American Economic Association The Investigation of Economic Expectations Author(s): Holbrook Working Source: The American Economic Review, Vol 39, No 3, Papers and Proceedings of the Sixty first Annual Meeting of the American Economic Association(May, 1949), pp. 150-166 Published by: American Economic Association StableUrl:http://www.jstororg/stable/1831740 Accessed:11/09/201303:21 Your use of the JSTOR archive indicates your acceptance of the Terms Conditions of Use, available at JStOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support(@jstor. org American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The American economic revie 的d http://www.jstororg This content downloaded from 202. 115.118.13 on Wed, I I Sep 2013 03: 21: 32 AM All use subject to STOR Terms and Conditions
American Economic Association The Investigation of Economic Expectations Author(s): Holbrook Working Source: The American Economic Review, Vol. 39, No. 3, Papers and Proceedings of the Sixtyfirst Annual Meeting of the American Economic Association (May, 1949), pp. 150-166 Published by: American Economic Association Stable URL: http://www.jstor.org/stable/1831740 . Accessed: 11/09/2013 03:21 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. . American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The American Economic Review. http://www.jstor.org This content downloaded from 202.115.118.13 on Wed, 11 Sep 2013 03:21:32 AM All use subject to JSTOR Terms and Conditions
THE INVESTIGATION OF ECONOMIC EXPECTATIONS By HoLBROOK WORKING Food Research Institute, Stanford University Fifty years ago economists rather generally allowed themselves to ignore economic expectations by concentrating attention on the static state. Today we talk a great deal about economic expecta tions, but how much do we know about their formation and their behavior? How much can we say about them that might not have been said by Alfred Marshall or by John Stuart Mill, or even by Adam Smith, if they had chosen to discuss expectations? I. The General problem There are two broad lines of possible attack on the subject of economic expectations: the individualistic and the aggregative. In fundamentals, the problem of individual expectations may be one for psychology rather than for economics, but the economist must not hesitate to go as far as he profitably can in considering the individual a good deal of information has been collected from time to time on economic expectations of individuals. Now and then someone becomes interested in the opinions of economists and makes a survey of their expectations regarding prices. Surveys of the expectations of businessmen have been made frequently by various agencies. If these surveys had seemed to produce results of much practical or scientific value, they could and should be organized more systematically and made the business, perhaps, of the federal Department of Commerce. Thus far, at least, it seems that the information gathered falls more in the category of news of current interest than in that of significant economic information Perhaps the information has lacked scientific value because the nets of inquiry have not been cast widely enough, or in the right quarters. Or perhaps the absence of recognized scientific significance reflects inadequacy of our perception rather than absence of real signincance I have not made a search for serious anal- lyses of data on individual economic expectations, but I think it safe to say that there have been few of them There is one which comes to mind, however and the results are worth noting. Some years ago Alfred Cowles made a The annual sur er finances instituted a few years ago by the board of of the Federal Reserve System are providing some information on consumer expectations and could easily provide more if the effort should seem worth whil content down ed stroe 2 2S R e os wecoldsep2
THE INVESTIGATION OF ECONOMIC EXPECTATIONS By HOLBROOK WORKING Food Research Institute, Stanford University Fifty years ago economists rather generally allowed themselves to ignore economic expectations by concentrating attention on the "static state." Today we talk a great deal about economic expectations, but how much do we know about their formation and their behavior? How much can we say about them that might not have been said by Alfred Marshall or by John Stuart Mill, or even by Adam Smith, if they had chosen to discuss expectations? I. The General Problem There are two broad lines of possible attack on the subject of economic expectations: the individualistic and the aggregative. In its fundamentals, the problem of individual expectations may be one for psychology rather than for economics, but the economist must not hesitate to go as far as he profitably can in considering the individual. A good deal of information has been collected from time to time on economic expectations of individuals. Now and then someone becomes interested in the opinions of economists and makes a survey of their expectations regarding prices. Surveys of the expectations of businessmen have been made frequently, by various agencies. If these surveys had seemed to produce results of much practical or scientific value, they could and should be organized more systematically and made the business, perhaps, of the federal Department of Commerce.' Thus far, at least, it seems that the information gathered falls more in the category of news of current interest than in that of significant economic information. Perhaps the information has lacked scientific value because the nets of inquiry have not been cast widely enough, or in the right quarters. Or perhaps the absence of recognized scientific significance reflects inadequacy of our perception rather than absence of real significance. I have not made a search for serious analyses of data on individual economic expectations, but I think it safe to say that there have been few of them. There is one which comes to mind, however, and the results are worth noting. Some years ago Alfred Cowles made a 'The annual surveys of consumer finances instituted a few years ago by the Board of Governors of the Federal Reserve System are providing some information on consumer expectations and could easily provide more if the effort should seem worth while. This content downloaded from 202.115.118.13 on Wed, 11 Sep 2013 03:21:32 AM All use subject to JSTOR Terms and Conditions
