Recently, however, reports regarding the financial difficulties experienced by U.S. agricultural cooperatives have been much more common than news of their successes. In particular, the 2002 bankruptcy of Farmland Industries — a federation of 1,700 independent Midwestern cooperatives and the nation’s largest agricultural cooperative — received considerable media attention. In California, news about cooperatives has centered on the bankruptcy of Tri Valley Growers in 2000; the dissolutions of Blue Anchor and the Rice Growers Association of California in 2000; and the conversions of Calavo in 2001 and Diamond Walnut Growers in 2005, to publicly traded, investor-owned corporations. Such news has raised concerns among producers and lenders regarding the viability of the cooperative form of agricultural business. In the agricultural sector, producers use cooperatives to market and process their crops and livestock, purchase supplies and services, negotiate terms of trade with processors of their raw product, and provide credit for their operations. An international management consulting firm, McKinsey & Company,plastic gutter issued a report in 2002 alleging that agricultural cooperatives “destroy value” because few cooperatives “have changed the way they operate” .
This report received considerable attention from the management and boards of numerous large cooperatives, despite the fact that its analysis was based on only 2 years of data. Some cooperative researchers also noted other technical limitations. Was McKinsey & Company’s claim that agricultural cooperatives destroy value justified? Or do cooperatives benefit California’s agricultural producers? What is the future for agricultural cooperatives in California? Cooperatives have been part of the agricultural sector in the United States for approximately 200 years. They can benefit their members in several different ways. In the Midwest, cooperatives were formed primarily to maximize the welfare of their individual members. These cooperatives handle the entire output of their members regardless of market needs, and are clearly extensions of their members’ farming businesses. Conversely, many of the marketing cooperatives formed in California during the first quarter of the 20th century were designed to create market power by improving product quality and restricting raw product flows. Such market power–oriented cooperatives seek to maximize the profitability of the firm, rather than the welfare of individual members. These different objectives can have vastly different impacts on the operations of cooperatives. A cooperative with a market-power structure could operate in niche markets with a strong brand identity and handle limited volumes of member product to maximize its profitability as a firm. This type of cooperative would then distribute some or all of its earnings to its members. Some of these cooperatives, such as Mountain States Lamb, require members to buy enough delivery rights to match their delivery volumes.
Members must invest in a delivery right for each lamb they deliver annually to Mountain States Lamb for processing and marketing. The delivery rights control the amount of raw product delivered by members; they depend on the processing capacity of the cooperative’s plant. Investment in delivery rights is part of a producer marketing agreement. If a producer is unable to deliver the agreed amount of raw product, purchase of commodities is authorized by the cooperative for undelivered obligations. Such delivery rights are marketable and can appreciate in value if the cooperative is successful. For example, the founding members of Dakota Growers Pasta paid $3.85 in 1991 for a right to deliver a bushel of durum wheat annually to the cooperative. By 1998, the cooperative’s strong earnings enabled retiring members to sell a delivery right for $7.50. In contrast, a Midwestern-style marketing cooperative could maximize benefits to its members by accepting their deliveries up to its break-even point, which would provide as much of a home for their product as possible without incurring losses. While this decreases the members’ potential earnings from the cooperative, it also reduces the risk they face. It is inappropriate to assume that all cooperatives are seeking to maximize their profitability as firms. Nonetheless, various national studies were conducted during the late 1980s that compared the financial performance of agricultural cooperatives and investor-owned firms . The findings from these studies varied widely . These financial performance studies used ratio analysis, including profitability measures. Ratio analysis is a tool used to evaluate a firm’s financial performance by taking data from its financial statements and comparing the ratios over time, and/or with those for other firms or the industry. However, Sexton and Iskow pointed out how analyses of cooperatives based upon financial ratios, although popular, were not based on economic theory. Specifically, they noted that since cooperatives are extensions of their members’ businesses, a cooperative could be less profitable than an investor-owned firm and still be beneficial to a member — as long as the member’s discounted stream of returns from the cooperative was greater than those from marketing the commodity directly or through an investor-owned firm.
