Mobile Edition
Print
Get the Mag
Weekly UpdatesStructural changes taking place on dairy farms are an important policy concern in the United States and elsewhere. Dairy farm herd sizes and cow productivity have exhibited significant increases during the last twenty years. The demand for dairy products has only grown slowly, however, leading to an imbalance between supply and demand and a consequent reduction in the number of dairy farms. Despite the general trend of increasing farm size, a very heterogeneous pattern of structural change appears across regions that relates to costs of production, technology, weather, and geography among other factors (MacDonald et al. 2007; Wolf 2003). Blayney and Normile (2004) contend that the main drivers of these changes are a mixture of technological, efficiency, and scale changes, and they note a lack of empirical evidence on such key technology indicators as scale economies and their variation across geographical areas in the United States, specifically. This research seeks to fill that gap.
This study uses a data set for 619 dairy farms drawn from the 2000 Agricultural Resource Management Survey (ARMS2000) (USDAERS 2000), (1) a national survey of U.S. dairy producers, to estimate scale economies in such a way as not to confuse them with economic inefficiency or other influences in the cost-output relationship. It builds on analyses of scale economies (e.g., Alvarez and Arias 2003; Kumbhakar 1993; Moschini 1988; Tauer and Mishra 2006) and estimated efficiency in the dairy sector (e.g., Maietta 2000; Stefanou and Saxena 1988).
Costs can decrease when efficiency improves due, for example, to better training, more time on the farm, or better technology that lowers costs of production. Our study builds on earlier ones that have tackled the question of performance by either using nonfrontier approaches to modeling inefficiency (e.g., Maietta 2000; Stefanou and Saxena 1988) or frontier using various approaches. Examples include Tauer and Mishra's (2006) single-equation cross-section; Alvarez and Arias' (2003) panel data single-equation models with time-varying and cross-sectional variation in inefficiency; and system approaches (where first-order conditions and economic objectives are employed) as, for example, in Kumbhakar (1993) and Moschini (1988). The latter employed a non-frontier framework to tackle economies of scale, one of our main concerns in this study.
Several earlier studies have analyzed scale economies and efficiency in dairy farms while investigating key variables that we also find important. Moschini (1988) found increasing returns to scale for Canada's Ontario dairy producers, with large levels of milk output but decreasing returns for the very largest ones. Factors affecting scale economies in that study were location, debt/equity ratio, milking techniques, building quality, cow type, education, and horsepower of the largest tractor. Tauer and Mishra (2006) also found increasing returns to scale, but they were quite small once accounting for inefficiency. In their study the higher cost of production on most small farms was caused by inefficiency rather than technology. They represented technology by number of cows and state dummies rather than by a production, cost, or profit function, however. Kumbhakar (1993) found that large farms have exploited short-run returns to scale to a greater extent than medium or small dairy farms. His study also noted that large farms were more technical, allocative, and scale efficient than the others. In addition, he found that off-farm income is negatively related to performance but least so for large farms, and the farmer's level of education contributes more for medium and large farms than for small ones. Unlike in our present study, he did not precisely disentangle allocative and technical inefficiency because of inconsistent distributional assumptions. Alvarez and Arias (2003, p. 141) found a U-shaped average cost curve and contended that "even if there are observed diseconomies of size, a large enough increase in managerial ability could outweigh the rising part of the average cost curve." They did not control for variation in allocative inefficiency. Using panel data, Maietta (2000) estimated long-run returns to scale as decreasing on average, and without providing specific sources for investigating allocative or technical inefficiency, she found a decrease in allocative inefficiency and an increase in technical inefficiency for larger farms. Finally, Stefanou and Saxena (1988) also contributed to the research on some key variables by finding that education and experience matter quite a bit for dairy farm profit efficiency. They employed a shadow profit system of equations that did take into account and explain both allocative and technical inefficiency variation.
Following and expanding on this literature, in this article we analyze farm-level data, allowing for variation in cost efficiency and incorporating variables commonly thought to influence farm performance. Our estimates are not plagued by the distributional inconsistencies and misspecifications of inefficiency faced by previous studies that have examined scale economies in dairy farming.