RY AND MEASUREMENT OF PRICE EXPECTA Itical study of the economic expectations recorded by professional stock-market forecasters. The main conclusion which emerged was that these expectations had characteristics substantially identical with those of random guesses. We shall see before we finish that such an observation is not necessarily evidence of poor forecasting Anyone who contemplates analysis of existing data on individual economic expectations, and especially anyone who proposes large-scale collection of more data, should consider certain difficulties that must be faced in their interpretation. First is the question of validity of the expressions of expectation. Many people will state expectations with great confidence, but evince no confidence at all when faced with a challenge to act on their stated opinions. Second is the problem of classifying persons queried in relation to the economic significance of any expectations which they may have. Some expectations of some people carry great weight in economic affairs, yet most expectations of most people carry little weight. I do not mean these comments as deprecation of the study of indi vidual expectations, but only to suggest that any other feasible approach needs consideration also I. The aggregative Approach The aggregative approach to study of economic expectations takes for analysis some recorded economic variable which reflects a sort of consensus of expectations. Many economic quantities are strongly fluenced by expectations and some are little influenced by anything else. Prices of industrial stocks, for example, are predominantly reflec tions of economic expectations. They may be regarded as expectations capitalized at going rates of interest, and in that sense they appear to reflect both expectations and existing rates of interest-or should I say that the capitalization is at expected rates of interest? The price of a commodity future is even more clearly and specifically a composite measure of economic expectations When we undertake study of the behavior of futures prices as a means toward knowledge of economic expectations, we must consider the objection, or qualification, that a futures price is not precisely a composite measure of expectations. British writers, including such distinguished economists as Keynes, Kaldor, Hicks, and Hawtrey, have given the question a good deal of attention since Keynes advanced eory rmal backwa ardation"in 1930. They have disagreed mewhat as economists are wont to do but i think that they are Alfred Cowles, " Can Stock Market Forecasters Forecast?"Econometrica, July, 1933 J. M. Keynes, Treatise on Money (New York, 1930), Il, p. 14 content down ed stroe 2 2S R e os wecoldsep2
THEORY AND MEASUREMENT OF PRICE EXPECTATIONS 15 1 critical study of the economic expectations recorded by professional stock-market forecasters. The main conclusion which emerged was that these expectations had characteristics substantially identical with those of random guesses.2 We shall see before we finish that such an observation is not necessarily evidence of poor forecasting. Anyone who contemplates analysis of existing data on individual economic expectations, and especially anyone who proposes large-scale collection of more data, should consider certain difficulties that must be faced in their interpretation. First is the question of validity of the expressions of expectation. Many people will state expectations with great confidence, but evince no confidence at all when faced with a challenge to act on their stated opinions. Second is the problem of classifying persons queried in relation to the economic significance of any expectations which they may have. Some expectations of some people carry great weight in economic affairs, yet most expectations of most people carry little weight. I do not mean these comments as deprecation of the study of individual expectations, but only to suggest that any other feasible approach needs consideration also. II. The Aggregative Approach The aggregative approach to study of economic expectations takes for analysis some recorded economic variable which reflects a sort of consensus of expectations. Many economic quantities are strongly influenced by expectations and some are little influenced by anything else. Prices of industrial stocks, for example, are predominantly reflections of economic expectations. They may be regarded as expectations capitalized at going rates of interest, and in that sense they appear to reflect both expectations and existing rates of interest-or should I say that the capitalization is at expected rates of interest? The price of a commodity future is even more clearly and specifically a composite measure of economic expectations. When we undertake study of the behavior of futures prices as a means toward knowledge of economic expectations, we must consider the objection, or qualification, that a futures price is not precisely a composite measure of expectations. British writers, including such distinguished economists as Keynes, Kaldor, Hicks, and Hawtrey, have given the question a good deal of attention since Keynes advanced his "theory of normal backwardation" in 1930.3 They have disagreed somewhat, as economists are wont to do, but I think that they are 2Alfred Cowles, "Can Stock Market Forecasters Forecast?" Econom,etrica, July, 1933, pp. 309-324. 'J. M. Keynes, Treatise on Money (New York, 1930), II, p. 143. This content downloaded from 202.115.118.13 on Wed, 11 Sep 2013 03:21:32 AM All use subject to JSTOR Terms and Conditions
AMERICAN ECONOMIC ASSOCIATION united in the opinion that the relation of a futures price to expectations may be expressed by the equation Pt=E (1) where Pt is a futures price, E is some sort of combination of effects of individual expectations, and r is some sort of combination of (effects of?) individual risk premiums This equation serves to indicate that in studying P, we may learn facts about r rather than about E. That need not trouble us. If we find significant behavior characteristics, we shall know that they derive from corresponding characteristics in either E or r, or from ome combination of both. We shall in fact find that the main char acteristics derive from E expectations We may now profitably adopt the term "market expectation synonymous with"futures price. " Firmly established habits of thought make it difficult to think of any price as resting primarily on expecta tions. By using“ market expectation” as an alternative to“ futures price, we may keep attention focused on the predominantly expecta tional quality of such a price Because the term"market expectation has no strongly fixed connotations, it should not be difficult to reme ber