Working Paper Abstracts – 1993
Empirical evidence suggests that banks play a unique role in the savings-investment process, affecting firms’ cost of capital and the level of investment. We argue that bank uniqueness is related to how the design of bank loan contracts allows banks to affect borrowers’ choice of project risk. Unlike corporate bonds, bank loans are typically secured senior debt which contain embedded options allowing the bank to “call” the loan. The option allows the bank to control borrowers’ risk-taking activity via renegotiation of the loan. We analyze the renegotiation outcomes and show that: (1) debt forgiveness occurs; (2) monitoring by the bank is not always successful in preventing the borrower from increasing risk; (3) renegotiated interest rates are not monotonic in borrower type; (4) inefficient liquidation can occur. In renegotiation seniority and collateral are crucial because they allow the bank to threaten the borrower and liquidate inefficient projects. We show that when a prepayment option is included in the bank loan contract, bank debt is more valuable (ex ante) to borrowing firms than corporate debt; it lowers the cost of capital.
In the last two decades U.S. banks have become systematically less profitable and riskier as nonbank competition has eroded the profitability of banks’ traditional activities. Bank failures, insignificant from 1934, the date the Glass-Steagall Act was passed, until 1980, rose exponentially in the 1980s. The leading explanation for the persistence of these trends centers on fixed-rate deposit insurance: the insurance gives bank shareholders an incentive to take on risk when the value of bank charters falls. We propose and test an alternative explanation based on corporate control considerations. We show that managerial entrenchment, more than moral hazard associated with deposit insurance, explains the recent behavior of the banking industry.
In a generalized-least-squares (GLS) regression of mean returns on betas, the slope and R-squared are determined uniquely by the mean-variance location of the market index relative to the minimum-variance boundary. In contrast to ordinary-least-squares, GLS gives a zero slope only if the mean return on the market index equals that of the global minimum-variance portfolio. When fitted mean returns from a cross-sectional regression on any variables serve as inputs to standard portfolio optimization, GLS regression provides the optimal inputs, and that regression’s R-squared depends on the relative efficiency of the optimized portfolio.
A Bayesian approach is used to investigate a sample’s information about a portfolio’s degree of inefficiency. With standard diffuse priors, posterior distributions for measures of portfolio inefficiency can concentrate well away from values consistent with efficiency, even when the portfolio is exactly efficient in the sample. The data indicate that the NYSE-AMEX market portfolio is rather inefficient in the presence of a riskless asset, although this conclusion is justified only after an analysis using informative priors. Including a riskless asset significantly reduces any sample’s ability to produce posterior distributions supporting small degrees of inefficiency.
This paper presents a simple general equilibrium model of asset pricing in which profitable informed trading can occur without any “noise” added to the model. It shows that models of profitable informed trading must restrict the portfolio choices of uninformed traders: in particular, they cannot buy the market portfolio. In this model, profitable informed trading lowers the welfare of all agents when compared across steady states.
In efficient markets the price should reflect the arrival of private information. The mechanism by which this is accomplished is arbitrage. A privately informed trader will engage in costly arbitrage, that is, trade on his knowledge that the price of an asset is different from the fundamental value if: (1) his order does not move the price immediately to reflect the information; (2) he can hold the asset until the date when the information is reflected in the price. We study a general equilibrium model in which all agents optimize. In each period, there may be a trader with a limited horizon who has private information about a distant event. Whether he acts on his information, and whether subsequent informed traders act, is shown to depend on the possibility of a sequence or chain of future informed traders spanning the event date. An arbitrageur who receives good news will buy only if it is likely that, at the end of his trading horizon, a subsequent arbitrageurs’ buying will have pushed up the expected price. We show that limited trading horizons result in inefficient prices because informed traders do not act on their information until the event date is sufficiently close.
Oil futures markets frequently exhibit backwardation whereby more distant oil futures prices are below the current spot price. This is inconsistent with Hotelling’s theory that the net price of an exhaustible resource rises over time at the interest rate. We characterize an oil well as a call option and show that backwardation is necessary to induce production. Production is shown to be non-increasing in the riskiness of future prices. The empirical analysis indicates that U.S. oil production is directly related to the backwardation and inversely related to implied volatility. Backwardation is positively related to implied volatility and to the at-the-money put option price.
