Recent studies indicate that the widespread assumption of parameter stationarity in empirical applications of asset pricing models may be inappropriate. This paper investigates the feasibility of modeling parameter instability as a sequence of persistent stable regimes. Recursive residual and log likelihood techniques are combined to detect and locate shift points. The results indicate that regime shifts are widespread, frequent, and often large enough to significantly effect empirical findings. The nature of the shifts appears to be a rotation of the regression line, indicating that correction of both alpha and beta parameters is required.

The current (Summer 1985) world oil price, and changes since 1973, cannot possibly be explained by scarcity, or by changes in scarcity. The level and dispersion of marginal costs, and the pattern of investment behavior since 1973, prove that supply is being restricted to maintain the price, which can be maintained so long as the low-cost oil is dammed up. If the dam breaks, so will the Statement of the Problem M X ANY public and private investment decisions, affecting a significant fraction of world income, depend on the expected price of crude oil. Since the price explosion of 1973, and especially since the second explosion of 1979, there has been a wide consensus: Oil has become, and will continue to be, in increasingly short supply for the rest of the century. Further price increases are necessary and inevitable. We need cite only a fraction of even the post1978 predictions of rising oil prices, which were and are used as a basis for private and public investment, and taxation. (U.S. CIA, 1979, p. iii; Fesharaki, 1980; U.S. Senate, 1982; OGJ, 1982, pp. 118, 210; OGJ, 1984a, p. 32; Canada, 1980, 1981a, 1981b.) The price decline since 1980-81 has not weakened the consensus, but has lowered the rate of expected increase. Prices are expected to be weak or stable for a few years, then begin the inexorable increase. (EMF 6, 1982; US. GAO, 1983; World Bank, 1983, p. 28; and 1985, pp. 37-38; New York Times, 1984; DOE, 1985; Erickson, 1985; Saunders, 1984.) The oil-company acquisitions of 1983-84 were obviously based on the consensus view. An expert panel convened by the CIA in April 1985, like one convened by the California Energy Commission, expected that prices would rise, at an increasing rate, starting around 1990. (Wall Street Journal, 1985; California, 1985.) The International Energy Workshop of the International Institute of Applied Systems Analysts compiled a consensus forecast in December 1981, July 1983, and July 1985. The first consensus was that the price would nearly equal $60 in 1990; the most recent has it approaching that number only by 2010. A twenty-year postponement is no small change, of course. But in each case, the consensus was that the current price of oil was approximately equal to, or mildly above, the scarcitydetermined price. In the long run, it would have to rise above the current level. (Manne and Nordhaus, 1985; it should not be assumed that they themselves share the consensus, or that they do not.) The Economic Theory of the Consensus In general, absent monopoly restriction of output, a rising price registers increasing scarcity as consumption puts increasing strain upon productive resources. Some formal oil price models are explicitly competitive. (Mead, 1979, 1985; Cremer and Salehi-Isfahani, 1980; MacAvoy, 1982; Roumasset et al., 1983). More often, competition is an implicit major premise. Prices are said to reflect not monopoly but forces or deeper forces. Or it is said that if the cartel of producing nations (not the unimportant organization OPEC) disappeared, oil prices would not be strikingly different from what they are. Or that discoveries had been shrinking, and demand increasing up to 1973, hence prices had to rise, so at most the cartel pushed them up a bit faster. Other models are non-committal, but implicitly competitive, in that they have price driven by total consumption pressing against reserves. Figure 2 does not, of course, disprove this proposition but it does suggest that if scarcity pushed up prices, scarcity must have been very sudden and strong. There have also been attempts to model the oil market as a monopoly, in the generic sense: a few Received for publication June 24, 1985. Revision accepted for publication November 19, 1985. * Massachusetts Institute of Technology. The research for this paper has been supported by the National Science Foundation, grant SES-8412971, and by the Center for Energy Policy Research of the M.I.T. Energy Laboratory. I am obliged to Michael C. Lynch for valuable assistance. For many helpful comments and criticisms, I am indebted to Paul G. Bradley, Harry G. Broadman, Richard L. Gordon, William W. Hogan, Gordon M. Kaufman, Stephen Martin, James W. McKie, Joe Roeber, James L. Smith, G. Campbell Watkins, and to two anonymous referees. But any opinions, findings, conclusions or recommendations expressed herein are those of the author, and do not necessarily reflect the views of the NSF or any other person or group. Copyright ? 1986 [ 387 ] This content downloaded from 157.55.39.17 on Wed, 31 Aug 2016 04:37:42 UTC All use subject to http://about.jstor.org/terms 388 THE REVIEW OF ECONOMICS AND STATISTICS sellers restraining output and raising prices. (For an excellent survey, see Gately (1984).) But these models are also based on rising scarcity. They aim to show the difference between the competitive response and the monopoly response. Perhaps, given the increasing scarcity inherent in an exhaustible natural resource... there is a very limited scope for the monopolist ... indeed, under the natural 'first approximation' of constant elasticity demand schedules, with zero extraction costs, monopoly prices and competitive equilibrium prices will in fact be identical (Stiglitz,

The impact of uncertainty in future income on the demand for education from both theoretical and empirical points of view is analyzed. Theoretical results deviate substantially from their counterparts in models of human capital formation with certain future income. The theory is tested with a sample of high school graduates that contains data on subjective expectations. Empirical evidence from binomial logit analysis does not entirely support the behavioral implications of the theoretical model.