that, writing Em for market expectation, its definition leads equation(1)to Em=E (2) and therefore that market expectation is not precisely equivalent to E unless r=0 Ill. Bias in Market Expectations Expectations may be defective in either or both of two respects: chey may be randomly inaccurate or they may be biased. Let us consider first the evidence under the head of bias At the outset, however, I want to mention, and to set aside, one kind of aberration in futures prices which might be treated as bias namely, the effects of corners consequences of an excessive freedom of enterprise which seems largely a thing of the past in American futures markets. The British grain trade has never permitted either corners or significant squeezes in its futures markets. Squeezes continue to occur in American futures markets, though they can and should be eliminated. Like corners hey involve gaining and exploiting a temporary monopolistic position The monopolistic element produces effects of quite a different nature gn in equation (1) because we shall usually want to use that equation in interpreting statistical evidence use a positive sign, recognizing that r may take either a positive or a negative value. to r, if r is not equal to o, is to give P,<E. On purely algebraic grounds it would be bette content down ed stroe 2 2S R e os wecoldsep2
152 AMERICAN ECONOMIC ASSOCIATION united in the opinion that the relation of a futures price to expectations may be expressed by the equation, Pf =Er. ......... (1) where Pf is a futures price, E is some sort of combination of effects of individual expectations, and r is some sort of combination of (effects of?) individual risk premiums.4 This equation serves to indicate that in studying Pf we may learn facts about r rather than about E. That need not trouble us. If we find significant behavior characteristics, we shall know that they derive from corresponding characteristics in either E or r, or from some combination of both. We shall in fact find that the main characteristics derive from E expectations. We may now profitably adopt the term "market expectation" as synonymous with "futures price." Firmly established habits of thought make it difficult to think of any price as resting primarily on expectations. By using "market expectation" as an alternative to "futures price," we may keep attention focused on the predominantly expectational quality of such a price. Because the term "market expectation" has no strongly fixed connotations, it should not be difficult to remember that, writing EM for market expectation, its definition leads, by equation (1) to EM - Er, ........ (2) and therefore that market expectation is not precisely equivalent to E unless r- 0. III. Bias in Market Expectations Expectations may be defective in either or both of two respects: they may be randomly inaccurate or they may be biased. Let us consider first the evidence under the head of bias. At the outset, however, I want to mention, and to set aside, one kind of aberration in futures prices which might be treated as bias; namely, the effects of corners and squeezes. Corners are, or were, consequences of an excessive freedom of enterprise which seems largely a thing of the past in American futures markets. The British grain trade has never permitted either corners or significant squeezes in its futures markets. Squeezes continue to occur in American futures markets, though they can and should be eliminated. Like corners, they involve gaining and exploiting a temporary monopolistic position. The monopolistic element produces effects of quite a different nature 4I give r a negative sign in equation (1) because we shall usually want to use that equation in interpreting statistical evidence arising under conditions where the effect of r, if r is not equal to 0, is to give Pf < E. On purely algebraic grounds it would be better to use a positive sign, recognizing that r may take either a positive or a negative value. This content downloaded from 202.115.118.13 on Wed, 11 Sep 2013 03:21:32 AM All use subject to JSTOR Terms and Conditions
THEORY AND MEASUREMENT OF PRICE EXPECTATIONS 153 from the others we have to consider and i choose not to try to deal Bias itself may be of several sorts. I shall use a classification involv ng four categories of bias which will be defined as we proceed. 1. The kind of bias in market expectations which has received most attention in theoretical discussion and in statistical studies is what may be called"general bias. There has been a prevalent view in agricultural circles that futures prices of a crop tend to be depressed in the period of heavy marketing soon after harvest, and thereafter to advance, perhaps tending to be highest shortly before the next harvest. Theoretical analysis, moreover, has led to the opinion that some such behavior should be expected as a reflection of risk premium Such was the argument of J. M. Keynes's"theory of normal back wardation, and similar conclusions were reached by Kaldor and other British writers who extended and improved on Keynes's theoretical analysis. Keynes himself held that the statistics of organized markets show that 10 per cent per annum is a modest estimate of the amount f this backwardation in the case of seasonal crops..., That statement is significant mainly as evidence of Keynes's opinion that an indication of the effective risk premium is obtainable from statistical calculations of general bias in futures prices. His acquaintance with the pe evidently superficial, for it is only data for short and unrepresentative periods which appear to support high al bias or“ norma backwardation"" cannot be estimated properly at over 2 or 3 per cent per annum. For Chicago corn futures, it may approach 5 per cent. nd Economic Stability, "Review of Economic Studies, on the id, 1939-40, pp 196-205 and Gerda Blau, "Some T Holbrook Working, " Theory of the Inverse n bring out some curious minor characteristics of general bias. For his, however, is probably a spurious indication of general bias, arising from the fact that Wheat Futi e1885, : month when many hedges are being transferred from that future to more distant ones. This content down ed stroe 2 2S R e os wecoldsep2
