Review on three Pricing journals

New Facts on the Marketing Performance of U.S. Farmers (Joy L. Harwood, USDA-Economic Research Service, Presiding)

IMPLICATIONS OF COMMODITY PRICE BEHAVIOR FOR MARKETING STRATEGIES

WILLIAM G. TOMEK AND HIKARU HANAWA PETERSON

Prices of agricultural commodities have sys- tematic, dynamic components, and many agricultural economists have thought that forecasts of this behavior ought to aid mar- keting decisions, thereby raising producers’ re- turns. But, if commodity markets are pricing efficient, then noncompetitive profit opportu- nities should be quickly arbitraged away. Thus, a debate exists about the magnitude of the likely benefits of various marketing strategies. Herein we address the question, what does the literature about commodity price behav- ior contribute to our understanding of the ex- pected outcomes from alternative marketing strategies?

Our brief answer is as follows. Spot prices in U.S. commodity markets typically behave in a way that is consistent with efficient mar- kets; systematic behavior persists because of the costs of arbitrage. Hence, price forecasts in the public domain cannot be used to produce returns which are consistently above the com- petitive norm, although low-cost strategies to manage price risk exist. Evaluating the evi- dence on these topics is not easy, however, because the price generating processes are complex. Sometimes prices seem to behave as if a market is inefficient. Advice based on anal- yses using short samples should be used with great caution.

The remainder of the article elaborates: the first section reviews conceptual models of price behavior, while the next section covers

William G. Tomek is professor emeritus at the Department of Applied Economics and Management, Cornell University. Hikaru Hanawa Peterson is assistant professor at the Department of Agri- cultural Economics at Kansas State University.

The authors acknowledge Wade Brorsen’s helpful suggestions. The usual caveats apply.

This article was presented in a principal paper session at the AAEA annual meeting (Providence, Rhode Island, July 2005). The articles in these sessions are not subjected to the journal’s standard refereeing process.

empirical analyses that link price behavior to marketing strategies. Then, the performance of selected marketing programs is compared, using simulated prices from a model of a grain market. The discussion abstracts from differ- ences in farmers’ risk tolerances, opportunity costs, tax management strategies, and other idiosyncratic factors, which influence farmers’ marketing decisions. Also, individual farmers can own grain with attributes that command a premium over a representative price series, such as number two yellow corn, while this ar- ticle is necessarily about such prices.

Models of Price Behavior

Commodity price behavior over time is a mix- ture of systematic and random factors. Spot prices exhibit autocorrelation with occasional spikes. For the grains, the variance and skew- ness of (say) monthly probability distributions tend to increase as the marketing season pro- gresses and inventories decline. This behavior is more or less accommodated by concep- tual models in the literature (for a synthesis relevant to grain markets, see Williams and Wright), and as noted in the next section, many empirical models of this behavior exist.

Peterson and Tomek (2005a) provide a model, in which decision makers have ra- tional expectations, that mimics the behav- ior of monthly corn prices. Storage is carried out by rational arbitragers whose expected profit depends on the expected appreciation in price and storage costs. Prices are autocor- related, consistent with the incentives needed to carry inventories from month to month and from one crop year to the next. The model al- lows prices to decline as the new harvest ap- proaches, even though inventories are positive, by including convenience yield.

Amer. J. Agr. Econ. 87 (Number 5, 2005): 1258–1264 Copyright 2005 American Agricultural Economics Association

Tomek and Peterson Price Behavior for Marketing Strategies 1259

The random components of prices are as- sociated with the arrival of new information about current and expected supply and de- mand conditions, but the monthly probability distributions have systematically changing mo- ments, associated with the seasonality of inven- tory changes and of information about growing conditions. An upward spike in prices can occur given a combination of changes in expec- tations related to a larger demand, smaller cur- rent inventory, and/or decline in the planned crop. In this case, the resulting (unexpected) price appreciation is far larger than the cost of carry, but this is the outcome of a combination of unforeseeable events and not of a market imperfection.

