sale and buy back?"

No, I didn't. But I should. Two of the goods in a repo are: money; an promise to pay money next month. And I'm trying to figure out whether that is an essential and interesting feature of repo or just a simple corrollary of the fact that almost all exchanges are monetary rather than barter.

"Don't bid ask spreads also exist in barter transactions?"

Yep, and they are presumably bigger, which is why we use money.

"And I'm not sure you can separate the issue of transaction costs from that of risk."

Yes and no. We can have risk without transactions costs. And the sort of uncertainty that creates transactions costs (e.g. "does he know this car he's selling me is a lemon?") may be very different from market risk ("Car prices may rise or fall next month"). Market risk may motivate trade. The risk the car is a lemon reduces the amount of trade.


The 'general theory of repo' seems to me to be currently pointing at :

1) Working capital financing of dealers (which has always been on a 'real bills' and hence risk-free basis), where the short term of the repo matches the frequent inventory turnover of the dealer.

2) Maturity transformation by investors.

3) Splitting of the 'funding' and 'taking risk' portions of the act of lending, so that we get to a model of banking where the 'bank' does not need to have 'assets' and we can focus on its liabilities. You thus get the true risk-free 'time preference' rate of the economy. (I should note that many central banks call the interest they pay on excess reserves the 'reverse repo' rate, including the Reserve Bank of India)

4) Money demand, in its modern form.

I don't know if you agree, and if you are able to map this to your initial three propositions.

To your point about how both parties benefit, I think it's important to come back to what SRW recently said is the essential nature of a bank - that its liabilities are widely circulated as money even though it does not have enough capital. The ultimate repo originator is the 'bank'. A repo loan, thus, is the only asset that a Friedman-ite bank holds.

All this while we must keep in our mind that out there in the money/ capital market, funds and investment banks often play the role of bank, dealer and investor at the same time, creating an ungodly concoction.

JKH: In fact, "would there be repo in a barter economy" is an interesting question. And if the answer is "no", then *part* of the answer to the question "why does repo exist?" must be "because we live in a monetary economy".

Ritwik: "1) Working capital financing of dealers (which has always been on a 'real bills' and hence risk-free basis), where the short term of the repo matches the frequent inventory turnover of the dealer.

2) Maturity transformation by investors."

I can see those as being empirically important, but not theoretically important. (Howls of outrage!). Because we also pawn/mortgage/repo real assets like watches and houses, as well as bonds, and a general theory would see bond dealers and maturity transformation as just particular examples (albeit common and important examples) of a more general explanation.

"3) Splitting of the 'funding' and 'taking risk' portions of the act of lending,..."

When I mortgage my house, rather than selling shares in my house, or renting and buying shares in a real estate investment trust, it's true I am taking the risk on my house rising or falling in value due to market changes. But I am also resolving the moral hazard/principal-agent problem that renters may not take care of a house.

"All this while we must keep in our mind that out there in the money/ capital market, funds and investment banks often play the role of bank, dealer and investor at the same time, creating an ungodly concoction."

Yep. Which suggests that yet another counterfactual to "why repo?" is "why not banks instead?"


- And if the answer is "no", then *part* of the answer to the question "why does repo exist?" must be "because we live in a monetary economy" -

Yes. That logic was roughly the reason for my question. Off the top, I'm guessing that the repo question generalizes to both barter and monetary, as does the related bid ask spread question. (And maybe there's just a single generic version of collateralized borrowing in a barter economy.) So it's a pretty big question.

I'll be watching for that 4 a.m. post.

Instead of repo, I could sell my bonds to the bank, and buy shares in the bank. Wouldn't something like that be equivalent, in some sort of Modigliani-Miller sense?


I like the question of whether there'd be repo in a barter economy (I already took a stab at it yesterday at 4:33pm, but I'll try again). As in the real world there would be young people with little wealth and old people with lots of wealth. The old people would have portfolios made up entirely of real-estate and production goods. The old people have no use for the real estate so they rent it to the young people in exchange for consumption goods. That's not ideal since the old people have no desire for the market volatility of an asset they never intend to consume, nor do they like the variable rents. Rather, what they want is a guaranteed steady stream of consumption goods. So they arrange an exchange with the young people: the young people get the real-estate and promise to pay the old people the steady consumption stream. The young people agree to commit the real estate as collateral in case they fail on the consumption guarantee. This works well for the young people since they don't care that much about the market volatility of real estate since they intend to use it for a very long time. Also, it's easy for them to exchange their labour for the consumption goods they have to pay the old people.

