Full text of Financial Crisis Inquiry Commission : Documents Relating to the Financial Crisis of 2007-2009 : Session 3 : Risk Taking and Leverage : Solving the Present Crisis and Managing the Leverage Cycle

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

Solving the Present Crisis and Managing the Leverage Cycle1
John Geanakoplos2
December 22, 2009

Abstract: The present crisis is the bottom of a recurring problem that I call the
leverage cycle, in which leverage gradually rises too high then suddenly falls much too
low. The government must manage the leverage cycle in normal times by monitoring
and regulating leverage to keep it from getting too high. In the crisis stage the
government must stem the scary bad news that brought on the crisis, which often will
entail coordinated write downs of principal; it must restore sane leverage by going around
the banks and lending at lower collateral rates (not lower interest rates), and when
necessary it must inject optimistic capital into firms and markets than cannot be allowed
to fail. Economists and the Fed have for too long focused on interest rates and ignored
collateral.
Key Words: Leverage, collateral, margins, leverage cycle, externality, principal

1

On October 3, 2008 I presented to Ben Bernanke and the Board of Governors at the Federal Reserve the
substance of the proposal embodied in this paper. One addition to my original proposal that is included
here is evidence on the necessity of principal reductions, taken from my March 5, 2009 New York Times
op-ed with Susan Koniak. I am grateful to her for very helpful comments and advice on this paper. I am
also appreciative of the very fine comments I received from the two referees of the paper.
2
James Tobin Professor of Economics, Yale University, and also a Partner in the hedge fund Ellington
Capital Management, which trades primarily in mortgage securities.

1

Introduction
The present crisis is the bottom of a leverage cycle. Understanding that tells us
what to do, in what order, and with what level of urgency. The government has acted
aggressively, but because its actions were not rooted in (or explained with reference to) a
solid understanding of the causes of our present distress, we have started in the wrong
place and paid insufficient attention and devoted insufficient resources to matters, most
notably the still growing tidal wave of foreclosures and the sudden de-leveraging of the
financial system, that should have been first on the agenda.
In short and simple terms by leverage cycle I mean this. There are times when
leverage is so high that people and institutions can buy many assets with very little
money down and times when leverage is so low that buyers must have all or nearly all of
the money in hand to purchase those very same assets. When leverage is loose, asset
prices go up because buyers of them can get easy credit and spend more. Similarly, when
leverage is highly constrained, i.e., when credit is very difficult to obtain, prices
plummet. This is what happened in real estate and what happened in the financial
markets. Governments and economists have long monitored and adjusted interest rates
as if doing that would suffice to ensure that credit did not freeze up and thereby threaten
the economic stability of a nation. It will not. Leverage (equivalently collateral rates)
must also be monitored and adjusted, if we are to avoid the destruction that the tail end of
an outsized leverage cycle can bring.
Economists and the public have often spoken of tight credit markets, meaning
something more than high interest rates, but without precisely specifying or quantifying
exactly what they meant. A decade ago I showed that the collateral rate or leverage is an
equilibrium variable distinct from the interest rate.3 The collateral rate is the value of
collateral that must be pledged to guarantee one dollar of loan. Today many businesses
and ordinary people are willing to agree to pay bank interest rates, but they cannot get
loans because they do not have the collateral to put down to convince the banks their loan
will be safe.
Huge moves in collateral rates, which I have called the leverage cycle, are a
recurring phenomenon in American financial history.4 The steps we must take at the end
of the current cycle emerge from understanding what makes a leverage cycle swing up,
sometimes to dizzying extremes, and then come crashing down, often with devastating
consequences.
All leverage cycles end with (1) bad news that creates uncertainty and disagreement,
(2) sharply increasing collateral rates, and (3) losses and bankruptcies among the
leveraged optimists. These three factors reinforce and feed back on each other. In
particular, what begins as uncertainty about exogenous events creates uncertainty about
endogenous events, like how far prices will fall or who will go bankrupt, which leads to
3

Geanakoplos (1997), Geanakoplos (2003)
The history of leverage is still being written, because until recently it was not a variable that was explicitly
monitored. But work by Adrian and Shin (2009) and others is helping to restore the historical record.
4

2

further tightening of collateral, and thus further price declines and so on. In the aftermath
of the crisis we always see depressed asset prices, reduced economic activity, and a
collection of agents that are not yet bankrupt but hovering near insolvency. How long the
aftermath persists depends on how deep the crisis was and how good government
intervention is.
Once the crisis has started the thematic solution is to reverse the three symptoms
of the crisis: contain the bad news, intervene to bring down margins, and carefully inject
“optimistic” equity back into the system. As with most difficult problems, a
multipronged approach is generally the most successful. To be successful any
government plan must respect all three remedial prongs, and their order. The unusual
government intervention in this cycle has in many respects been quite successful in
averting a disaster, precisely I would argue, because the Fed and the Treasury have begun
applying some of the novel leverage cycle principles I describe here. But effective as the
intervention was in some ways, it did not respect the priorities of the problem, and as a
result it is losing public confidence even as it succeeds in some dimensions.
In what follows I explain what happens in the leverage cycle and why it is so bad for
the economy that it must be monitored and controlled by the government. I show how
this last cycle fits the pattern and I further explain why this leverage cycle is worse than
all the others since the Depression. I point out that the now-famous counterparty risk
problem, which has received so much attention of late, is also a matter of collateral. Next
I present details on how to intervene to pull out of a leverage cycle crisis like the one
were are passing through now. I conclude with a list of recommendations about
managing the leverage cycle in its ebullient period that might prevent the next cycle from
reaching such a devastating crisis stage.

I. Margins, the Leverage Cycle, and Asset Prices
Traditionally, governments, economists, as well as the general public and the press,
have regarded the interest rate as the most important variable in the economy. Whenever
the economy slows, the press clamors for lower interest rates from the Fed, and the Fed
often obliges. But sometimes, especially in times of crisis, collateral rates (equivalently
margins or leverage) are far more important than interest rates. The Fed should be
managing collateral rates all through the leverage cycle, but especially in the ebullient
and the crisis stages.
The use of collateral and leverage is widespread. A homeowner (or a big investment
bank or hedge fund) can often spend $20 of his own cash to buy an asset like a house for
$100 by taking out a loan for the remaining $80 using the house as collateral. In that case
we say that the margin or haircut is 20%, the loan to value is $80/$100 = 80%, and the
collateral rate is $100/$80 or 125%. The leverage is the reciprocal of the margin, namely

3

the ratio of the asset value to the cash needed to purchase it, or $100/$20 = 5. All of
these ratios are different ways of saying the same thing.
Leverage is important for three reasons. At the macro level, it enables a small
group of buyers with little cash to own and control a vast amount of assets. At the single
investor level, as every trader knows, a buyer who leverages his purchase λ times makes
λ% return on his cash for every 1% rise in the asset price; but he loses λ% of the cash he
put down for every 1% decline in the asset price. (If the home price above rises to $101,
the buyer can sell it, return the borrowed $80, and pocket the $1 profit as 5% return on
his $20 investment.) Finally, when seizing the collateral is the only recourse the lender
has for default, the borrower effectively has a “put” option to walk away if the collateral
falls in value below the debt.
In standard economic theory, the equilibrium of supply and demand determines the
interest rate on loans. But in real life, when somebody takes out a loan, he must negotiate
two things: the interest rate, and the collateral rate. A proper theory of economic
equilibrium must explain both. Standard economic theory has not really come to grips
with this problem for the simple reason that it seems intractable: how can one supplyequals-demand equation for a loan determine two variables, the interest rate and the
collateral rate? There is not enough space to explain the resolution of this puzzle here,
but suffice it to say that supply and demand do indeed determine both. Moreover, the
two variables are influenced in the equilibration of supply and demand mainly by two
different factors: the interest rate reflects the underlying impatience of borrowers, and the
collateral rate reflects the perceived volatility of asset prices and the resulting uncertainty
of lenders. 5
A second critical insight is that for many assets there is a class of natural buyers or
optimists who are willing to pay much more than the rest of the public. They may be
more risk tolerant. Or they may simply be more optimistic. Or they may like the
collateral (for example, housing) more.6 If they can get their hands on more money
through borrowing, they will spend it on the assets and drive those asset prices up. If
they lose wealth, or lose the ability to borrow, they will be able to buy less of the asset,
and the asset will fall into more pessimistic hands and be valued less.7

5

In Geanakoplos (1997) I showed how supply and demand can indeed simultaneously determine the
interest rate and the collateral rate. In Geanakoplos (2003) I showed how intertemporal changes in volatility
lead to changes in the equilibrium leverage over time as part of what I called a leverage cycle.
6
Two additional sources of heterogeneity are that some investors are more expert at hedging assets, and
that some investors can more easily obtain the information (like loan level data) and expertise needed to
evaluate the assets.
7
It is useful to think of the potential investors arrayed on a vertical continuum, in descending order
according to their willingness to buy, with the most enthusiastic buyers at the top. The higher is the
leverage, the smaller is the number of buyers at the top required to purchase all the available assets. As a
result the marginal buyer, who is the agent on the cusp of selling or buying and whose opinion determines
the price, will be higher in the continuum and therefore the price will be higher.

4

It is well known that a reduction in interest rates will increase the prices of assets like
houses. It is less appreciated, but more obviously true, that a reduction in margins will
raise asset prices. Conversely, if margins go up, asset prices will fall. A potential
homeowner who in 2006 could buy a house by putting 3% cash down might find it
unaffordable to buy now that he has to put 30% cash down, even if the Fed managed to
reduce mortgage interest rates by 1% or 2%. This has diminished the demand for
housing, and therefore housing prices. What applies to housing applies much more to the
esoteric assets traded on Wall Street (such as the mortgage-backed investments) where
the margins (i.e. leverage) can vary much more radically. In 2006 the $2.5 trillion of socalled toxic mortgage securities could be bought by putting $150 billion down and
borrowing the other $2.35 trillion. In early 2009 those same securities might collectively
be worth half as much, yet a buyer might have to put nearly the whole amount down in
cash. In section IIa I illustrate the connection between leverage and asset prices over the
current cycle.
Economists and the Fed ask themselves every day whether the economy is picking the
right interest rates. But one can also ask the question whether the economy is picking the
right equilibrium margins. At both ends of the leverage cycle, it does not. In ebullient
times the equilibrium collateral rate is too loose; that is, equilibrium leverage is too high.
In bad times equilibrium leverage is too low. As a result, in ebullient times asset prices
are too high, and in crisis times they plummet too low. This is the leverage cycle.
The policy implication of the leverage cycle is that the Fed should manage systemwide leverage, seeking to maintain it within reasonable limits in normal times, stepping
in to curtail it in times of ebullience, and propping it up as market actors become anxious,
and especially in a crisis. To carry out this task, of course, the Fed must do a much better
job monitoring leverage. The Fed must collect data from a broad spectrum of investors,
including hitherto secretive hedge funds, on how much leverage is being used to buy
various classes of assets. Moreover, the amount of leverage being employed must be
transparent. The accounting and legal rules that govern devices, such as SIV’s, that were
used to mask leverage levels must be reformed to ensure that leverage levels can be more
readily and reliably discerned by the market and regulators alike.
The leverage cycle is no accident, but a self-reinforcing dynamic. Declining margins,
or equivalently increasing leverage, are a consequence of the happy coincidence of
universal good news and the absence of danger on the horizon. With markets stable and
the horizon looking clear, lenders are happy to reduce margins and provide more cash.
Good, safe news events by themselves tend to make asset prices rise. But they also
encourage declining margins which in turn cause the massive borrowing that inflates
asset prices still more.
Similarly, when the news is bad, asset prices tend to fall on the news alone. But the
prices often fall further if the margins are tightened.
Sudden and dramatic increases in margins are relatively rare. They seem to happen
once or twice a decade. Bad news arrives much more often than that, so it is not bad or

5

even very bad news alone that drastically raises margins. Bad news lowers expectations,
and like all news usually clarifies the situation.
Every now and then bad news, instead of clarifying matters, increases uncertainty and
disagreement about the future. It is this particular kind of scary bad news that increases
margins. For example, three years ago people disagreed whether losses from defaults on
prime mortgages would be 1/4% or 1/2%, and whether losses on subprime mortgages
would be 1% or 5%. By contrast, after the scary news of last year, people disagreed about
whether some subprime losses would be 30% or 80% and there is still substantial
disagreement. Even from their current low prices, many lenders are afraid many assets
could lose even more value, maybe all. The present is worse, and the future more
uncertain.
The upshot of the increased uncertainty and disagreement is that margins go up
drastically. Lenders are typically more pessimistic than buyers. Otherwise they too would
be buying, instead of lending. Even if the optimists are not much worried about more
losses, the lenders are, and they will demand high margins. When the lenders are worried
about 80% losses from current levels, they will lend only if margins are at least 90%, or
not lend at all.
As we have just witnessed, the rapid increase in margins always comes at the worst
possible time. Buyers who were allowed to massively leverage their purchases with
borrowed money are forced to sell when bad news drives asset prices lower. But when
margins rise dramatically, more modestly leveraged buyers are also forced to sell.
Tightening margins themselves force prices to fall further. We enter the crisis stage I
discuss below.
The dynamic of the leverage cycle cannot be stopped by a tongue lashing of greedy
Wall Street investors or over-ambitious homeowners in the ebullient stage of the cycle, or
by exhortations not to panic in the crisis stage. The cycle emerges even if (in fact
precisely because) every agent is acting rationally from his individual point of view. It is
analogous to a prisoner’s dilemma, where individual rationality leads to collective
disaster. The government must intervene.
The intervention becomes all the more necessary if agents are irrationally exuberant
and then irrationally panicked, or are prone to short-sighted greed, or to the keeping up
with the Jones syndrome. If greedy investors want higher expected returns, no matter
what the risk, competition will force even conservative fund managers into leveraging
more. For example, an investor comes to a hedge fund and says “the fund down the
block is getting higher returns.” The fund manager counters that the competitor is just
using more leverage. The investor responds, “well whatever he's doing, he's getting
higher returns.” Pretty soon both funds are leveraging more. Housing prices can rise in
the same way. When some families borrow a lot of money to buy their houses, housing
prices rise and even conservative homeowners are forced to borrow and leverage so they
too can live in comparable houses, if keeping up with their peers is important to them. If
I had to name one source of irrationality that exacerbated this leverage cycle, I would not

6

point to homeowners who took out loans they could not really afford, but rather to
lenders who underestimated the put option and failed to ask for enough collateral.
The observation that collateral rates are even more important outcomes of supply and
demand than interest rates, and even more in need of regulation, was made over 400
years ago. In the Merchant of Venice, Shakespeare depicted accurately how lending
works; one has to negotiate not just an interest rate but the collateral level too. And it is
clear which of the two Shakespeare thought was the more important. Who can remember
the interest rate Shylock charged Antonio? But everybody remembers the pound of flesh
that Shylock and Antonio agreed on as collateral. The upshot of the play, moreover, is
that the regulatory authority (the court) intervenes and decrees a new collateral level -very different from what Shylock and Antonio had freely contracted -- a pound of flesh,
but not a drop of blood. The Fed too should sometimes decree different collateral levels
(before the fact, not after as in Shakespeare).
The modern study of collateral seems to have begun with Kiyotaki and Moore (KM
1997), Bernanke, Gertler, and Gilchrist (BGG 1996 and 1999), Holmstrom and Tirole
(HT 1997), and Geanakoplos (1997), (2003), and Geanakoplos-Zame (GZ unpublished).8
BGG and HT emphasized the asymmetric information between borrowers and lenders as
the source of limits on borrowing. For example, HT argued that the managers of a firm
would not be able to borrow all the inputs necessary to build a project, because lenders
would like to see them bear risk, by putting their own money down, to guarantee that they
exert maximal effort. KM and I studied the case when the collateral is an asset such as a
mortgage security, where the buyer/borrower using the asset as collateral has no role in
managing the asset, and asymmetric information is therefore not important. The key
difference between KM (1997) and Geanakoplos (1997) is that in KM there is no
uncertainty, and so the issue of leverage as a ratio of loan to value does not play a central
role; to the extent it does vary, leverage in KM (1997) goes in the wrong direction,
getting higher after bad news, and dampening the cycle. In Geanakoplos (1997, 2003) I
introduced uncertainty and solved for equilibrium leverage and equilibrium default rates;
I showed how leverage could be determined by supply and demand, and how under
certain conditions, volatility (or more precisely, the tail of the asset return distribution)
pinned down leverage. In Geanakoplos (2003) I introduced the leverage cycle in which
changes in the volatility of news lead to changes in leverage.. During periods of scary bad
news, asset prices fall because of the bad news, because leverage plummets, and because
leveraged optimists go bankrupt. This line of research has been pursued by Gromb and
Vayanos (2002), Fostel-Geanakoplos (2008), Brunnermeier and Pedersen (2009), and
Adrian and Shin (2009), among others.

