Mobile Edition
Print
Get the Mag
Weekly UpdatesFarm households often receive income from off-farm employment and government support payments in addition to farm production revenues. If income is fungible, then marginal consumption should not vary by income source. However, Carriker et al. (1993) show that farm household consumption does vary at the margin by income source. They find that off-farm income and income from government subsidies are consumed at a higher marginal rate than income from farm production revenues. If farm household consumption varies at the margin by income source, different types of government payments may affect consumption differently. Closely following Carriker et al. (1993), this analysis further disaggregates sources of farm household income and compares how different types of government payments affect farm household consumption. It is shown that government payments that are relatively "decoupled" from farm production decisions are consumed at a much higher marginal rate than government payments that vary according to market conditions. (1)
Since the study by Carriker et al. (1993), the structure of U.S. agricultural policy has undergone significant changes. One change in particular was the creation of a new type of government subsidy under the 1996 Federal Agricultural Improvement and Reform (FAIR) Act intended to be "decoupled" from (independent of) farm production decisions. The purpose of this payment was to increase the consumption, savings, and investment of farm households without affecting farm production decisions (Burfisher and Hopkins 2003: Westcott and Young 2004). These payments continued under the 2002 Farm Security and Rural Investment (FSRI) Act and are included in the Food, Conservation, and Energy Act of 2008. They are paid to qualifying farm households with few restrictions as to what or how much they produce. Other types of government payments are more volatile, being paid only in times of poor agricultural production or low commodity prices. Still others are paid for the conservation of agricultural land that is environmentally sensitive. If these different types of payments affect farm household consumption in different ways, some payments may be better suited to achieving a specific policy objective than others.
Data for the analysis are taken from the U.S. Department of Agriculture's (USDA) Agricultural Resource Management Survey (ARMS) for the years 1998 to 2004. (2) Because ARMS data do not observe individual farm households over time, a pseudo-panel of data is created that observes groups of relatively homogeneous farm households over time. An empirical analysis, based on the theoretical model developed by Carriker et al. (1993), employs advances by McKenzie (2004) in the econometric theory of dynamic pseudo-panel analysis.
Sources of Farm Household Income
Farm households rely on a variety of revenue streams. The instability of income from farm production leads total farm household incomes to vary more from year to year than the incomes of other U.S. households (Mishra et al. 2002). Income from farm production is volatile over time due to changing weather conditions, fluctuations in agricultural prices, changes in the price of production inputs, or changes in farm size. Besides revenue from farm production, most farm households receive income from off-farm and government sources. Nearly 76.4% of all family farms have at least one person, either the principal operator, spouse, or both, laboring off of the farm for a wage or salary (Hoppe and Banker 2006). Farm households may also receive unearned income from off-farm financial or business investments.
Farm households receive billions of dollars in U.S. government subsidies every year from a variety of policy instruments. One type of payment is largely decoupled from production or commodity price outcomes. (3) These decoupled payments are tied to agricultural land (called "base" acres under the 2002 FSRI Act) that has a history of producing certain field crops: wheat, rice, corn, sorghum, barley, oats, upland cotton, soybeans, peanuts, and other oilseeds. They are paid to the operators of base acreage generally without regard to what crops they produce in the current period, or even if they produce any crops at all. (4) They are also paid independent of commodity prices in the current period. For farm households, decoupled payments are fixed by law and relatively stable between periods of farm legislation, varying only when base acres are sold, rented, or taken out of agriculture (Burfisher and Hopkins 2003; Westcott and Young 2004).
Some agricultural subsidies are paid out only under unfavorable market conditions. Marketing loan programs such as the loan deficiency payment (LDP) and marketing loan gains (MLG) essentially provide a price floor when market prices fall below legislated per-unit loan rates (Westcott, Young, and Price 2002). In an attempt to reduce farm household income variability, these payments are designed to supplement farm household income in years when revenue from farm production is low due to low commodity prices.
Various types of environmental conservation program payments are also available to help farmers address specific environmental problems. These payments are made mostly to smaller rural-residence and retired farm operators. The bulk of these are Conservation Reserve Program (CRP) payments to farm operations that remove land from production for a period of ten to fifteen years (Lambert et al. 2006). Other payments include Wetland Reserve Program (WRP) payments and Environmental Quality Incentives Program (EQIP) payments.