Now I was going to tell the same long story about the production goods (claims on corporate assets), and how repo solves the problem, but I don't have to. You know exactly what I am going to say.


If I take out a one year personal loan with my house as a mortgage, I have engaged in maturity transformation. When the auto-dealer mortgages his inventory of cars for a loan, he is financing his working capital.

So I see working capital and maturity transformation (financial or 'real') as the two essential planks for a general theory of repo, with bond dealers and securities investors being the empirically most relevant forms.

I don't quite get your counterfactual. The repo originator IS a bank. Its ability to extend a repo loan freely makes it a bank. Or, more accurately, only banks can extend repos. It doesn't matter if they are formally incorporated as a bank or not.

K: "I think you are not properly closing the loop in the monetary system."

Ok, I get it.

K and JKH: Yep, my top-of-the-head response was also that repo generalises to barter. And i was thinking through examples like yours.

But then I thought: we would only have barter in a world of zero transactions costs. One important transactions cost is that people may not deliver what they promised to pay (you can build a theory of "why money" on the premise that monetary exchange is a substitute for personal trust). If people always delivered what they promised to pay, we wouldn't need securitised loans, and so wouldn't need repo either.

So now I've confused myself even more, and made the question even more undefined!

Ritwik: "If I take out a one year personal loan with my house as a mortgage, I have engaged in maturity transformation. When the auto-dealer mortgages his inventory of cars for a loan, he is financing his working capital."

I don't get the distinction. Both I and the autodealer mortgage our real asset for a one-year loan. I am financing my house; he is financing his cars. His reason for wanting to own cars is different from my reason for wanting to own a house, but does it matter?

On repos and banks: is the only difference between repos and banks an artefact of legislation? (If an outside observer didn't know the regulations and names of things, would he actually observe any difference?)


I don't think "maturity transformation" introduces anything new, and I think it's a potentially very misleading category. What it suggests, to me, is you create a corporation, issue $1m of short term paper and buy $1m of 30-year bonds, magically removing the bonds from the market and replacing them with the equivalent of tbills. This doesn't work (though it's a nasty scam that's sometimes performed by finance for (very) short term gain). What can *legitimately* be done is to create the same
corporation, but issue $500k of short term paper and $500k of equity and buy the same $1m of 30 year bonds. For the shareholders this has the same effect as buying $1m of 30-year bonds using their own $500k plus $500k from a repo loan against the $1m of bonds with a 50% haircut. That is not a scam. But I wouldn't call it maturity transformation. You've transformed the 30-year bond risk into short term paper risk (negligible) and equity risk. The equity holds essentially all of the market risk of the 30-year bonds. The duration risk is still all there and can't be "maturity transformed" away.

Nick: Agreed, banks and repos are the same thing. Bank equity = repo haircut.

A better question would be why do unsecured loans exist. As other commenters have explained, repo is just a legally safe way of securing a loan of money (although sometimes the driving motivation is to borrow an asset offering money as security). I dare say most borrowers have assets comfortably in excess of their liabilities, so if I were lending someone money, my inclination would be to ask for security, and if I sensed a reluctance to provide security, I would be more wary about making the loan. I would only expect unsecured loans to be made when the cost of establishing this security relative to the size of the loan was prohibitively high. And then I would expect the lender to impose covenants and closely monitor the borrower's financial health, plus charge a higher rate of interest to actuarially cover the default risk. The financial crisis was in part a shock to lenders' ideas of the state of borrowers' balance sheets, so it is not surprising that the demand for security and hence the total value of collateral pledged in repos has increased. Naturally, in the face of increased demand for value-stable collateral from private sector lenders, borrowers would be grateful if central banks demanded less or less safe collateral, but that does not mean that the central banks are causing the shortage of safe collateral. The central bank's choice is not so much as whether to borrow or buy assets as whether to buy secured loan assets from banks or other kinds of assets from a potentially wider range of market counterparties. Central banks will typically borrow a wider range of assets than they will buy, because they hope not to have to take possession of the assets they borrow.

One word: rehypothecation.

The comments to this entry are closed.

Search this site

  • Google

Blog powered by Typepad