We relate wealth redistribution, asset pricing, and trade in financial assets by introducing heterogeneous agents into a Lucas tree-model. Heterogeneity of agents causes trade in financial assets and dynamic wealth redistribution. When consumers have time-separable, constant elasticity utilities with constant time-discount factors, the price-representative consumer has declining temporal relative risk aversion and intertemporal discount factors. Resulting asset prices “over-react”: Adverse aggregate consumption shocks cause wealth redistribution towards more risk averse consumers, reinforcing the adverse market value effect. Interest rates, risk premia, return volatility, and trade volume exhibit time-variance.
We derive a closed-form expression for the differences between forward and futures prices in the framework of a Lucas (1978) equilibrium model. We calculate this difference for fixed-income securities in two ways: 1. Using historic interest rate data to calibrate the matrix of nominal state price, and 2. By testing a large sample of randomly-generated state price matrices. In both cases we find few meaningful differences between futures and forward prices. Larger differences are generated for highly diagonal state price matrices. We conclude that in most economically relevant circumstances the costs of marking to market for fixed income securities are negligible.
Inventory and asymmetric information models of price formation, as well as the stabilization rules of the NYSE, all have different predictions for the autocorrelation function of bid or ask changes and the relation between bid or ask changes and order imbalances. The actual autocorrelation function of bid or ask changes for NYSE common stocks in 1990 does not conform to any of these predictions: The first order autocorrelation is negative, the second order positive, the third order negative, and so on. There is a weak relation between a particular measure of order imbalance and subsequent quote revisions, consistent with the predictions of some theoretical models of price formation.
We develop routines in Mathematica for pricing various European and American options using the binary option model and Monte Carlo methods. As might be expected, Mathematica permits parsimonious programming of the option pricing expressions.
This paper derives closed-form solutions for the investment and market value, under uncertainty, of competitive firms with constant returns to scale production and convex costs of adjustment. Solutions are derived for the case of irreversible investment as well as for reversible investment. Optimal investment is a non-decreasing function of q, the shadow value of capital. The conditions of optimality imply that q cannot contain a bubble; thus, optimal investment depends only on fundamentals. However, the value of the firm may contain a bubble that does not affect investment behavior. Relative to the case of reversible investment, the introduction of irreversibility does not affect q, but it reduces the fundamental market value of the firm.
Two hypotheses concerning firms issuing debt for the first time are tested. The first is that new firms’ debt will be discounted more heavily by lenders, compared to firms which have credit histories (but are otherwise identical), and that this excess discount declines over time as lenders observe defaults. The declining interest rate corresponds to the formation of a “reputation”, a valuable asset which provides an incentive for firms to not choose risky projects. The second hypothesis is that prior to the establishment of a reputation new firms issuing debt are monitored more intensely. The sample studied consists of new banks issuing bank notes for the first time during the American Free Banking Era (1838-1860). The presence of a reputation effect in debt prices is confirmed: the debt of new banks is discounted more heavily than banks with credit histories. Note holders are then motivated to monitor new banks because the excess discount provides an incentive for notes of new banks to be redeemed. As lenders learn that new banks can redeem their notes, the discount declines as predicted for surviving banks. The precision of learning increases during the period due to technological improvements in information transmission, namely, the introduction of the telegraph and the railroad. The results explain why the pre-Civil War system of private money issuance by banks was not plagued by problems of overissuance (“wildcat banking”).
This paper extends the theory of investment under certainty to incorporate fixed costs of investment, a wedge between the purchase price and sale price of capital, and potential irreversibility of investment. In this extended framework, investment is a non-decreasing function of q, the shadow price of installed capital. The optimal rate of investment is in one of three regimes (positive, zero, or negative gross investment) depending on the value of q relative to two critical values. In general however, the shadow price q is not directly observable, so we present two examples relating q to observable variables.
