Cash marketing styles and performance persistence.

One decision that has long troubled crop producers is when to sell their commodities. Producers can choose to sell based on either a mechanical or an active marketing style. With a mechanical style, producers sell grain in the same week every year, regardless of market information. The active style is to sell grain in different weeks of different years depending on price expectations. Producers may use an active style if they believe they can use market information to get a higher price. Alternatively, farmers may use a mechanical strategy if they believe that markets are efficient and there is no gain in trying to predict prices.

This article determines the extent to which farmers use active or mechanical styles in marketing wheat in Oklahoma. We also test whether producers using an active or mechanical style receive a higher price and find no significant difference. Furthermore, we test for performance persistence and find no evidence of any producers consistently outperforming or underperforming other producers. We use data containing individual transactions of grain sales for nine crop years. The data allow measuring for each producer, the annual average week sold, number of weekly transactions, and total bushels sold. The activeness of a marketing style is measured as the standard deviation of mean week sold. Price received is then regressed against activeness to test the hypothesis that grain farmers using an active style receive lower prices than farmers using a mechanical style.

Following recent literature in finance (Agarwal and Naik 2000; Blake and Timmermann 2003; Brorsen and Townsend 2002; Harri and Brorsen 2004), we include performance persistence tests. Persistence performance means identifying winners that follow winners or losers that follow losers within a particular industry. The null hypothesis with the performance persistence test is that the past ranking of a farmer's price received does not help predict the farmer's future ranking. If there are farmers who have consistent performance, then these farmers' actions can be used to identify styles with superior performance.

Unlike most previous literature, we study what farmers do rather than try to determine what they should do. For example, past literature has focused on optimal hedging strategies (e.g., Shi and Irwin 2006; Zuniga, Coble, and Heifner 2001), tests of market efficiency (e.g., Kastens and Schroeder 1996; Zulauf and Irwin 1998), and price forecasting (e.g., Just and Rausser 1981; O'Brien, Hayenga, and Babcock 1996). The main finding of these studies is that there is little chance of increasing profit through trying to predict prices. Yet, farmers still demand considerable market information from both private market advisory services and extension economists (Ortmann et al. 1993).

This dichotomy between research results and farmer actions suggests a need to determine the actual pricing performance of producers.

Both McNew and Musser (2000) and Slusher (1987) have looked at individual farmer marketing decisions, but only Slusher (1987) studied actual producer transactions. McNew and Musser (2000) used data from a hedging game to evaluate producer marketing decisions. Slusher used recall data from 129 soybean producers in comparing returns from using different marketing alternatives such as spot sales, storage, contracts, and basis contracts. Slusher found that the choice of the marketing alternative did not affect the price received by farmers, and storing soybeans beyond one month led to negative returns. Our study differs from Slusher's study in two important areas. First, we use transaction-level data rather than producer recall data. Further, our wheat data series is considerably longer and more recent than Slusher's soybean data. Second, while Slusher compares returns among farmers using different marketing outlets, we focus on active versus mechanical strategies.

Another related area of research has been the numerous surveys that determine what marketing information producers are using (Ford and Babb 1989; Gloy, Akridge, and Whipker 2000; Ortmann et al. 1993; Patrick and Ullerich 1996). Most of these surveys show that farmers used on-the-farm data and information from paid sources such as magazines, consultants, and market advisory services. Pennings et al. (2004) found that few producers exactly followed recommendations from market advisory programs so knowing the information sources that producers use may give little indication about how they actually make decisions.

The next section describes the theory of optimal storage, highlighting the differences between active and mechanical strategies. Next, we describe the data and discuss how the limitations imposed by the data are overcome. Then, we describe the procedures used to measure activeness and persistence performance. The procedures include linear regression, style indicators, and performance persistence tests. The discussion of study results precedes the last section summarizing the main findings and implications.

Theory

The theory of optimal storage by crop producers was derived by Fackler and Livingston (2002). Because of the irreversibility of the storage decision, there is an option value from continuing to store. They set up an optimal stopping problem that must be solved numerically. To illustrate the concepts considered here, it is sufficient to consider only their expectation rule that ignored the option value. With the expectation rule, producers will sell when the current price is greater than or equal to the maximum expected return from holding stocks:

(1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where [P.sub.t] is price per bushel at time t, h is days after harvest, T is the end of the marketing period, r is the daily interest rate, [s.sub.t] is daily per-bushel storage costs, and [[OMEGA].sub.t] is the information available at time t. If the option value were added to (1), it would also be an expectation conditional on [[OMEGA].sub.t].

In the absence of transportation costs and with efficient markets, equation (1) would always hold as an equality and thus producers would be indifferent as to when to sell their crops. In Oklahoma, we observe June harvest prices being roughly equal to returns from selling in September through November (Klumpp, Brorsen, and Anderson 2005). There is little return to storage past November, presumably due at least partly to convenience yield (see Working 1949; Fama and French 1987). Oklahoma prices in July and August tend to be lower than June prices, presumably because of large harvest sales in more northern states.

The key difference between active and mechanical traders is the way that expectations are formed in (1). A mechanical strategy assumes that the solution to (1) is the same every year. Thus, this strategy assumes that arbitrage is always sufficient to bring prices into line. On the other hand, an active strategy is based on solving (1) each day of each year based on the available information. An active strategy could profit either from responding to market signals or from being able to predict prices. An active strategy could also be worse than a mechanical strategy if producers suffer from behavioral biases such as myopic loss aversion (see Kahneman and Riepe 1998) that would cause them to hold onto their crop even in the presence of negative expected returns to storage.

The Fackler and Livingston theory does not consider risk so the solution is to sell the entire crop at one time. Fackler and Livingston (2002, p. 657) argue that a risk-averse producer might want to space out sales in order to reduce risk. The theory is also incomplete because producers time marketing for reasons other than price expectations. Producers may sell because they need cash to meet financial obligations. They could also postpone sales until the next calendar year to postpone paying income taxes. Wheat may be held for seed and sold when not needed. Also, any possible effects of government payments such as loan deficiency payments (LDP) are not considered.

In the finance literature, any strategy that is not buy and hold is considered active and activeness is determined by transaction frequency. Barber and Odean (2000) found individual investors using more active strategies earn substantially lower returns, but this is due primarily to trading costs. Results with mutual funds are mixed (Allen et al. 2003) with some indication that more active managers earn slightly higher returns, but still not enough to cover the fees typically charged. This definition of activeness is not appropriate here as there are no charges for trading frequency and storing wheat indefinitely is clearly not the optimal strategy.

The first null hypothesis of interest is that returns are the same for active and mechanical strategies. Note that the optimal solution in (1) will result in producers with lower interest costs storing longer. Also, prices will vary by year. Regression is used to separate the effects of activeness from the effects of year and the average time after harvest when the crop is sold.

Data

Data are from three grain elevators located in the north, south, and center of western Oklahoma. The data are from the harvest of 1992 through the spring of 2001 (nine crop years). The data contain all individual transactions of wheat sales at each elevator. Each transaction has the seller, number of bushels, price per bushel, and date. However, each seller's name was not always spelled correctly and some sellers operated under a variety of names. To remedy this problem, elevator managers were asked to identify the primary marketing decision maker for each sale.