A tale of two banks
What does it mean to say something is risky? How would we know? Would a failure confirm the view that a particular activity was risky?
Consider two people gambling at roulette. Joe puts $100 on each number from 1 to 35. Jane puts $3500 on number 36. Both have bet $3500, and both bets have negative expected values (assuming a 36-1 payoff on the winning number, and 38 total numbers.)
To me, Joe’s bet looks less risky. There’s more than a 90% chance he’ll win $100, although the expected value of the bet is still negative due to the fact that he loses $3500 if number 36, 0 or 00 comes up. Jane has more than a 90% chance of losing all $3500, but will win very big if number 36 comes up. That seems riskier.
Now let’s assume that both people make their bets, and the little ball lands on number 36. Does that improbable outcome mean that Jane’s bet was actually less risky than Joe’s. I’d say no; she just got lucky.
When I speak with people, I often get the impression that they conflate “risky” with “failure”. That’s not how I interpret the term. Consider two banks:
1. Silicon Valley Bank (SVB) takes deposits and invests them in Treasury bonds. It is a fast growing bank.
2. Bank OZK (formerly Ozark) rapidly grows from a small Arkansas bank to a major lender for real estate projects in America’s largest cities.
Which bank’s assets seem risker? Based on this evidence, I would say that Bank OZK was far riskier.
Now assume that SVB goes bankrupt, while Bank OZK is doing great. Does that impact your view as to which bank engaged in a riskier strategy? Should that fact influence your view as to which bank engaged in a riskier strategy?
If failure is evidence of riskiness, what does that imply about the roulette example discussed above.
In 2018, I did a post on Bank OZK, citing it as an example of the sort of risk-taking bank encouraged by the moral hazard in our banking regime. In retrospect, it looks like SVB would have been a better example. But is that true? Was SVB actually a riskier bank? Or did number 36 come up on the roulette wheel?
My failure to spot the bank that actually failed illustrates a problem faced by regulators. In 2018, I was presumably looking back at the banking crises of the 1980s and 2007-10, and noticing that real estate lending often led to banking distress. At that time, Treasuries had been in a bull market from almost 4 decades. We tend to estimate risk based on past performance, especially the recent past. Regulators are unlikely to spot risk that comes from an area that was not previously a major problem. (Recall that in 2006, MBS investors were lulled by the fact that the US had never experienced a large nationwide decline in house prices.)
In recent years, real estate has done surprising well, as inflation tends to boost the value of hard assets like land and buildings. On the other hand, inflation reduces the value of T-bonds. It’s quite possible that, ex ante, SVB’s approach was less risky (perhaps even profit-maximizing!), but these unpredictable macro trends hurt SVB and helped Bank OZK. (To be clear, I suspect that there were other differences as well, perhaps Bank OZK has superior management.)
I don’t believe we’ll ever be able to fix the banking system through regulation. Regulators will always be like generals fighting the pervious war. Instead, we need to remove the underlying problems—moral hazard and a lack of diversification. Trump likes to talk about “Making America Great Again”. How about “Make America’s banking system more Canada’s”?
PS. David Beckworth has an excellent piece in Barron’s discussing how the rise in interest rates has helped long-term borrowers (including the Treasury) while hurting bondholders.
Mar 23 2023 at 8:19pm
Not sure why you’d call the Beckworth piece excellent if it’s the mark to market value of bonds that went down, when the Fed has no intention of buying back on the open market.
Mar 23 2023 at 10:49pm
Can you be more specific?
Mar 23 2023 at 8:36pm
Since you’re so very focused on moral hazard, we can eliminate it by adopting Cochrane’s narrow banks. They don’t have to be required, merely allowed.
Mar 23 2023 at 10:49pm
Yes, I’d support that.
Mar 25 2023 at 9:25pm
Actually, merely not actively opposed by the FED
Mar 23 2023 at 11:44pm
Hasn’t banking always involved risk from maturity mismatch? Also, duration risk has always been a problem with fixed investments. Almost every fixed income fund publishes its duration.
On your question of which bet is riskier, I ask a related question. Which would you prefer, a $126,000 payoff with a chance of 1 in 38, or a $3,600 payoff with a chance of 35 in 38?
Mar 24 2023 at 8:34am
They both have an expected value of $3,316, so a risk-neutral agent would be indifferent. But a lot of economics and finance is based on the idea of risk aversion, and any model of expected utility would conclude the second payoff is worth more.
Mar 24 2023 at 9:36am
The standard deviation on the second is much lower, so would be less risky.
Which is what you said, but the SD is why.
Knut P. Heen
Mar 24 2023 at 12:42pm
No, decreasing marginal utility of wealth is why. That leads to risk-aversion. The longer explanation is that you spend your first dollar on something that brings a lot of utility, the second dollar on something that brings slightly less utility and so on. Standard deviation is a measure of risk, it does not explain why people don’t like risk.