The price of a futures contract at current time t is the expected value—in essence a forecast—of the cash price at contract matu- rity, time T, conditional on the information available at t.1 Since the conditioning informa- tion can change by large amounts with the pas- sage of time, the time t price can change rad- ically as maturity approaches; futures prices can be unbiased, but imprecise forecasts. The variance of futures price changes is not a con- stant; rather the price changes have time-to- maturity and seasonal components (e.g., An- derson; Fackler and Tian), and like spot prices, they are likely not normally distributed.

Prices of old- and new-crop futures are usu- ally correlated because inventories link the crop-years, but when inventories are tiny, a disconnection can occur (Tomek and Gray). High current prices reflect small supplies rel- ative to demand. Low prices for future de- livery of the new crop reflect large expected production relative demand. It is not possible in this situation for producers to obtain high prices for the forthcoming crop by selling old- crop futures and rolling over these positions into new-crop contracts (Lence and Hayenga). Again, these varied price relationships in dif- ferent years need not be the consequence of market imperfections.

At harvest, the prices of a constellation of fu- tures contracts, for different delivery months, can be compared with the current cash price. The differences are prices of storage, which can be assured by hedging, subject to basis risk.

1 Futures prices, in contrast to cash prices, can follow a martin- gale process; the expected price change is zero in an efficient fu- tures market. (Risk premiums in grain markets appear to be zero or tiny.) Testing for this property is tricky, however. A price series may be constructed inappropriately, if the analyst uses only the nearby contract prices, which are essentially the spot prices. The martingale property pertains to prices for individual contracts.

Prices of storage vary with the passage of time, depending on changes in the supply and de- mand for storage. Regional prices also differ and adjust to allocate inventories over space (Benirschka and Binkley). The fact that a mar- ket is providing incentives to store in some lo- cations, but not in others, is consistent with the operation of a competitive market. The returns to storage can be less than the costs of storage, in most locations, in an efficient market.

Analogous models exist for animal agricul- ture (e.g., Mundlak and Huang; Rosen). Prices are autocorrelated because of the dynamics inherent in managing herds and flocks. At a point in time, the economic problem is to allo- cate the existing stock of animals between cur- rent consumption and future use. The rate of culling from the breeding herd can be varied as can the allocation of young female animals be- tween the breeding herd and slaughter. These decisions depend on current prices relative to (discounted) expected prices in the future, tak- ing account of the relevant costs of maintaining the herd, analogous to a cost of carry. Given the price risk associated with arbitraging over long-time periods, it is not surprising that cy- cles in hog and cattle prices cannot be arbi- traged away. An added complexity is that the costs of arbitrage themselves vary with the pas- sage of time.

Clearly, a variety of price behaviors are con- sistent with pricing efficiency. Of course, some spot and futures markets may be inefficient. At a minimum, a few traders can have superior private information and can earn a return on the investment in this information. Such mar- kets are semistrong, but not strong, form effi- cient (Fama). In addition, the information flow among traders may be slow (Grossman and Stiglitz); a high proportion of traders may be uninformed about new information; and con- cepts from behavioral finance imply that prices may have “psychological runs” (for a review, see Park and Irwin). Perhaps profit opportuni- ties are not always quickly arbitraged away.

In sum, spot prices for agricultural com- modities have systematic components,2 but in an efficient market price forecasts cannot be used to produce returns that consistently ex- ceed arbitrage costs. Prices in efficient futures markets follow a martingale process and can- not be forecast (footnote 1); rather, they are a

2 An implication is that spot prices for agricultural commodities should not follow a random walk. Research that claims to have found nominal price series that are integrated of order one is likely mistaken (Wang and Tomek).

1260 Number 5, 2005 Amer. J. Agr. Econ.

type of forecast. But, markets may not be fully efficient. Nonetheless, if they are semistrong form efficient, arbitrage opportunities will be short-lived.