Model
There are several models that demonstrate the relationship between consumption and income, including the permanent income (Friedman 1957), life-cycle (Ando and Modigliani 1963), and behavioral life-cycle (Sheffrin and Thaler 1988) hypotheses (PIH, LC, and BLC, respectively). This article uses a theoretical model set forth by Carriker et al. (1993), who modify the traditional LC model to allow for different sources of income to account for a fixed percentage of total consumption. They show that if incomes are fungible, the LC model can be written as
(1) [C.sub.t] = [[beta].sub.0] + [[beta].sub.1] [z.summation over (s=1)] [Y.sub.st] + [[beta].sub.2][C.sub.t-1] + [[beta].sub.3] [W.sub.t]
where [Y.sub.st] is income in time period t from source s and [[beta].sub.1] is the short-run marginal propensity to consume income from all sources. The variable C is consumption and W is a measure of wealth.
Carriker et al. (1993) state that this specification of consumption is incorrect if incomes are not fungible. They create a system of consumption equations in which a fixed percentage of consumption is assigned to each source of income [Y.sub.st].
(2) [[lambda].sub.s][C.sub.t] = [[beta].sub.0s] + [[beta].sub.1s][Y.sub.st] + [[beta].sub.2s][[lambda].sub.s][C.sub.t- 1] + [[beta].sub.3s] [W.sub.t].
Equation (2) is a representative equation from this system where s indexes one of z sources of income. The share of total consumption ([C.sub.t]) purchased with income from source s is [[lambda].sub.s], where [[summation].sup.z.sub.s=1] [[lambda].sub.s] = 1.
Carriker et al. (1993) also state that because [[lambda].sub.s] is unknown, individual equations for each source of income cannot be empirically estimated. By summing across the s equations, they derive equation (3)
(3) [z.summation over (s=1)] [[lambda].sub.s][C.sub.t] = [C.sub.t] = [z.summation over (s=1)] ([[beta].sub.0s] + [[beta].sub.1s][Y.sub.st] + [[beta].sub.2s][[lambda].sub.s][C.sub.t-1] + [[beta].sub.3s][W.sub.t])
which can be rewritten as
(4) [C.sub.t] = [[beta].sup.*.sub.0] + [z.summation over (s=1)] ([[beta].sub.1s][Y.sub.st]) + [[beta.sup.*.sub.2][C.sub.t-1] + [[beta].sup.*.sub.3][W.sub.t]
where [[beta].sup.*.sub.0] = [[summation].sup.z.sub.s=1] [[beta].sub.0s], [[beta].sup.*.sub.2] = [[summation].sup.z.sub.s=1] ([[beta].sub.2s][[lambda].sub.s]), and [[beta].sup.*.sub.3] = [[summation].sup.z.sub.s=1] [[beta].sub.3s]. Equation (4) is an estimable equation in which the short-run marginal propensity to consume income from source s ([[beta].sub.1s]) is preserved for all z sources of income.
Empirical Considerations
Data on farm household income, wealth, and expenditures are taken from the ARMS survey. This survey is conducted each year by the USDA's National Agricultural Statistics Service (NASS) and collects data from thousands of farm operators. The ARMS data set does not constitute a true panel of data. Rather, it is a cross-sectional survey repeated each year where observations present in one year are not known to be present in other years. A pseudo-panel of data is therefore constructed by creating cohorts of observations within each cross-section whose characteristics are unlikely to change over time (Deaton 1985; Verbeek and Nijman 1992). These cohorts are treated as cross-sections that are observed over time, using cohort means as the relevant observation data points. Because equation (4) contains a lagged dependent variable, the estimation of a dynamic model using a pseudo-panel of data must be considered. Relatively few theories have been developed to address the issues involving the estimation of dynamic pseudo-panels (Moffit 1993; Collado 1997; McKenzie 2004; Verbeek and Vella 2005). This article follows McKenzie (2004).
Because ARMS data do not observe the same farm operation over time, each operation i is indexed with a time period t to indicate the period in which the individual farm operation is observed. Following McKenzie (2004), equation (4) is rewritten in its econometric form as
(5) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
where [u.sub.i(t),t] is the error term. The intercept term ([[beta].sup.*.sub.0i(t),t]) for each farm operation is assumed to be normally distributed around some cohort mean, or [[beta].sup.*.sub.0i(t)] = [[beta].sup.*.sub.0c] + [[mu].sub.i(t)], where farm operation i(t) belongs to cohort c(t) and no other cohort. The cohort intercept term ([[beta].sup.*.sub.0c]) is specific to each cohort and is time invariant. The individual variation from the cohort mean ([[mu].sub.i(t)]) represents individual farm operation fixed (time invariant) effects.
FEED