Ia. Investor Heterogeneity, Equilibrium Leverage, and Default

8

Minsky (1986) was a modern pioneer in calling attention to the dangers of leverage. But to the best of my
knowledge, he did not provide a model or formal theory. Tobin and Golub (1998) devote a few pages to
leverage and the beginnings of a model.

7

Without heterogeneity among investors, there would be no borrowers and lenders, and
asset prices would not depend on the amount of leverage in the economy. It is interesting
to observe that the kind of heterogeneity influences the amount of equilibrium leverage,
and hence equilibrium asset prices, and equilibrium default.
When investors differ only in their optimism about future events in a one dimensional
manner, then the equilibrium leverage will consist of the maximum promise that does not
permit default.9 For example, suppose an asset will be worth either 1 or .2 next period.
Suppose further that risk neutral investors differ only in the probability h that they assign
to the outcome being 1. The most optimistic investor h = 1 is sure that the asset will be
worth 1, and the most pessimistic investor h = 0 is sure the asset will be worth .2. At any
asset price p the investors with h big enough that h*1 + (1-h)*(.2) > p will want to buy
the asset, while the rest will want to sell the asset. The buyers with high h will want to
borrow money in order to get their hands on what they regard as cheap assets, while the
sellers with low h will not need the money and so will be willing to lend. How much
will the borrowers be able to promise using the asset as collateral, assuming the promise
is not contingent on the state? The answer is .2, precisely the maximum promise that
does not lead to default in either state.10
Thus when the heterogeneity stems entirely from differences in opinion, equilibrium
leverage entails no default. A consequence of this is that the loans will be very short
term. The longer the maturity of the loan, the more that can go wrong in the meantime,
and therefore the smaller the loan amount can be if it avoids any chance of default.
Investors who want to borrow large amounts of money will be driven to borrow very
short term.
The Repo market displays these characteristics of short, one-day loans, on which there
is almost never any default, even in the worst of crises.
Much the same analysis holds when investors differ only in their risk aversion. For
the most risk-averse investors, an asset that pays 1 or .2 will be regarded as too
dangerous, while investors with greater risk tolerance will find it attractive at the right
price. These risk-tolerant investors will leverage their purchases, by borrowing money to
buy the asset, using it as collateral for their loan. Once again the equilibrium leverage
will rise to the point that the promises made will be (.2,.2) but no more. To be more
concrete, suppose contrary to the previous case, that all the agents regard the outcomes 1
and .2 as equally likely. But suppose that untraded endowments rise and fall together
with the asset payoffs. Then risk-averse agents on the margin will regard an extra penny
when the asset pays 1 as less valuable than an extra penny when the asset pays .2; on the
9

See Geanakoplos (2003).
At first glance it would seem that the most optimistic buyers might be willing to promise say .3 in both
states, in order to get more money today to invest in a sure winner of an asset. But since this promise will
deliver .3 in the good state but only .2 in the bad state (assuming no-recourse collateral), the lenders will
not want to pay much for this debt: this risky debt is very much like the asset they do not want to hold, and
so they will pay very little more for it than the (.2,.2) promise. Since the borrowers would have to give up
.3 > .2 in the state they think is likely to occur, they will choose to use their scarce collateral for the (.2,.2)
promise instead of the (.3,.3) promise.

10

8

margin they would prefer a penny when the asset pays .2. Hence they will behave as if
they regarded the payoff of 1 as less likely, exactly the same way the pessimists behaved,
despite having the same beliefs as the risk-tolerant agents. Equilibrium leverage with
heterogeneous risk aversion becomes the same as with heterogeneous beliefs.
The situation changes when some investors simply like owning the asset for its own
sake in the period they buy it, such as when a homeowner likes living in the house. A
similar situation arises if a producer can get more output from the asset than can be
recovered if the lender takes it over. Somewhat surprisingly, in these cases the optimal
leverage might be to promise (1,1) even when the asset will only deliver (1,.2) with
probabilities everyone agrees on. If there are multiple states, and a cost of seizing the
collateral, then the equilibrium promise will be somewhere between the maximum and
minimum delivery. Contrary to the previous two cases, equilibrium leverage will involve
a distinctly positive probability of default.
Traditionally the mortgage market has always recognized a substantial probability of
default. And mortgage loans are traditionally of much longer duration than Repo loans.
Ib. The crisis stage
The crisis stage of the leverage cycle always seems to unfold in the same way. First
there is bad news. That news causes asset prices to fall based on worse fundamentals.
Those price declines create losses for the most optimistic buyers, precisely because they
are typically the most leveraged. They are forced to sell off assets to meet their margin
restrictions, even when the margins stay the same. Those forced sales cause asset prices
to fall further, which makes leveraged buyers lose more. Some of them go bankrupt. And
then typically things shift: the loss spiral seems to stabilize—a moment of calm in the
hurricane’s eye. But that calm typically gives way when the bad news is the scary kind
that does not clarify but obscures the situation and produces widespread uncertainty and
disagreement about what will happen next. Suddenly lenders increase the margins and
thus deliver the fatal blow. At that point even modestly leveraged buyers are forced to
sell. Prices plummet. The assets eventually make their way into hands that will take
them only at rock bottom prices.
During a crisis, margins can increase 50% overnight, and 100% or more over a few
days or months. New homeowners might be unable to buy, and old homeowners might
similarly be unable to refinance even if the interest rates are lowered. But, holding long
term mortgages, at least they do not have to put up more cash. For Wall Street firms the
situation is more dire. They often borrow for one day at a time on the Repo market. If the
next day the margins double, then they immediately have to double the amount of cash
they hold for the same assets. If they don't have all that cash on hand, they will have to
sell the assets. This is called de-leveraging.
All this would happen even if traders were completely rational, processing information
dispassionately. When we add the possibility of panic and the turmoil created by more
and more bankruptcies, it is not surprising to see lending completely dry up.

9

Ic. The Aftermath of the Crisis
After the crisis ends, many businesses and individuals will be broke and unemployed.
Parts of the economy will be disrupted, and some markets may be on the verge of
shutting down. The government will then face the choice of who to bail out, and at what
cost. The bailout will typically be very inefficient, causing further losses to economic
productivity. Doubts about which firms will survive will create more uncertainty,
contributing to a difficult lending environment.
Id. What is so Bad about the Leverage Cycle
The crisis stage is obviously bad for the economy. But the leverage which brings it on
stimulates the economy in the good times. Why should we think the bad outweighs the
good? After all, we are taught in conventional complete markets economics that the
market decides best on these sorts of trade-offs. In Geanakoplos (2009) I discussed nine
reasons why the leverage cycle may nevertheless be bad for the economy.
First, very high leverage means that the asset prices are set by a small group of
investors. If agent beliefs are heterogeneous, why should the prices be determined
entirely by the highest outliers? In the current crisis, as I said earlier, the $2.5 trillion of
toxic mortgage securities were purchased with about $150 billion in cash and $2.35
trillion in loans. As of 2006 just two men, Warren Buffet and Bill Gates, between them
had almost enough money to purchase every single toxic mortgage security in the whole
country! So few people should not have so much power to determine crucial prices.
Leverage allows the few to wield great influence.11
Second, asset prices can have a profound effect on economic activity. As James Tobin
argued with his concept of Q, when the prices of old assets are high, new productive
activity, which often involves issuing financial assets that are close substitutes for the old
assets, is stimulated. When asset prices are low, new activity might grind to a halt.12
When asset prices are well above the complete markets price, because of the expectation
by the leveraged few that good times are coming, a huge wave of overbuilding usually
results. In the bad state that overbuilding needs to be dismantled at great cost.
Third, a large group of small business people who cannot buy insurance against
downturns in the leverage cycle can easily sell loans to run their businesses or pay for
their consumption in the good times at the height of the leverage cycle, but have a hard
time at the bottom. Government policy may well have the goal of protecting these people
by smoothing out the leverage cycle.13
11

Standard economics does not really pay any attention to the case where agents have different beliefs, and
median beliefs are closer to the truth than extreme outliers. This consideration motivates reasons 1 and 2.
12
See Tobin and Golub (1998).
13
Here I am relying on Tobin’s Q and the absence of insurance markets. The small businessmen cannot
insure themselves against the crisis stage of the leverage cycle. In conventional complete markets
economics, they would be able to buy insurance for any such event.

10

Fourth, the large fluctuations in asset prices over the leverage cycle lead to massive
redistributions of wealth and changes in inequality. A buyer who leverages his purchase
λ times makes λ% return on his cash for every 1% rise in the asset price; but he loses λ%
of the cash he put down for every 1% decline in the asset price. When leverage λ = 30,
there can be wild swings in returns and losses. In the ebullient stage, the optimists
become rich as their bets pay off, while in the down states they might go broke.
Inequality becomes extreme in both kinds of states.14
This brings us to the fifth potential cost of too much leverage. Instead of regarding the
optimists as crazy, let us think of them as indispensable to the economy, or at least more
valuable to the economy even than they are to themselves. That is probably what is meant
by the term "too big to fail". If their marginal contribution to society is bigger than what
they are paid, then their bankruptcy results in an externality, since they internalize only
their private loss in calculating how much leverage to take on. This happens for example
when managers of a firm calculate their own loss in profits in the down states, but neglect
to take into their calculations the disruption to the lives of their workers when they are
laid off in bankruptcy. If in addition the bankruptcy of one optimist makes it more likely
in the short run that other optimists will go bankrupt, perhaps starting a chain of defaults,
then the externality can become so big that simply curtailing leverage can make
everybody better off.
Sixth, debt overhang destroys productivity. Banks and homeowners and others who
are underwater often forego socially efficient and profitable activities. For example, a
homeowner who is underwater loses much of the incentive to repair a house, even if the
cost of the repairs is less than the gain in value to the house, since increases in the value
of the house will not help him if he will be foreclosed eventually anyway.
Seventh, seizing collateral often destroys a significant part of its value. The average
foreclosure of a subprime loan leads to recovery of only 25% of the loan, after all
expenses and the destruction of the house are taken into account.
Reasons six and seven show that default leads to efficiency losses. If it were possible
to arrange contingent loans, it would be better to have loans that automatically got
smaller in bad states, eliminating these losses. We shall come to this later. But one
might argue that in a world absent such loans, it may be so important to get the borrower
the money, and the crisis might ex ante be so unlikely, that it is “second best” to go ahead
with the big leverage and bear the cost of the unlikely crisis. After all, the losses in six
and seven are losses the borrowers and lenders should have anticipated and internalized.
Thus I augment reasons six and seven with two more reasons that compound them.
Eighth, in an effort to mitigate the crisis, the government often intervenes in
inefficient ways. In the current crisis the government is propping up the banks by
holding interest rates inefficiently low at zero. And worse still, the government has
botched the mortgage foreclosure relief situation, which I shall discuss in detail later.
14

This is a purely paternalistic reason for curtailing leverage.

11

The agents in the economy do not take into account that by putting the economy more at
risk, they create more inefficient government interventions. And of course the
expectation of being bailed out by the government, should things go wrong, causes many
agents to be more reckless in the first place.15
Ninth, a key externality that borrowers and lenders on both the mortgage and repo
markets at the high end of the leverage cycle do not recognize is that if leverage were
curtailed, prices would rise less in the ebullient stage and fall much less in the crisis.
Foreclosure losses would then be less, and so would inefficiencies caused by agents being
so far underwater. Limits to leverage in the good times in effect would provide insurance
for investors in the bad times who we could imagine need to sell promises in order to
start new building, but who are unable to buy the insurance directly because of the
missing markets.

II. The Leverage Cycle of 2000-2009 Fits the Pattern
IIa. Leverage and Prices
By now it is obvious to everybody that prices soared from 1999 (or at least post the
disaster period after 9/11/2001) to 2006, and then collapsed from 2007-9. My thesis is
that this rise in prices was accompanied by drastic changes in leverage, and was therefore
just part of the 1999-2006 upswing in the leverage cycle after the crisis stage in 1997-98
at the end of the last leverage cycle. I do not dispute that irrational exuberance and then
panic played a role in the evolution of prices over this period, but I suggest that they may
not be as important as leverage; certainly it is harder to regulate animal spirits than it is
leverage.
Let us begin with the housing bubble, famously documented by Robert Shiller. In the
graph below I display Shiller’s national housing index for 2000-2009. It begins at 100 in
Q1 2000, reaches 190 in Q2 2006 and falls to 130 by Q1 2009, as measured on the right
vertical axis. But I superimpose on that graph a graph of leverage available to
homeowners each month. This is measured on the left vertical axis and labeled by %
downpayment, which is 100% - LTV. To compute this I began by looking house by
house each month from 2000-2009 at the ratio of all the outstanding mortgage loans
(usually a first and sometimes a second lien) to the appraised value of the house at the
moment a first mortgage was issued for every subprime and Alt A house available in the
Loan Performance Data Base. I then averaged over the 50% houses with the highest loan
to value levels. 16 In this way I get a robust estimate of leverage offered to homeowners.
By leaving out the bottom 50% I am ignoring homeowners who clearly chose to leverage
less than they could have, and by including all homes in the top 50% I am ensuring that
the leverage measure was really available and not just a special deal for a few outliers. If
anything my numbers underestimate the offered leverage.
15

This mechanism has been formalized in Farhi-Tirole (unpublished).
This data was compiled and analyzed by the research team at the hedge fund Ellington Capital
Management.

