Why does the economy bounce back from some financial crashes but not others? In 2001, the stock market lost a huge proportion of its value during the dot.com crash, but there were few lasting effects on the economy at large. In contrast, after the 2008 crash, recovery was very slow and the effects on output and employment were large and persistent. Research has shown that one reason the latest recession was so deep was that the financial system itself was badly damaged.
This is important, because contrary to many conventional theories about market behavior, it is financial intermediaries—broker-dealers who trade across all asset classes—and not households that are responsible for a large share of transactions. Shocks to the financial health and strength of these intermediaries thus have a big impact on markets and asset prices, and those effects can ripple through the economy.
In important new work, University of Chicago researchers Zhigou He and Bryan Kelly, along with Asaf Manela of Washington University in St. Louis, are reexamining asset pricing and risk premia from the perspective of these financial intermediaries, which helps show how the financial sector affects the larger economy.
They build on research from He and Krishnamurthy (2013) that starts with a dynamic, continuous time model with an explicit role for intermediaries. In their model, the financial sector is a crucial determinant of the behavior of the economy over business cycles. When intermediaries find that their capital ratio is low, they become more cautious in their investments in the risky asset, with adverse consequences for the economy as a whole. One implication is that drops in the intermediaries’ capital ratios are an important source of risk in financial markets.
He, Kelly, and Manela (2017) test the empirical implications of the He and Krishnamurthy (2013) model by focusing on the primary dealer counterparties of the New York Federal Reserve. They construct a measure of the banks’ capital ratios, and use this measure to test whether some assets are more sensitive to this risk than others. If banks are short of capital, some assets will be harder to hold than others. For instance, some asset types might subject the bank to higher financing needs in a volatile market. They call this sensitivity a “capital beta.”
This model suggests that the expected return on assets is not just a function of their market betas (see box). Their empirical results show that their measure of the banks’ capital stress sets the returns on a wide variety of assets. Banks will only hold assets with a high capital beta if compensated for the capital risk.
He, Kelly, and Manela present convincing empirical evidence that the capital beta is an important determinant of asset returns for a broad set of asset classes, ranging from equity, bonds, and derivatives. By regressing the average returns on both capital betas and market betas, they find that the capital beta is a statistically and economically significant determinant of average returns.
That’s one of the reasons that this paper is important. The financial risk factor performs well empirically, especially for the asset classes that are sophisticated (say, options and credit default swaps) and hence are mostly traded by these primary dealers. Compared to other factors that are used in asset pricing literature, this capital risk factor is also well grounded in a coherent theoretical model. This new class of models brings hope of a firmer theoretical foundation for asset pricing.
Theoretical models with a role for intermediaries:
Most forms of the Capital Asset Pricing Model assume that price differences can be effortlessly arbitraged away. Newer asset pricing models allow for limits to arbitrage, such as an inability to borrow or lend large amounts of money at the risk-free rate. Some of these models build in a role for intermediaries as well as consumers. In these newer models, exemplified by Brunnermeier and Sannikov (2013) and He and Krishnamurthy (2013), the intermediaries are subject to financial constraints: there may be a limit to the amount of equity finance, or to the ability to borrow through the issuance of debt. In these models, the intermediaries are the marginal (price-setting) traders, not the consumers as in earlier models.
These models are dynamic, so as to capture behavior over the business cycle. Prices, investments, and the business cycle itself are endogenous. Shocks to the variables, such an exogenous drop in equity, generate changes in behavior. One observed phenomenon that these models strive to capture is the sharp increase in risk premia during a financial crisis, and the long slow recovery afterwards.
Many earlier dynamic models relied on changes in consumer behavior to generate economic dynamics, but it’s hard to generate the kind of economic dynamics we observe in practice based on reasonable parameters for consumer behavior. Using a continuous time setting allows them to focus on the behavior of the economy during crises. Earlier methods based on log-linear approximation around a steady-state equilibrium do not allow the sort of sharp spikes in risk premia that are observed in practice during a financial crisis.
Recent models have explicitly introduced an intermediary sector. In He and Krishnamurthy’s model, consumers do not invest directly in risky assets, but in the intermediaries. Only the intermediaries participate in the risky asset market. (See figure 1). One crucial determinant of an intermediary’s investment behavior is their capital ratio, their total equity divided by total equity plus debt. If the capital ratio drops, they will invest more cautiously: a further drop in equity may put the firm into distress. As a result, the equilibrium risk premium rises when intermediaries have low capital ratios.