Empirical Evidence

Econometric analyses of agricultural prices attempt to capture the systematic, dynamic be- havior of cash prices (Tomek), but few pub- lished studies relate forecasts to marketing decisions. Not surprisingly, private consultants do not publish the precise foundation for their advice. Extension economists do provide fore- casts and marketing advice through newslet- ters and websites, but the evaluation of their forecasts is not easy (see below). It is even more difficult to evaluate whether farmers have benefited from using the advice (Brorsen and Irwin).

Both point and interval forecasts have not been particularly precise (Allen; Isengildina, Irwin, and Good), and the risk associated with an imprecise forecast is a component of the cost of using it. An imprecise forecast may be the result of using conditioning informa- tion that subsequently turns out to be wrong rather than the result of a poor model and/or sampling error. Further, a precise forecast in one time period does not necessarily imply a precise forecast in a subsequent period. Fore- casts from econometric models apparently cannot outperform those implied by futures market quotes (Just and Rausser), and techni- cal analyses that seem profitable in one period probably are not in subsequent periods (Park and Irwin).

While market imperfections might persist that can be forecast and provide superior arbi- trage profit opportunities, the evidence to this effect is small. An important exception is the analysis of futures prices for corn and soybeans by Wisner and colleagues (e.g., Wisner, Blue, and Baldwin). They argue that the prices of soybean and corn futures contracts for harvest- time delivery, observed the prior winter and spring, are biased upward. This is particularly true in a year when the prior harvest is rela- tively small and current prices are high. Thus, “. . . hedge positions were placed routinely in the fourth week in February [in December corn] after short-crop years . . . ” (p. 294) on the expectation that the price would decline. The “forecast” is qualitative, based on historical av- erage performance of prices from a twenty-two year sample, but it is linked to specific advice.

Whether or not this advice will continue to be profitable is uncertain (Zulauf and Irwin); the rule would have resulted in losses in 2003, but gains in 2004.

Other studies of pricing inefficiency exist in the futures markets literature, and they can be linked to speculative strategies. Tests of effi- ciency have problems, and results are mixed. For example, in applying a particular trad- ing system to a 1986–93 sample for different futures contracts, Olszewski found that “[the system] generated the best profits for cotton, corn, and Japanese yen” (p. 701). But, based on additional analysis for corn, Olszewski con- cluded that “no statistical evidence [existed] that the trading system generates profits other than by chance” (p. 702). Schaefer, Myers, and Koontz found that the live cattle futures market was not strong form efficient, but in concluding they write, “ . . . the predictability of price changes dies out quickly . . . , and we have not demonstrated that this predictabil- ity can be used to generate profitable trading rules that take transactions costs into account” (p. 450). Zulauf et al. obtain diverse results for soybean prices, which are influenced by the model specification; a price-level model found significant bias, a percentage price change model did not.3 We are predisposed to con- clude that the seeming inefficiencies found in empirical analyses are likely the result of id- iosyncrasies of the sample and model spec- ifications and hence that futures markets in the United States are likely semistrong form efficient.

Although the underlying foundation for marketing advice provided by private consul- tants is not available, the AgMAS project at the University of Illinois has evaluated the out- comes of the advice given by many consultants. One conclusion is that it is difficult for any advi- sor to provide above-average returns on a con- sistent year-to-year basis; good performance in one year does not predict good performance in the next year. This is an expected outcome if markets are pricing efficient.

A similar evaluation of marketing programs developed by extension economists would be

3 Park and Irwin provide a survey of the profitability of techni- cal analyses for many different asset prices, although relatively few applications relate to agricultural commodities. Results reviewed by Park and Irwin are mixed: technical rules sometimes appear profitable, but it is usually unclear whether the estimated returns exceed transaction costs. Typically the rules did not remain prof- itable beyond the sample period. Also, if markets are not strong form efficient—a few traders have superior, private information— such results by definition are not in published studies.