THEORY AND MEASUREMENT OF PRICE EXPECTATIONS 153 from the others we have to consider, and I choose not to try to deal with them here. Bias itself may be of several sorts. I shall use a classification involving four categories of bias which will be defined as we proceed. 1. The kind of bias in market expectations which has received most attention in theoretical discussion and in statistical studies is what may be called "general bias." There has been a prevalent view in agricultural circles that futures prices of a crop tend to be depressed in the period of heavy marketing soon after harvest, and thereafter to advance, perhaps tending to be highest shortly before the next harvest. Theoretical analysis, moreover, has led to the opinion that some such behavior should be expected as a reflection of risk premium. Such was the argument of J. M. Keynes's "theory of normal backwardation," and similar conclusions were reached by Kaldor and other British writers who extended and improved on Keynes's theoretical analysis.5 Keynes himself held that "the statistics of organized markets show that 10 per cent per annum is a modest estimate of the amount of this backwardation in the case of seasonal crops..."2 That statement is significant mainly as evidence of Keynes's opinion that an indication of the effective risk premium is obtainable from statistical calculations of general bias in futures prices. His acquaintance with the pertinent statistical studies was evidently superficial, for it is only data for short and unrepresentative periods which appear to support so high an estimate of general bias. For Chicago wheat futures, the average general bias or "normal backwardation" cannot be estimated properly at over 2 or 3 per cent per annum. For Chicago corn futures, it may approach 5 per cent.7 'Nicholas Kaldor, "Speculation and Economic Stability," Review of Economic Studies, 1939-40, pp. 1-27; J. C. R. Dow, "A Theoretical Account of Futures Markets," ibid., 1939-40, pp. 185-195; Nicholas Kaldor, J. C. R. Dow, and R. G. Hawtrey, "A Symposium on the Theory of the Futures Market," ibid., 1939-40, pp. 196-205; and Gerda Blau, "Some Aspects of the Theory of the Futures Market," ibid., 1944-45, pp. 1-30. ' Treatise on Money (New York, 1930), p. 143. 'Holbrook Working, "Theory of the Inverse Carrying Charge in Futures Markets," Journal of Farm Economics, February, 1948, pp. 8-12. (The citation in footnote 18 there should read "p. 214.") Detailed inquiry can bring out some curious minor characteristics of general bias. For example, the reference above examines an apparent general tendency for corn futures prices to rise slightly from May to August and to decline similarly from August to October. This, however, is probably a spurious indication of general bias, arising from the fact that exaggerative bias (to be considered below) operates mainly in the upward direction and occurs most often in July and August. Chicago wheat futures prices show an apparently true general tendency to decline slightly from the latter part of February to late March. ("Price Relations between May and New-Crop Wheat Futures at Chicago since 1885," Wheat Studies of the Food Research Institute, February, 1934, pp. 214-216.) There is also a quite prevalent tendency for the price of any future to rise slightly (perhaps about 1 per cent) during two or three weeks before the beginning of a delivery month, when many hedges are being transferred from that future to more distant ones. This This content downloaded from 202.115.118.13 on Wed, 11 Sep 2013 03:21:32 AM All use subject to JSTOR Terms and Conditions
154 The small size of general bias indicated by a proper interpretation of the statistics raises an interesting theoretical question: Is the effective risk premium really so small; or is there actually a positive bias in expectations proper (the E of the equation Em=E-r) which partially offsets the effects of risk premium? A good argument can be made for either interpretation. On the one hand, it is clear that speculators tend to be people who are not particularly averse to taking risks; many of them, like many people who enjoy gambling may willingly accept even positive risk premiums On the other hand prevalence of the mistaken impression that futures prices tend to be strongly depressed in the period shortly after harvest undoubtedly produces a real positive bias in the price expectations of many indi- idual speculators. Their bias could form the basis for a small positive bias in the consensus of expectations, E In any case, it is clear that any errors in market expectations which arise from general bias must be exceedingly small in comparison with errors from other sources 2. We now turn to the sorts of bias which may result in large errors of market expectation. One of these may be called"conservative bias, "and defined as a tendency for market expectations to delay in adjusting fully to new substantive information. when I was studying potato prices a good many years ago, i thought that I perceived such a bias, involving a tendency for prices to be too low early in the season following a short harvest and too high in the early part of a season following a large harvest, these errors being corrected more or less gradually as the season progressed. Those observations were based, of course, on spot prices of a commodity in which there was then no futures trading So far as i know there has been no demonstration that such a tendency to conservative bias exists in prices of any commodity which has a prominent futures market, with one rather obscure exception That exception is in wheat and is associated with what i called long cycle. 8 Market expectations in wheat seem to have tended tendency is chiefly noteworthy as a of mistaken conclusions; it permits a combina on of data for different futures in such a way as to give a much exaggerated applying the term ycles in Wheat Prices, "Wheat Studies, November, 1931 People with a good deal of mathematical of cycles, I think. Economists have commonly used the term in the broad sense in which n following a disturbance tick by the monheratically mindes s fluctuations swill not terre because it is too broad, content down ed stroe 2 2S R e os wecoldsep2
