Two Contracts That Are Central to the Financial System

In the old days, there were penalties such as debtors’ prison. Presumably there was no great joy for a lender in sending anyone to debtors’ prison, but the fact that you might face imprisonment served to increase your incentive to repay. The good news about foreclosure is that it makes the borrower repay what’s owed, but the bad news is that imposing this penalty is tremendously inefficient. As an investor you don’t want to do it very often, and there are huge ex-post costs when you choose this option, reducing what you can recover.

The second option for overcoming conflicts of interest involves monitoring. Each lender can continuously monitor a borrower’s business to make sure that these conflicts don’t arise, and then the borrower will repay without the threat or act of foreclosure. But it takes a lot of time and resources to continuously monitor a borrower. There are large up-front costs of monitoring.

When there is no monitoring, the optimal financial contract is one that specifies an amount to repay, essentially saying, “If you pay me this amount, I won’t foreclose. Pay me any amount less, and I will.” If you’re the borrower, you will repay whenever you can. You would rather pay the €1.1 million that you owe than spend many years in debtors’ prison. But if you owe €1.1 million and only have €1 million, off to debtors’ prison you go, which is bad for you and inefficient overall. In other words, the optimal financial contract without monitoring is a debt contract.

Imagine there’s some uncertainty about how much the borrower can actually repay. Say you lend €1.1 million to a firm whose project will either produce €2 million or €1 million, and the odds are 50/50 for each. Because you need the borrower to pay you back more than €1 million, you’re going to have to impose foreclosure or liquidation half the time. That would either cause credit to be tremendously expensive, or you might recover so little from foreclosure half the time that you wouldn’t be willing to lend in the first place.

The average value of the cash that the borrower has, the expected value, is €1.5 million. If the borrower always has that amount, debt contracts enforced by foreclosure would work perfectly. You could say, “Pay me €1.1 million,” the borrower would pay out of the €1.5 million, and nothing inefficient would ever happen.

But the borrower does not always have €1.5 million. Because the borrower has to pay more than €1 million, you would foreclose inefficiently half of the time. Suppose you recovered almost nothing from foreclosure. The other half of the time, you’d get €2 million, at most. If you, the investor, require a €1.1 million expected return, you would not be willing to lend. So the best contract without monitoring is debt with foreclosure (debt with penalties). It works well if the borrower doesn’t have much volatility in ability to pay, but otherwise it works poorly.

But suppose you monitor the borrower. Doing this has up-front costs, but when the borrower has €2 million, you can collect full payment. When the borrower has €1 million, you can take that entire amount without going through a foreclosure process. Monitoring—which could mean finding out what the actual cash flow is and renegotiating bond covenants, or making sure the borrower isn’t stealing money, or making sure a brother-in-law is not involved—avoids the need for inefficient foreclosure penalties.

Now, imagine that there are 10,000 small investors, and each one makes a €110 loan (adding up to €1.1 million), but it costs €110 to monitor each loan. It would be pretty stupid for the investors to monitor the debt. Direct finance with monitoring would be too expensive, so the best possible contract would be an unmonitored debt contract. But if you’re the borrower and there’s enough uncertainty about how much you will be able to repay, you’re going to end up finding yourself unable to borrow. That’s because although the best contract for this situation is unmonitored debt, that tactic works poorly here.

Pooling and tranching

This is the idea behind financial intermediation. Instead of having all 10,000 investors monitor their loans, we delegate the monitoring to one other party that I will call a banker. The firm will know its cash flows, and the banker will know those too. This delegated monitoring will ensure that the firm repays the banker.

The question is, how do we get the bank to repay the investors? The banker can see whether the borrower and the brother-in-law are doing a deal, but the investors can’t see that. Imagine the banker, the brother-in-law, and the borrower get together and say, “Well, let’s just tell them we only have €1 million this period. We’re not going to pay them more than that.” The investors will only know how much the bank repays them, not the actual amount the borrower has. Without monitoring, we had a conflict of interest between the borrower and investors. With delegated monitoring, we add the banker, who also has a conflict of interest.

Just as in my earlier example, there is a way to resolve this added conflict of interest. In that case, the borrower had to repay the investors, and if the borrower paid too little, there was a foreclosure. In this case, the bank writes a debt contract that specifies that if it pays too little, the investors foreclose on the bank. This threat of bank failure turns out to be a great incentive for the bank to monitor borrowers and to repay investors.

The optimal contract between the bank and investors (in this context, depositors) to provide incentives for delegated monitoring is a debt contract issued by the bank promising to pay the depositors. The bank will fail if it defaults on its contract and pays too little. This forces the bank to pay depositors. In addition, if the bank doesn’t monitor, it won’t have enough money to pay investors, and there will be a default. The bank will monitor.

But it turns out that for this to work, the bank has to be large and diversified. To understand why diversification is beneficial, consider this extreme example of perfect diversification. Suppose the bank makes lots of loans, and assume that exactly half of the borrowers’ projects will produce €2 million and the other half will produce €1 million. There’s no uncertainty about the amount the average borrower is going to pay, and the amount of cash that the bank could collect from the loans will always be up to €1.5 million. What we don’t know is which borrower will have €2 million, and which borrower will have €1 million.