Tomek and Peterson Price Behavior for Marketing Strategies 1261

interesting, inter alia, because some of them suggest that technical analysis is a useful de- cision tool. Moreover, websites for these pro- grams contain quotes from producers stating that the marketing program has increased their returns. But, given the seeming inconsistency of technical analysis with price theory, one wonders whether the above-average results are sustainable.

Forecasts, however, have the potential to help decision makers earn a competitive profit in an efficient market. An example is to use a forecast of basis convergence over a stor- age interval to make a decision about storing and hedging (Working). At harvest, the cur- rent price of a distant futures contract can be compared with the cash price; this basis should narrow as contract maturity approaches; and the magnitude of convergence can be forecast from relatively simple models. The standard error of forecast is an estimate of basis risk. This information can be an input into a deci- sion to store and hedge, or not. A few such models exist in the literature (e.g., Heifner; Lence, Hayenga, and Patterson; Hranaiova and Tomek), but we are unaware of economic evaluations of them.

An extensive line of research provides esti- mates of optimal marketing portfolios, based on the relationships among prices in cash, futures, and options markets (Tomek and Peterson). The optimum is usually defined as minimizing risk, where risk is measured by the variance of returns, and empirical estimates find optimal portfolios that shift risk, at low cost, though they do not increase average re- turns. But, most farmers’ marketing decisions differ from the estimated optima (Harwood et al.). This is likely because the objective func- tions used in the analyses are inappropriate specifications relative to individual farmer’s situations. The functions often exclude realis- tic potential alternatives in the portfolio and their costs, and the assumption of variance minimization is likely inappropriate (Collins). Yet, if objective function specifications could be made more realistic, this research could be useful to decision makers.

A somewhat related line of research has used short samples (perhaps ten to twenty years of prices) to compute returns from differ- ent marketing strategies with the objective of finding the alternative that gave the largest av- erage return. The “optimal” alternative is then recommended to farmers as the best choice to be used in future (out-of-sample) years. Such an analysis is analogous to data mining, which

in our view, is a dangerous basis for mak- ing marketing recommendations. The results in the next section help illustrate the danger and more generally the problems of inter- preting the results from alternative marketing strategies.

Simulating Marketing Strategies in an Efficient Corn Market

As implied above, the academic literature on pricing efficiency often conflicts with the mar- keting advice given by private consultants and extension economists. These differences are difficult to reconcile using relatively short sam- ples. One approach to a better understanding, if not a reconciliation, of the different views is to analyze large samples obtained by simulat- ing market prices.

Thus, to compare marketing strategies, monthly prices of corn were simulated to rep- resent central Illinois (Peterson and Tomek 2005a). The model assumes an efficient mar- ket, and the monthly probability distributions of the simulated prices have changing mode, variance, and skewness similar to the distribu- tions of cash prices in the 1990s. Prices of May and December futures contracts are simulated under the assumption that they are the con- ditional expected values of the spot prices at maturity. Prices are generated for forty-year “lifetimes” to illustrate the varying behavior that could be faced in different lifetimes. The long run, or expected, outcomes are based on 10,000 forty-year periods.

Alternative Strategies

The base strategy is marketing the entire crop at harvest. This is a common base for compar- ing alternatives. The individual farmer is as- sumed to be a price taker, who controls the quantities marketed per month.

In this context, strategy 1a diversifies by sell- ing the corn crop in equal amounts in the cash market over nine months (see Peterson and Tomek 2005b). Strategy 1b diversifies by pre- selling 40% of the crop in December futures and the remainder in the cash market after harvest.