Anyway, the problem is the combination of highly levered debt financing and limited liability in banks. Suppose you borrow $3,500 and do as Jane (place everything on 36). In 37 out of 38 cases, you will lose, go bankrupt, and pay nothing. In 1 out of 38 cases, you win $126,000 and pay back $3,500. This is a free lottery ticket handed out to Jane by the government. Joe will win $3,600 and pay back $3,500 in 35 out 38 cases. In 3 out of 38 cases, Joe also goes bankrupt. This is also a free lottery ticket handed out to Joe by the government. In the real world, the stockholders of the banks must put up around 10 percent of the bet to have some skin in the game, but management need not. This is the moral hazard problem. Management gets big bonuses if they win and pays nothing if they lose. Nice casino.
Mar 24 2023 at 6:51pm
Interesting. I approach it from a similar angle, but come up with the opposite conclusion. Winning $3,600 doesn’t mean a whole lot to me, it won’t change my life in any meaningful way. $126,000 would though, that’s enough that it could flip some decisions one way or the other. Maybe I leave a crappy job before I have a new one lined up. Maybe I take a flyer on starting a business. Maybe I have a good down payment on a house or decide to start a midlife crisis early and buy a sports car. And if I lose, well we’ve already established that’s not life changing money.
It brings me to a larger point, I tend to be risk tolerant and generally speaking I’d prefer a higher variance if it led to a higher expected payoff. On a societal level, I’d much prefer that banks took bigger risks, even if it led to more financial crises, If I expected that to lead to a higher overall rate of RGDP growth. I’m not sure that’s the case, but I’m also not sure that even if it was, that’s a trade more risk averse people would be comfortable with.
Mar 24 2023 at 12:12pm
Allow me to rephrase the question. What would you pay to play each game?
Knut P. Heen
Mar 24 2023 at 12:59pm
You don’t pay. That is what makes the casino nice.
Mar 24 2023 at 3:00pm
How much one would pay to play was just a tool to quantify preference. Despite the same expected payoff, most would pay a higher price for the option that had the lower maximum payoff.
As you wrote, the issue is diminishing marginal utility. As Garrett wrote, it’s risk aversion. As robc wrote, SD shows the low payoff is less risky.
Mar 25 2023 at 6:45am
Given the premise of an “infinite” number of plays the answer is obvious that the expected values are identical. But is there a difference between an infinite number of “plays” versus , for example, 2 plays? I think there might be. The percent of wins —-or “risk aversion” (risk aversion is not mathematical——-its merely psychological) ——will clearly make one bet on the latter as Garrett states.
If I needed money, I would play the latter. I believe many traders “win” because they play the latter method. If they play long enough, they will blow up. Like selling out of the money puts. It is an odd concept as they are not identical ideas. It’s not that dissimilar from derivatives or Calculus in general.
Knut P. Heen
Mar 27 2023 at 8:32am
Dylan: The most popular models are based on perfectly divisible goods. If goods are not divisible, strictly diminishing marginal utility of wealth may not be the case. That is one reason why casinos/lotteries may exist. It allows you exchange a cup of coffee for a small probability of winning a sports car. You get that behavior if the utility of wealth is S-shaped (which may be plausible if some goods are not divisible, the counter-argument is that you may rent a sports car for a time and that it therefore really is divisible). Friedman and Savage wrote a paper about it in the Journal of Political Economy back in 1948.
Vince: The fundamental forces are preferences (utility) and the properties of the goods (divisible/indivisible). Risk aversion and the interest in standard deviation are rational responses to these. If you really want a specific house and cannot afford it unless you gamble, it may be rational to take the gamble. We need a theory which explains why some people buy both a lottery ticket and fire insurance. Risk aversion cannot explain both. Standard deviation is not necessarily a good measure of risk if pay-offs are not normally distributed like in this casino example. Skewness is important too. Positive skewness is often preferred over no skewness or negative skewness. That is what casinos and lotteries offer. The reason people fear climate change is probably because it is a story about negative skewness (small probability of a very bad outcome). People hate those bets.
Michael: If you play the same game a large number of times, you will achieve the expected outcome with almost certainty. That is the point of diversification. If you do this at a casino, however, you will pay $3,500 per round and win back $3,316 per round. It will take you about 550 rounds to lose $100,000 with almost certainty. You will be broke long before you hit infinity. Finally, it is shorting uncovered calls that kills you in the long-run (unlimited losses will happen with certainty if you continue long enough).
Mar 24 2023 at 2:27pm
Vince, You said:
“Hasn’t banking always involved risk from maturity mismatch?”