154 AMERICAN ECONOMIC ASSOCIATION The small size of general bias indicated by a proper interpretation of the statistics raises an interesting theoretical question: Is the effective risk premium really so small; or is there actually a positive bias in expectations proper (the E of the equation Em -E - r) which partially offsets the effects of risk premium? A good argument can be made for either interpretation. On the one hand, it is clear that speculators tend to be people who are not particularly averse to taking risks; many of them, like many people who enjoy gambling, may willingly accept even piositive risk premiums. On the other hand, prevalence of the mistaken impression that futures prices tend to be strongly depressed in the period shortly after harvest undoubtedly produces a real positive bias in the price expectations of many individual speculators. Their bias could form the basis for a small positive bias in the consensus of expectations, E. In any case, it is clear that any errors in market expectations which arise from general bias must be exceedingly small in comparison with errors from other sources. 2. We now turn to the sorts of bias which may result in large errors of market expectation. One of these may be called "conservative bias," and defined as a tendency for market expectations to delay in adjusting fully to new substantive information. When I was studying potato prices a good many years ago, I thought that I perceived such a bias, involving a tendency for prices to be too low early in the season following a short harvest and too high in the early part of a season following a large harvest, these errors being corrected more or less gradually as the season progressed. Those observations were based, of course, on spot prices of a commodity in which there was then no futures trading. So far as I know, there has been no demonstration that such a tendency to conservative bias exists in prices of any commodity which has a prominent futures market, with one rather obscure exception. That exception is in wheat and is associated with what I called a "long cycle."8 Market expectations in wheat seem to have tended tendency is chiefly noteworthy as a source of mistaken conclusions; it permits a combination of data for different futures in such a way as to give a much exaggerated indication of general bias over a period of a year. 8 In applying the term (in "Cycles in Wheat Prices," Wheat Studies, November, 1931, pp. 1-66), I intended no implication of regularity of repetition, or periodicity. People with a good deal of mathematical training accustomed to associate "cycles" with the perfect regularity of a sine curve, have often misinterpreted economists' discussions of cycles, I think. Economists have commonly used the term in the broad sense in which it may designate only the effects of a tendency toward oscillation following a disturbance, including oscillation so highly damped that one disturbance produces only a single perceptible "wave." It is hard to find a good alternative term to avoid possible misinterpretation by the mathematically-minded. "Fluctuations" will not serve because it is too broad, lacking the connotation of progress from one phase to another. This content downloaded from 202.115.118.13 on Wed, 11 Sep 2013 03:21:32 AM All use subject to JSTOR Terms and Conditions
HEORY AND MEASUREMENT OF PRICE EXPECTATIONS 155 usually to respond very tardily to emergence of a large world wheat surplus, and sometimes also to respond tardily to shortage of supplies when low stocks at the beginning of a crop year have aggravated the seriousness of a poor crop On this interpretation, the "long cycle in wheat prices is, or was a reflection of conservatism in response to the stocks position. Possibly heat market expectations will behave differently in the future, with the benefit of better information on actual wheat stocks. If so . the phenomena of the long cycle cannot continue to be taken as evidence f true conservatism of market expectations. They will have to be egarded as effects of bias related to inadequacy of information, to which we now turn 3. Bias in market expectations arising from inadequacy of informa tion is probably the main source of reliable profits for speculators. Many traders in futures markets give a great part of their attention to acquiring information which has not become generally available and thus has not been reflected in market expectations. These traders, or the commission houses which serve them, gather crop information ahead of its reporting by public agencies; they study the weather eports and seek to predict effects of the weather on the crops, thus trying actually to base expectations on crop developments which have not yet occurred; and they have even employed a long-range weather forecaster to predict the weather several weeks ahead The tendency in futures markets, therefore, is for most classes of pertinent information to be brought to bear on market expectations ith extraordinary promptness. Errors arising from failure to obtain or to use available information of known pertinence are small and very short-lived. There are, nevertheless, certain recognizable tend encies toward bias arising from failure to give due attention to available information. The bias associated with level of wheat stocks entioned above, may be an example. a clearer example again from the wheat market, appears in the tendency for market expectations in the United States to give too little attenion to much news from overseas Critical studies of price behavior in other futures markets are needed before confident interpretation can be the tendency of American wheat markets to ignore pertinent fe ews. Possibly there is a general tendency for any market to ignore events that are distant; but this conclusion should not be drawn too hastily on superficial evidence. There is much evidence to the contrary. It may Perhaps the best evidence on this point is in Robert D, Calkins, "Price Leadership mong Major Wheat Futures Markets, " Wheat Studies November, 1933 pp.35-70 content down ed stroe 2 2S R e os wecoldsep2
THEORY AND MEASUREMENT OF PRICE EXPECTATIONS 155 usually to respond very tardily to emergence of a large world wheat surplus, and sometimes also to respond tardily to shortage of supplies when low stocks at the beginning of a crop year have aggravated the seriousness of a poor crop. On this interpretation, the "long cycle" in wheat prices is, or was, a reflection of conservatism in response to the stocks position. Possibly wheat market expectations will behave differently in the future, with the benefit of better information on actual wheat stocks. If so, the phenomena of the long cycle cannot continue to be taken as evidence of true conservatism of market expectations. They will have to be regarded as effects of bias related to inadequacy of information, to which we now turn. 3. Bias in market expectations arising from inadequacy of information is probably the main source of reliable profits for speculators. Many traders in futures markets give a great part of their attention to acquiring information which has not become generally available, and thus has not been reflected in market expectations. These