Strategy 2a routinely sells the entire ex- pected crop in May, using the December futures contract, completing the hedge in December when the entire crop is marketed. An alternative, 2b, is a conditional hedge, which is intended to approximate the advice of Wisner, Blue, and Baldwin. The crop is presold

1262 Number 5, 2005 Amer. J. Agr. Econ.

Table 1. Returns from Alternative Marketing Strategies, 10,000 Forty-Year Lifetimes

Difference from Basea Proportion

Mean SD Above Triggeredc

Strategy (c/ per Bushel) (c/ per Bushel) Baseb (Percent)

1. Diversification a. Spot only −3.0 −2.4 31 NA b. Spot/futures −7.0 −11.9 31 NA

2. Preharvest hedges a. Routine in May −1.1 −12.0 51 NA b. Conditional −1.2 −17.3 52 50

3. Storage hedges (Nov–May) a. Routine −2.4 −12.1 45 NA b. Conditional 3.8 −3.5 52 33

4. Speculative −0.1 −1.4 19 6 Source: Adapted from Peterson and Tomek, (2005b), table 2. a Base (harvest-time sales only): mean = 256 c/ per bu.; SD = 45 c/ per bu. b Proportion of years within a forty-year period where the return exceeded the base return, averaged over 10,000 replications. c Proportion of years within a forty-year period when the strategy is triggered, averaged over 10,000 replications.

in February, using December futures, when supply is small relative to demand. Otherwise, the crop is sold in May, as in 2a.

Strategy 3a routinely stores the crop and sells May futures in November, completing the hedge in May. This is done whether or not the basis in November provides an incentive to store. An alternative, 3b, is to store and hedge only when the basis signals that it is profitable to do so; otherwise, the crop is sold at harvest.

Finally, a speculative strategy (number 4) is considered. The crop is sold at harvest, but positions are taken in May futures selectively based on forecasts of the May price made in November using a regression model. Although the simulated prices are based on an efficient market, they sometimes appear to have a sig- nificant upward or downward bias in regres- sion models fitted to ten-year subperiods. Thus, if a statistically significant bias is found in a par- ticular period, a speculative position is taken in year eleven in May futures based on the ex- pected price change. Returns from the crop sale are adjusted for the speculative gain or loss.

Empirical Results

The long-run results are consistent with the performance of an efficient market, as they should be. Because of the time-to-maturity ef- fect of the variance of futures prices, the pre- harvest sale of all or a part of the crop reduces the variability of returns (strategies 1b, 2a, 2b, and 3a in table 1). Returns are adjusted for hedging costs, and hence average returns usu- ally are lower than for the base case. Also, for

strategy 3a, hedging locks in basis convergence that does not cover the cost of storage in some years.

If the harvest-time basis provides an incen- tive for storing (which occurred in about 33% of the years, on average), then hedging assures the return (strategy 3b).4 This strategy has little influence on the variability of returns, because the crop is sold at harvest whenever fore- cast basis convergence does not make storage profitable.

One possibly surprising result is that diver- sification of cash sales (strategy 1a) has only a small effect on the variability of returns. This occurs because sales are being deferred from harvest time, when prices are less variable, to months later in the marketing year, when prices are relatively more variable. We are re- minded that diversification does not “automat- ically” reduce the variability of returns.

Deeper insights are obtained by examining the results from the individual forty-year peri- ods. Although space constraints limit our abil- ity to characterize the diversity of results, those presented demonstrate that a farmer can face very different outcomes from a given market- ing strategy, depending on the particular cir- cumstances that occur within the time period in which the strategy is used. Prices occurring in one finite period are only a sample of the many possibilities that could have been gener- ated by a given market structure.

4 Recall, prices need not provide for a positive return to storage in every year in every location. The simulated prices should be interpreted as pertaining to a location (Illinois).

Tomek and Peterson Price Behavior for Marketing Strategies 1263

For example, although diversification of cash sales is a poor alternative when appraised by the average performance over 10,000 dif- ferent lifetimes, the mean return from diversi- fication exceeded the base return in over 30% of the years in the average forty-year period (table 1). Since this is the average outcome over 10,000 replications, a nontrivial probabil- ity exists that diversification would “pay” in thirteen or more years of a farmer’s forty-year lifetime. Also, since diversification produces a wide range of standard deviations, as does the base case (not shown), it is possible to find forty-year periods in which diversification does reduce variability relative to the base strategy. The opposite is also true; diversification can perform poorly in the large majority of years faced by a farmer. On average, it is not partic- ularly helpful in managing risk.

