Knowledge that Transforms

ECONOMIC DECISION MAKING UNDER UNCERTAINTY often takes place in the presence of multiple and in markets that are less than complete. As a consequence, choices about endogenous sometimes must be made while simultaneously facing one or more immutable exogenous risks that are not under the control of the agent, and that are independent of endogenous risks. It is somehow natural to assume that an exogenous deterioration in background wealth will cause an individual to take more care elsewhere. If we define a deterioration, for example, as making the individual poorer by removing a fixed amount of initial wealth, we know from Pratt (1964) that decreasing absolute risk aversion (DARA) of an individual's von Neumann-Morgenstern utility function yields this natural result. If, on the other hand, background wealth becomes riskier due to the addition of a zero-mean risk, that is also statistically independent of all other risks, behavior will be more risk averse if and only if preferences are risk vulnerable as defined by Gollier and Pratt (1996). Risk vulnerability (described below in Section 4) is a stronger notion than DARA and includes proper risk aversion (Pratt and Zeckhauser (1987)) and standard risk aversion (Eeckhoudt and Kimball (1992), Kimball (1993)) as particular cases. But a deterioration in background wealth may encompass more complicated distribution changes than the introduction of another statistically independent risk. In this paper, we examine background wealth deteriorations that take the form of both general firstand second-degree stochastic dominance changes in risk (FSD and SSD respectively). In particular, we determine conditions that are both necessary and sufficient for each of these two types of background risk changes to imply more risk-averse behavior on the part of the individual. For the case of FSD changes, this condition turns out to be Ross' stronger characterization of decreasing absolute risk aversion. In the case of general SSD changes in the distribution of background wealth, the condition derived is a stronger version (in Ross' sense) of the conditions characterizing preferences that are locally risk vulnerable in the sense of Gollier and Pratt. The necessary and sufficient conditions derived are fairly restrictive upon preferences. However, if we take as positive behavior that individuals act in a more risk-averse manner whenever the distribution of background wealth deteriorates, these conditions place canonical limits upon appropriate utility representations. At the very least, they

This paper derives the asymptotic power envelope for tests of a unit autoregressive root for various trend specifications and stationary Gaussian autoregressive disturbances. A family of tests is proposed, members of which are asymptotically similar under a general 1(1) null (allowing nonnormality and general dependence) and which achieve the Gaussian power envelope. One of these tests, which is asymptotically point optimal at a power of 50%, is found (numerically) to be approximately uniformly most powerful (UMP) in the case of a constant deterministic term, and approximately uniformly most powerful invariant (UMPI) in the case of a linear trend, although strictly no UMP or UMPI test exists. We also examine a modification, suggested by the expression for the power envelope, of the Dickey-Fuller (1979) t-statistic; this test is also found to be approximately UMP (constant deterministic term case) and UMPI (time trend case). The power improvement of both new tests is large: in the demeaned case, the Pitman efficiency of the proposed tests relative to the standard Dickey-Fuller t-test is 1.9 at a power of 50%. A Monte Carlo experiment indicates that both proposed tests, particularly the modified Dickey-Fuller t-test, exhibit good power and small size distortions in finite samples with dependent errors.

Our general subject is model determination methods and their use in the prediction of economic time series. The methods suggested are Bayesian in spirit but they can be justified by classical as well as Bayesian arguments. The main part of the paper is concerned with model determination, forecast evaluation, and the construction of evolving sequences of models that can adapt in dimension and form (including the way in which any nonstationarity in the data is modelled) as new characteristics in the data become evident. The paper continues some recent work on Bayesian asymptotics by the author and Werner Ploberger (1995), develops embedding techniques for vector martingales that justify the role of a class of exponential densities in model selection and forecast evaluation, and implements the modelling ideas in a multivariate regression framework that includes Bayesian vector autoregressions (BVAR's) and reduced rank regressions (RRR's). It is shown how the theory in the paper can be used: (i) to construct optimized BVAR's with data-determined hyperparameters; (ii) to compare models such as BVAR's, optimized BVAR's, and RRR's; (iii) to perform joint order selection of cointegrating rank, lag length, and trend degree in a VAR; and (iv) to discard data that may be irrelevant and thereby reset the initial conditions of a model.

This paper explores a one-agent Bayesian model of learning by doing and technological choice.To produce output, the agent can choose among various technologies.The beneficial effects of learning by doing are bounded on each technology, and so long-run growth in output can take place only if the agent repeatedly switches to better technologies.As the agent repeatedly uses a technology, he learns about its unknown parameters, and this accumulated expertise is a form of human capital.But when the agent switches technologies, part of this human capital is lost.It is this loss of human capital that may prevent the agent from moving up the quality ladder of technologies as quickly as he can, since the loss is greater the bigger is the technological leap.We analyze the global dynamics.We find that a human-capital-rich agent may find it optimal to avoid any switching of technologies, and therefore to experience no long-run growth.On the other hand, a human-capital-poor agent, who because of his lack of skill is not so attached to any particular technology, can find it optimal to switch technologies repeatedly, and therefore enjoy long-run growth in output.Thus the model can give rise to overtaking.

We consider the situation where a single consumer buys a stream of goods from different sellers over time. The true value of each seller's product to the buyer is initially unknown. Additional information can be gained only by experimentation. For exogeneously given prices the buyer's problem is a multi-armed bandit problem. The innovation in this paper is to endogenize the cost of experimentation to the consumer by allowing for price competition between the sellers. The role of prices is then to allocate intertemporally the costs and benefits of learning between buyers and sellers. We examine how strategic aspects of the oligopoly model interact with the learning process. All Markov perfect equilibria (MPE) are efficient. We identify an equilibrium which besides its unique robustness properties has a strikingly simple, seemingly myopic pricing rule. Prices below marginal cost emerge naturally to sustain experimentation. Intertemporal exchange of the gains of learning is necessary to support efficient experimentation. We analyze the asymptotic behavior of the equilibria.

mutation.) Kandori, Mailath, and Rob (henceforth KMR) first provide a useful general theorem concerning the stationary distribution of strategies under Darwinian dynamics. They then divide the analysis of the 2 x 2 game into three cases: dominant strategy games (e.g., prisoners' dilemma), coordination games, and games with no symmetric pure strategy equilibrium (e.g., battle of the sexes). We refer to these as DS, C, and NP games. In each case, KMR claim that, as the rate of mutation vanishes, the stationary distribution of strategies converges to a symmetric Nash equilibrium. They emphasize C games, which have two symmetric Nash equilibria, and characterize the conditions under which the distribution converges to the risk dominant equilibrium. In this note, we argue that while their formal conclusions for C games are correct, their results for DS and NP games are valid only for large populations of players. In small populations, Darwinian dynamics may produce non-Nash outcomes in these two cases. Section 1 summarizes the KMR model, and Section 2 provides examples of 2 x 2 games in which Darwinian dynamics generate non-Nash outcomes. A theorem in Section 3 describes the Darwinian equilibrium of any 2 x 2 game.