traders, or the commission houses which serve them, gather crop information ahead of its reporting by public agencies; they study the weather reports and seek to predict effects of the weather on the crops, thus trying actually to base expectations on crop developments which have not yet occurred; and they have even employed a long-range weather forecaster to predict the weather several weeks ahead. The tendency in futures markets, therefore, is for most classes of pertinent information to be brought to bear on market expectations with extraordinary promptness. Errors arising from failure to obtain or to use available information of known pertinence are small and very short-lived. There are, nevertheless, certain recognizable tendencies toward bias arising from failure to give due attention to available information. The bias associated with level of wheat stocks, mentioned above, may be an example. A clearer example, again from the wheat market, appears in the tendency for market expectations in the United States to give too little attenion to much news from overseas.9 Critical studies of price behavior in other futures markets are needed before confident interpretation can be given the tendency of American wheat markets to ignore pertinent foreign news. Possibly there is a general tendency for any market to ignore events that are distant; but this conclusion should not be drawn too hastily on superficial evidence. There is much evidence to the contrary. It may 9Perhaps the best evidence on this point is in Robert D. Calkins, "Price Leadership and Interaction among Major Wheat Futures Markets," Wheat Studies, November, 1933, pp. 35-70. This content downloaded from 202.115.118.13 on Wed, 11 Sep 2013 03:21:32 AM All use subject to JSTOR Terms and Conditions
156 AMERICAN ECONOMIC ASSOCIATION be that the wheat market is a somewhat exceptional case and that its tendency to underestimate the importance of news from overseas arises largely from misapprehension, fostered by the fact that such news is usually unimportant for other grain markets in the United States, and persisting because of inadequacy of basic economic studies and education regarding the major price factors in wheat. When painstaking statistical research uncovers bias in market expectations the results may deserve to be taken mainly as evidence that the research was needed to lay a basis for better functioning of the markets 4. The main objectionable bias commonly charged against specula tive markets is exaggerative biasa supposed tendency for market expectations to respond excessively to day-to-day news and rumors and to generate unwarranted price fluctuations. As with so much of economic thought today, the current prevalence of this idea must be attributed in considerable part to the gifts for vigorous and picturesque expression of John Maynard Keynes. His statement, applied specifi- cally to the stock market, that "we have reached the third degree where we devote our intelligences to anticipating what average opinion expects average opinion to be, 0 is a gem. It takes no great amount of conversation with traders, when one can get on a basis of frank expression of ideas to learn that many of them are mainly interested in anticipating the swings of opinion of others. This, however, is not necessarily objectionable; it may repre- sent merely an extension over somewhat longer time intervals of the clearly useful function of the professional"scalper, whose activities certainly tend to diminish price fluctuation and make a better market To pass judgment, we need to know the effects on price behavior One consequence to be expected from an excess of activity toward anticipating what average opinion expects average opinion to be' is that price movements should develop certain recogniz ible patterns Apparent confirmation that this occurs is readily found among A good many of them are firm believers in the significance of and- shoulder formations;”“ resistance levels,”’ and the like. accepting their opinions as valid evidence however, one may wish to reflect on the fact that these traders rarely wear an air of prosp and that their ideas are generally scorned by more substantial ticipators in the ma ket General Theory of Employment, Interest, and Money(New York, 1936), p 156 Professor Norton took his stand with those who think that the activities of scalpers hay he need for objective determination of price effects of speculative trading, including scalp- petent to settle. a fairly clear objective answer should be furnished by the " error-time test"proposed near the end of this paper. content down ed stroe 2 2S R e os wecoldsep2
156 AMERICAN ECONOMIC ASSOCIATION be that the wheat market is a somewhat exceptional case, and that its tendency to underestimate the importance of news from overseas arises largely from misapprehension, fostered by the fact that such news is usually unimportant for other grain markets in the United States, and persisting because of inadequacy of basic economic studies and education regarding the major price factors in wheat. When painstaking statistical research uncovers bias in market expectations, the results may deserve to be taken mainly as evidence that the research was needed to lay a basis for better functioning of the markets. 4. The main objectionable bias commonly charged against speculative markets is exaggerative bias-a supposed tendency for market expectations to respond excessively to day-to-day news and rumors and to generate unwarranted price fluctuations. As with so much of economic thought today, the current prevalence of this idea must be attributed in considerable part to the gifts for vigorous and picturesque expression of John Maynard Keynes. His statement, applied specifically to the stock market, that "we have reached the third degree where we devote our intelligences to anticipating what average opinion expects average opinion to be,"'" is a gem. It takes no great amount of conversation with traders, when one can get on a basis of frank expression of ideas, to learn that many of them are mainly interested in anticipating the swings of opinion of others. This, however, is not necessarily objectionable; it may represent merely an extension over somewhat longer time intervals of the clearly useful function of the professional "scalper," whose activities certainly tend to diminish price fluctuation and make a better market." To pass judgment, we need to know the effects on price behavior. One consequence to be expected from an excess of activity toward "anticipating what average opinion expects average opinion to be" is that price movements should develop certain recognizable patterns. Apparent confirmation that this occurs is readily found among traders. A good many of them are firm believers in the significance of "headand-shoulder formations," "resistance levels," and the like. Before accepting their opinions as valid evidence, however, one may wish to reflect on the fact that these traders rarely wear an air of prosperity and that their ideas are generally scorned by more substantial participators in the markets. 