Speculation has little effect, on average, over 10,000 simulated lifetimes, but the returns to speculation ranged from −$2.41 per bushel in one year to $2.99 in another. Speculative rules can appear profitable, even over forty years, al- though not in the long run. The probability of finding “bias” can exceed the nominal proba- bility of type I error, associated with commonly used t-tests, as the error terms of the regression models are not normally distributed.

The simulations also illustrate the difficulty of evaluating the Wisner, Blue, and Baldwin strategy, which is conditional on supply– demand information known in February. Their condition was triggered in 50% of the years in an average forty-year sample. This contrasts with 31% in their 1975–96 sample, a frequency which is near the low that can occur in our simulations. We find that the early sale of the expected crop reduces the returns variability (strategy 2b), but with a slight decline in av- erage returns. The returns from this strategy, however, exceeded those from the base case in about half of the forty years in the aver- age sample. This strategy can seem successful a high percentage of the time in an efficient market.

Concluding Comments

An efficient commodity market can generate highly diverse price behavior with the pas- sage of time, and consequently the relative performance of alternative marketing strate- gies can vary over different samples. A strategy that is inferior, on average, can perform rela- tively well in a particular period. Indeed, this

happened in 30%–50% of the years (depend- ing on the strategy) in an average simulated forty-year lifetime, and outcomes in particular periods vary substantially around this average.

Another conclusion is that it is exceedingly difficult to discriminate between efficient and inefficient markets and to determine whether or not profitable arbitrage opportunities exist that will persist. The sampling error associated with estimates from short data series, com- bined with the consequences of data mining, raise serious qualifications about the value of technical analyses and of comparisons of mar- keting programs. If grain markets in the United States are semistrong form efficient, then rec- ommendations based on seeming market inef- ficiencies have little validity.

The evidence is clear, however, that routine hedging can often reduce the variance of re- turns, at a relatively small cost. In addition, it is possible to develop models to forecast ba- sis convergence and basis levels. Such analyses should be viewed as helping producers earn a competitive return to their production or stor- age decisions, and our challenge is to make these analyses more relevant to producers.

Our skepticism about the robustness of em- pirical results has grown steadily with our ex- perience as price analysts. Nonetheless, we hope that this article will not only discourage some past practices, but encourage fresh think- ing about analyses underlying recommended marketing programs. Useful results require careful scholarship.

References

Allen, P.G. “Economic Forecasting in Agricul- ture.” International Journal of Forecasting 10(1994):82–135.

Anderson, R.W. “Some Determinants of the Volatility of Futures Markets.” Journal of Fu- tures Markets 5(1985):331–48.

Benirschka, M., and J.K. Binkley. “Optimal Stor- age and Marketing over Space and Time.” American Journal of Agricultural Economics 77(1995):512–24.

Brorsen, B.W., and S.H. Irwin. “Improving the Relevance of Research on Price Forecast- ing and Marketing Strategies.” Agricultural and Resource Economics Review 25(1996):68– 75.

Collins, R.A. “Toward a Positive Economic Theory of Hedging.” American Journal of Agricultural Economics 79(1997):488–99.

1264 Number 5, 2005 Amer. J. Agr. Econ.

Fackler, P.L., and Y. Tian. “Volatility Models for Commodity Markets.” In T.C. Schroeder, ed. Applied Commodity Price Analysis, Forecast- ing, and Market Risk Management. NCR-134 Conference Proceedings, pp. 247–56, 1999.

Fama, E.F. “Efficient Capital Markets: A Review of Theory and Empirical Work.” Journal of Fi- nance 25(1970):383–423.

Grossman, S.J., and J.E. Stiglitz. “On the Impos- sibility of Informationally Efficient Markets.” American Economic Review 70(1980):393– 408.

Harwood, J., R. Heifner, K. Coble, J. Perry, and A. Somwaru. Managing Risk in Farming. USDA ERS Agricultural Economic Report 774, 1999.