For example, membership in an almond marketing cooperative that is averaging a 6% operating margin while one of its investor-owned competitors is averaging a 10% operating margin could still be beneficial to the cooperative’s members. Members could receive a higher price for their almonds from the cooperative than if they sold their crop to the investor-owned firm; the investor-owned firm strives to minimize its costs, including the price it pays for its almonds. John Hicks is credited with advancing the conjecture that changes in relative prices induce technical progress . This conjecture implies that relative factor prices serve a dual function, as signals of resource scarcity and as determinants of the firm’s technology choice. Hayami and Ruttan revitalized Hicks’ conjecture and made important contributions to the explanation of the magnitude and direction of TP in the American and Japanese agricultural sectors using the relative price hypothesis. Over the past thirty years, many authors have attempted to test this hypothesis using aggregate data and obtaining mixed results. In these studies, the consensus approach to the econometric estimation and testing of the hypothesis that technical progress is induced by relative prices has been to regress the ratio of some factors of production over a distributed lag series of their price ratios and other similar series of extension,blueberry container public and private R&D expenditures. Thirtle, Schimmelpfennig and Townsend summarized several significant studies of this kind and produced one of their own. The sample information about output quantity and output price is remarkably absent in many of these studies. This omission seems in contrast to the conjecture advanced by several economists according to whom the choice of techniques is determined, to a large extent, by profitability considerations. In this paper, therefore, we attempt to recast the price-induced technical progress hypothesis into a framework that utilizes all the available theoretical and sample information, including output price and quantity. This approach leads to a novel set of comparative statics conditions of the economic theory of the firm undergoing technical progress that provides an exhaustive scaffolding for testing the PITP conjecture.When dealing with technical progress, it is convenient to distinguish the innovation phase from the adoption phase. The majority of price-taker firms self select into the adoption phase. In general, the choice of available techniques made by those firms is guided mainly by expected profitability considerations. When price-taker firms are aggregated into an industry, such as the US agricultural sector, the R&D and extension expenditures may become determinants of the industry technical progress. Griliches , Arrow , Hirsch and other economists have suggested that expected profitability objectives may be a determinant of adoption rates. The expected profitability conjecture relating expected profits to TP leads to a model where expected output and input prices enter the production function as shifters of the technology frontier. As originally suggested by Paris and re-elaborated more recently by Paris and Caputo and by Caputo and Paris , we incorporate expected relative factor prices explicitly into the production function and assume a cost-minimizing behavior of the individual entrepreneur.
The introduction of expected relative prices into the production function invalidates the traditional comparative statics relations of the competitive firms but leads—by necessity—to a more general model of the cost-minimizing/profit-maximizing entrepreneur. The novel set of comparative statics conditions depends on both primal and dual relations and is expressed in the form of a symmetric and negative semidefinite matrix of estimable terms. It follows that the empirical implementation of the PITP conjecture developed in this paper requires the joint estimation of the derivatives of the cost function with respect to relative input prices, the production function and the first order necessary conditions.Traditionally, aggregate models of TP based upon time series data have been specified using a distributed lag representation of either quantities or prices, or both. This approach seems to have been taken for two main reasons: to capture, somehow, a dynamic aspect that is assumed to be inherent in a process of technical progress, and to represent some process of expectation formation of the entrepreneur about quantities and prices. Often, the two aspects are confounded. With respect to the PITP model presented above, we would like to point out that the expectation process is taken into consideration explicitly and there is no need to formulate a distributed lag representation of expected quantities and prices. We acknowledge that the dynamic aspect of TP requires an explicit theory, akin to the static theory formulated above: a distributed lag specification without theory is only an ad-hockery. In general, it will be wise to postulate that the theoretical relations expressed in equations – are represented by flexible functional forms. Such forms are not self-dual in the way that the Cobb-Douglas and the CES functions are. Hence, the implementation of the above model requires the statement of a cost function that has entirely different parameters from those of the production function. The coherent link between the primal and the dual frameworks is represented by the unknown expected quantities and prices that must be estimated along with the parameters. The discussion of how to estimate the model given by equations – will be the subject of the following sections. We would like to advance here that, in principle, a Bayesian approach along the lines presented by Zellner would produce consistent estimates. But, as we are not comfortable with elaborate and multi-dimensional integration techniques, we will propose a two-phase approach based upon a nonlinear least-squares estimator.The sample input data for the present analysis were made available by Thirtle, Schimmelpfennig and Townsend and are described in their paper. The time series consist of four input quantity and price indices relating to machinery, labor, fertilizer and land, from 1880 to 1990; public and private R&D and extension expenditures are also from 1880 to 1990. Additionally, aggregate output quantity and price indices from 1910 to 1990 were derived from the US Historical Statistics and USDA databases and provided by Spiro Stefanou. All the index series are defined with base 1967 = 100. Because the primal-dual model of PITP developed in this paper uses also the output quantity and price series, the usable sample data range from 1910 to 1990 with 81 observations. In this paper we chose to deal with the single aggregate of output for the US agriculture. All the data were scaled by a factor of 100 so that the average of most series is close to 1.Phase I of the PITP model was estimated using the GAMS programming package and unitary € λ weights for the objective function . This choice was dictated by a lack of knowledge of the true weights. The selection of these weights transforms the given problem into a nonlinear Total Least Squares model, originally described by Gulob and Van Loan , and by a vast literature since then. The model constraints, represented by equations , and , are highly nonlinear and non-convex. Hence, the solution achieved is only locally optimal. The problem was solved several times with different initial values. A serial correlation of order 1 was implemented during the estimation procedure. The use of the GAMS 21.6 programming package requires a careful choice of upper and lower bounds for all parameters. Still, the solution of the problem is a non-trivial enterprise. The phase I PITP model has 1495 constraints and 1721 unknown parameters. In a typical run, the CPU time to achieve a locally optimal solution was about 20-30 minutes on a Supermicro machine .