Structural Change and Scale Economies in U.S. Dairy Farming
The transformation of dairy operations is usually analyzed through changes in location, production system, herd size, total and per-cow production, and organizational shifts through time (see, e.g., Blayney and Normile 2004; MacDonald et al. 2007). Here, we focus instead on structural changes that have occurred during the last twenty years. Figure 1 shows the inverse relationship between the number of cows in the national herd and production of milk per cow. Given that demand growth for dairy products has not kept pace with the increase in milk production per cow, the national herd has declined 11% from 1987 to 2007. During this same period, milk production per cow increased 47%. These production trends have led to total milk production increasing by 30% (U.S. Department of Agriculture, National Agricultural Statistical Service [USDA-NASS] 2008).
Simple correlation analysis provides some evidence that scale economies are important determinants of productivity. There is a wide variation in milk produced per cow across states. The correlation between milk produced per cow and the number of milk cows per operation across dairy farms is strong and positive, indicating a potential role for scale economies in determining productivity. A simple correlation analysis using publicly available USDANASS data at the state level shows a correlation of 0.436 between milk produced per cow and cows per establishment in 1987 and 0.564 for 2007 (USDA-NASS 2008).
Figure 2 shows further evidence of scale economies. From 1998 to 2007, the number of dairy operations in the United States decreased by 39%. This decline, however, was not symmetrical across farm sizes, resulting in fewer small and a considerable increase of large dairy farms. The cow inventory of dairy farms with herd sizes between one and forty-nine declined from 14.1% to 7.4% of the total, and that of operations with 50-199 milk cows decreased from 43.6% to 28.8%. In contrast, dairy operations with 200-1,999 head increased their cow inventory from 35% to 40.17%, and operations with 2,000 or more head, from 7.3% to 23.1% (USDA-NASS 2008).
[FIGURE 1 OMITTED]
[FIGURE 2 OMITTED]
The change in size structure has not affected all regions of the country equally. A sense of the regional shifts that have occurred lately can be grasped by looking at the ranking of milk-producing states in 1987 and 2007. In 1987, the ten largest milk-producing states were, in order: Wisconsin, California, New York, Minnesota, Pennsylvania, Michigan, Ohio, Texas, Iowa, and Washington; in 2007, they were: California, Wisconsin, New York, Idaho, Pennsylvania, Minnesota, Michigan, Texas, New Mexico, and Washington. In 1987, the top ten states produced 68% of the national milk supply, while in 2007 the top ten produced 73% (USDA-NASS 2008).
These regional changes also imply a shift in types of production systems. (2) Many operations in states like California, Idaho, and New Mexico, for example, have seen the emergence of so-called dry-lot systems (i.e., they rely on purchased feed). These emerge when low capital requirements and large herd sizes that enable exploitation of scale economies lower cost per unit of output. In 1987, in California, for example, the average number of milk cows per operation was 226, while in Idaho and New Mexico, it was fifty-three and fifty head, respectively. By contrast in 2007, California had an average of 824 cows per operation, Idaho had 684, and New Mexico had 814 cows. More traditional states increased their average size of operation but by a much smaller percentage. In Wisconsin, New York, and Pennsylvania, the average number of cows per operation in 1987 was 49, 57, and 39 cows, respectively; in 2007, it was 87, 101, and 65 cows. These states do not rely so much on purchased feed as on homegrown feed or pasture (USDA-NASS 2008); that is, they use capital differently.
Given the heterogeneity of the changes in dairy operation size across production technologies and regions, the question of the nature of scale economies becomes crucial. According to Chavas (2001), in general the average cost curve for the agricultural sector in developed countries tends to be L-shaped; while scale economies tend to exist for small farms, no strong evidence indicates that diseconomies of scale exist for large farms (i.e., there is a wide range in which scale economies are constant). In addition, Chavas emphasizes the importance of taking into account variables like the shadow value of unpaid labor. Morrison et al. (2004) have also examined the trend toward consolidation in U.S. agriculture generally and have found significant scale and efficiency advantages of large farms. For dairy specifically, Jones (1999) presents a similar picture in which scale economies are exhausted quickly. The variation in dairy operation sizes, moreover, can be explained by a myriad of variables internal and external to the dairy farm such as pecuniary economies, transaction costs, tax policy, regulation, and risk. Wolf (2003) argues that dairy farms in traditional areas such as Wisconsin, New York, and Pennsylvania face higher adjustment costs (because of high sunk costs) than emerging regions, a condition that constrains their growth and adoption of technology.
FEED