Heifner, R.G. “The Gains from Basing Grain Stor- age Decisions on Cash-Futures Spreads.” Jour- nal of Farm Economics 48(1966):1490–95.

Hranaiova, J., and W.G. Tomek. “Role of Delivery Options in Basis Convergence.” Journal of Fu- tures Markets 22(2002):783–809.

Isengildina, O., S.H. Irwin, and D.L. Good. “Evalu- ation of USDA Interval Forecasts of Corn and Soybean Prices.” American Journal of Agricul- tural Economics 86(2004):990–1004.

Just, R.E., and G.C. Rausser. “Commodity Price Forecasting with Large-Scale Econometric Models and Futures Markets.” American Jour- nal of Agricultural Economics 63(1981):197– 208.

Lence, S.H., and M.L. Hayenga. “On the Pitfalls of Multi-year Rollover Hedges: The Case of Hedge-to-Arrive Contracts.” American Jour- nal of Agricultural Economics 83(2001):107– 19.

Lence, S.H., M.L. Hayenga, and M.D. Patterson. “Storage Profitability and Hedge Ratio Estimation.” Journal of Futures Markets 16(1996):655–76.

Mundlak, Y., and H. Huang. “International Com- parisons of Cattle Cycles.” American Jour- nal of Agricultural Economics 78(1996):855– 68.

Olszewski, E.A. “Assessing Inefficiency in the Fu- tures Markets.” Journal of Futures Markets 18(1998):671–704.

Park, C-H., and S.H. Irwin. The Profitability of Tech- nical Analysis: A Review. AgMAS Project Re- search Report 2004-04, University of Illinois at Urbana-Champaign, 2004.

Peterson, H.H., and W.G. Tomek. “How Much of Commodity Price Behavior Can a Rational Expectations Storage Model Explain?” Agri- cultural Economics: Journal of the Interna- tional Association of Agricultural Economics 33(2005a):289–303.

——. “Marketing Strategies and Lifetime Events in Grain Markets.” Department of Agricultural Economics, Kansas State University, 2005b.

Rosen, S. “Dynamic Animal Agriculture.” Amer- ican Journal of Agricultural Economics 69(1987):547–57.

Schaefer, M.P., R.J. Myers, and S.R. Koontz. “Ra- tional Expectations and Market Efficiency in the U.S. Live Cattle Futures Market: The Role of Proprietary Information.” Journal of Futures Markets 24(2004):429–52.

Tomek, W.G. “Commodity Prices Revisited.” Agri- cultural and Resource Economics Review 29(2000):125–37.

Tomek, W.G., and R.W. Gray. “Temporal Rela- tionships among Prices on Commodity Fu- tures Markets: Their Allocative and Stabilizing Roles.” American Journal of Agricultural Eco- nomics 52(1970):372–80.

Tomek, W.G., and H.H. Peterson. “Risk Manage- ment in Agricultural Markets: A Review.” Journal of Futures Markets 21(2001):953–85.

Wang, D., and W.G. Tomek, “Commodity Prices and Unit Root Tests.” Paper presented at the American Agricultural Economics Association Meetings, Denver CO, 1–4 August 2004.

Williams, J.C., and B.D. Wright. Storage and Com- modity Markets. Cambridge: Cambridge Uni- versity Press, 1991.

Wisner, R.N., E.N. Blue, and E.D. Baldwin. “Pre- harvest Marketing Strategies Increase Net Returns for Corn and Soybean Growers.” Re- view of Agricultural Economics 20(1998):288– 307.

Working, H., “Hedging Reconsidered.” Journal of Farm Economics 35(1953):544–61.

Zulauf, C.R., and S.H. Irwin. “Market Efficiency and Marketing to Enhance Income of Crop Producers.” Review of Agricultural Economics 20(1998):308–31.

Zulauf, C.R., S.H. Irwin, J.E. Ropp, and A.J. Sberma. “A Reappraisal of the Forecast- ing Performance of Corn and Soybean New Crop Futures.” Journal of Futures Markets 19(1999):603–18.