16

12

Housing Leverage Cycle
Margins Offered (Down Payments Required) and Housing Prices

190

2%
4%

170

6%
8%

150

10%
12%

130

14%
16%

Case Shiller National HPI

Down Payment for Mortgage -- Reverse Scale

0%

110

18%
90
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2

20%
2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Avg Down Payment for 50% Lowest Down Payment Subprime/AltA Borrowers
Case Shiller National Home Price Index (right axis)
Observe that the Down Payment axis has been reversed, because lower down payment requirements are correlated with higher home
prices.
Note: For every AltA or Subprime first loan originated from Q1 2000 to Q1 2008, down payment percentage was calculated as
appraised value (or sale price if available) minus total mortgage debt, divided by appraised value. For each quarter, the down payment
percentages were ranked from highest to lowest, and the average of the bottom half of the list is shown in the diagram. This number is
an indicator of down payment required: clearly many homeowners put down more than they had to, and that is why the top half is
dropped from the average. A 13% down payment in Q1 2000 corresponds to leverage of about 7.7, and 2.7% down payment in Q2
2006 corresponds to leverage of about 37.
Note Subprime/AltA Issuance Stopped in Q1 2008.

21

It is striking how correlated prices and leverage are, rising and then falling together.
Especially noteworthy is that the leverage peaks in 2006, with less than 3% down, exactly
when housing prices peak, and heads down much faster than housing prices.
In the next graph I present the history of the J.P. Morgan AAA prime floater mortgage
index from about 2000-2009. The index is measured on the right vertical axis. These
prime mortgages are taken out by investors with pristine credit ratings, and they are also
protected by some equity in their deals. For most of its history this index stays near 100,
but starting in early 2008 it falls rapidly, plummeting to 60 in early 2009. The
cumulative losses on these prime loans even today are still in the single digits; it is hard
to imagine them ever reaching 40%. It is of course impossible to know what people
were thinking about potential future losses when the index fell to 60 in late 2008 and
early 2009. My hypothesis is that leverage played a big role in the price collapse.
On the left vertical axis I give the loan to value, or equivalently the down-payment or
margin, offered by Wall Street banks to the hedge fund Ellington Capital Management on
13

a changing portfolio of AAA mortgage bonds.17 As I said earlier, it is astonishing that
the Fed itself does not have such historical data, and is apparently not keeping it even
now. Fortunately the hedge fund Ellington that I have worked with the past 15 years
does keep its own data. The data set is partly limited in value by the fact that the data
was only kept for bonds Ellington actually followed, and these changed over time. Some
of the variation in average margin is due to the changing portfolio of bonds, and not to
changes in leverage. But the numbers, while not perfect, provide substantial evidence for
my hypothesis and tell a fascinating story. In the 1997-98 emerging markets/mortgage
crisis margins shot up, but quickly returned to their previous levels. Just as housing
leverage picked up over the period after 1999, so did security level leverage. Then in
2007 leverage dramatically fell, falling further in 2008, and leading the drop in security
prices. Very recently leverage has started to increase again, and so have prices.

100.0

10
20

90.0

30
80.0
40
50

Price

Margin % (Down Payment Required to Purchase Securities) Reversed Scale

Securities Leverage Cycle
Margins Offered and AAA Securities Prices

70.0

60
60.0
70
80
6/1/98

10/14/99

2/25/01

7/10/02

11/22/03

4/5/05

8/18/06

12/31/07

50.0
5/14/09

Average Margin on a Portfolio of CMOs Rated AAA at Issuance
Estimated Average Margin
Prime Fixed Prices
Note: The chart represents the average margin required by dealers on a hypothetical portfolio of bonds subject to
certain adjustments noted below. Observe that the Margin % axis has been reversed, since lower margins are
correlated with higher prices.
The portfolio evolved over time, and changes in average margin reflect changes in composition as well as changes
in margins of particular securities. In the period following Aug. 2008, a substantial part of the increase in margins is
due to bonds that could no longer be used as collateral after being downgraded, or for other reasons, and hence
count as 100% margin.

1

IIb. What Triggered the Crisis?
The subprime mortgage security price index collapsed in January of 2007. The stock
market kept rising until October 2007, when it too started to fall, losing eventually around
50% of its value before rebounding to within 25% of its October peak. What, you might

17

These are the offered margins, and do not reflect the leverage chosen by Ellington, which since 1998 has
been drastically smaller than what was offered.

14

wonder, was the cataclysmic event which set prices and leverage on their downward
spiral?
The point of my theory is that the fall in price from scary bad news is naturally going
to be out of proportion to the significance of the news, because the scary bad news
precipitates and feeds a plunge in leverage. A change in volatility, or even in the
volatility of volatility, is enough to prompt lenders to raise their margin requirements.
The data show that that is precisely what happened: margins were raised. But that still
begs the question, what was the news that indicated volatility was on the way up?
One obvious answer is that housing prices peaked in mid 2006, and their decline was
showing signs of accelerating in the beginning of 2007. But I do not wish to leave the
story there. Housing prices are not exogenous; they are central to the leverage cycle. So
why did they turn in 2006?
IIc. Why did Housing Prices Start to Fall?
Many commentators have traced the beginning of the subprime mortgage crisis to
falling housing prices. But they have not asked why housing prices started to fall.
Instead they have assumed that housing prices themselves, fueled on the way up by
irrational exuberance and on the way down by a belated recognition of reality, were the
driving force behind the economic collapse.
I see the causality going in the other direction, starting with the turnaround in the
leverage cycle. The leverage cycle was of course greatly exacerbated by the terrible
consequences of falling housing prices, which then fed back to cause further housing
declines.
As I hope I have made clear, in my view housing prices soared because of the
expansion of leverage. Greater leverage enabled traditional buyers to put less money
down on a bigger house, and therefore pushed up housing prices. It also enabled people
to buy houses who previously did not have enough the cash to enter the market, pushing
housing prices up still further.
There is, however, a limit on how much leverage can increase, and on how many
new people can enter the market. Though negative amortizing loans pushed the
envelope, no money down is a natural threshold beyond which it is hard to move. And as
more and more households with less and less money entered the market, lenders began to
become apprehensive that these people were less reliable. The rapidly expanding supply
of new housing demand, fueled by access to easy mortgages, began to slow for
completely rational reasons, not because of a sudden pricking of irrational exuberance.
This naturally led to a peak in housing prices in 2006.
In my view the trigger that started the market falling from its peak was the creation of
standardized CDS contracts on mortgages in 2005, at the very height of the market,
which enabled CDS to be traded in large quantities on subprime mortgages in 2006. The

15

CDS is like an insurance contract paying if a particular bond (or index of bonds) defaults
on part of its principal: if a homeowner mortgage default leads to a loss of capital of $1 in
some BBB bond, then a single CDS contract on the BBB bond requires a payment of $1.
The CDS payment, however, could be much more, because one speculator might have
written 4 contracts to another speculator or series of speculators, effectively agreeing to
pay $4 for every $1 somebody else lost on the BBB security. Thus there can be a
tremendous magnification of gross losses by firms making bad CDS bets.
As I mentioned in Section Id, one important consequence of a leveraged purchase of
assets is that when the underlying asset moves 1% in value, the buyer leveraged λ >> 1
times suffers a gain or loss in his cash position of λ%. Exactly the same multiplier
applies to the CDS contract when the insurance amount is a multiple λ of the bond
principal: a 1% change in the probability of default, which means basically a 1% change
in the value of the bond, causes a λ% swing in the CDS contracts. The difference is that
now instead of just the buyers of bonds bearing the big risk, both buyers and sellers of
CDS are exposed to swings. In this Section we shall focus on the implication of the
buyers leveraging through CDS; in Section IIIc we shall focus on the sellers leveraging
through CDS contracts.
CDS allowed pessimists to leverage their views on assets for the first time, by taking
out CDS insurance contracts on bonds they did not even own. For the optimists taking
the other side of the CDS market this was not a new opportunity; they were essentially
just increasing their leverage ratios on the underlying assets. (Many optimists were
delighted to do so, with the result that they were more vulnerable to bad news, which
would lead to steeper declines once the bad news hit, as I discuss more in Section IIIc.)
But the point here is that some optimists, who felt they had already leveraged enough,
switched from leveraging through asset purchases to leveraging by writing CDS
insurance contracts on those same assets. To the extent that there were such switchers,
the emergence of the CDS market induced some asset holders to sell and thus put
downward pressure on asset prices even before any bad news hit.
In the leverage cycle prices soar because the most optimistic investors can leverage
their views through huge borrowing. The pessimists cannot express their views because
it is difficult to sell short. Once they can, in effect, sell short via the CDS, prices must
reflect their views and not just the views of the leveraged optimists. I do not have access
to data documenting the size of the CDS market, and how rapidly it grew in 2005 and
2006, but I suspect there lies a key to the crisis that has heretofore been overlooked.
The downward pressure on bond prices from CDS meant that the same
securitizations became more difficult to underwrite. Securitizers of new loans looked for
better loans to package in order to continue to back bonds worth more than the loan
amounts they had to give homeowners. They asked for loans with more collateral. Thus
not only did the trend to more housing leverage stop, it began to reverse in Q2 2006.
This had enormous implications. First, it meant that potential new homeowners
began to be closed out of the market, which of course reduced home prices. But more

16

insidiously, increasing margins kept homeowners from refinancing because they simply
did not have the cash to make the dowpayment on a new loan. Until 2007, subprime
bond holders could count on 70% or so of subprime borrowers refinancing between their
second and third years in the loan. These homeowners began in pools that paid a very
high rate of interest because of their low credit rating. But after two years of reliable
mortgage payments, they would become eligible for new loans at better rates, which they
traditionally took in vast numbers. Of course a prepayment means a full payment to the
bondholder. Once downpayment requirements tightened, refinancing plummeted and this
sure source of cash disappeared; the bonds became much more at risk and their prices fell
more. Mortgagees who had anticipated being able to repay were trapped in their original
loans at high rates; many became delinquent and entered foreclosure.
Foreclosures obviously lead to forced sales and downward pressure on housing
prices. And falling home prices are a powerful force for further price reductions, because
when house values fall below the loan amount, homeowners lose the incentive to repay
their loans, leading to more defaults, foreclosures and forced selling. All this leads to
falling security prices, making it even harder to issue new loans unless they involve
higher and higher downpayments, which makes new home buying more difficult, which
again forces housing prices down. And of course this leads back to tighter margins on
securities.
The feedback from waffling security prices to higher margins on housing loans to
lower house prices and then back to tougher margins on securities and then lower security
prices is what I call the double leverage cycle.
Bond prices for 2006 vintage subprime loans began to fall in November of 2006
with the smallest trickle of bad news about homeowner delinquencies, then spiked
downward in January 2007 after the year end delinquency report. This collapse of
pricing is a powerful illustration of the potency of market expectations. The actual losses
on subprime loans at that point were about 1% or less, yet the market was already
anticipating huge losses on the order of 10%. Recall that for a pool to lose 10% of its
value, the market must anticipate something like 30% of the homeowners will be thrown
out of their houses, with 30% losses on each home sold. This first market crash should
have been enough to alert our government to the looming foreclosure disaster, but three
years later we still have not taken decisive action to mitigate foreclosures.

III. Why this Leverage Cycle is the Worst since the Great Depression
Every leverage cycle has the problems just listed. The crisis stage of every leverage
cycle is bad. But the current crisis is far worse than the crises we saw in the two previous
leverage cycles. There are a number of reasons why this cycle is worse than all previous
cycles since the Depression, but the unifying theme behind all of them is a failure to put
up enough collateral to back promises.
IIIa. Securities Leverage got Higher then Fell Farther than ever Before

17

In this cycle, leverage on traditional collateralizable assets increased to more than the
highs from the previous cycle. One reason for this is that the lull since the previous
leverage crisis included a period of especially low volatility, which led margins to sink
lower and lower. That can be seen in the history of one mortgage hedge fund’s margins
(haircuts) over the last 11 years, given in the diagram in Section IIa. Note that before the
crisis of 1997-1998 that ended the last leverage cycle, leverage was about 10 to 1
(margins were about 10%). During the 1998 crisis margins jumped to 40%, staying there
about two months, before returning to their previous levels of 10%. In the “great
moderation” in the 9 years afterward, leverage increased from about 10 to 1 to about 20
to 1 (the margins fell from 10% to 5%).
In 2007 leverage collapsed, with margins going from 5% to 70% on average. Two
years after the collapse leverage is still low, whereas in 1998 the crisis was all over in two
months.
The most dramatic change in margins has come from assets that were rated AAA, and
which have been, or are about to be, downgraded. Previously one could borrow 90 or
even 98.4 cents on a dollar’s worth of AAA assets, and now one cannot borrow anything
at all with these assets as collateral. According to Moody’s, AAAs are supposed to have
a 1 in 10,000 risk of default over a 10 year period. We are now seeing over 50% of all
Alt-A and subprime AAA bonds partially defaulting, and we will see virtually 100% of
all AAA CDOs partially default. Even when some assets have little or no chance of
losing more than a few percent of their value, the market no longer trusts the AAA rating,
and lenders will not lend even one percent of their current price.
In the runup to the present crisis, many new kinds of assets became usable as
collateral. Thus, even if margins had not declined on old collateral, the leverage of the
economy as a whole would have increased because there was new borrowing backed by
previously unusable collateral, which brings us to pooling and securitization.
The process of pooling and securitization has been a crucial source of new collateral
and increased leverage. Imagine a single subprime mortgage loan. Even in the days
when it was believed that the expected loss from such a mortgage was between 1% and
4%, people still recognized that there was a non-trivial chance of a much bigger loss on a
single loan. Lenders, inherent pessimists, would not have considered lending using a
single subprime mortgage as collateral. But now consider a pool of subprime mortgages
from around the country. If one believed that the loans were independent, so that a
housing price decline in Detroit did not imply a housing price decline in California, then
on a big enough pool of loans, the chance for more than 30% losses might be considered
less than 1 in 10,000. Even a very pessimistic lender who believed in a 4% expected loss
per loan would be willing to lend 70% of the value of the entire pool, provided that he got
paid before anyone else. Thus a buyer of the pool of mortgages could imagine borrowing
70% of their collective value, when it would have been impossible to borrow anything on
the individual loans.