Testing the Theory Empirically:
In their empirical work, He, Kelly, and Manela focus on the role of primary dealers, who are the counterparties of the New York Federal Reserve in implementing US monetary policies. Primary dealers, such as Goldman Sachs, Deutsche Bank, and J. P. Morgan, trade in many capital markets, so it is plausible that they are marginal traders in these markets.
Primary dealers participate in the markets for all the asset classes that they consider. If they are the same marginal agents across all the asset classes, this could result in the measured capital risk premium being the same across asset classes. If there are specialist primary dealers (e.g. most foreign exchange is done by one company) then the price of capital risk for that asset class might be different without rejecting the model.
In order to measure the primary dealers’ capital at the holding company level, the researchers tie each primary dealer to its parent holding company. The capital of a trading subsidiary may be buffered by inter-firm movements of capital. If one subsidiary loses money but has good future investment opportunities, the parent company may infuse it with cash. If the parent company is short of cash, the subsidiary may be constrained in its ability to invest.
Since holding companies are mostly publically traded, their equity can be measured based on the market information at a monthly or even weekly frequency, rather than relying on book value data, which is available only quarterly and often reflects stale information. Since they are trying to capture what happens during relatively short-lived crises, it is particularly important to have up-to-date data, and there is not much movement in book value. (Debt prices are not readily available, so they use the book value of debt in place of market value.)
This approach allows the researchers to construct the capital ratio—the value of equity divided by the total value of an intermediary’s debt plus equity. A low capital ratio indicates high stress for the banks. When the capital ratio drops, banks’ participation in markets drops. Figure 1 (reproduced from He et.al. 2017), shows the history of the capital ratio and its innovations from 1970 to 2010.
Their work improved upon a similar measure offered by Adrian et.al. (2014). That work focused on broker-dealers, who are smaller and more specialized than primary dealers and relied on book value of broker-dealer subsidiary’s capital, which tends to be stale. If capital transfers between subsidiaries and holding companies are empirically important, then measuring the holding company’s market capital ratio is the better measure of distress.
He, Kelly, and Manela test the relationship between the capital ratio and seven asset classes: stocks, US government and corporate bonds, other sovereign bonds, options, credit default swaps (CDS), commodities, and foreign exchange (FX). This is a broader array of assets than previous work, which mostly tested stocks and US government bonds. Earlier work argued that intermediaries were the marginal investors for more sophisticated assets like CDS or FX but not for stocks and bonds. However, few investors participate directly in stock and bond markets either; perhaps intermediaries are the marginal investors for all asset classes.
They find that the premia rewarding capital risk for five out of the seven asset classes range from seven to eleven percent, with FX and options being somewhat higher. Their data is consistent with a capital risk premium of 9 percent for all asset types, with the observed deviations being the result of measurement error. The primary dealers are active in all the markets studied. This result suggests that the dealers do, in fact, act as the pivotal players, setting asset prices on the margin.
In equilibrium, the behavior of intermediaries connects asset markets. A shock to the banking system can be buffered or transferred between markets. A small shock to one market can be buffered by the capital of the holding companies. A large shock, or one that hits several markets at once, so that it deals a serious blow to the intermediaries’ capital, may be transferred to other markets, affecting the system as a whole.
If financial stress is an important source of risk and cannot be diversified away, then assets which are more sensitive to that risk should, on average, earn a risk premium. The authors test an asset pricing model in which expected return is explained by the standard market beta and the capital beta. The capital beta is statistically significant and economically large.
Adrian, Tobias, Erkko Etula, and Tyler Muir. 2014. “Financial Intermediaries and the Cross-Section of Asset Returns.” The Journal of Finance 69 (6). Wiley Online Library: 2557–96.
Brunnermeier, Markus K., and Yuliy Sannikov. 2016. “Chapter 18 - Macro, Money, and Finance: A Continuous-Time Approach.” In Handbook of Macroeconomics, edited by John B. Taylor and Harald Uhlig, 2:1497–1545. Elsevier.
Brunnermeier, Markus K. and Yuliy Sannikov. A Macro Model with Financial Sector, American Economic Review.
He, Zhiguo, and Arvind Krishnamurthy. 2013. “Intermediary Asset Pricing.” The American Economic Review 103 (2): 732–70.
He, Zhiguo, Bryan Kelly, and Asaf Manela. 2017. “Intermediary Asset Pricing: New Evidence from Many Asset Classes.” Forthcoming in Journal of Financial Economics, available at http://faculty.chicagobooth.edu/zhiguo.he/research/intermediaryCapital.pdf.