Content uploaded by Steven C. Blank
Author content
All content in this area was uploaded by Steven C. Blank on Oct 24, 2017
Content may be subject to copyright.
Profit and Productivity Patterns from the Farm-Gate to the Global Market-
place: Implications for American Agricultural Competitiveness (Steven C.
Blank, University of California at Davis, presiding)
AGRICULTURAL PROFITS AND FARM
HOUSEHOLD WEALTH
STEVEN C. BLANK,KENNETH W. ERICKSON,CHARLES B. MOSS,
AND RICHARD NEHRING
Recent decades (especially since 1973) have
been an era of decreasing production prof-
its that threaten the survival of many mid-
and small-sized American farms (Blank 2003).
Normally, the survival of a firm depends on
its profitability, both in absolute and relative
terms. To remain viable, a firm must offer
returns that are both sufficient to cover the
owner’s financial obligations and competitive
with returns from alternative investments. If
a firm is profitable, the wealth of its owners
can increase over time. An unprofitable firm,
on the other hand, reduces owners’ wealth.
Yet, American agriculture is full of firms
that routinely earn low or negative returns
on equity from production operations (Blank
2002), thus complicating the evaluation of
the industry’s economic health and prospects.
This suggests that macro-level forecasts of
American agriculture’s future structure and
performance require a micro-level under-
standing of the relationship between farm
profits and owner wealth. This paper addresses
that relationship.
Steven C. Blank is extension economist, Agricultural and Re-
source Economics Department, University of California, Davis,
and member, Giannini Research Foundation. Kenneth W. Er-
ickson is economist, Farm Sector Performance and Well-Being
Branch, U.S. Department of Agriculture, Economic Research Ser-
vice. Charles B. Moss is professor, Food and Resource Economics
Department, University of Florida. Richard Nehring is economist,
U.S. Department of Agriculture, Economic Research Service.
Theauthors thank the reviewers at the USDA-ERS for their
input, and Charles Hallahan of the USDA-ERS for his invaluable
assistance in estimating the ARMS pseudo-panel equations. The
views expressed here are not necessarily those of the Economic
Research Service, U.S. Department of Agriculture.
This article was presented in a principal paper session at the
AAEA annual meeting (Denver, Colorado, August 2004). The ar-
ticles in these sessions are not subjected to the journal’s standard
refereeing process.
Assessing financial stress within American
agriculture involves identifying which groups
are more or less profitable. It also involves
assessing farmers’ well-being in the context
of income, wealth, and consumption at the
household level (Mishra et al.). Previous stud-
ies (e.g., Dodson) raise expectations of prof-
itability differences due to resources available
(and quality) across locations, economies of
scale across farm sizes, and supply/demand
differences across commodity markets caused
by comparative advantage (i.e., competitive-
ness) issues. However, economic theory says
that returns converge over time as resources
flow into more profitable industries and out of
less profitable industries, causing factor price
changes (O’Rourke and Williamson, Caselli
and Coleman). Both traditional growth and
trade theories say factor markets will ad-
just to equalize commodity returns over time
(Andres, Bosca, and Domenech; Ben-David;
Gutierrez; Schott).
Assessing farm owner–operator wealth in-
volves understanding that farmers are making
production decisions based on total household
wealth, not just on farm production profitabil-
ity (Carriker et al., Schmitt). We want to ex-
plain wealth patterns across regions, farm sizes,
and commodity specialization to derive in-
sights into the future financial prospects for
American agriculture as a whole, or at least
for some agricultural industries. We also want
to test the relationships between farm size and
productivity, and productivity and profitability.
There are three general objectives of the pa-
per. The first objective is to determine whether
or not there is convergence of rates of return
on farm assets across states over time. The sec-
ond objective is to derive a system of equations
Amer. J. Agr. Econ. 86 (Number 5, 2004): 1299–1307
Copyright 2004 American Agricultural Economics Association
1300 Number 5, 2004 Amer. J. Agr. Econ.
that explains interlinkages between the vari-
ous components of a farm household’s wealth
at some point in time. The third objective is
to use those equations to empirically assess in-
come and wealth patterns across regions, farm
sizes, and commodity specializations.
Theoretical Relationship between Income
and Wealth
The three components of income (i.e., eco-
nomic gains) contributing to wealth are profits
from farm output, off-farm income, and capital
gains on assets. Total “wealth” (W)isusually
expressed as equity at time t:Wt=Wt−1+
Wt.“Wealth changes” during a time period
ending at tequals “farm income” (FInc) plus
“off-farm income” (OFInc) plus some function
of “capital gains” (K) minus “consumption”
(C), or
Wt=FInct+OFInct+Kt−Ct.(1)
In this regard, capital gains (even unrealized
gains) immediately improve a farmer’s abil-
ity to borrow, and thus they aid in financing
a larger operation.