18

Securitization took this borrowing on pools one step further by converting the
loans into long term loans. The underwriter of the pool typically issued different bonds,
whose payments depended on the homeowners’ payments on their loans. Consider for
example a bond structure with just two “tranches’ or bonds. The senior tranche might
pay interest slightly above the riskless government rate on the best 70% of the loans. As
long as losses on the pool are below 30%, the senior tranche holder continues to get paid
his interest and eventually his principal. The junior bondholder receives what is left from
the pool after the senior bond holder is paid. The whole securitized structure can be
interpreted as if the buyer of the junior piece actually bought the whole pool, using a long
term loan from the buyer of the senior piece, collateralized by the whole pool. Once one
understands the juniors as effectively borrowing from the seniors, it becomes clear how
the rapid spread of securitization over the last 30 years, but especially over the last 10
years, dramatically increased the leverage in the system.
IIIb. Housing and the Double Leverage Cycle
Leverage on houses got to be much higher in this leverage cycle. In the recent
leverage cycles, ending in 1994 and 1998, homeowner leverage did not get remotely as
high as it did in the recent cycle. In 2006 many homeowners were borrowing with
basically no money down, or as little as 3%.18 New mortgages like option arms were
invented which abetted this mad rush to loan homeowners all or nearly all of the purchase
price.
Thus the current cycle is really a double leverage cycle: not only are the mortgage
securities subject to the leverage cycle, but their ‘fundamental” cash flows (namely
homeowner mortgage payments) are also subject to the leverage cycle. These two cycles
feed off each other. When margins are raised on homeowners, it becomes more difficult
to get a new mortgage and home prices fall, jeopardizing mortgage securities backed by
houses. But more importantly, it becomes more difficult for homeowners to refinance
their old loans, putting these loans and the securities they back in much more jeopardy of
defaulting. Similarly, when margins on securities are raised and their prices fall, then in
order to sell the securities for higher prices, underwriters demand better underlying
mortgages, i.e., more money down for homeowners.
IIIc. Credit Default Swaps (CDS)
The current cycle has been more violent because of the creation of the derivative
credit default swap (CDS) market for mortgages in 2005, just at the top of the leverage
cycle. One reason, as we discussed earlier, is that CDS allowed pessimists to leverage at
just the worst time. Once CDS emerged, they were bound to put downward pressure on
18

It is deeply troubling that the government thru the FHA is now again offering mortgages with less than
3% down, which is the effect of the FHA minimum of 3.5% downpayment combined with the tax credit for
homebuyers passed by the Congress. This is the wrong way to go about supporting home prices, a goal I
support but not through methods that recreate the leverage problem that got us here. In Section IV I discuss
methods of pre
venting and reversing an overcorrection in home prices that do not rely on overleveraging.

19

prices, because they allowed pessimists to express their views and indeed leverage those
views. Had the CDS market for mortgages been around from the beginning, asset prices
might never have gotten so high. But their appearance at the very top of the cycle
guaranteed that there would be a fall.
A second reason that CDS made the fall much worse is because they allowed
optimists to leverage even more. To the extent that CDS did not lower prices before any
bad news, it was because leveraged optimists increased their leverage by taking the other
side of the CDS, on top of their leveraged purchases of the underlying assets. But this
made the optimists lose more and the crash much bigger once the bad news hit. CDS is a
kind of insurance market for bond defaults, but instead of cushioning losses, it made them
much worse because the buyers of the bonds did not buy the insurance, they sold it.19
One might mistakenly think that CDS should just wash out. In other words, for
every dollar lost on the insurance, there should be a dollar gained by the recipient. But
the optimistic writers of insurance are very different from the pessimistic buyers of
insurance. When the bad news hits, the former lose and must reduce their purchases of
assets; the latter gain, but still won’t buy the assets. Writers of CDS insurance expose the
economy to the same problems of excessive leverage I described earlier.
This brings us to the question of just how much leverage one could obtain via the
CDS market. One difference is that a leveraged buyer of an asset has to make his
downpayment in cash; this is what limits his leverage. With CDS it now appears that
many firms, like AIG, were allowed to make naked bets, without any credible showing of
collateral to back up their promise to pay in the event the default they were “insuring”
against came to pass. If enough collateral had been put up by AIG, there would have
been no reason to bail it (or more to the point, all its counterparties) out. In retrospect it
seems that in some critical markets leverage knew no bounds.
The failure of some buyers of CDS insurance to insist on proper collateral from
the writers of the insurance was made far worse because the gains and losses from CDS
are not netted. A firm F that was neutral, betting one way against firm A on tranche
BBB, and betting the opposite way on the same tranche against firm C could come out a
loser anyway. If firm A defaults on its insurance payment, then F will be unpaid by A
but still on the hook for paying C. So instead of just one firm A going bankrupt and
another firm C going unpaid in the absence of collateral, as would happen with netting,
another firm F might also go bankrupt, closing shop, firing workers, and creating other
social costs.
Losses by leveraged buyers of assets can cause a chain reaction when a margin
call forces a leveraged buyer to sell, which lowers the price forcing another leveraged
buyer to sell and so on. But with uncollateralized CDS the chain reaction is more direct,
unmediated by market price. The implication I shall draw later is that CDS should be
19

Of course there were undoubtedly some hedge funds who bought bonds they thought were undervalued,
and bought insurance on similar bonds in order to hedge their position against the risk of a market
downturn. These are the leveraged buyers who survived the crisis without a bailout.

20

traded on an exchange instead of in bilateral contracts, both to ensure that collateral is
always posted by the writer of the insurance, and to make sure losses are netted.20
There is another reason that there should be a limit on the size of the CDS positions
people hold. In theory rational agents should be allowed to make bets of arbitrary size on
exogenous events. But the CDS events are far from exogenous. Consider a credit default
swap for a trillion dollars on a corporate bond promising a billion dollars. The writer of
the insurance has every incentive to buy the whole failing corporation and pay off its
bond holders one billion dollars rather than pay the trillion dollars of default insurance.
Thus the holders of insurance can never be sure they will get their money.
In traditional insurance law, as I understand it, there is a prohibition against overinsuring by taking out insurance for more than the underlying asset, precisely because of
the moral hazard such practices entail. Similar prohibitions should be adopted for CDS.

IIId. Counterparty Risk
In bilateral CDS contracts it was often the case that the insurer did not post
enough collateral to guarantee payment. This CDS problem illustrates a more general
flaw in the whole system of contracting on Wall Street. These contracts to a great degree
were written in such a way that only one side of every transaction was presumed liable to
default, so that only the other side needed protecting. For example, in the Repo market a
hedge fund borrower gets a loan from an investment bank, and puts up collateral at the
bank worth more than the loan. The investment bank is protected against the potential
default of the hedge fund, because in that event the collateral can be sold to recover the
loan amount. But the contract does not contemplate the bankruptcy of the investment
bank. What recourse does the hedge fund have if the investment bank goes out of
business, shutting its doors and swallowing the collateral security? Traders who never
before had to give a second thought to these counterparty risk questions suddenly had to
reevaluate all their contracts, with disastrous effects on liquidity and price discovery.
Now, this unplanned-for counterparty risk has become the primary rationale for
the government’s seemingly-unending commitment to inject capital into “too-big-to-fail”
institutions. “We can’t afford another Lehman,” is the common refrain; we had to bail
AIG out not because it was so vital, but because if it defaulted a chain reaction might
ensure. I am not at all sure that the chain reaction would have been so devastating, or that
policymakers made a serious calculation of the costs of allowing the defaults to continue.
The unstated and quite questionable underlying premise is that we must be able to afford
whatever it now will cost to protect counterparties completely from the pain that would
otherwise have attended their failure to protect themselves. That is not just doubtful, but
ignores the moral hazard that such a policy creates. I will return to the question of moral
hazard before I conclude.

20

I note with dismay that as of this writing the measures before the House and Senate do not do nearly
enough to regularize CDS trading, allowing significant continuation of over the counter CDS trading.

21

The prospective solution to the counterparty risk problem is to ensure that both
sides put up enough collateral. Of course people are now more alert to their counterparty
vulnerability than they were before, and thus pressure will grow, for example, on Repo
lenders to warehouse the collateral at a third site that would not be compromised by the
bankruptcy of the lender. This raises questions about whether there is enough collateral
in the economy to back all the promises people want to make, which I discuss at length in
Geanakoplos (1997) and Geanakoplos-Zame (unpublished). But I believe there should be
a government initiative to push as many bilateral contracts onto exchanges as possible;
agents trading with the exchange will be forced to put up collateral, and the netting
through the exchange will economize on collateral. As for any finance-related bilateral
contracting so particular that it could not be moved to an exchange, the parties could
either accept strict disclosure requirements and limits on how much of this contracting
they could engage in or accept doing without the instruments altogether. CDS in
particular should be traded on an exchange.
IIIe. Government Laxity, Deregulation, and Implicit Guarantees Increased
Leverage
The mildness and shortness of the crisis stage of the last two leverage cycles, in 1994
and 1998, led many people, perhaps including the regulators, to ignore leverage
altogether. The abrupt tightening of margins in 1998 was explained by the supposed
irrationality of lenders, who it was said, reacted by raising margins after the fact, i.e. after
the fall in prices had already occurred. It appears that virtually no lenders lost money on
loans against mortgage securities in that crisis. The run-up in asset prices and home
prices during the current cycle was attributed mostly to irrational exuberance, instead of
being understood, first and foremost, as an inevitable consequence of the increase in
leverage. Partly as a result of this faulty narrative, the Fed and the other branches of
government did nothing to curtail the dramatic growth in homeowner leverage, or
consumer leverage more generally, or corporate leverage, or securities leverage.21 Banks
were allowed to move assets off their balance sheets and thus avoid capital requirements,
further increasing their leverage.
Not only did the Fed (and everyone else) turn a blind eye to the rising leverage
pervading the system, it encouraged the deregulation that unleashed the leverage inherent
in outsized credit default swaps. As I mentioned earlier, CDS contracts seem on their
face to be either gambling or writing insurance in excess of the value of the property
being insured. Under either interpretation they would have run afoul of state laws
prohibiting gambling or over-insurance. So it took a positive act of Congress, prodded by
Treasury and Fed Chairman Greenspan, to pass legislation exempting CDS from those
limitations.
Perhaps the most important and unwitting government stimulus to the increased
leverage was the implicit government guarantees for entities that were considered too big
21

I note that the Fed was not alone in its lackadaisical approach to leverage. The SEC lifted what weak
leverage limits it had in 2006 and no arm of government took any steps to stop the leverage orgy that led to
the present crisis.

22

to fail. Fannie Mae and Freddie Mac grew bigger and bigger. The presumed government
guarantee on their promises enabled them to leverage their assets to 30 or more, and still
issue debt just above Treasury rates. Without this implicit government backing, they
would never have been able to borrow so much with such little capital. Even more
important than Fannie and Freddie, many investment banks were allowed to write CDS
and give Repos without collateralizing their implicit promises. It seems virtually
inexplicable that Wall Street ignored this counterparty risk, unless it assumed that these
banks were backstopped by the Fed. And indeed after some doubts when Lehman
collapsed, that expectation proved correct.
IIIf. The Rating Agencies effectively increased Leverage
The expansion of the mortgage market into less credit worthy households made it
more likely that a shock would someday be big and bad and scary, creating more
uncertainty and more disagreement. The anticipation of that, however remote the
possibility seemed, should have made lenders nervous and caused them to put a brake on
leverage. This rational concern was dramatically reduced by a foolish faith many
investors had in the rating agencies and their default models, which were widely relied
upon by market participants (and the ratings agencies themselves), but which failed to
account adequately for the probability that defaults in certain circumstances would be
highly correlated. Some investors forgot the incentives of the rating agencies and the
incentives of many market actors to downplay seriously the probability of highly
correlated defaults. In the face of a long history of low defaults and with billions of
dollars of deals waiting on the blessing of a small handful of rating agency actors it
would have been astonishing if ratings had been as tough as they should have been. The
same lesson applies to the mortgage brokers who were able to collect fees for signing up
borrowers without facing any losses themselves if the borrowers defaulted.
IIIg. Global Imbalances increased Leverage
An important factor documented by Ricardo Caballero and others is that the enormous
savings glut coming from Asia increased the demand for safe assets. This presented a
profit opportunity to American financiers, who were thus stimulated to engineer the
securitizations which created apparently safe bonds out of risky assets. It is hard to
assess how important this factor is, but surely a gigantic demand for safe bonds would
indeed give a big incentive to create those bonds and thus inevitably to concentrate more
risk in other bonds. On the other hand, one must acknowledge that the Chinese did not
buy the AAA toxic mortgage securities that have been at the center of the crisis.
IIIh. All Upside Down
The upshot of the huge credit boom and the plunging prices was that an extraordinary
number of households, businesses, and banks ended up upside down or underwater, that
is, with debt exceeding their assets. It is estimated that between 6 and 10 million
homeowners will be thrown out of their houses; the government has assumed trillions of

23

dollars of mortgage debt, spent hundreds of billions of dollars bailing out banks and firms
like AIG, and on account of the huge number of failing banks, the FDIC is on the verge
of running out of money.

IIIi. Why didn’t Wall Street Risk Managers Anticipate the Collapse?
Having discussed many of the factors that exacerbated the crisis of 2007-9, we are
now in a position to assess the widely held view that nobody saw it coming.
Nobody doubts that Wall Street understood that there was considerable risk in
subprime mortgage pools. That is why they were tranched into different tiers, called
AAA, AA, down to BBB. And these bonds were all senior to residual pieces and
overcollateralization, which together provided another 8% of protection. Recall that if
25% of the loans result in homeowners being thrown out of their houses, with 25% losses
on each foreclosed home, that amounts to losses of just 6.25% for the pool as a whole,
which would leave the rated bonds unscathed. So the question is really not whether Wall
Street overlooked the risk, but rather how it came to be that Wall Street so badly
underestimated the size of the risk?
The answer I believe is that it was nearly impossible to foresee the devastating
consequences of the multiple feedbacks between securities and houses embodied in the
double leverage cycle. Complex adaptive systems are notoriously hard to predict.
Contrary to the myth that nobody imagined that housing prices could go down as well as
up, I suspect that virtually every large bank and hedge fund considered a scenario in
which housing prices went down at least 10%. But how many anticipated that at the
same time mortgage downpayments would rise to the point that subprime refinancing
virtually stopped, as did origination, causing further house declines? And that at the
same time servicers and banks would refuse to write down principal, leading to more
foreclosures and further house declines? And that the government would stand by doing
close to nothing about the foreclosure problem for what is now approaching three years?

