Do Production Contracts Raise Farm Productivity? an Instrumental Variables Approach
Key, Nigel, McBride, William D., Agricultural and Resource Economics Review
Estimating how the use of production contracts affects farm productivity is difficult when unobservable factors are correlated with both the decision to contract and productivity. To account for potential selection bias, this study uses the local availability of production contracts as an instrument for whether a farm uses a contract in order to estimate the impact of contract use on total factor productivity. Results indicate that use of a production contract is associated with a large increase in productivity for feeder-to-finish hog farms in the United States. The instrumental variable method makes it credible to assert that the observed association is a causal relationship rather than simply a correlation.
Key Words: productivity, production contracts, instrumental variables, sample selection
Production contracts, in which the contractor exercises control over some production decisions and farmers are paid a fee for services rendered, are becoming increasingly common in the United States (MacDonald et al. 2004). In the U.S. hog industry, growth in use of production contracts has been particularly rapid: the share of all hog production under contract increased from about 5 percent in 1992, to about 40 percent in 1998, to 67 percent in 2004 (Key and McBride 2007). Production contracts offer several potential advantages over independent production that could explain their growing use: contracts can reduce information asymmetries between growers and processors, improve coordination and timing of product delivery, and lower income risk for growers. In addition to these benefits, production contracts could raise farm productivity by improving the quality of farm management decisions or speeding the transfer of technical information to growers (e.g., by requiring growers to attend training courses or follow integrator-provided procedures), improving growers' access to credit, and facilitating the adoption of more efficient technologies.
Using data from a 1998 survey, Key and McBride (2003) found that production contracts were associated with greater hog farm productivity. However, between 1998 and 2004 the number of hog farms in the United States fell from 113,830 to 69,500 (USDA, various issues), the average number of finished hogs removed per farm increased from 2,590 to 4,650, and the share of output under contract increased from 40 percent to 67 percent (Key and McBride 2007). Given these recent structural changes, it is reasonable to suppose that since 1998 a number of less efficient independent operations have exited, and that some independent operations have begun to contract (and therefore experienced a resulting increase in productivity), so that the productivity of the surviving independent operations increased. Hence, it is an open question whether production contracts continue to be associated with greater farm productivity, and if so, whether this effect has diminished. How contracts affect farm productivity will be an important factor influencing the future organizational decisions of farmers. Understanding how contracts affect productivity is important for evaluating legislative efforts to regulate contract production.
To measure the effect of contracting on farm productivity it is necessary to control for differences between those farmers who use contracts and those who produce independently. Farmers who contract might have larger operations, tend to be located in certain regions, be more creditconstrained, be more risk-averse, value autonomy less, or have less managerial or entrepreneurial ability-characteristics that could be positively or negatively correlated with farm productivity. Unfortunately, some factors correlated with the decision to contract and with farm productivity are not observable. Consequently, a simple regression of productivity on exogenous factors and a contract indicator could produce biased parameters. For example, if "management ability" is unobservable and negatively correlated with contracting but positively correlated with productivity, then an OLS regression would underestimate the effect of contracting on productivity. …