There are at least four components of wealth
changes. Those components, on the right-hand
side of (1), are themselves functions of other
factors:
FInct=Rt−PCt−OKt
(2)
OFInct=Salt+Invt
(3)
Kt=(Kt−Kt−1)LTVtand
=f(LTV)
(4)
and
Ct=CLt+QLt.(5)
Each of these four equations is explained be-
low, beginning with farm income.
For this analysis of American farms, farm
income comes from three sources. Only total
revenues from farmers’ and ranchers’ sales of
production output (R) are considered part of
farm income; government transfers and other
nonfarm income sources are excluded to test
the true sustainability of farm production as an
income source. Yet, to many farm households,
government payments may be significant. This
will be captured in an error term in equations
(1) and (2). In equation (3), government pay-
ments could come from various sources, such
as unemployment benefits. Therefore, the in-
dustry’sfarm income from all agriculturalcom-
modities (i=1, 2, ...,n)attime tis expressed
in (2), where
Rt=
n
i=1
PitQit
(2a)
Qit =Yit Ait
(2b)
PCt=ijpcjt xijt
(2c)
OKt=ihokht ziht
(2d)
and Pi,pcj,okh>0; Yi,Ai,xj,zh≥0. The num-
ber of commodities produced by the industry
is n.The average unit price of commodity iat
time tis Pit.The quantity of commodity ipro-
duced during the period ending at time tis Qit.
Yiis the average yield per acre of commodity
i.Aiis the total acreage devoted to commod-
ity i.PCtis the total production costs of all
commodities at time t. Unit costs of jvariable
inputs is pcj. Quantities of jvariable inputs to
be applied in the production of commodity i
is xij.OKtis the total ownership costs of all
commodities at time t. Unit costs of hcapital
inputs (land, improvements, equipment, etc.)
is okh. Quantities of hcapital inputs used in
the production of commodity iis zih.
Equation (3) states that off-farm income
for a period ending at time tconsists of the
sum of off-farm salary or wages earned (Sal)
and nonfarm investment or “unearned” in-
come (Inv). Off-farm employment is the pri-
mary source of nonfarm income for a major-
ity of farm and ranch households, representing
over 90% of average farm household income
in recent years (Mishra et al.). Nonfarm invest-
ment includes income sources such as interest
income on savings, Social Security and other
retirement benefits, and capital gains and div-
idends on nonfarm assets such as stocks and
bonds.
Equation (4) specifies how farmers’ change
in wealth is influenced by capital gains. To be-
gin, capital gains are simply the change in value
of a farmer’s capital from one period to the
next (i.e., Kt−Kt−1). Capital gains are only
realized if the asset is sold. However, some
portion of unrealized capital gains can be used
to improve a farmer’s operation. Lenders will
usually loan a farmer up to some specific por-
tion of the market value of the assets, referred
to as the “loan-to-value” ratio (LTV). In (4),
is an estimate of how much of unrealized capi-
tal gains are immediately converted into cash,
Blank et al. Agricultural Profits and Farm 1301
and is assumed to be a function of the LTV. In
all cases, 1 >LTV ≥≥0.
The capital variable (K)in(4) can be ex-
pressed as the sum of the market values for all
assets (real estate, nonreal estate, and nonfarm
assets) held by a farm at time t,
Kt=LV
t+MV
t+FV
t
(4a)
where LV is the “value of land” and improve-
ments (buildings, irrigation systems, etc.), MV
is the “value of nonreal estate assets” (e.g.,
machinery and other equipment), and FV is
the “value of nonfarm financial assets” (stocks,
bonds, etc.). Farmland has historically repre-
sented about 75% of assets held by farm house-
holds. Also, farmland values vary much more
than do the other agricultural assets because
they are a function of numerous variables
(Drozd and Johnson). A simple model for the
expected price of farmland can be specified
as
E(LV
ft)
=E(Rft −CKft +TFPft +Dft ).
(4b)
An empirical version of (4b) is
LV
ft =1Rft −2CKft +3TFPft
+4ft +vf+εt
(4c)
where LVft is the value of farmland and build-
ings in state or farm fat time t,Rft is the cash
flow (revenue, as specified in (2a)) from agri-
cultural production in state or farm fat time t,
CKft is the average “cost of capital” at time t,
TFPft is a technology variable (state-level “to-
tal factor productivity”), we also use a farm-
level estimate of “productivity” (PROD), the
“population density” (people per acre) in
county or state fat time tis Dft,is a coef-
ficient to be estimated, and vfand εtare errors
for, respectively, state or farm for for time tif
a random effects model is used.
