Kenneth Arrow Quotes

Krugman's whole attack is directed at a statement made neither by Arthur nor by Cassidy. Krugman has not read Cassidy's piece with any care nor has he bothered to review what Arthur has in fact said.

Child care has grown up under different circumstances than education and probably for a mixture of reasons, good and bad. There are many systems of child care, some private, some public. As compared with primary and secondary education, there is clearly less need for coordination. The sequencing of classes is much less important. It would appear that the ability of parents to monitor the conduct of the child care activity is much greater because the activity is much closer to everyday experience and knowledge. Most of the informational and structural arguments for the public supply of education are absent in the case of child care. Reputation and experience may suffice for adequate monitoring.

For the voucher system to work, it would be necessary to have informed parents. One cannot be dogmatic without empirical evidence, but I would be surprised if the average parent has the time or patience or competence to digest the relevant information. Indeed, one wonders where the information is to come from and in what form it should exist. Do we use test scores, themselves affected by the selection processes of the students? Impressions of individual teachers or of the physical appearance of the school will tend to dominate.

The conflict of incentives and information is increasingly leading to a diffusion of responsibility for medical decisions. The simple picture of the physician making decisions for the patient has certainly become more complicated. Patients always had a role in choosing to seek medical advice and from whom. Their choices are becoming increasingly restricted as medical practice becomes more organized. The problems of cost control in an insured world are partly met by the increasing use of control of medical services through review by health maintenance organizations and insurance carriers.

The slippery role of information as an economic good is of deep significance to economic behavior, especially in the relatively information-rich modern economy. It is an economic good in the traditional sense; it is valuable, and it is costly. But it has a peculiar algebra. Adding one ton of steel to another permits more to be done; repeating the same item of information does not add anything useful. On the other hand, supplying a ton of steel to another reduces the steel available to the supplier; supplying information to another does not reduce the information available to the supplier.

I think one of the things we learned from the physicists and also the theoretical biologists is the idea that when you're dealing with very complex systems you're going to get a large variety of behavior which can be interpreted as hill climbing, but hill climbing with a lot of modifications, hill climbing with big jumps occasionally. This is an elaboration of the idea of the learning model. The learning model story takes off from psychology, but the adaptive processes take off from biology and physics. They have the same story. One thing it does suggest in some sense is that we have to be more modest in what we claim.

Unfortunately, the word 'learning' is a very general word. It isn't a very specific theory and we can have a lot of learning models, and it's unlikely that any one is going to track. When you talk about learning, you talk about the human mind adapting to conditions, and we haven't nailed that down very well. This is always an objection to the whole idea of bounded rationality. Not that it's wrong, but if it's right, it doesn't actually tell you what to do. Rationality is unique. That isn't really quite true, but at least under many circumstances it is. To say that we're not at the top of a hill gives you a lot of variety as to where you might be. So the problem with bounded rationality is not that it's wrong. On the contrary, I think it's very apt to be correct. It's just that its predictions are a lot more vague than those implied by rationality. At the moment, I don't know what to do about that.

"Information and the Organization of Industry", Rivista Internazionale di Scienze Sociali, Anno 102, No. 2 (aprile-giugno 1994)

Information, one of the fundamental determinants of production, laps over from one firm to another, yet the firm has so far seemed reasonably sharply defined in terms of legal ownership. It seems to me there must be increasing tensions between legal relations and fundamental economic determinants. Small symptoms are already appearing in the legal and economic spheres. There is continual difficulty in defining intellectual property. The United States courts, at least, have come up with some strange definitions of property.

I am old-fashioned enough to retain David Hume’s view that one can never derive “ought” propositions from “is” propositions. The two issues, method and value, are distinct.

As a general rule, the greater the uncertainty, the better to avoid large and irreversible commitments, to the extent that it is possible. When the famous 1930s gangster, Dutch Schultz, was dying, his incoherent last remarks were taken down by a stenographer. One of them was. "Don't make no bull moves." His words are a lesson for the kind of future that one might choose. Maintaining flexibility or keeping ones options open, is key in these matters.

The uncertainty of the future is inescapable, one must think about it and arrive at plans for action. A statement attributed to a number of thinkers is, "Prediction is very difficult, especially of the future." Postdiction, knowing what went on in the past, is also difficult. The past, however, is our basis for understanding the future.

Even Ricardo's most famous accomplishment, the law of comparative advantage in foreign trade, is incomplete, though not wrong.

The implications for planning are ambiguous. On the one hand, the optimal allocation under increasing returns will not be obtained under free markets. In fact, as I have already remarked, the competitive equilibrium is not even viable, and the outcome will be some kind of imperfect competition. But optimal allocation under increasing returns is difficult. For one thing, the optimisation requires the use of integer programming, a procedure intrinsically more complex than linear programming, and impossible to carry out for large systems even with the most powerful computers. Even more serious are the data demands. The information needed is widely dispersed among the industries and cannot be effectively communicated. It is for these reasons that decentralised decision-making with some element of monopoly is likely to be more efficient.

The tension between chaotic behavior and perfect foresight was observed. Start with an equilibrium dynamics of a standard type derived from the hypothesis that future prices are predicted perfectly. Suppose that the solution to the difference equations characterizing the solution exhibits chaotic behavior. Is it realistic to assume that the future, even though deterministic, is in fact predictable? Clearly, part of the lessons drawn by natural scientists, especially meteorologists, from nonlinear dynamics is precisely the opposite; chaotic behavior implies that small errors of observation in the starting position may lead to virtually total unpredictability after some period of time. This creates no difficulties of consistency when the predictor is not part of the system being predicted. But when the predictors are the economic agents being examined, there is a fundamental inconsistency. This epistemological antinomy is reinforced by the empirical observation that actual behavior of prices of assets such as securities could never reasonably have been predicted; if it had, there would have been much more buying or selling at earlier stages.