The paper analyzes the effect of tax-deferred individual retirement accounts (IRAs) in the United States on net individual saving. The results are based on a model of constrained optimization with the limit on tax-deferred saving the principle constraint. The estimates suggest that contributions to IRAs represent substantial net saving increases. Were the IRA limit to be increased, only about 10 to 20% of resulting increase in IRA contributions would be taken from other savings. About 50% would come from reduced consumption and about 35% from reduced taxes.

Yields on short-term prime-grade municipals vary through time in relation to after-corporate-tax yields on short-term U.S. Treasury securities. The pattern is not related to the default premium in municipal yields or to the historical ceiling on bank deposit rates (Regulation Q). However, there is a strong link to the default premium in corporate yields and to municipal holdings by large commercial banks. These findings suggest that taxable and tax-exempt markets are linked both by the capital-structure decisions of firms and by the tax-arbitrage activities of banks.

We use a translog production function approach to examine the effects of medical staff physicians on hospital production, and how their effects differ in teaching and nonteaching hospitals. In teaching hospitals, we also focus on the special role of medical residents. We find that physicians have a strong positive influence on the productivity of other inputs, and that they are substitutes for other resources. Controlling for patient casemix causes significant changes in estimated marginal products; those of labor inputs increase and that of capital declines. The implications of our findings for policy are explored.

A model of aggregate economic activity is formulated which enmphasizes the effects of borrowing constraints in the presence of uninsurable risk. An important determinant of current income level is shown to be the cross-sectional distribution of wealth. As this distribution evolves endogenously, the model is capable of producing rich dynamics from a simple specification of exogenous shocks. The model shows that this phenomena can contribute to observed price volatility. IT IS COMMONLY THOUGHT that individuals have only limited opportunities to borrow against future labor income and cannot totally insure all types of risk. It has also been suggested that such departures from the presumptive norm of frictionless, complete information capital markets may have implications for aggregate economic activity. AlthLough there has been some work analyzing the implications of borrowing constraints for individual savings behavior (18, 2, 8), there has been no systematic analysis of how such borrowing constraints will affect the time series properties of output, prices, and interest rates. In this paper, we present a completely specified infinitely lived two agent equilibrium model which emphasizes the roles of borrowing constraint and uninsured risk for affecting aggregate outcomes. Specifically we assume that agents are prohibited from ever having negative nonhuman wealth. The model has the central feature that there is no aggregate uncertainty, but each agent's own productive opportunities are stochastic. If there were a full set of Arrow- Debreu contingent claim markets each agent could attain a certain consumption stream and the resulting allocation and (implicit) relative prices would be constant through time. However, we assume that such markets do not exist. Rather, we assume that at each point in time agents may trade only the single durable asset for the single perishable consumption good. This may be interpreted either as fiat mon ay with a fixed own nominal return of zero, or as claims to productive capital which emits a fixed exogenous flow of the consumption good. We assume also that output may be produced by labor. However, only one of the two agents is productive at any instant in time. The duration of time over which a single agent is productive is assumed to be random, and, for analytical simplicity, is assumed to be generated by a Poisson counting process. The resulting allocation has the property that the agent who is not productive exchanges some of his

[This paper develops a model of labor supply for married women which takes into account both the joint decision on participation and hours, and the nonlinear shape of the budget constraint due to taxation. The model can explain the absence of observations of the tax kink by assuming the existence of optimization errors in addition to errors capturing taste variation. The estimates of the model, which are obtained with British micro data, suggest that the overall wage elasticity is about 2 and that participation is more responsive to wages than hours of work.]

Empirical studies of consumer demand or of factor demand have now moved far beyond the Cobb-Douglas functional form and elasticities of interest are no longer estimated as parameters of the system. Instead, such elasticities are typically non-linear functions of the parameters that have been estimated and it is natural to want to be able to say something about the statistical properties of such elasticities. One way of dealing with this is to linearly approximate the elasticity formulas (in terms of the estimated parameters) and use classical statistical procedures to get approximations to the underlying variances. If y-f(x) and x has a variance covariance matrix V, the linear approximation is given by: Var(y) (8f/8x)V(8f/8x). The data needed for such an approximation are estimates of the parameters and of the associated variance-covariance matrix. Some of the earliest references that we have found to uses of this approximation technique in the elasticity context are to Griffin and Gregory (1976), Griffin (1977), and Fuss (1977), while the earliest references to