IV. The Solution to the Crisis: A Multi-Pronged Approach
Once the economy is plunged into circumstances as dangerous as we saw last year, the
government has no choice but to act boldly. The correct course of action is to reverse the
final stages of the crisis and thus stop the panic. At the outset of this crisis I
recommended the three prong approach I present here, a thematic solution to the crisis,
which addresses in order of importance, all aspects of the final stages of the leverage.
As I explained above all leverage cycles end with (1) bad news creating
uncertainty and disagreement, (2) sharply increasing collateral rates, and (3) losses and
bankruptcies among the leveraged optimists. These three factors reinforce and feed back
on each other. In particular, what begins as uncertainty about exogenous events creates
uncertainty about endogenous events, like how far prices will fall or who will go

24

bankrupt, which leads to further tightening of collateral, and thus further price declines
and so on. In the aftermath of the crisis we always see depressed asset prices, reduced
economic activity, and a collection of agents that are not yet bankrupt but hovering near
insolvency. How long the aftermath persists depends on how deep the crisis was and the
quality of the government’s response. Whether we find ourselves in a similar crisis in the
future depends on whether, understanding how leverage got us here, we adopt reforms
that require the Fed to monitor and regulate leverage in the good times. First, I take up
what government actions should have been taken and in what order to address the final
stage of the double leverage cycle that the government was called on to address in 2007.
The thematic solution once the crisis has started is to reverse the three symptoms
of the crisis: contain the bad news, intervene to bring down margins, and carefully inject
“optimistic” equity back into the system. To be successful any government plan must
respect all three remedial prongs, and should be explainable and explained to the public
in terms that it can understand. Without public confidence, which can only flow from
public understanding, any government plan undermines its own objectives and limits its
prospects for success. The government’s actions thus far have not addressed all three
prongs adequately and policymakers have thus far largely failed to explain how their
various solutions are tied to the roots of the crisis we face.
Unfortunately, the government’s first bailout plan was not clearly thought through
and neither it, the ostensible solution, nor the problem that required a solution were
clearly explained. The public, confused, afraid and suspicious, lost faith and prices fell
further. But even now, after the panic has subsided, we must ask who or what is the
government trying to save? Many in the public have come to believe it is merely trying to
save banks, or some big banks, from failure because somehow their failure would signal a
catastrophe for the American brand, to be prevented at all costs. The confusion about the
government’s goals has created its own set of problems, which we can ill afford.
Clarifying the government’s goals will be harder now, but it remains an indispensible
step.
IVa. Step One: Addressing the Precipitating Cause of the Crisis: Scary Bad News
(Massive Uncertainty) About Housing and the Assets Built on Housing
To foster recovery from the dramatic final stage of a leverage cycle as large as the one
we have just experienced, the government must address the cause of the uncertainty that
triggered the end stage. Without that, the efforts taken thus far to bring margins down
and recapitalize banks, even had they been perfectly implemented, would not be enough
to reverse the cycle and restore the economy to health. In this crisis with its roots in
housing that means: doing something for housing prices and homeowners. This makes
undeniable sense in this crisis, not just because addressing the cause of the uncertainty
and disagreement (the scary bad news) is critical to reverse any leverage cycle, but
because the biggest social losses will probably come from the displaced homeowners.
And, of course, the biggest reason for the tumbling mortgage security prices, and the
resulting insolvency of the banking sector, is fear that housing prices will keep falling.

25

IVa1. Saving the Homeowners: Stemming the Tsunami of Foreclosures to Come
One of the saddest stories in this financial meltdown is that millions of homeowners
are being thrown out of their homes for defaulting on their mortgages. Throwing
somebody out of his home is tragic for the homeowner, but also very expensive for the
lender. One of the shocking aspects of the foreclosure crisis is how low the recoveries
have become on foreclosed properties, after expenses. (Interestingly, the mortgage bond
index markets anticipated these bad recoveries.) Nobody gains when the homeowners
are thrown out and the banks and/or investors collect pennies on the dollar for the money
they loaned. Nevertheless, nearly 2 million homeowners have already been evicted,
another 3.2 million are seriously delinquent and almost surely will be evicted in the near
future, and at least another 3 million more will eventually default and be evicted if trends
continue. Without much bolder action than has thus far been taken by the government,
the stream of evictions and bad recoveries for lenders will continue and accelerate,
becoming a torrent that will further depress housing prices and impede economic
recovery.
The single most important reason homeowners are defaulting is not job loss; it is that
their houses are underwater. The following study done at Ellington Capital Management
in February 2009 shows the monthly default rate for homeowners (with various loan
types), as a function of the ratio of the loan amount to the current value of the house.
The study examined every home in the Loan Performance data base, taking the appraisal
value of the house at the moment the first loan was given, and then assuming thereafter
that the house changed in value according to the Case Shiller index for houses with the
same zip code.
As can be seen, homeowners who have positive equity in their homes default
infrequently. But for homeowners with negative equity, the rate of default is staggering.
For subprime borrowers with a 160% loan to value ratio (that is, the ratio of all the
mortgages on the home divided by the current home price), the default rate is 8% per
month!

26

Monthly Net Flow (Excluding Modifications)
from <60 Days to ≥60 Days Delinquent
Based on Performance from Nov 08 - Jan 09 for all Deals Issued in 2006
14.0%

12.0%

10.0%

Subprime
Option ARM
AltA
Prime

8.0%

6.0%

Bold circles indicate
median CCLTV by
Product

4.0%

2.0%

0.0%
0

50

100

150

CCLTV (%)

200

250

300

CCLTV =
Combined Current
Loan-to-Value Ratio

These findings seemed surprising when I first presented them in a New York
Times editorial written with Susan Koniak on March 5, 2009. But nowadays many other
researchers are reaching the same conclusion. 22 The conclusion is an inescapable matter
of incentives. It is economically foolish for a homeowner to continue to pay off a
$160,000 loan when his house is only worth $100,000.23 Mortgage loans have turned out
to be no-recourse – after seizing the house, the lender almost never comes after the
borrower for more payments. The only other thing the homeowner loses by defaulting is
22

The Congressional Oversight Panel cited negative equity as the single greatest predictor of default in its
report of March 6, 2009. It included the data I provide here as evidence of this fact, data which I supplied
to the Panel in advance of its report, as well as data from an array of government agencies, all of which
corroborated the Ellington data presented here. That is not to say that joblessness is not now having a
significant effect on default rates. It is. But even now negative equity is the best predictor of default and
many Americans with jobs are defaulting, and will continue to default, not just the unemployed. See
generally the Congressional Oversight Panel’s October 9, 2009 report on the continuing foreclosure
problem and the unimpressive results from government foreclosure prevention efforts taken thus far.
Finally, to the extent job loss has become (it was not at the start of this crisis) a significant cause of
defaults, strong effective measures to eliminate the scary bad news, i.e., efforts to stabilize the housing
market, will help the economy recover faster and thus help the employment rate.
23
The implication of this statement is that the Obama plan of reducing interest rates to lower mortgage
payments to homeowners who are underwater is, at least for those seriously underwater, an invitation or
encouragement to those homeowners to act in a manner that may make no or little economic sense, i.e.,
stretching to make mortgage payments, albeit lowered from their highs, on homes those people will never
own when many of them might be able to rent more cheaply.

27

his credit rating, but especially for a non-prime borrower with a low credit rating to begin
with, how much can that be worth?
Foreclosures are horribly expensive. At the present time, subprime lenders collect
about 25 cents per dollar of loan when they foreclose. For example, if the loan is for
$160,000 and the house has fallen in value to $100,000 and the homeowner defaults and
is evicted, the lender can expect to get back $40,000. It takes 18 months on average to
evict a homeowner, and during that time the house is often left empty and vandalized. Of
course the main reason the average recoveries are so low is that the defaulters are the
homeowners who are furthest underwater.
In a rational world, these foreclosure losses would never happen. The lenders
would renegotiate the loans by reducing the principal so the homeowners could pay less
and stay in their homes, and the lenders would actually get more by avoiding the losses
from legal fees and bad home price sales. If the above loan were written down to
$80,000, the homeowner would likely find a way to pay it, or else fix up the house and
sell it for $100,000. Either way the lender would get $80,000 instead of $40,000. That
would have the further benefit of keeping many homes off the market and thereby aid in
the stabilization of home prices..
The Obama plan of paying servicers to temporarily reduce interest payments
was doomed to fail from the start, and so far it has utterly failed. Cutting monthly
interest payments by half will temporarily reduce the homeowners’ payments by the same
amount that cutting principal by half would. But under the government’s plan the cut is
temporary, not permanent, and thus is sure to lead to many more defaults in the long run
than cutting principal would as soon as the interest rate goes back up. In fact, since the
homeowner will still be underwater, he won’t in any meaningful sense own his house.
He will be less likely to make repairs, he will not be able to give the house to his
children, he will not be able to sell it if he gets a job in another city. In short, there is
every reason to think he will likely default even before the interest rate goes back up.
And indeed for loan modfications where there is no principal reduction, the redefault rate
is above 50% within nine months.24 The government’s present plan allows servicers to
increase principal while cutting interest by adding fees and other things to the old
principal amount. The plan is likely to leave more homeowners underwater than there
would be absent the plan and others more deeply underwater, i.e., with even less chance
of ever owning their homes and thus less incentive to keep up with mortgage payments,
than they would have without this government “rescue” plan.
The Obama plan has already wasted billion of dollars, and is slated to cost $75
billion -- all on a foolhardy mission that in the end will hurt homeowners and
bondholders, but enrich servicers, as I will soon explain. In the first six months of the
plan, according to the Congressional Oversight Panel’s October 2009 report, only 85,000
mortgages had been modified, and of those only 1,711 were “permanent” modifications
(i.e. permanent/temporary since interest rate reductions under the plan are designed to
24

See OCC/OTS Mortgage Metrics Report, Q2 2009

28

end in a few years time), and of those only 5 involved principal reductions.25 The
administration apparently still does not understand that the servicers have incentives that
put them at odds with bondholders and homeowners, so that they actually prevent
modifications that would help lenders and homeowners but hurt servicers.
In the case of many non-prime borrowers, the loans have been pooled in a trust,
and their principal has been tranched into many different bonds, each held by a different
investor. The lenders are the bondholders, but they are numerous and dispersed and by
contract have given up the legal right to renegotiate with homeowners, delegating that
right to an agent.26 That agent is the servicer, who has a fiduciary responsibility to act in
the interests of the bondholders in the trust. In “normal” times this arrangement worked
tolerably enough. But in this crisis, with so many mortgages in or near default, it has
failed miserably for at least least four reasons, all traceable to a misalignment of interests
between servicers and those whose interests they are supposed to protect, which has now
ruptured with terrible effects.
First, modifying loans is a time consuming and expensive operation. The servicers
who have the legal right to make modifications do not get paid directly for improving the
cash flows to loans. It is generally cheaper for them to move into foreclosure. In
particular they have no incentive to set up the huge infrastructure and to hire and train the
extra staff required to make sensible modifications on a grand scale.
Second, modifying the loans has different effects on different bondholders. It has
proved difficult to modify loans in a way that pleases everyone. Writing down principal
immediately would make more money for the trust as a whole. But it would immediately
wipe out the BBB bonds and possibly other lower level bondholders. Letting the
homeowners sit in their houses without paying for a year or two means that during all that
time all the bondholders, including the BBB, get their coupons paid in full from servicer
25

On December 10, 2009, as I was completing this paper, the Treasury issued a press release stating that
“permanent” modifications had grown to just over 31,000 and that just over 700,000 modifications were
now ‘under way” across the country. But my criticism of the plan is not based on the number of the timelimited “permanent” modifications completed, but rather is centered on the near-exclusive concentration on
interest reduction and, as I explain in the text below, on leaving the servicers in charge of the modification
decision. I could find no updated information in the report on how many, if any, of the trial or permanent
modifications involved principal reduction as opposed to interest reduction, and I have no reason to assume
that the percentage of modifications with principal reductions as increased. It is also worth noting that in
the Congressional Oversight’s Report of October 2009, see p. 127, the Panel notes that the apparent rise in
modifications due to the administration’s plan might be overstated as there was some evidence of a
‘substitution effect,” i.e., the number of “voluntary” modifications by servicers (or modifications made
outside of the administration’s plan) went down in the first six months of the plan, suggesting that the gross
number of modifications attributable to the plan itself might be exaggerated. The new report by the
government does not provide data from which one can assess any substitution effect.

26

It should be noted that this right was given up to avoid the collective action problems inherent when the
lenders are numerous and dispersed, and thus was given to a third party (the servicer) to be exercised on the
lenders behalf, the servicer acting as a fiduciary for the lenders. It was not given to the servicer to be used
to benefit the servicer’s interests at the expense of the principals (the lenders) and using the discretion to
modify or foreclose that way is self-dealing on the part of servicers and a breach of their obligation to the
lenders.

29

advances. The servicers then recoup their advances, at the expense of the trust (which by
then is mostly the AAA bonds), when the house is finally sold. The servicers say they
are terrified of lawsuits from the bondholders if their modifications help most
bondholders but hurt others. In reality servicers are simply using this as an excuse to
keep their fees coming. That was revealed when Congress passed legislation that freed
servicers from lawsuits by bondholders.27 Whether or not servicers are or were afraid of
lawsuits, there is a complex negotiation involving many bondholders and the servicer
which is not being resolved efficiently, and the government needs to intervene to break
an impasse for the public good.
Third, now that the government mortgage plan based on interest reductions has
given the servicers cover to reduce interest instead of principal, they can be counted on to
do the former and eschew the latter. Cutting the principal by half for example
immediately reduces the servicer’s fee by half (since the fee is computed as a percentage
of principal), while cutting interest does not. Moreover, cutting principal increases the
likelihood the homeowner will sell or refinance, which would cause the servicer to lose
his fee entirely.
Fourth, the biggest servicers happen to be owned by the biggest banks, who in
turn own a huge number of second loans. Cutting principal on first loans almost implies
cutting the principal drastically, if not to zero, on second loans. But that would mean that
the banks could no longer hold the second liens on their books at the inflated prices that
they are holding them now. The banks want desperately to postpone the write-down of
those second liens, which is to say, they have yet another powerful motive not to do what
is in the interest of lenders, homeowners and the economy as a whole, reduce principal on
the first loans they are servicing. By contrast, cutting interest on first loans makes it
easier to justify carrying the second liens on bank balance sheets at inflated values for the
near term (which is what matters to the banks) as homeowners are more likely to be able
to make the lower monthly payments (from lower interest rates) than their original
payments, at least in the short run.
Another proof that servicers have bad incentives is that when the big banks who
own the servicers hold the same kind of loans in their private portfolios, they do reduce
principal. During the second quarter of 2009, 30% of all loan modifications done to
loans directly held in bank portfolios involved some principal reduction. During that
same quarter the servicers reduced principal on 0% of their loan modifications, as did the
government owned agencies Fannie Mae and Freddie Mac.28
Loans that have not been securitized and are held entirely by banks (whole loans)
are also not being written down fast or far enough.29 The pathology this time is if
27

See Section 201 of the Helping Families Save Their Homes Act of 2009, preventing lender/bondholders
from suing servicers who modify mortgages under a qualified mortgage modification plan, which is defined
in the Act broadly enough to include all economically sensible modifications, i.e.,, those with a reasonable
prospect of returning more money to the lenders than a foreclosure.
28
See OCC/OTS Mortgage Metrics Report, Q2 2009
29
At first, it appeared that they were not being written down at any greater rate than securitized loans,
although the data is not perfect on this. Foote et al (2009) argued that this showed there was no real