0 General Theory of Employment, Interest, and Money (New York, 1936), p. 156. " It came to me as a surprise, in discussion following presentation of this paper, that Professor Norton took his stand with those who think that the activities of scalpers have a bad rather than a good effect on price behavior. I welcome his remarks as evidence of the need for objective determination of price effects of speculative trading, including scalping. When opinions of competent observers differ so much as they do regarding the effects of scalping, it becomes clear that the issue is one which "informed judgment" is incompetent to settle. A fairly clear objective answer should be furnished by the "error-time test" proposed near the end of this paper. This content downloaded from 202.115.118.13 on Wed, 11 Sep 2013 03:21:32 AM All use subject to JSTOR Terms and Conditions
THEORY AND MEASUREMENT OF PRICE EXPECTATIONS 157 If we turn to a study of actual price behavior for evidence of unwarranted fluctuations, we easily find many cases in which prices fell quickly after a rise or reacted promptly after a decline. This evidence of unwarranted fluctuations must also be regarded with suspicion. It is valid evidence only if accompanied by some proof that he fluctuations were not in fact warranted. Such proof has seemed ry hard to come by at least I know of very little published evidence on the subject which bears up under critical examination This inadequate sketch must serve for the present introduction of a large and difficult problem. i pass now to some specific evidence of exaggerative bias in market expectations. The best possible evidence of unwarranted price fluctuation in a futures market is that frequently occurring price movements as sociated th simple objectively specified conditions have invariably been followed promptly by reverse movements. 2 For Chicago wheat futures prices, a specification involving only magnitude and rate of price hange, with two theoretically reasonable qualifications regarding time of occurrence, serves to identify a class of price movements which ave regularly been followed by approximately equal price reaction. There are four clear characteristics 4 of the evidence on exaggerative bias in the United States wheat market which have special theoretical interest:(1) Evidence of exaggerative bias has been found only in cor lection with upward price influences, not with downward influences; (2) exaggerative bias accounts in large part for most of the con spicuous advances of United States wheat prices during a period of over sixty years;(3 )most of the very largest price advances show no evidence of the presence of exaggerative bias;(4)the patterns of price reaction conformity with the degree of confidence or uncertainty necessar, "o after emergence of exaggerative bias have varied systematically in clo attending the formation of expectations at the time On the important question of explanation of the occurrence of exaggerative bias in the wheat market, the facts are obscure. The ready inference that it is a general characteristic of prices in speculative markets and the more limited inference that it arises from the par ch iA fact which will carry special weight with statisticians is that criteria of scovered as in application to the historical d. the confirmation included ls, which must be omitted here may be in Holbrook Working, "Cycles in Wheat Prices, "Wheat Studies, November, 1931, pp. 18-24 content down ed stroe 2 2S R e os wecoldsep2
THEORY AND MEASUREMENT OF PRICE EXPECTATIONS 157 If we turn to a study of actual price behavior for evidence of unwarranted fluctuations, we easily find many cases in which prices fell quickly after a rise or reacted promptly after a decline. This evidence of unwarranted fluctuations must also be regarded with suspicion. It is valid evidence only if accompanied by some proof that the fluctuations were not in fact warranted. Such proof has seemed very hard to come by; at least I know of very little published evidence on the subject which bears up under critical examination. This inadequate sketch must serve for the present introduction of a large and difficult problem. I pass now to some specific evidence of exaggerative bias in market expectations. The best possible evidence of unwarranted price fluctuation in a futures market is that frequently occurring price movements associated with simple objectively specified conditions have invariably been followed promptly by reverse movements.'2 For Chicago wheat futures prices, a specification involving only magnitude and rate of price change, with two theoretically reasonable qualifications regarding time of occurrence, serves to identify a class of price! movements which have regularly been followed by approximately equal price reaction.'3 There are four clear characteristics'4 of the evidence on exaggerative bias in the United States wheat market which have special theoretical interest: (1) Evidence of exaggerative bias has been found only in connection with upward price influences, not with downward influences; (2) exaggerative bias accounts in large part for most of the conspicuous advances of United States wheat prices during a period of over sixty years; (3 ) most of the very largest price advances show no evidence of the presence of exaggerative bias; (4) the patterns of price reaction after emergence of exaggerative bias have varied systematically in close conformity with the degree of confidence or uncertainty necessarily attending the formation of expectations at the time. On the important question of explanation of the occurrence of exaggerative bias in the wheat market, the facts are obscure. The ready inference that it is a general characteristic of prices in speculative markets and the more limited inference that it arises from the par- 12The qualifications of frequency of occurrence and simplicity of the specifications are necessary to exclude significant possibility that the observed association is mere coincidence. Better evidence would be furnished by direct proof that the initial price movement in each instance was not accompanied by conditions which warranted it, but in view of the variety and complexity of legitimate price influences, such proof is rarely possible, at least in the present state of our knowledge. 13 A fact which will carry special weight with statisticians is that the criteria of identification have worked as well in current use during the eighteen years since they were discovered as in application to the historical record. The confirmation has included evidence of validity of the qualifications regarding timing. 4Details, which must be omitted here, may be found in Holbrook Working, "Cycles in Wheat Prices," Wheat Studies, November, 1931, pp. 18-24. This content downloaded from 202.115.118.13 on Wed, 11 Sep 2013 03:21:32 AM All use subject to JSTOR Terms and Conditions