Equation (5) specifies farm household con-
sumptionduring a period endingat time tas the
sum of the basic cost of living (CLt), such as the
cost of providing a minimum level of food and
shelter to members of the household, and the
extra expenditures made by household mem-
bers to raise the quality of life to the desired
level (QLt).
Industry sales and profit totals are simply
the sum of results from decisions made by the
individual firms that constitute the industry.
In American agriculture, individuals are as-
sumed to make production decisions based on
the goal of maximizing expected profits. This
study follows Klepper in recognizing that re-
sults are influenced by both the innovation ex-
pertise and capital available within an indus-
try. Thus expected profit, for firm fat time t,is
specified as
E(ft)=E[Rft −PCft −OKft
+(mf)g(crft )G−(crft)]
(6)
where R,PC, and OK are defined as above,
but are for firm fonly. E(·)isthe expected
value of (·). The innovation expertise of firm
fis denoted mfand influences the firm’s suc-
cess at improving productivity. The probabil-
ity of firm fimproving its productivity in pe-
riod tis (mf)g(crft), where crft is defined as the
firm’s cumulative investment in human capi-
tal and productive resources through time t
and is some function of profits earned in all
prior periods. The function g(crft ) reflects the
opportunities for improving productivity. The
potential increase in profits earned by an in-
novation that improves productivity is G.This
can result from either reduced input costs per
unit (PC/Q and/or OK/Q)orincreased rev-
enue from a higher yield (Y). Gis defined
to equal (Rft −PCft −OKft )−(R∗
ft −PC∗
ft −
OK∗
ft), where the asterisk indicates a value
that would have existed for firm fin period t
without the innovation. The change in cumula-
tive investment during period t[(crft)] equals
crft −crft−1, and it is constrained to be ≥0.
A firm’s expected sales revenues are
E(Rft)=Rft−1
+E[(mf)g(crft )G+(crft)].
(7)
Current revenues are expected to equal the
previous year’s revenues plus expected im-
provements from a productivity component
[(mf)g(crft)G] and an investment component
[(crft)].
Procedure
We use state-level data from the USDA Eco-
nomic Research Service (ERS) for 1960–2002
to test for convergence of rates of return
on farm assets and over time. Next, we use
farm-level data from the USDA’s Agricul-
tural Resource Management Survey (ARMS)
tohelp explain the inter-linkagesbetween farm
household wealth, returns, and productivity
(USDA). We construct a unique pseudo-panel
data set from pooled ARMS data for 1996–
2002 over three regions: the Lake States, the
1302 Number 5, 2004 Amer. J. Agr. Econ.
Corn Belt, and the Southeast. These regions
were selected to represent production sectors
dominated by dairy and wheat, corn, and spe-
cialty crops, respectively. Then, we estimate
the equations using a two-way fixed effects
approach (Baltagi, Chapter 3), examining fac-
tors affecting profitability and the change in
wealth across regions.
Model of Convergence
In this study, we determine whether the rate
of return on agricultural assets is converging
across regions. A typical formulation of con-
vergence (Sala-i-Martin) can be expressed as
ln yit
y∗t=0+1ln yi,t−1
y∗,t−1
+2zit +εit
(8)
where ln (·) denotes the natural logarithm, yit is
the level of income per capita in region or state
iin time t,y∗tis the index income per capita at
time t,zit is a vector of other economic vari-
ables (such as initial capital) in region or state
iat time t,εit is an error term, and 0,1, and 2
are estimated coefficients. In this formulation,
if 0→0,1<1, and 2=0, the income in re-
gion iconverges over time toward the income
of the index. Further, this convergence is un-
conditional, or does not depend on other vari-
ables (such as initial capital). The convergence
is conditional if 0→0,1<0, and 2= 0.
Given that the rates of return on agricul-
tural assets are stationary (results not reported
here), we reformulate convergence in (8) into
ln(yit )−ln(y∗t)
=0+1[ln(yi,t−1)−ln(y∗,t−1)]
+2zit +εit
dit =0+1di,t−1+2zit +εit
(9)
where dit is the logarithmic difference between
returns in state iand the index state at time t.
Since the rate of return data for agricultural as-
sets in (9) are stationary, convergence can be
estimated directly. Unfortunately, the formula-
tion in (9) cannot be directly applied to agricul-
tural returns because negative rates of return
are sometimes observed in the data. Thus, we
redefine (9) so that dit =r∗t−rit, where r∗tis
the maximum rate of return to agricultural
assets in each of ERS’s 10 regions. Finally,
we estimate 0and 1in (9) using maximum
likelihood.