Let’s say the debt (deposit) contract the bank issues to depositors has a promised payment of €1.1 million per borrower. The bank can always collect up to €1.5 million per borrower. The bank will never fail. In this best case, this diversification means foreclosure will never be needed, but will only be a threat. The bank issues perfectly safe deposits.

This is perfect diversification, but in the real world and my model, we have banks with imperfect diversification. Banks are going to fail sometimes—maybe because they’re small or exposed to a single industry, or maybe because they’re very large but still extremely exposed to macroeconomic aggregate risks.

It’s important for bank supervisors and regulators to look at more than just bank credit standards and banker competence. The need for diversification implies that it is just as important to examine a bank’s exposure to aggregate risks such as national income, interest rates, and wars. Around the world, there’s always been a focus on this in supervision and regulation, but since the 2008–09 financial crisis, stress tests have explicitly looked at how exposed a financial institution is to large aggregate macroeconomic shocks that might bring down a diversified bank or even the whole financial system. Substantial diversification within a large bank without excessive exposure to aggregate macro shocks is essential for banking to work well.

A more modern phrase for the contract structure I have discussed, involving diversification of assets and the issuance of debt claims to investors, is pooling and tranching. Pooling is just another name for diversification: make a bunch of loans and put them together in a pool from which to pay claims to investors. Tranching refers to giving investors senior claims and having the banker retain some junior claims to finance bonuses. This structure resolves many conflicts of interest in addition to those involving monitoring. Stanford’s Peter DeMarzo has a nice 2005 paper on this, “The Pooling and Tranching of Securities: A Model of Informed Intermediation.”

As I mentioned before, the pooling part doesn’t work perfectly, particularly when there’s increased correlation across returns of the assets. We saw this in the housing crisis of 2008–09: When house prices went down worldwide and in the US in particular, all of a sudden, the correlation between mortgage defaults, whether on subprime or other kinds of loans, increased dramatically. The diversification effect from pooling almost disappeared, and we saw senior mortgage-backed securities become, unexpectedly, much riskier.

The self-fulfilling prophecy

Pooling and tranching is one important financial contracting technology. A second one is in a paper, “Bank Runs, Deposit Insurance, and Liquidity,” that I wrote with Washington University’s Philip H. Dybvig. It again explains why banks and intermediaries are good to have in the middle, and why they use certain financial contracts. The financial contract we have in mind here is short-term debt. Why do banks use so much short-term debt to finance long-term illiquid assets, which leaves them subject to bank runs? Why choose to write contracts that leave you subject to runs if runs are so bad?

If you hold assets through a bank, you are better off than if you hold the bank’s assets yourself. Dybvig and I assume that the bank’s assets are long term and safe if you hold them to maturity, but illiquid. We show why the liabilities, deposits of this bank, should be short-term debt and liquid.

We take solvency off the table as a reason for bank failures, and that’s not because it’s off the table in practice. We assume that everybody knows that if you hold the long-term assets to maturity, the bank will be solvent. Because insolvency is not the reason for a run, we can see that bank failures can be caused by potential runs.

Bank deposits are more liquid than the loans they hold. This means if you hold a bank deposit, you get a higher return for holding it for one period and then getting rid of it quickly versus holding the underlying bank loan yourself for the same amount of time. Bank deposits then provide some insurance against needing to get out fast. That’s the first part of our model.

We also demonstrate that banks should do this because investors like liquid assets better than illiquid assets. That’s due to the fact that investors don’t know how long they’re going to want to hold those assets. They might hold them for one period, or they might not need them right away so hold them for two periods. There’s an important liquidity risk left: even though these assets are safe, you as a borrower face the risk that you might need to get out early. Liquidity is a form of insurance against this early need for funds.

Here’s a super-simple illustration of our model: There’s an illiquid but safe asset that you or the bank could invest in. If you directly invest €1 in this asset at date zero, you can hold it for either two periods and get €2 out, or one period and get your €1. If you hold this asset to maturity, you double your money, and if you get out of it early, you destroy half of that value. There’s a big loss if you get out in a hurry.

Suppose that the bank is going to issue more-liquid, short-term deposit claims backed by this illiquid, safe asset. The bank deposit is going to offer you a choice between €1.28 at date one or €1.81 at date two. The key thing is that if you hold the illiquid asset directly, you only get €1 at date one, whereas if you hold the bank deposit, you get €1.28. What’s important in this example is that 1.28 is a number bigger than 1. All these things that I am about to say are true for any number bigger than one.

This is creating liquidity because you’re giving people a bigger return over the short horizon. It’s not free for the depositor, who has to give up something over the long horizon, but it is a form of insurance against the need for liquidity.

