John Geanakoplos has been a trailblazer in modelling the influence of leverage over the economic cycle. His work has provided a tremendous push forward in both theory and policy as the Great Recession has taught us that we need to both understand and regulate leverage in the economy, especially the financial sector.
For a couple of years now I have been intrigued by the work of John Geanakoplos (hereafter JG), from YaleUniversity. Thanks to Yale placing his lectures (as well as Robert Shiller and others) online, I have been able to look into his work more closely. It is quite remarkable that this work began well before the crisis, while the crisis itself has added even more relevance to this work.
A full description of the model is beyond this post, but his work has important ramifications for both theory and policy.
Basics of the model
JG uses a time tree to denote different states of nature, as well as heterogeneous agents in the model. Similar to the agents in the Cubic IS curve , these agents are distributed along a spectrum by their optimism/pessimism on the asset in question. Optimists have a rosier view of the outlook for the asset, which results in a lower probability that they place on the down state. There is a marginal buyer on this spectrum, where the ‘natural buyers’ are those above this marginal buyer. Geanakoplos states that macroeconomics has ignored this ‘natural buyers’ hypothesis, but he sees this as crucial to explain the loss and downward spiral of the crisis.
As these most optimistic buyers are also those that take on the greater leverage, which pushes prices higher, when the bust ensues, they are the ones that either go bankrupt or have to deleverage severely. This exacerbates the fall in prices. So in JG’s model, these natural buyers are very significant for asset prices, both on the way up and down.
Crucially, JG’s model allows asset prices to rise without a change in the fundamentals. This is a fascinating point: it’s the loosening of lending standards, or the increasing of available leverage, that drives the prices higher – not any change in the fundamentals. He also demonstrates how the same demand and supply functions can determine both the interest rate and the leverage ratio, thus endogenising leverage. It’s the use of different states of nature allows JG to endogensise leverage – as each state has a leverage ratio and an associated interest rate. As he summarises:
I showed that the right way to think about the problem of endogenous collateral is to consider a different market for each loan depending on the amount of collateral put up and thus a different interest rate for each level of collateral. A loan with a lot of collateral will clear in equilibrium at a low interest rate, and a loan with little collateral will clear at a high interest rate. A loan market is thus determined by a pair (promise, collateral), and each pair has its own market clearing price.
Lessons for theory and modelling in this area
JG’s modelling has challenged the fundamental pricing of assets theory, as it is more than just the discounted flow of funds from the asset that determines its price. His use of heteregenous agents and endogensing leverage is also important.
JG’s modelling also highlights the question as to what drives the change in lending standards and leverage? JG comments in his lectures that he thinks Shiller is partly right in his story of irrational exuberance, but the other part of the story is leverage. I would frame these two together rather than as separate forces. It is the animal spirits/irrational exuberance that drives house price expectations higher for both lenders and borrowers, this lowers lending standards and increses leverage. This leads to higher house prices as in JGs model, and the cycle feeds itself until the Minsky moment.
I also wonder about the marginal buyer in JG’s model. In his model, the marginal buyer increases higher with higher leverage. In reality and in the basic Cubic model, increasing house price expectations drive leverage and bring more people into the housing market. This seems reasonable with the introduction of sub-prime borrowers into the US housing market in the mid 2000s.
Other theoretical work from Shin has used balance sheet classifications and institutional “frictions” to address this issue. In Shin’s model everyone has the same beliefs, but there are institutional differences – some are leveraged buyers and some are not. It’s a change in the fundamentals that causes an increase in demand for the asset, where the price is then amplified by the leveraged buyers. Hence, it is possible to tell a similar story without the microfoundations of JG, but I would like heterogeneous agents with respect to beliefs or expectations to drive the leverage.
JG on the current crisis and policy issues
On the current crisis, JG summarises as follows:
One of the main causes of the severity of the current leverage cycle is that there are two of them, in the housing market (via mortgages) and in the mortgage securities market (via the repo market), and the two reinforce each other in a destructive feedback. Houses back mortgage securities, hence a crash in housing prices has ramifications for the securities market. But a crash in the price of mortgage securities affects the loans that homeowners can get, which in turn affects the housing market.
This analysis, along with his model, lead to some very interesting suggestions for both policy and regulation. Central banks should be looking at collateral rates as well as interest rates, which ties in with current theme of macro-prudential supervision and using banking guidelines to smooth the economic cycle. It also chimes with my piece on financial market regulation. In a crisis, the central bank should look at restoring leverage to a reasonable level, which could also include the writing down of principle. This is a point also addressed by the IMF.
While not against credit default swaps (CDS), JG warns about the timing of their introduction:
In my view, an important trigger for the collapse of 2007–9 was the introduction of CDS contracts into the mortgage market in late 2005, at the height of the market. Credit default swaps on corporate bonds had been traded for years, but until 2005 there had been no standardized mortgage CDS contract. I do not know the impetus for this standardization; perhaps more people wanted to short the market once it got so high. But the implication was that afterward the pessimists, as well as the optimists, had an opportunity to leverage.
Finally, he has an intriguing discussion about the reasons for the contagion of the current crisis:
Fostel and Geanakoplos (2008b) gave one possible explanation for contagion.
We argued that if the same investors were the leveraged optimists in many markets (called crossover investors), then bad news in just one sector could cause price drops in other markets with totally independent payoffs. Once the scale of leverage is recognized, it becomes apparent that the pool of risk‐taking capital is small compared to the size of global asset markets; once it shrinks, and once deleveraging starts, prices fall in unrelated sectors.
While it is difficult to do justice to this work in one blog post, the theory of the leverage cycle will continue to be very influential for both theoretical and applied economics, including my own work.
 See The Leverage cycle by John Geanakoplos Cowles foundation paper no. 1304