Evidence of deviations from the Capital Asset Pricing Model (CAPM) has accumulated over the past decades. The source of these deviations is often assumed to be missing risk factors, leading researchers to look to multifactor asset pricing models as alternatives. The analysis in this paper suggests that consideration of different alternatives may be fruitful, since existing evidence against the CAPM is also evidence against multifactor alternatives being the whole story. Data-snooping biases, market failure, or market inefficiencies are more likely explanations for the deviations.
The framework creates for the analysis provides additional insights. These insights relate to the role of residual risk in explaining the cross-section of expected asset returns and to the formation of prior distributions for Bayesian analysis of portfolio efficiency.
Recent data released pursuant to the 1989 amendments to the Home Mortgage Disclosure Act (HMDA) which show large disparities in mortgage lending between minority and non-minority neighborhoods, have refocused the attention of policy makers, lenders, community advocates and academics on possible racial discrimination in the home loan market. In this paper, we review the existing literature on redlining. Many of the methodological shortcomings of the previous studies can be remedied by using post-1989 HMDA data to examine whether lender acceptance or rejection of mortgage applications is related to racial and ethnic neighborhood composition. We test two models of the lender’s decision to accept or reject loan applicants, one including and one without variables that proxy for neighborhood risk using data for Boston and Philadelphia. With proxies for neighborhood risk included, the results do not support the hypothesis that financial institutions redline neighborhoods in these two cities.
Asymptotic variances of estimated parameters in models of conditional expectations are calculated analytically assuming a GARCH process for conditional volatility. Under such heteroskedasticity, OLS estimators of parameters in single-period models can possess substantially larger asymptotic variances than GMM estimators employing additional multiperiod moment conditions – an approach yielding no efficiency gain under homoskedasticity. In estimating models of long-horizon expectations the VAR approach provides an efficiency advantage over long-horizon regressions under homoskedasticity, but that ordering can reverse under heteroskedasticity, especially when the conditional mean and variance are both persistent. In such cases, the VAR approach maintains a slight efficiency advantage if the OLS estimator is replaced by an alternative GMM estimator. Heteroskedasticity can increase dramatically the apparent asymptotic power advantages of long-horizon regressions to reject constant expectations against persistent alternatives.
This paper examines empirically the behavior of institutional traders using unique data on the equity transactions of 21 institutions of differing investment styles during 1991-1993. The data provide a detailed account of the anatomy of the trading process, and include information on the number of days needed to fill an order and types of order placement strategies employed. We analyze the determinants of trade duration and the decisions regarding order type. Our analysis provides some support for the predictions made by theoretical models, but suggests that these models fail to capture important dimensions of trading behavior.
This paper documents the frequency of non-trading for NYSE and AMEX stocks based on information in the CRSP monthly and daily data files. We find a declining pattern of non-trading over the 1926 to 1990 period: 23.4 percent of NYSE stocks do not trade on an average (end-of-month) day during the 1926 to 1945 period, compared with 1.29 percent on average over all days during the 1973-1990 period. In the 1973-1990 period, non-trading averaged more than 15 percent for AMEX firms. We find that the average amount of non-trading is larger for smaller stocks, is lowest at the end of the year, and tends to be lowest at the beginning of the week and is highest at the end of the week. We also find substantial heterogeneity in the amount of non-trading across the stocks within each size decile. For example, while 10 percent of the stocks in the smallest decile trade virtually every trade day, 10 percent of the stocks in that decile do not trade on 51 percent of the trade days during the year, and one percent do not trade on 76 percent of the trade days during the year. Finally, we briefly discuss some implications of our non-trading evidence for measured autocorrelations.
In this paper, a genetic algorithm is used to find technical trading rules for Standard and Poor’s Composite Stock Index in 1963-89. Compared to a simple buy-and-hold strategy, these trading rules lead to positive excess returns in the out-of-sample test period of 1970-89. In addition, the rules appear to reduce the variability of the returns. The results are compared to benchmark models of a random walk, an autoregressive model, and a GARCH-AR model. Conventional statistical tests and bootstrapping simulations are carried out to study the robustness of the results. It is found that the excess returns are both statistically and economically significant, even when transaction costs are taken into account.