30

anything more distressing. It appears that the banks, abetted by the suspension of mark to
market rules, are unwilling to take losses.30 It is better for them to keep the mortgage on
their books at $160,000, even though it will eventually bring them only $40,000, than it is
to reduce the principal to $80,000 and mark it there. The suspension of mark to market
rules has also fed the pathology discussed above on second liens. Abetting the banks in
their efforts to hide losses is bad government policy. It is sad to say that while the
government’s foreclosure plan has failed to stem the avalanche of defaults and looming
foreclosures, it has succeeded (with help from the suspension of mark to market
accounting) in obsuring the value of bank assets, many of which are being guaranteed by
the government, and thus in turn obscuring the value of mortgage assets now owned by
the government. In my terms, this only ensures the continuation of scary bad news
(uncertainty) when the goal should be for government plans to clarify the situation (the
value of assets) and thus help dispel what I call “scary bad news” that keeps leverage
severely constricted.
As in the past, the solutions the government has proposed and tried have thus far
all concentrated on interest rates, here quite destructively, and once again have ignored
the equity/collateral problem that must be addressed to reverse the impending avalanche
of foreclosures we now face. (For more details see my op-ed with Susan Koniak,
Principal Matters, on why the administration’s plan to stop foreclosures by chiefly
relying on interest rate reductions was doomed to fail.. NY Times March 5, 2009.)
Nonetheless, the government still has a remarkable opportunity to clean up the process
of reworking loans, but it must change course dramatically to do so. In October of 2008
(see my proposal with Susan Koniak summarized in the op-ed, Mortgage Justice is Blind,
NY Times, October 30, 2008) I urged the government to take the reworking process out
of the hands of the servicers and put the decision into the hands of government hired
trustees. All the failed efforts to stop the ever-growing number of foreclosures have only
demonstrated the need for just such a bold plan. In my approach, the government-hired
trustees would be told only about the homeowners, and would be blind to the bonds built
atop the loans. Their job would be to choose modifications or foreclosure, whichever
they judged would lead to the greatest recovery on the original loan. They would thus be
carrying out the duties of the servicers exactly as they were intended, but free from the
conflicts of interest and perverse incentives which have prevented the servicers from
carrying out their mission.31
incentive to write down loans. Now, again based on imperfect data, there seems to be some evidence that
principal on whole loans, at least at some banks, is being written down more often than principal on
securitized loans (which effectively never see reductions in principal), although reductions in principal on
whole loans is still much less frequent and much less widespread than one would expect to see given the
economics of the situation, i.e., that reducing principal for many underwater homeowners will yield much
more money than foreclosure or (over the long term) than interest reductions.
30

Banks may also still be holding out for some more direct government subsidy for their failing whole
loans, either through government assumption of the mortgage risk or some other form of direct payment for
anticipated whole loan losses.
31
Under this plan the servicers would still collect the servicing fees they do now. They would continue
their duties of sending letters to homeowners, collecting the monthly payments and distributing them to
bondholders, evicting homeowners who did not pay, selling their homes, and so on. The only change is

31

For a vast number of homeowners now upside-down in their mortgages, i.e.,
owing more than their home is presently worth, this process would likely result in a
reduction of principal. Why? Because reducing principal would yield investor/lenders
vastly more money than foreclosing, as we have seen.
If the government handled this correctly, most homeowners who were unable to pay
the original loan but were willing and able to pay a modestly lesser amount would get to
stay in their homes, the bondholders collectively would get more payments than they are
currently expecting (though some tranches would be hurt), and the government would not
have to invest any capital.
This plan is not the same as “cramdown” in bankruptcy, which Congress has thus
far rejected and which entails costs and creates some perverse incentives that my plan
avoids. To get a reduction in principal through bankruptcy (assuming the law were
changed to allow that) would encourage homeowners now current on their mortgages but
underwater and thus likely to default sometime in the future to default immediately to
support their petition for bankruptcy relief. That might well precipitate a sudden and
severe fall in housing prices and the value of mortgage securities. On the other hand, my
plan, as originally conceived, builds in a presumption that underwater homeowners now
current on their mortgages would not qualify for principal reduction if they went into
default upon the adoption of this government plan unless they could show the sudden
default was caused by some exogenous hardship, such as job loss or medical emergency
with attendant high costs. That would give underwater homeowners now holding on for
the short term a continued incentive to keep paying until the government trustees could
evaluate their loans and circumstances for a reduction in principal. Second, my plan
differs from bankruptcy in that it does not subject homeowners, many of whom are
underwater through no fault of their own or through, at most, fault equally shared
between themselves and the irresponsible mortgage hawkers who sold them a financial
product that made little economic sense, to the shame and devastating harm to future
credit and thus to their economic circumstance that a bankruptcy proceeding entails.
Third, my plan contemplates putting experts in local housing markets, community
bankers, in place as the government trustees, not bankruptcy judges who are neither
numerous enough to handle the number of defaulting homeowners who should justifiably
qualify for principal reduction, nor as knowledgeable as the personnel I would put in
charge. Indeed, bankruptcy judges would, as they commonly do, have to hire the kind of
personnel I am advocating the government hire directly, to advise the courts on the
appropriateness and nature of any mortgage modifications the courts were to order.
Indeed, it is highly doubtful that our bankruptcy courts could handle the job Congress
would be giving them if so-called cramdown legislation were adopted, at least not if it
were adopted without first having a plan like the one I propose up and running to handle
the vast majority of underwater homeowners. If my plan were indeed up and running,
that the mortgage loan modification would be taken out of their hands and put into the hands of the
government trustees. This reassignment of a particular duty in the contract is not a “takings” from the
servicer, among other reasons because the servicers have failed to carry out their fiduciary obligations to
the bondholders who employ them to get the most possible value out of the loans. See Dana (2010).

32

bankruptcy might be something worth considering as a true last resort for those already
deeply in default. Finally, bankruptcy involves all kinds of hidden costs, like lawyer fees
and trustee expenses (on top of the experts required to advise the bankruptcy judges) that
are unnecessary and wasteful for the vast majority of homeowners and lenders who
should be able to make a win-win deal without incurring those costs.
My plan envisions the government paying for the trustees (community bankers) to
decide on whether principal modification would bring in more for bondholders than
foreclosure, but I estimate that government expenditure should come to under $5 billion.
The current government plan allocates $75 billion to pay incentives to servicers, a wasted
allocation in that the incentives pulling servicers in the other direction are more powerful
than the promised government payouts. My plan would eliminate the expert fees
bankruptcy would entail and because my plan involves no court proceeding, the hidden
costs to our court system and legal fees would also be saved.
My original plan called for legislation to cut through the agency-problem mess in
securitized pools of mortgages by eliminating contract provisions in pooling
arrangements that now enable servicers to act contrary to the interests of the investors the
provisions were originally designed to protect. Thus, I envisioned that the government
trustees would only be empowered to modify securitized mortgages. This would leave
unsolved the problem of whole loans that banks are still refusing to modify sensibly, by
writing down principal for underwater homeowners. I am tempted now to advocate an
extension of this plan to the whole loans as well.
I believe, however, that once a government program of modifications for
securitized loans proved its worth by resulting in more recovery for investors, banks
would be likely to adopt similar standards to modify whole loans. In the first place, they
are already starting to reduce principal on their whole loans, even when they refuse to do
so as servicers for securitized loans. As I explained earlier, I believe that banks are
reluctant to make sustainable mortgage modifications by reducing principal on their
whole loans because they fear the massive write downs that would entail; particularly
when no competitor is making that move, each bank is loathe to endanger public and
investor confidence in its financial condition and also potentially jeopardize its
government support. It typically takes 18 months to evict a defaulter, and sometimes
longer if the homeowner temporarily resumes paying. Apparently, banks believe that by
waiting to the bitter end during a period in which they are making huge profits (on the
back of Fed monetary policy), they will be in a much stronger position to take losses.
Nonetheless, a solid government plan to force sensible principal reductions for
securitized loans would, I believe, go a long way toward convincing the banks that no
better deal from the government was forthcoming, particularly if the government clearly
articulated that this was so, and would make it more difficult for regulators to turn a blind
eye to the inflated value of the whole loans and second liens as now priced on the banks
balance sheets. Obliging the banks to mark to market would, of couse, also push them to
get the most value out of their loans by writing down principal for underwater homes.
But assuming the continued suspension of mark to market accounting, if the government
program to reduce principal on securitized loans (with the bondholders, not the

33

government bearing the loss) was not enough to convince banks to follow suit with their
whole loans, more direct government coercion should be considered.
For some lenders, recognizing the losses of whole loans on their books might
indeed put them out of business. I shall come to them later. Note that for the loans that
have been securitized, such as the subprime loans, these losses have already been
recognized.
Finally, what if homes prices vastly appreciate by the time the homeowner sells
his home? To prevent unwarranted windfall profits to homeowners, the government plan
could easily require the homeowner to share 50/50 with the lenders any appreciation in
home price up to the full price of the original mortgage and might even provide that for
houses sold for more than the original loan price that lenders receive a greater percentage
of the original loan, depending on how much above the original loan amount the selling
price was.
IVa2. A Floor to Housing Prices and Restarting Private Lending on Mortgages:
Government Equity Stake in Homes
There are at least four reasons to support housing prices directly, in addition to
doing so through effective foreclosure relief. First, if housing prices held firm, fewer
homeowners would be under water, and so more would have an incentive to make their
payments. That would keep them in their homes. Second, firm housing prices would
staunch the losses on mortgage securities even if there were foreclosures. Third, once
there is a floor to housing prices, pessimistic lenders would be relieved of the disaster
scenario for many mortgage securities, and margins on mortgage securities would come
down significantly, enabling optimistic buyers to purchase them using leverage, pushing
up the price of mortgage securities.32
Fourth, the leverage cycle is less severe for housing than for mortgage securities, and
so can more easily be fixed by government intervention, since home buyers generally
lock in their loans and leverage for 30 years. Only new buyers of homes, and those who
want to change homes or are forced out of their current homes in foreclosures, need to
confront the tougher margins. Old buyers sitting in their homes cannot be forced to put
more money down, whereas mortgage security holders who borrowed on one-day Repos
have found that they now face tougher margin requirements that require putting more
money down. Thus there are fewer homes in play than there are mortgage securities.
The Obama administration has recognized this by cobbling together a series of ad
hoc measures to prop up housing prices. I am afraid these measures will expose the
government to billions of dollars of future losses, in addition to substantial current costs,
while leaving private mortgage lending dead in the water. We simply cannot sustain a
situation where all mortgage lending is done by the government. The plan I propose

32

Again, as I will discuss below, margins must in the future be monitored by the Fed to assure that they do
not once again get excessively low, precipitating another massive and dangerous leverage cycle.

34

helps to stabilize housing prices and to reinvigorate private lending. And in the long run
it may cost the government much less, possibly even making money.
Current government FHA policy is to make mortgage loans with as little as 3.5%
down. These homeowners start with little incentive to continue payments, particularly in
rough economic times, and any further decline in housing prices will find them
underwater and thus a new source of future defaults. This policy is a repetition (albeit on
a smaller scale) of the bad practices that got us here. It exposes the government to a huge
risk of default, and does nothing to stimulate private mortgage lending.
The government has also tried to stabilize housing prices through its efforts to
keep mortgage interest rates low and thereby encourage purchases and refinancing. To
this end the government has bought a trillion dollars of agency mortgage securities. This
choice reflects once again a lamentable concentration on interest rates and the neglect of
collateral (leverage) effects that is the core of my argument. The government’s interestrate-centric approach has not been terribly successful in encouraging new purchases; in
fact, after a temporary drop, for a long time the plan did not even succeed in lowering the
mortgage rates. In the end it finally lowered mortgage interest rates, but surprisingly few
homeowners were able to take advantage of the lower rates by refinancing because they
could not find the downpayment. There are already signs that as the purchases wind
down, mortgage rates are going back up.
A third government initiative is to give an $8000 tax credit to buyers of homes.
This tax credit does appear to have been more successful at stimulating home purchases.
Indeed, the push to adopt and then expand it demonstrates just how ineffective the
purchase of securities was. But the tax credit has no upside for taxpayers, it does nothing
to reinvigorate private lending, and it asks those members of the public not benefiting
from the tax credit and their children to absorb the long term costs of this government
plan, precisely the kind of policy that undermines public support for the government’s
rescue efforts. If $8000 were spent on 7 million homes, this deadweight cost would
come to $56 billion. By contrast the equity stake plan I propose below is a purchase of
value for value; in the long run it may cost nothing and actually have upside for
taxpayers. It should also stimulate demand, and it would reinvigorate private mortgage
lending.
As we said earlier, toughening margins have affected housing prices, because
many homeowners can no longer put up the cash payment needed to buy new homes.
New homeowners are being asked to put as much as 30 to 40% down if they cannot get a
government loan. The government could stimulate demand for new purchases, and also
mitigate the margin problem, by offering to buy a 20% equity stake in any new home
purchase (under some maximum price, as with agency conforming loans). Thus suppose
a house is purchased for $100. The government pays $20 and gets a 20% equity piece,
which it collects whenever the homeowner sells. If down the line the house sells for
$200, the government gets $40. The government is thus earning the home price
appreciation on its piece, without having to bear the expense of maintaining the house.
The homeowner gains because he gets to live in the whole house while paying for only

35

80% of it. If the home buyer needs a loan to get the house, the government equity piece
reduces the downpayment the buyer must make, and the ongoing mortgage payments he
must make. And if we make the government’s equity piece the second loss piece, it
leaves the lenders in a very, very safe position, encouraging lending. In effect it lowers
the margin to the borrower, and raises the margin of safety to the lender. Here is how it
works.
Under the plan the home buyer who wanted a loan to purchase the house would be
allowed to borrow at most 80% of the $80 of the house he bought, or $64. He would
have to put up 20% x $80 = $16 of his own cash. The homeowner would then have a big
incentive to make his payments. If he walks away from his debt, he can save $64, but he
has to give up living in a $100 house on which he had an $80 ownership share. But if the
borrower does default, and if the lender has to foreclose, the lender would be able to
collect his debt out of the house sale proceeds ahead of the government equity piece. The
government would collect next, and lastly the buyer would get any left over cash. If the
house sold in foreclosure (net of expenses) for $82, the lender would get his $64, the
government would get $18, and the homeowner nothing. The effective margin for the
homeowner is thus 16% on the asset price of $100, but the margin of safety for the lender
is 36%. This should make the lender feel very safe and encourage private lending on
mortgages. The homeowner’s downpayment of 16% on the total home price is about half
the downpayment many non-government lenders are demanding now. On top of that, the
new buyer’s mortgage payments would be 20% lower than before, because he would be
paying on a loan of $64 instead of $80.
What about the costs of my plan? Last year there were 5.5 million new home
purchases, down from a high of 7 million. Even if the government had to buy the equity
in the whole 7 million, at an average home price of $200,000, it would cost $280 billion.
But the government would own equity, and be protected by the homeowner’s
downpayment. Housing prices would need to fall another 16% before the government
lost equity value. As housing prices stabilized, the government would gradually phase
out the program, in all likelihood in a year, at most two, after adoption.
A much smaller version of the same plan might accomplish nearly as much. Simply
restrict the government equity plan to home buyers who do not currently live in houses.