Suppose there are 100 investors in the world and that we know for sure 25 of them will need their money on date one (call them early investors) and the remaining 75 will need their money on date two (call them late investors). We know the proportion will be 25 and 75, but what we don’t know yet is who’s going to need their money early and who’s going to need it later. There’s uncertainty about when you’re going to need your funding. You’d like to buy insurance against this uncertainty, and that’s what the deposits do in this setup.

Suppose that there’s not going to be a bank run and that the only people who pull their money out of the bank on date one are the 25 early investors who actually need their money then. In this case, we give 25 people €1.28, and we have to liquidate 32 assets to do so. That’s going to leave just enough assets (or actually a few extra assets because I did some rounding) to pay €1.81 to the people who leave their money in until date two. If everybody is clear on this point, there’s not going to be a run on this bank. The bank can create more-liquid deposits out of less-liquid assets. We can actually create liquidity here, which is good. If everyone forecasts that 25 will withdraw, this is a self-fulfilling prophecy.

If this is the outcome, everybody is fine. They would indeed put all of their money into the bank. The trouble is that if the bank creates liquidity—which is to say, it pays more than €1 to the people who take their money out at the early date—there’s a self-fulfilling prophecy that the bank is going to fail.

How does the bank pay €1.28 to the people who withdraw? It has to liquidate a bigger fraction of the assets than the fraction of people who withdraw. If 25 percent of the depositors withdraw, you liquidate 32 percent of the assets. If 100 percent of the people withdraw, though, you can’t liquidate 128 percent of the assets because you only have 100 percent to start with. You can’t give 100 depositors €1.28, and you can’t even give 79 depositors €1.28. So if the depositors think that they’re all going to come and ask for their money on date one, everybody making this prediction will rush to get their money out of the bank—because anyone who leaves money in the bank will get zero on date two.

If you’re a depositor and get there quickly enough on date one, you’ll get €1.28, but if everybody else has the same plan, there will soon be nothing left. If a run is expected, the bank will fail. If the bank fails, that’ll cause a run. This is a self-fulfilling prophecy, a Nash equilibrium. It’s why short-term-debt runs can bring down a solvent financial intermediary.

If you expect a run, you’ll want to get out early, and if you don’t expect a run, you won’t. These multiple self-fulfilling prophecies are one way of thinking about why the financial system is somewhat unstable. You could eliminate bank runs (by paying €1 to depositors at date 1), but you’d do that by removing the value banks add by creating liquidity.

With this view of what the financial system is doing and creating, you can see that problems are hard to fix because you don’t want to throw out the baby with the bathwater. But there are only a few ways of getting the best of liquidity creation without runs.

One point I like to make is that private financial crises are everywhere and always due to the problems of short-term debt. The government has lots of ways to cause problems, and sovereign financial crises can also come from fiscal policy, for example. Bank runs are not just about banks, and in fact, banks are just an example of the many types of financial institutions financed by short-term debt. We saw runs on money market mutual funds in both 2008 and March 2020, on Lehman Brothers, on the stablecoins Terra and LUNA, and on short-term securitizations called asset-backed commercial paper—all of these things involve borrowing short and lending long and illiquid. All had runs; none were banks.

So if you borrow short and lend long and illiquid, you might create liquidity but you will be subject to a run. This instability from runs is not just about the fact that there are only fractional reserves in the banking system. We already understood that if the bank has to give you currency when you take your money out, and there’s not enough currency in the vault for everybody to take their money out, there’s the possibility of a run. That’s the story in It’s a Wonderful Life. Like the film’s protagonist, George Bailey, explains to his neighbors, their money is in Joe’s house, not back in the vault. That’s true. If everybody needs all their currency, the bank’s going to fail.

But that is the wrong model for policy makers and investors to think about when they are wondering where there may be a self-fulfilling run. It’s not just about the bank and currency. It’s about short-term debt. Lehman Brothers did not have to give out currency to its investors, yet it had a run that brought it down.

So how can you stop bank runs? If you have a very solvent government (not all governments are), it can provide deposit insurance that will stop this feedback loop. Another way is for a central bank to commit to lend to allow banks to survive a run. The commitment to lend is necessary, but absent in practice. The US Federal Reserve didn’t lend to banks in the 1930s, as it turns out. In addition, even these solutions are not perfect because this run risk migrates out of the regulated and insured sector to shadow banks.

To summarize, I described two contract forms that are optimal contracts written by intermediaries in the financial system. One is pooling and tranching to resolve conflicts of interest. The other is creating liquidity by borrowing short term and lending long term and illiquid. I would say that a large fraction of what the financial system has done in creating new products, for good or for evil, has involved some combination of these two contracting technologies. Many institutions use these two contract forms, not just banks. To understand a new product, you should ask, “OK, how does it mix these two? Does it use too much of one or the other, or too much of both?”

Douglas W. Diamond is the Merton H. Miller Distinguished Service Professor of Finance at Chicago Booth and is the recipient of the 2022 Nobel Prize in Economic Sciences. This essay is adapted from his live Nobel Prize lecture, given December 8, 2022, in Stockholm, and is distinct from his published Nobel Prize Lecture document © The Nobel Foundation 2022.