P 97 - 2 January 1997 STAFF PAPER SERIES ARE LARGE FARMS MORE EFFICIENT ?

ARE LARGE FARMS MORE EFFICIENT? Accurate estimates of returns to scale require that inputs and output are measured without error and that environmental and managerial differences among firms of varying sizes are taken into account. Measurement problems affecting estimates of returns to scale in agriculture include: (1) combining the farm dwelling with capital inputs, (2) correlation of environmental and management characteristics with size and (3) the effect of off-farm employment on small farm output and production costs. Estimates of long run average total cost curves for farms in the corn belt reveal that after the above factors are taken into account, estimated scale economies in agriculture disappear, while there is evidence of diseconomies as farm size increases. 1 The proposition that large farms are more efficient than small ones, namely that economies of scale exist in U.S. agriculture appears to be widely accepted both by the public and the profession. The growth in size and decrease in number of farms over the past half century is consistent with this hypothesis. Moreover numerous empirical studies have reported scale economies. (See for example, Ahearn, Whittaker, and El-Osta; Batte and Sonka; Cooke and Sundquist; Griliches; Hallem; Hayami and Ruttan; Kumbhakar, Biswas, and Bailey; Madden and Partenheimer; Moschini). Recently, questions have been raised about the validity of the scale economies hypothesis (Kislev and Peterson, 1996). We are reminded here that scale economies must be a temporary, disequilibrium phenomenon since, according to Euler's theorem, payments to factors will exceed output. Obviously this situation cannot persist indefinitely. Yet conventional estimates of returns to scale in U.S. agriculture have not decreased over the past 40 years. Moreover, returns to scale estimates did not decline during the 10-year period mid 1970s to mid 1980s even though farm size remained relatively constant during this time. It is argued, therefore, that estimates of returns to scale have been biased upward. Accurate estimates of returns to scale require that inputs and output be measured without error and that environmental and managerial differences which may exist for firms of varying sizes are taken into account. Labor quality is often cited as a source of bias. If large farms are managed by more capable people, as seems likely, and this is not reflected in the labor measure, the labor input on large farms will be understated causing an upward bias of returns to scale estimates. Use of covariance analysis to account for management or labor skills differences among farms has resulted in reduced estimates of returns to scale (Kislev; Mundlak). The 2 evidence also suggests that education, a proxy for managerial ability, is associated with larger farm size (Sumner and Leiby). The overall objective of this study is to assess the impact of several measurement problems inherent in agricultural census data. Much of what we know about returns to scale in agriculture is based on these data. These problems include (1) the addition of the farm dwelling to capital inputs, (2) differences among farm sizes in land and management quality, and (3) differences in off-farm employment. MEASUREMENT PROBLEMS 1. Farm dwelling. To estimate the value of land and buildings on farms, the agricultural census asks respondents to report the value of land and all buildings, including the value of the farm dwelling. As a result, the service flow, or implicit rental value of the house, does not show up as farm output but the house is included among the inputs. More importantly, the house accounts for a larger share of the inputs on small farms than on large farms. The house also accounts for a larger share of output on small farms. Indeed, for small part-time and hobby farms the service flow of the house and yard probably is the main output of the farm, i.e. the reason for living there. Since the upward bias of inputs, and downward bias of output, increase as farm size decreases, inclusion of the house in capital inputs leads to an upward bias of estimated returns to scale, at least using census or other data where the house and yard are included in the value of land and buildings. 2. Land Quality. As indicated in the next section, yields on large farms are substantially higher than on small farms. This should be cause for concern; a plant or an animal does not know if it is growing on a large or small farm. Under similar management and environmental conditions, 3 biological efficiency should not vary among size groups. Differences in yields are likely to stem mainly from land quality differences. And evidence will be presented which seems to indicate large farms are situated on higher quality land than small farms, at least in the corn belt. Failure to take into account land quality will make it appear that large farms are more productive than small farms, i.e. that economies of scale prevail, when in fact we may be observing economies of better quality land and higher yields. 3. Off-Farm Employment. This activity which is more prevalent on small farms has several effects on the farm operation. Because the principal operator and/or family members now have less time to allocate to the farm, there can be an economic decision to shift towards a less intensive management level. Management is not only knowing what to do and when to do it, but it also has an opportunity cost. For example, more timely planting and harvesting that jeopardizes the off-farm job, or results in a reduction in off-farm earnings in excess of what is gained on the farm, will not be economically justified. Like other inputs, there is an optimal level of management for each farm depending on its cost and returns. A lower level of economically justified management on small, part-time farms along with lower quality land can explain their apparent lower level of technical efficiency, and apparent scale economies. This does not mean that off-farm work of small farm operators is socially wasteful. As long as what is gained from off-farm work exceeds what is given up on the farm, which must be the case if willingly done, there is a net gain to society by off-farm work. This is not to say that managerial ability, however defined, of small farm operators is equal to those who operate large commercial operations. Presumably large farms are more difficult to manage than small ones. In the long run, a manager will increase the size of the farm until the 4 long run average total cost curve begins to turn up. For better managers this quantity will be larger than it is for their less skillful counterparts. Moreover, one would expect the payoff to investment in acquiring management skills to be greater among capable managers of large farms than it is for their less capable, small farm counterparts. More investment in human capital will take place by those who