Yes, although it obviously varies from one bank to another. It was a big factor in the S&L crisis.
I’d take the safer gamble—35 numbers.
Mar 24 2023 at 2:51pm
Right, the S&L crisis was a perfect example of maturity mismatch. The problem is not new. The regulators have no excuse.
Thomas Lee Hutcheson
Mar 25 2023 at 12:37pm
But term transformation need not imply interest rate mismatch risk The FI can invest in VR instruments
Mar 24 2023 at 9:30am
I’ve often found the essay _The Ethics of Belief_ (https://people.brandeis.edu/~teuber/Clifford_ethics.pdf) instructive on this point.
Clifford offers the hypothetical of a shipowner who convinces himself without evidence that a ship is seaworthy, and then is distraught when the ship wrecks: “…he did sincerely believe in the soundness of his ship; but the sincerity of his conviction can in no wise help him, because he had no right to believe on such evidence as was before him.” He then flips the outcome: “suppose that the ship was not unsound after all; that she made her voyage safely…The man would not have been innocent, he would only have been not found out.”
The distinction between “having gotten it right” and “having gotten away with it” is nearly impossibly to discern, or even to remember to check, after the fact; this is why IMO we should weight before-the-fact transparent and disclosed reasoning much more heavily than we weight outcomes which are after all mostly a matter of happenstance.
Mar 24 2023 at 12:18pm
Nice. It’s like the maxim, you’re not guilty unless you’re caught. Many years ago, a cop who ticketed me for an illegal U-turn told me that. Later, I saw another car do the same thing. It was a cop.
Thomas Lee Hutcheson
Mar 24 2023 at 6:20pm
SVB and OZK were taking very different kinds of risk. SVB was taking on interest rate mismatch risk, OZK was taking on credit risk. Which one took on MORE risk I cannot say. One question would be how much of each kind of risk did each take on relative to its uninsured deposits.
Mar 24 2023 at 10:39pm
First, the fact that you got away with it doesn’t mean you didn’t take any risk. Someone posted this meme on the ‘net and I send it out frequently … a woman standing next to a rhinoceros, with one hand on the friendly beast’s shoulder, smiling for the camera.
Second, SVB had zero Treasuries in their portfolio. It was all agency MBS and CLO. Amazing to me how the “they just invested in treasuries” lie persists. Is it really too much effort to take a 30 second look at their 10k? It doesn’t change the credit risk of the assets (not by much, anyway), but it significantly increases the asset-liability duration mismatch.
Third, here’s my favorite illustration of risk. Which is riskier, a bet with a guaranteed loss of $100 or a bet with a 10% chance 0f winning $1000 and a 9o% chance of losing $222? Answer: the second one is riskier. There is zero risk in any guaranteed outcome. Not convinced? Think about how much you’d be willing to pay(receive) to take each of those bets.
Mar 25 2023 at 2:49pm
The DIDMCA turned the thrifts into banks. In order to survive the thrifts had tact like banks, but their portfolios were different (a higher percentage of longer-term real-estate assets).
A credit crunch occurs when there is an outflow of funds or negative cash flow.
The last period of disintermediation for the commercial banks, prior to the monetary policy blunders during the Great Recession, occurred during the Great Depression, which had its most force in March 1933.
Ever since 1933, the Federal Reserve has had the capacity to take unified action, through its “open market power”, to prevent any outflow of currency from the payment’s system. That’s what the discount window is for.
BAGEHOT’S DICTUM: the central banks should lend early and ‘without limits’ to solvent firms at a ‘higher interest rate’ with ‘good collateral’. Discounting was made a penalty rate on January 6, 2003
But Volcker did the opposite where discounting was not contractionary.
And: “In 2002, the Federal Reserve began to set the discount rate above the federal funds rate, reversing its previous practice of keeping the discount rate below the funds rate.”
Bank credit contraction is cumulative and reinforcing. Contraction is usually volatile and disorderly.
Mar 25 2023 at 9:37pm
If the question is, was SVB just unlucky?
The answer is, if you put yourself in the position of going bankrupt if interest rates go back to a level they have been in for most part of financial history and then the interest rates do go back there and you do go bankrupt, you can NOT call that being “unlucky”.
In your roulette example it will be akin of risking losing it all betting red just because the little ball has landed red the last 50 times and then getting black. This is not bad luck by any means. Just a profound ignorance of how odds work.
When it comes to bank you can call it “manage by ear” or “manage with the flow”. It is always risky betting big on “this time is different”
Mar 25 2023 at 11:10pm
But that wasn’t the only reason they went bankrupt.
Mar 26 2023 at 9:13am
You very unlikely go bankrupt just for one reason.
For instance, not having a chief risk officer is, I am sure, good for business … until it is not.
Comments are closed.