Pre-harvest price expectations for corn: the information content of USDA reports and new crop futures.

Futures prices play an important price discovery role in the marketing of storable commodities. This function requires that agents (1) have informed expectations regarding the size of the future crop and futures prices reflect that available information. Hence, two interesting questions that arise are: What sources of information contribute to rational agents' harvest-time price expectations? And do commodity futures prices fully reflect these expectations? One potentially useful source of public information that is available to agents several months prior to harvest is United States Department of Agriculture (USDA) crop production reports. The efficient markets hypothesis (EMH) would assert that as new fundamental information becomes available, futures prices should immediately adjust to the "news" to reflect a change in rational agents' price expectations. The two main objectives of this article seek to establish whether USDA crop reports contain valuable information and determine whether new crop futures prices adhere to a semi-strong form version of EMH by embodying this information. These objectives are achieved by employing more powerful methods than the traditional event study approach to estimate rational agents' harvest-time price expectations. If USDA crop reports contribute to rational agents' price expectations, then it is assumed that reports have informational value. If futures prices react to and embody rational agents' price expectations, then it is assumed that futures markets are efficient.

A large body of agricultural economics research has utilized the event study approach to test whether or not USDA crop production reports contain new information that was unanticipated by the market prior to their release. This body of research hypothesizes that significant changes in market prices following a report's release are an indication that the report was "newsworthy." This hypothesis implicitly supposes that market prices conform to EMH. Early research found significant price reactions in corn and soybean markets following report release days, suggesting at least the possibility that reports might have economic value (Fackler 1985; Milonas 1987; Sumner and Mueller 1989).

Subsequent studies have extended previous research by incorporating private crop forecasts in tests designed to gauge the informational content of USDA crop reports (Garcia et al. 1997; Egelkraut et al. 2003; Good and Irwin 2006). It is assumed in these studies that those crop forecasts provided by private agencies, such as Sparks Companies, Inc. and Conrad Leslie, which have been historically released several days prior to USDA report announcements are good proxies for market expectations. These studies tested the forecast accuracy of USDA reports relative to private forecasts and concluded that production forecast errors for USDA and private agencies were highly correlated, suggesting that at least some information contained in USDA reports is already anticipated by private agencies. Egelkraut et al. (2003) also noted that over time, private agencies' forecasts made in August have improved and are marginally more accurate than the USDA August report.

Theoretically, the EMH states that market prices should only respond to new information. To address this issue, Garcia et al. (1997) used private crop forecasts to model market expectations and distinguish between anticipated and new information in crop reports-an approach initially used by Coiling and Irwin (1990) with respect to USDA Hogs and Pigs Reports (HPR's). Using data from 1971 to 1992, they concluded that corn and soybean futures prices react to new information, but reactions are smaller post-1984. They also employed a willingness-to-pay test--a concept first developed by Carter and Galopin (1993) with respect to HPR's--to determine the informational value of crop reports. In a similar vein to the their price reaction tests, Garcia et al. (1997) found that a hypothetical futures trader would be willing to pay for advanced knowledge of the USDA crop report for their full sample period (1971-1992) but not for the later subsample period (1985-1992). More recently, Good and Irwin (2006) have shown that corn and soybean futures prices continue to react to the release of new information in August crop reports. Plots in Good and Irwin indicate that if anything, the reaction of new crop corn and soybean futures prices after release of the reports has actually increased in recent years.

In summary, the two modal conclusions of recent research appear contradictory. First, forecast accuracy of USDA corn and soybean reports relative to private forecasts clearly show that private market forecasts released just prior to August reports have been at least as accurate as USDA forecasts since the mid-1980s (Garcia et al. 1997; Egelkraut et al. 2003, Good and Irwin 2006). Second, futures prices continue to react to the release of new information contained in these same USDA forecasts (Garcia et al. 1997; Good and Irwin 2006). A number of possible explanations for this apparent contradiction are discussed in Good and Irwin (2006) and Garcia and Leuthold (2004). They include the possibility that agents regard USDA crop forecasts as less risky than private forecasts and hence still induce price reactions (Garcia et al. 1997). Another potential explanation promotes the idea that the release of public information may play a role in coordinating the beliefs of market agents, even if the pubic announcement contains no valuable information with respect to market fundamentals in itself. This explanation is based upon the theoretical model developed by Morris and Shin (2002).

Good and Irwin (2006) assert that one of the most plausible explanations for the contradictory findings of recent research is derived from the theoretical model of Falk and Orazem (1985). In this model market agents rationally update the weights placed on crop production forecasts as new information becomes available to the market. By combining information from private and USDA forecasts, agents do not necessarily place a weight of 100% on the USDA forecast. However, combined forecasts will generally have lower forecast error variance than USDA forecasts alone. Therefore, even if USDA forecasts are no better than private forecasts, their release will still induce futures price reactions. Orazem and Falk (1989) use a signal extraction model to empirically investigate this concept with respect to USDA soybean forecasts. It is assumed that USDA forecasts do not fully reflect available market information, and a discrepancy exists between USDA and market agents' forecasts. Their results suggest that post-report movements in soybean futures prices are more responsive to agents' production forecasts, conditioned on both private and USDA information rather than on USDA forecasts alone.

This article sheds further light on the remaining puzzle of why futures prices continue to react to the release of USDA crop reports despite the fact that the reports appear to provide no better production forecasts than private forecasts. Using a modeling approach developed by Hamilton (1992), statistically optimal inferences are obtained--in a mean square error sense--about market agents' ex ante harvest-time corn price expectations. The approach exploits co-movements in USDA corn and soybean crop production forecasts with corn prices to uncover agents' price expectations. Soybean is a closely related commodity to corn, and its production is included in the model to incorporate potential price substitution effects (Garcia et al. 1997). Movements in harvest corn prices that are correlated with the release of "news" contained in USDA reports are assumed to have been anticipated by agents at that time. "News" is modeled as the difference between USDA production forecasts and previously released private market forecasts. Movements in harvest corn prices that are correlated with rational agents' production forecast errors are assumed to have been unanticipated by agents at the time. Rational agents' production forecasts are modeled as USDA forecasts adjusted for other market information that would be useful in forecasting final production. These assumptions make it possible to decompose harvest corn price changes into anticipated and unanticipated components and allow inferences to be drawn about rational agents' ex ante price expectations. Hamilton's (1992) original approach exploited co-movements between aggregate prices and commodity futures prices to uncover rational agents' aggregate price expectations during the deflationary period of the "Great Depression." Holt and McKenzie (2003) subsequently built on Hamilton's approach and analyzed co-movements between broiler prices, broiler production, and related commodity futures prices to infer agents' broiler price expectations.