reap a higher return. The less skillful managers can maximize income by keeping the farm small while supplementing earnings from off-farm employment where management decisions are made by someone else. Family background also makes a difference. The operator who takes over or inherits a family farm consisting of 1000 acres of prime cropland is more likely to end up with a large commercial farm then one whose parents owned 160 acres of marginal land. The latter will more likely be a part-time farmer, at least if jobs are available within commuting distance. Both could be of equal managerial ability. Where one ends up depends a lot on where one starts. Data and Methodology Agricultural census data, Volume II, "Government Payments and Value of Sales," (1987) are employed to measure efficiency differences among size classes. Nine sales classes are reported for each state ranging in size from $500,000 and above to $2500 and below. These data are well suited for measuring returns to scale because of the large variation in farm size. Also, cross-section data depict long run relationships and returns to scale is a long-run concept. Ten Corn Belt states are included in the sample making a total of 90 observations: IL, IN, IA, MI, MN, MO, NE, OH, SD, and WI. Data are from the 1987 agricultural census. Because 1987 was not an atypical year for prices or production, abnormal costs and/or inventory changes 5 should not be major problems, especially among size groups. Returns to scale are defined by the long run average total cost (LRATC) curve. A negative relationship between LRATC and output implies economies of scale while a positive relationship implies diseconomies. Because farms produce and sell a variety of products, output is measured in monetary terms. Thus LRATC is dollar(s) per dollar of output. The data are available for nine size classes based on sales. Each size class is represented by a 0-1 dummy; the middle size is the reference dummy. This allows the estimated LRATC curve to take on any shape dictated by the data. Use of continuous data with a specific functional form can yield results contingent on that form. The flexible dummies allow for any shaped LRATC curve including a horizontal line, a ski-jump shaped line, a U-shaped curve, or any combination of these. The LRATC curve is estimated by the following equation: LRATC = f(S ,...S , S ...S ,X , X ) 1 4 6 9 1 2 Size group S is the smallest ($2500 sales and less) and S is the largest ($500,000 sales 1 9 and over). Variables X and X represent two characteristics defined later, that can influence 1 2 LRATC other than size. Size Characteristics To facilitate presentation of the data in Table 1, the nine size classes are aggregated into three groups with three classes in each. The large group includes farms with $100,000 of sales and over, while the small group's farms have sales of less than $10,000. Although no direct information is available on land quality and topography in the census, indirect evidence suggests that large and small farms operate in different geographic 6 environments, at least in the corn belt. As shown in Table 1, small farms have a smaller proportion of land in crops and exhibit lower yields. Because of the relatively small percent of land in harvested crops, small farms appear to be located in areas where woods, hills, and sloughs are more common. Generally these correlate with lighter forest soils and a topography of hills and small, odd-shaped fields. In contrast, areas where cropland is a large percent of the total tend to be characterized by relatively level or, gently rolling terrain, heavier prairie soils, and large square or rectangular fields, at least in the states included in this study. Land quality and topography are expected to influence unit production costs. The per acre cost of tillage, planting, weed control, and harvesting does not vary in proportion to yields. Thus high quality land will entail lower costs per unit of output., Moreover, there is less travel and turning time per acre when fields are large and rectangular shaped. This also reduces unit costs. Because of the positive correlation of farm size with soil fertility and favorable topography for crop production, part of what appears to be economies of scale can instead be economies of fertile soil and large, level fields. Cropping patterns between large and small farms also differ. Corn and soybeans account for about two-thirds of harvested cropland on large farms, and only about one-third on small farms. The proportion of harvested cropland in small grains and hay is the opposite--almost twothirds on small and one-fourth on large farms. Since corn and soybeans are high value crops, the difference in cropping patterns can make it appear that large farms are more productive. In part, the difference in cropping patterns may stem from difference in land quality. In spite of the likelihood of higher quality land on large farms, per acre value of land and buildings is highest on small farms (Table 1). This could be due to the higher proportion of the 7 house in total assets of small farms. Also one might expect small farms to be located closer to population centers where off-farm employment is more readily available but where land prices are relatively high. Higher land prices that are unrelated to land quality for agricultural purposes also will make small farms appear less efficient. Information on days of off-farm work is available for principal operators only. Income from this source was estimated form manufacturing wages in each state. Although income from off-farm work typically is not added to the farm's output or sales, the separation of the two is not strictly correct. Because part-time farmers have less time to carry out non-routine tasks such as repairs and maintenance, animal health care, and harvesting, the purchase of professional services to perform such tasks will be more prevalent among small farms. These inputs add to the costs but off-farm earnings that require and help pay for these services are deleted from measured output, again biasing the results in favor of economies of scale. This phenomenon is similar to homemakers who work outside the home and use part of their earnings to purchase child care services, labor-saving appliances, convenience foods, and restaurant meals. The fact that repairs and maintenance expense per dollar of sales is nearly four times higher on small than large farms supports the contention that the latter rely more heavily on purchased services. Part of this difference could be due to the use of older machinery by small farms. By purchasing lower cost used machines and spending proportionately more on repairs and maintenance, small farms are able to convert fixed cost to variable cost, thereby diminishing the importance of lumpy inputs as a factor causing scale economies. Considering this added element of flexibility, machines can be purchased in an array of fixed and variable costs.