IVb. Step Two: A Fed Lending Facility to Help Restore Reasonable Leverage
The most easily implementable step and the second priority, after addressing the
source of the uncertainty (the scary bad news), in responding to the final stage of any
leverage cycle should be government action to decrease astronomical collateral rates.
Thus in October 2008 I suggested that the most immediate step the Fed could take was to
lend money using the so-called troubled assets (those that suddenly became near
worthless as collateral, as I explained earlier) as non-recourse collateral. I suggested 50%
margins on average, a sane halfway level between the 5% margins required at the peak of
the leverage bubble, and the 70-90% margin rate demanded in 2008. The TALF and

36

PPIP programs, announced in early 2009 at what turned out to be the bottom of the price
cycle, embody the spirit of my recommendation. The turnaround of prices after these
programs were announced seems to me to be some evidence for the wisdom of the
intervention. But in some important details those programs went awry, as I shall try to
explain, and in any case it now appears that having achieved their purpose, they have
been drastically attenuated.
Lending on risky collateral is a great departure from the traditional role of the
Fed. The orthodox view is that the Fed injects liquidity into the system by lending
money to banks and others with impeccable reputations for repaying so as to reduce the
riskless rate of interest on very short term loans. The banks would then presumably turn
around and re-lend that money to investors, at a lower interest rate than would have
obtained absent the Fed’s intervention. However, the great bulk of lending in the
investment world is not based on the reputation of the borrower but based instead on the
value of the collateral. The lesson of the leverage cycle is that when lenders demand too
much collateral for their loans, liquidity dries up. The Fed cannot undo this by making
riskless loans at a lower interest rate than the market, because in liquidity crises it is not
the interest rate the banks charge that impedes investor borrowing but rather the amount
of collateral they require. The Fed needs to make risky loans on less collateral than the
market, if it is to have the desired effect.
The mechanics of such a massive lending program require some careful thought, but
nothing compared to the difficulties of directly buying. The Fed could simply announce
that any arm’s length buyer of any designated security could, at the moment of purchase,
take that security to the Fed and receive a 5-year loan of 50% of the price in exchange for
putting the security up as collateral. The Fed would not need to price the security itself.
The market would have just done the pricing. With a 50% margin, the government
money is still quite safe. Remember, the 50% loan is against the price the securities will
be traded at, not against the original price when issued. The government should
thereafter monitor prices, periodically demanding more cash from the borrower to
maintain its 50% margin, which would make the government lending safer and more
responsible. Monitoring the collateral price is a much easier job then deciding the price
to buy, since there is a 50% margin of error: the price monitoring only has to be half
right. And the government should consider charging a slightly higher interest rate than it
does today, thereby potentially making a profit for taxpayers. That would also make the
program easier for the public and politicians to accept.
Buyers would then be able to purchase securities using only half the cash they
need to put up at the bottom of a cycle when margins might become 100%. Aside from
allowing their own cash to go further, this borrowing allows investors to earn leveraged
returns. If they think the security trading for 60 might only rise to 66 in the near future,
they can buy it with 30 down and earn a return of 20% when it rises to 66 instead of a
return of 10%. Again, with this potential for private profit, the program would make
more political sense if a somewhat higher interest rate for the loans were charged thus
building in a real chance for taxpayer profit.

37

The government might even arrange all this lending without having to come up
with the money. Under this alternative, the government could loan slightly less, say 40%,
and give up the right to make margin calls. The loan could then be securitized,
guaranteed by the government, and then sold off to the private sector. With the
government guarantee, the money would easily be raised. Or even more directly, for
some bonds where this makes sense, the government could simply guarantee a certain
percentage of the principal payments. Private lenders could then lend this much without
any risk of default. Of course, on some securities the government might be able to lend
much more than 40% and still regard the money as safe.
Lending is better than the government’s first (and quickly shelved) idea, as
proposed by then-Secretary Paulson, of buying up the “troubled assets.” As I explained in
October 2008, lending against collateral does not require the government to choose what
prices to pay, as it would have to if the Treasury directly bought securities. Moreover,
lending, unlike buying, is direct action to restore leverage and restoring leverage is the
thematic solution to the leverage cycle crisis. It is not some stop gap band aid invented
only under the pressures of the moment.
Further, lending puts taxpayer money at far less risk than buying does. Assuming
the Fed loans at 50% margins, every dollar the government lends using the targeted assets
as collateral will necessarily be matched by money the investor spends on those assets.
The government can say its money is being leveraged. The investors who avail
themselves of the government lending will still have their money at risk. Because these
investors will do the buying, and not the government, there is little, if any, chance that
this action will push prices to outrageous levels and enrich undeserving sellers.
The Fed has boldly gone a long way in this direction, further than any previous Fed.
Through TALF and PPIP the Fed and Treasury have indeed embodied many of these
ideas. PPIP lends at 50% margins on toxic mortgage securities, just as I recommended.
Its announcement I believe played a pivotal role in starting what is now more than a half
year rebound in security prices. But the two programs did not go far enough in general,
they took too long to get going, and in some cases TALF actually took leverage up
almost to the crazy levels it had been before.
In the emergency stages of the leverage cycle, the Fed should have (and still
should) extend lending on more kinds of collateral. TALF restricted leverage mostly to
new securities, or to securities that were still AAA rated. As more and more mortgage
securities get downgraded below investment grade status, they lose their ability to be
used as collateral even in the private sector. Lending against the most toxic securities is
actually necessary to maintain their value.33
33

Again, such lending would be much less risky if the government had adopted a sensible plan to staunch
foreclosures and stabilize housing prices, such as I’ve just outlined, because such a plan would reduce the
toxicity of the securities at issue. And the quicker the government moves to do that, the less risky such
lending will become, not to mention the good it would do for the value of the toxic securities the
government now owns through one program or another or now guarantees, representing continuing and
enormous government money still at considerable risk. This point is why I stress the importance of

38

The TALF program makes government loans on new credit cards, auto loans, and
college loans, and other securitizations at 20 to 1 leverage. In my opinion, this repeats
the error of FHA mortgage program, lending at the same inflated leverage that got us into
trouble in the first place. The Fed has rightly observed that propping up new security
values is more important than propping up legacy security values, because new securities
represent new activities. When new prices go down, new securities are not issued and the
underlying activity for which the securities would be issued (students going to school,
cars being purchased, new houses being built, consumers buying with credit cards) stops.
However, as I have argued more formally in Geanakoplos (2009), the Fed could raise the
price of these new securities further by leveraging them less, if it would also leverage the
legacy securities to modest levels. The reason is that buyers of these new securities are
tempted to put all their capital into the depressed legacy assets where they are nearly sure
of a high return. This indeed is one of the main reasons banks are not lending to
businesses or homeowners: they can get better returns by buying depressed legacy assets.
The only way TALF could redirect this private money into new securities was by giving
leverage on them at astronomical 20:1 rates. If instead the Fed would give two to one
leverage on all the assets, it would raise all asset prices, including even the new
securities, because it would remove the bargains investors are seeking in the legacy
assets.34 It would thus also go a long way to solving the bank lending problem. As I
showed again in Geanakoplos (2009) (in a stylized example, to be sure), despite lending
on a much larger scale, by allowing leverage at two to one on a wide array of assets
rather than at 20 to 1 on a narrow set of assets, the Fed could actually reduce its expected
defaults while increasing the prices of all the securities.
In the crisis stage the Fed needs to go around the banks and lend directly to more
investors. In theory the Fed could make no-recourse loans only to a few banks, who
would turn around and relend to everyone else. But the banks are nervous about showing
too much lending on their books, they ask for too much collateral, and now the Fed is
giving them more profitable ways to make money than by lending; so the Fed must reach
out directly to more borrowers. Curiously, PPIP has been restricted to 10 potential
borrowers/investors, making its scope and size in the end less than what was anticipated.
With only 10 investors taking government money the potential for conflicts of interest
seem very high. If one of the 10 becomes a big enough buyer with government money, it
could conceivably offer to rid a bank of toxic asssets, at favorable prices, in exchange for
favors like easier credit later. I, of course, realize that the Fed’s lending to many may still
be controversial. And there are legitimate reasons for concerns about lending to nontransparent borrowers. That, however, just means those problems should be fixed:
regulators, as all now acknowledge, need to be more vigilant and transparency (long
absent) must be restored by zero tolerance for hide-the-risk accounting tricks and
understanding the nature of the crisis in crafting sensible solutions and how failing to address one part of
the problem, in our case the failure to adequately address housing, limits the good that otherwise sensible
programs might make.
34

Another reason it actually could raise new security prices is that by leveraging the legacy securities at
two to one it will free some investor equity to put into the new securities.

39

“special” vehicles that obscure an institution’s true financial exposure. And if lending to
more borrowers means subjecting new entities (like investment banks and hedge funds)
to more regulatory review and greater transparency requirements, so be it.
The TALF and PPIP programs took too long to get up and running. Hopefully at
the bottom of the next leverage cycle, or even earlier, similar programs could be
implemented sooner. I recommend that the Fed keep a standing, permanent lending
facility up and running. In normal times it would lend a little bit across a wide range of
assets, to be ready to spring into action if private collateral rates became too high. This
facility could be administered directly by the Fed, by people it hired, or it could be run
through the Repo desks of the Wall Street Banks. In the latter case it would be wise to
insist that the bank put some of its capital at risk along with the Fed money. The
advantage of using Repo desks is that they are already staffed with trained personnel,
who have great expertise in making margin calls. Duplicating that expertise would be
expensive.35 The advantage of a permanent facility is that the Fed would be ready to
quickly lend on a grand scale, on many securities, and to many lenders, in the next crisis.
IVc. Step Three: Restoring Optimistic Capital
Lending will not by itself bring the prices of assets to their old levels (which is ok,
given that “old” values were inflated by excessive leverage, as I’ve explained). But that
means that the most optimistic buyers, unfortunately including some of the biggest and
most prominent financial institutions in America, have irretrievably lost a huge amount of
capital. Not only is their capital no longer available to spend on these securities, but
similarly the money they borrowed to spend on these securities has also disappeared.
The easiest way to address this problem of our still-ailing banks would be to sell stock
in optimistic American companies to foreigners. That started a while back, but the
foreigners saw that they were losing money and stopped investing.
The next most obvious thing the government could do, it did: inject money into failing
firms. The idea was that then the firms would continue to function as optimistic buyers
and their workers would not join the ranks of the unemployed. But the government
injected money with no strings, a mistake. In exchange the government should have
received shares of stock, beyond minimal interest payments, and commitments to spend
the money on distressed securities or new loans. And it should have demanded changes
in management and compensation practices that reward short-term risk taking at the
expense of long term stability But the main problem with the way the government
injected capital is not just the “no strings” approach, it is that this injection of capital was
not coordinated with vigorous programs to address the two other prongs of the end of any
leverage cycle: the source of the scary bad news (here housing) and the precipitous drop
in leverage, which I’ve just addressed in speaking about Fed lending.

35

I presented this proposal for a lending facility to the New York Fed in early 2009. Recently Pedersen
and Santos have made a similar proposal.

40

Put bluntly, it makes no sense to inject money into firms that will still go bankrupt
anyway or to support so-called zombie banks (too supported to fail, but too weak to
succeed without rolling the dice on highly speculative and risky bets). In the absence of
vigorous programs to address the first two prongs of any leverage crisis (the source of
gross uncertainty and outsize demands for collateral that few, if any, can meet) injecting
capital does nothing but push an ultimate reckoning down the road. And without steps
one and two, the true financial status of our financial institutions is unknown and
unknowable because there is no reliable way to price many of the assets they hold. As
long as no one knows whether and to what extent our biggest financial institutions are
sound, our economy cannot recover. To the extent our banks now appear to be “thriving”
or, at least enjoy, some market confidence, it is more a function of their access to cheap
money, their ability to make outsize bets with that money and the implicit government
bailout that investors now expect should one of the big boys suddenly stumble. This is
not a recipe for financial recovery or a stable economy.

IV.c.1 Bailouts with Punishment
After a double leverage cycle as outsized as we have just been through, it is likely that
even with a lending facility established, and capital injected properly into the system,
some, maybe many, firms would still fail. In general that is what we should want. The
government cannot afford to make good everybody’s debt. Some debt holders must lose
when a financial system is allowed to become bloated by artificially high prices
maintained by excess leverage from the ebullient stage of the leverage cycle. In the
ebullient phase of this cycle, too many people were drawn into the financial sector by the
resultant artificial profits. Failures will remove many of these excesses.
In such circumstances, many of the failures will be banks, like those in this crisis
that were or are currently holding individual loans that are internally marked 40 points
higher than they are actually worth. Raising the price of securities to what they are worth
will not save those institutions.
But what if those institutions are seen by the government as, in current jargon,
systemically important? For those firms the Treasury might want to intervene, as the Fed
did last year, on a case by case basis. But, if that approach is used, the shareholders
should have to give up their shares and the bondholders should lose too, maybe all their
value, and new management should be put in place. Even in cases where old
management is not that old, i.e., cannot be reasonably charged with responsibility for all
the excess, replacing management may be wise, if only to help bolster public support for
the government’s actions and expenditures of taxpayer funds. It is also imperative that
the government decide as quickly as possible after a crisis presents itself (and on grounds
that can be explained as fair and objective), who it will let fail, and then coordinate an
orderly liquidation.
Quite possibly the biggest mistake the government made, at least in the public
opinion, was bailing out too many firms on too generous terms.