AMERICAN ECONOMIC ASSOCIATION ticular class of public participation which occurs in the wheat market must apparently be rejected. I have looked for similar evidence of exaggerative bias in corn and oats prices and have failed to find it. s The best hypothesis which I can offer to account for the main facts now known is that exaggerative bias in wheat prices is related to another bias ously discussed: that it occurs largely b significant overseas news tends to be ignored or inadequately weighted in the American wheat market Pertinent in this connection are the acts that exaggerative bias in wheat market expectations has occurred most commonly in connection with crop scares, and that the crop damage involved has always been in North America and almost always in the United States There is one feature common to all the kinds of bias that we hay onsidered, except the first, which deserves to be noticed before we pass to the next section. General bias, being so small, probably should be regarded as evidence of a risk premium rather than of any true tendency to error in market expectations. The other kinds of bia all reflect errors in market expectation, even though bias arising from inadequacy of pertinent information may warrant no criticism of the market itself. These kinds of errors of expectation, moreover, all have the characteristic that a given error persists over a period of time and tends to diminish progressively after reaching its maximum This characteristic of the errors of bias is very clearly observed them, n exaggerative errors of crop-scare cycles in wheat prices. In them, prices rise rapidly and excessively and then decline more or less gradually. In a case of conservative bias, the price movement is quite different from that in a case of exaggerative bias but if one imagines the course which the price would have taken in the absence of bias it becomes apparent that the error due to bias must usually develop rapidly and then progressively diminish. Bias owing to inadequacy of information is like conservative bias in all respects except cause. IV. Accuracy of Market Expectations Concentration of attention on bias in market expectations involves emphasis on the imperfections of expectation without due considera- tion of their relative magnitude. Excessive attention to bias, if I may ay so, risks creating a biased impression. We should seek, therefore, to get a balanced view of the inaccuracies of market expectations The most perfect expectations possible in economic affairs must be subject to substantial error because the outcome depends on ur It should be recalled, nevertheless, that in earlier mention of a peculiarity in evidence of apparent "ge content down ed stroe 2 2S R e os wecoldsep2
158 AMERICAN ECONOMIC ASSOCIATION ticular class of public participation which occurs in the wheat market, must apparently be rejected. I have looked for similar evidence of exaggerative bias in corn and oats prices and have failed to find it."5 The best hypothesis which I can offer to account for the main facts now known is that exaggerative bias in wheat prices is related to another bias previously discussed: that it occurs largely because significant overseas news tends to be ignored or inadequately weighted in the American wheat market. Pertinent in this connection are the facts that exaggerative bias in wheat market expectations has occurred most commonly in connection with crop scares, and that the crop damage involved has always been in North America and almost always in the United States. There is one feature common to all the kinds of bias that we have considered, except the first, which deserves to be noticed before we pass to the next section. General bias, being so small, probably should be regarded as evidence of a risk premium rather than of any true tendency to error in market expectations. The other kinds of bias all reflect errors in market expectation, even though bias arising from inadequacy of pertinent information may warrant no criticism of the market itself. These kinds of errors of expectation, moreover, all have the characteristic that a given error persists over a period of time and tends to diminish progressively after reaching its maximum. This characteristic of the errors of bias is very clearly observed in the exaggerative errors of crop-scare cycles in wheat prices. In them, prices rise rapidly and excessively and then decline more or less gradually. In a case of conservative bias, the price movement is quite different from that in a case of exaggerative bias, but if one imagines the course which the price would have taken in the absence of bias, it becomes apparent that the error due to bias must usually develop rapidly and then progressively diminish. Bias owing to inadequacy of information is like conservative bias in all respects except cause. IV. Accuracy of Market Expectations Concentration of attention on bias in market expectations involves emphasis on the imperfections of expectation without due consideration of their relative magnitude. Excessive attention to bias, if I may say so, risks creating a biased impression. We should seek, therefore, to get a balanced view of the inaccuracies of market expectations. The most perfect expectations possible in economic affairs must be subject to substantial error because the outcome depends on un- 1 It should be recalled, nevertheless, that in earlier mention of a peculiarity in evidence of apparent "general bias" in corn prices, I inclined to attribute it to existence of some exaggerative bias. This content downloaded from 202.115.118.13 on Wed, 11 Sep 2013 03:21:32 AM All use subject to JSTOR Terms and Conditions