41

IV.c.2 Government Purchases of Assets
The government could replace the lost optimistic capital by buying distressed
securities directly. In effect the Treasury would take conservative and pessimistic
taxpayers’ money, that would never be invested in these securities, and invest it there,
assuming, of course, that it did so with the expertise necessary to make reasonably sound
judgments on which securities to buy and how much to pay for them. This was the plan
with which Secretary Paulson began and which has reared its head a number of times in
the months since then. The banks, as I suggested above, may in fact have been waiting
for just such a plan, including the government buyout of not just securities, but shaky
whole loans too, and that hope may have contributed to their failure to modify whole
loans in a rational manner.
Government buying plans are a risky approach, riskier than the steps I have laid out
above, and thus, if ever used, must be implemented with extreme care. An argument that
is often blithely made for government buying is that when security prices are terribly
depressed in “fire sales”, the government might make some good investments. It is
likely, the argument goes, that the general taxpayer is too conservative, and by
transforming pessimistic capital into optimistic capital, the government might even be
directly helping the taxpayer, while at the same time staunching the collapse of security
prices.
Forcing natural pessimists into purchases they fear, however much potential
financial upside, may well undermine public confidence in government, especially if the
investments start to go bad. But even if taxpayers were on board, caution should be the
watchword. The lending mentioned earlier (a much more direct approach to restoring
leverage) would probably raise security prices, so the government purchases would not
be at rock bottom prices. Private investors (naturally more agile and quicker than
government), knowing that the government would be buying, would rush to buy first,
reducing potential government profits. Of course that, in some sense, would be what the
government would want to happen because it would mean that security prices would rise
more quickly. But it might also result in taxpayers getting stuck with the worst assets,
causing public outrage and charges of foul play.
The biggest obstacle and the one that apparently stopped Secretary Paulson’s original
plan to buy the troubled assets is the enormous challenge of deciding what to buy, and at
what price. We must not forget that the downward swing in the leverage cycle is always
triggered by genuine bad news, which I called scary because it creates more uncertainty.
Private investors hold back for fear of “catching a falling knife”; the government has far
less expertise than these private investors. Since the distressed mortgages are very
heterogeneous, it is not at all clear how the government acting alone could figure out
what prices to pay. Indiscriminately buying assets at a fixed percentage premium over
the internal prices given to the assets by the firms that hold them is a terrible idea. It
rewards the firms with the largest number of bad assets and especially those with the

42

most distorted internal prices. But how else could the government decide what to buy,
and at what prices?
One suggestion is by reverse auction. The government would divide the securities
into different categories, and then buy from each category those securities that the current
asset holders are willing to sell for the lowest price. But how would the government
decide what the categories are, and how much to spend on each? The lobbying in
Washington would be, to put it mildly, intense. And worse than that, it is a sure thing
that the cheapest securities the government would get in each category would be of the
lowest quality. If the purchases were to be made by an auction mechanism, I would have
suggested a variation in which private bidders were allowed to enter the auction, not just
private sellers. I would have recommended that the government commit to buying half
the winners’ purchases, at their winning prices. That way the government could ride on
the expertise of the private buyers. Still, even that solution, could be gamed, particularly
given that some private buyers might hold other positions, I’m thinking of CDS here, that
made it worthwhile for them to overbid in a manner that might not be easy to deter or
discover.
The dangers of government buying look so profound that in October 2008 I
recommended that if the government were to buy at all, it would be better for the
government to invest through professional money managers, again piggy backing on the
choices they make to invest their own capital. Under the PPIP plan the government has
set up accounts with professional money managers in which each government dollar is
invested side by side in the same securities with a dollar of investor capital. The
government should be able to take advantage of its enormous size to negotiate small fees
for the managers, perhaps relying to some extent on the patriotic instinct of managers to
keep fee demands low. Of course it is absolutely crucial that the managers have
incentives to perform well. Otherwise they might be tempted to spend the taxpayers’
money buying portfolios sold by the failing companies of their cronies, in exchange for
favors later on. Or they might pay less attention to the government investments than the
investments of their fee paying clients. These conflict of interest become more acute to
the extent that the number of managers is small and to the extent they each have a huge
amount of government money to wield. Moreover, the most qualified money managers
are likely to hold many securities and other assets that might provide those managers with
incentives to buy for the government with an eye towards benefitting their private clients.
One way to check some of these conflicts and to help ensure money managers had the
right incentives would be to divide the government money up among a large number of
private managers and to make the investments and returns of these companies very
public. These managers would then be competing with each other on a world stage to see
how their investments performed. A more conventional incentive device would be to say
that a manager gets no fees until the return on the assets passes some hurdle. Only after
the taxpayers make money would the managers earn any fees. As originally designed,
however, the PPIP program sought to concentrate the government’s investment money in
just a few managers’ hands. Were it not for the small size of the investment that it now
appears these funds will manage, this could have been a recipe for disaster. I would

43

strongly encourage the government, if it continues with the program, to expand it to
include many more managers, each confined to a relatively small scale. That is, of
course, assuming a sufficient number of such competent, reliable and willing managers,
as I imagine would be necessary, could be found.
But even with all the caveats I have offered above, buying may still not be a wise
policy, particularly not as a substitute for an adequate lending program, such as I
described above. Buying would leave the government holding a large portfolio of
mortgage securities, even larger than that the government already holds as a result of the
rescue efforts and resolutions it has already facilitated. Those investments might well
lose the public money. And yet, by modifying loans and using others of its powers the
government could act to mitigate those losses, and there lies the rub. All other investors
might well anticipate the government doing just that. In short, the more securities the
government owned outright, the more uncertainty might adversely affect security prices.
Perhaps even worse, had the government bought the troubled assets outright at the start as
the original Paulson plan envisioned, every other effort it should have taken to stabilize
housing (modifying loans and taking equity stakes) might be viewed as the government
trying to inflate the value of its own “book.”
For these reasons, and because it would have a smaller effect on restoring leverage
than the lending route, I do not believe the direct buying approach is an optimal response.
V. Moral Hazard
It is often said that every bailout causes a moral hazard which leads to a bigger
problem the next time. The problem would be that bailing people out in this crisis would
lead to higher leverage in the next cycle. There really is only one reliable antidote to that,
and that is regulation of leverage.
One observation, however, which I have made with Felix Kubler (and which Jeremy
Stein has apparently come to independently), is that general system-wide intereventions
(like restoring sane leverage) in the crisis do not always create deleterious incentives in
the long run. Surviving a crisis means tremendous profit opportunities in the good phase
of the next cycle. If a systemic intervention gives a chance for the prudent firms to
survive, rather than everyone going under, those firms will have an increased incentive to
be prudent. Bailouts which rescue firms no matter how imprudent they have been (in fact
precisely because they in particular were imprudent) are the source of moral hazard.
Some have suggested that writing down principal on mortgage loans will also
cause a moral hazard. They say it will encourage homeowners to behave badly, and the
government to intervene in too many markets, and threaten the sanctity of contracts. I
disagree, because the writing down of principal should be done as a function of the
decline in some index of housing prices. The index is beyond the control of the
homeowner, so it does not distort homeowner incentives. Moreover, it should be done
first for homeowners who have not defaulted yet, and only later for homeowners who
have defaulted under some special hardship. It should only be done, as I have said, if it

44

promises to bring more money to the lenders. A good test of whether it is a good idea is
whether it would be written into the contract in the first place if people had thought of the
possibility of this much home price decline. I agree with Robert Shiller who suggests
that just these kinds of mortgages, with principal automatically reduced if some housing
index falls enough, should and will become the standard mortgages of the future.
VI. Managing the Ebullient Stage of the Leverage Cycle
After this crisis passes we must prepare for the next leverage cycle. The first step
is to constantly monitor leverage at the securities level, at the investor level, and at the
CDS level.
Every newspaper prints the interest rates every day, but none of them mentions
what margins are. The Fed needs to settle on a menu of different security classes,
monitor their haircuts daily by talking to all the big lenders and borrowers, and then
making averages public.
The leverage of money managers should also be public. Moreover, legislation
and regulations should contain strong and clear prohibitions against misleading the public
or regulators on the degree of leverage, notwithstanding any other legal rule or
accounting device employed to justify the masking of actual leverage fairly calculated.
One contributing factor to the outsized nature of the crisis we faced that I may not yet
have emphasized sufficiently was the ease with which capital requirements could be
evaded and leverage hidden, even by the most “regulated” institutions, banks that hold
deposits insured by the federal government.
We discussed at great length in Sections II and III how CDS contracts provide an
opportunity to leverage, so these must be monitored as well. Putting them on an
exchange would facilitate monitoring, as well as netting and ensuring enough collateral is
posted.
Transparency about actual leverage should bring a great deal of discipline to the
market, and warn investors of impending trouble. In my earlier leverage diagrams one
can see the tremendous spikes in margins during the crisis stages of the last two cycles.
One can also see a drift down in haircuts in the ebullient stage of the last cycle. The
combination of security leverage data, investor leverage data, CDS leverage data, and
asset price data should give the Fed tremendous information for managing future leverage
cycles that they did not have, or chose to ignore, in this and in past leverage cycles.
But transparency alone is not enough. Some investors will not curtail their
leverage, no matter how much the public scrutiny, and how far out of line with recent
practice they become. Put bluntly, the market alone will not take care of outsize leverage.
It is thus imperative that the Fed put outside limits on leverage.
The first step in controlling leverage is to make sure that even the little amounts
of collateral that the markets require of most borrowers are indeed put up by all

45

borrowers. All too often CDS insurance buyers allowed the writers of insurance to get
away without actually putting up the collateral. Repo borrowers put up collateral to
protect lenders, but had no protection themselves in case the lenders went bankrupt and
swallowed up their collateral. It is imperative that some sort of exchange be established
so that in conventional Repo transactions both sides of the deal are protected against
counterparty risk, and so that CDS contracts are properly collateralized and netted, as we
discussed earlier.
But even after all this is done, it will still be necessary to regulate leverage. The
lesson of the leverage cycle is that there are many externalities (nine that I listed) and we
should always expect cycles of too much leverage followed by too little leverage. Many
people have argued that setting margin limits is difficult because securities are so
heterogeneous. But I believe this problem will eventually be solved once the haircut data
history becomes more public. It was not obvious how to manage interest rates either.
But little by little the Fed has gotten better at it. The same will be true with leverage.
The critical thing is that with the data in hand, the Fed will be able to monitor dramatic
changes in leverage and asset prices, and therefore will easily recognize when we are
reaching either end of the cycle.
One way of controlling leverage is to tax firms that borrow excessively, or that
borrow excessively on their collateral, or that lend excessively on collateral. A very
small tax might go a long way to discouraging excessive leverage, and might also change
the maturity structure, inducing longer term loans, if it were designed properly. Another
advantage of the leverage tax is that revenues from it could be used to finance the lending
facility the Fed would need to keep at the ready in anticipation of the downside of future
leverage cycles.
Another way of controlling leverage is by mandating that lenders can only tighten
their security margins very slowly. Knowing they cannot immediately adapt if conditions
get more dangerous, lenders will be led to keep tighter margins in the good safe times.
It has become fashionable nowadays to say that leverage regulation should be
countercyclical, by which people mean that investor leverage should be allowed to go up
in bad times and down in good times. I myself would like to see much more focus put on
securities leverage. The leverage of an investor is often a meaningless number, since just
when things are getting bad, and margins on securities are tightening and the whole
economy is being forced to de-leverage, many firms will appear to be more leveraged
because their equity will be disappearing. Rather than being an argument in favor of
letting their leverage rise in bad times, this phenomenon suggests to me that it is a waste
of time trying to control the uncontrollable.
I would rather see regulation forbidding loans at too high leverage in ebullient
times. Banks would simply not be allowed to lend 97% of the value of the house.
VII. Conclusion

46

The leverage cycle brought us to the edge of a cliff. We have moved back from
the precipice, but unless we understand the features of the leverage cycle and design our
responses to address the specific problems that characterize the end stage of an outsize
leverage cycle, we are left hoping for a miracle to restore our financial prosperity.
Marking time and waiting for the miracle of things getting better appears to be part of the
current government policy, at least as it relates to housing and foreclosures. That
miracle, if it comes, will be nothing more than the start of another cycle, maybe one even
worse than the one we have just experienced. And that is a miracle we cannot afford.
Bibliography
Adrian, Tobias, and Hyun Shin "Liquidity and Leverage". 2009. Forthcoming in the
Journal of Financial Intermediation
Bernanke, Ben, Mark Gertler and Simon Gilchrist. 1996. "The Financial Accelator and
the Flight to Quality." Review of Economics and Statistics, 78(1): 1--15.
Bernanke, Ben, Mark Gertler and Simon Gilchrist. 1999. "The Financial Accelerator in
a Quantitative Business Cycle Framework", pp 1341-1393 in Handbook of
Maroeconomics, Volume 1, ed by J.B. Tayolor and M. Woodford, Elsevier.
Brunnermeier, Markus, and Llasse Pedersen 2009. "Market Liquidity and Funding
Liquidity."Review of Financial Studies, vol 22,6,2201-2238.
Caballero, Richardo J., and Arvind Krishnamurthy. 2001. "International and Domestic
Collateral Constraints in a Model of Emerging Market Crises." Journal of Monetary
Economics, 48(3): 513--48.
Dana, David 2010. “The Foreclosure Crisis and the Anti-Fragmentation Principle in
State Property Law”, University of Chicago Law Review, Forthcoming.
Foote, Christopher, Kristopher Gerardi, Lorenz Goette, and Paul Willen, 2009.
“Reducing Foreclosures”, Federal Reserve Bank of Boston Public Policy Working Paper
No. 09-02, unpublished.
Emmanuel Farhi and Jean Tirole, 2009. "Collective Moral Hazard, Maturity Mismatch
and Systemic Bailouts” NBER Working Papers 15138
Fostel A., Geanakoplos J. 2008. "Leverage Cycles and the Anxious Economy"
American Economic Review 2008, 98:4, 1211-1244.
Geanakoplos, John. 1997. "Promises, Promises." In The Economy as an Evolving
Complex System II, ed. W. Brian Arthur, Steven Durlauf, and David Lane, 285--320.
Reading, MA: Addison-Wesley.
Geanakoplos, John. 2003. "Liquidity, Default, and Crashes: Endogenous Contracts in
General Equilibrium." In Advances in Economics and Econometrics: Theory and
Applications, Eighth World Conference,Vol. 2, 170--205. Econometric Society
Monographs.
Geanakoplos, John and Herakles Polemarchakis 1986. "Existence, Regularity, and
Constrained Suboptimality of Competitive Allocations when the Asset Market is
Incomplete" in Uncertainty, Information, and Communication, Essays in Honor of Ken
Arrow, edited by W. Heller, R. Starr, and D. Starrett, 65-96, Cambridge, Cambridge
University Press, 1986.
Geanakoplos, John and Felix Kubler 2005. "Leverage, Incomplete Markets, and Pareto
Improving Regulation", unpublished.

47

Geanakoplos, John and William Zame (2009). "Collateralized Security Markets",
unpublished. Earlier versions 1997, 2002, 2005.
Gromb, Denis and Vayanos, Dimitri, 2002. "Equilibrium and welfare in markets with
financially consrained arbitrageurs" 66,361-407.
Harrison, J. Michael and David M. Kreps 1979 "Martingales and Arbitrage in Multiperiod Securities Markets", Journal of Economic Theory, 20, 381-408.
Holmstrom, Bengt and Jean Tirole, 1997. “Financial Intermediation, Loanable Funds,
and the Real Sector,” Quarterly Journal of Economics, 112, 663-692.
Kiyotaki, Nobuhiro, and John Moore. 1997. "Credit Cycles." Journal of Political
Economy, 105(2): 211--48.
Minsky, Hyman 1986. Stabilizing an Unstable Economy. Yale University Press, New
Haven.
Myers, Stewart. 1977. "Determinants of Corporate Borrowing" Journal of Financial
Economics, 5, 147-176.
Tobin, James and Stephen Golub 1998 Money, Credit, and Capital. Irwin/McGrawHill, New York.

48