Estimation of Income and Wealth Patterns
Ideally, one would like to take repeated cross-
section surveys of U.S. farm households over
time. However, it is impossible to track the
same farm household over time. Instead, we
construct a pseudo-panel data set. For empiri-
cal studies using such panel data, the temporal
pattern of a given farm’s production behavior
must be established. In the absence of genuine
panel data, repeated cross-sections of data
across farm typologies may be used to con-
struct pseudo-panel data (Deaton, Verbeek
and Nijman). A pseudo panel is created by
grouping individual observations into homo-
geneous cohorts, distinguished according to
time-invariant characteristics such as fixed as-
sets, geographic location, or land quality. The
empirical analysis is then based on the cohort
means rather than the individual farm-level
observations.
We assigned the farm-level data to cohorts,
based on the ERS farm typology (TYP) groups
(Hoppe and MacDonald). A cohort group is
formed for each state in the sample. There are
thirteen cohorts per state and fourteen states,
resulting in a total of 182 cross-sectional en-
tities per year. We refer to these entities as
“firms.”
The problem when using time series and
cross-sectional data is to specify a model that
will adequately allow for differences in behav-
ior over cross-sectional units as well as for
differences in behavior over time for a given
cross-sectional unit. Fixed effects regression is
a method of controlling for omitted variables
in panel data when the omitted variables vary
across“firms” but do not changeover time.The
fixed effects regression model has ndifferent
dummy intercepts, one for each firm.
To test if some omitted variables are con-
stant over time but vary across regions, while
other variables are constant across states but
vary over time, we include both location and
time effects. This is done by including both
n−1 state binary variables and T−1 time
binary variables in the regression, plus an in-
tercept. The combined time and firm fixed re-
gression model is
yit =0+i+Xit +vi+εt
(10)
where Xit is a vector of other variables to be
estimated (from (4c)). This model has an over-
all constant term as well as a “group” effect
for each group and a “time” effect for each
time period. The combined time and state fixed
effects regression model eliminates omitted
Blank et al. Agricultural Profits and Farm 1303
variables bias arising from unobserved vari-
ables that are constant over time and/or con-
stant across states.
We estimated reduced forms of equa-
tions (1), (2), (4c), and (6) for the three
regions using unbalanced panels. The equa-
tions were estimated by ordinary least squares
(OLS) since these equations constitute a re-
cursive system (Baltagi; Greene, p. 659). If
we take the i’s to be identical across loca-
tions, OLS provides consistent and asymptoti-
cally efficient estimations of and (Greene,
pp. 560–62).
Table 1. Estimated Autoregression Coefficients for Difference in Rate of Return on Assets,
1960–2002
State 10State 10
Connecticut 0.427 0.039 Kentucky 0.904 0.066
(0.140) (0.008) (0.062) (0.029)
Maine 0.165 0.054 Tennessee 0.946 0.093
(0.154) (0.007) (0.044) (0.040)
Maryland 0.548 0.038 Virginia 0.931 0.087
(0.131) (0.007) (0.050) (0.032)
Massachusetts 0.663 0.050 West Virginia 0.413 0.159
(0.114) (0.013) (0.141) (0.027)
New Hampshire 0.299 0.106 Alabama 0.338 0.018
(0.150) (0.023) (0.147) (0.003)
New Jersey 0.557 0.049 Georgia 0.485 0.002
(0.128) (0.009) (0.137) (0.004)
New York 0.722 0.051 South Carolina 0.629 0.030
(0.105) (0.014) (0.131) (0.008)
Pennsylvania 0.684 0.072 Louisiana 0.210 0.015
(0.111) (0.013) (0.153) (0.003)
Rhode Island 0.626 0.038 Mississippi 0.327 0.015
(0.119) (0.014) (0.147) (0.003)
Vermont 0.676 0.044 Oklahoma 0.466 0.006
(0.112) (0.014) (0.139) (0.002)
Michigan 0.191 0.030 Arizona 0.880 0.009
(0.153) (0.004) (0.075) (0.017)
Wisconsin 0.449 0.014 Colorado 0.880 0.014
(0.138) (0.005) (0.085) (0.015)
Illinois 0.365 0.014 Montana 0.824 0.022
(0.145) (0.003) (0.096) (0.014)
Indiana 0.297 0.023 Nevada 0.755 0.039
(0.149) (0.004) (0.102) (0.009)
Missouri 0.646 0.042 New Mexico 0.743 0.018
(0.118) (0.006) (0.101) (0.009)
Ohio 0.358 0.045 Utah 0.853 0.041
(0.147) (0.004) (0.082) (0.012)
Kansas 0.576 0.012 Wyoming 0.877 0.035
(0.127) (0.004) (0.078) (0.018)
North Dakota 0.516 0.011 Oregon 0.705 0.048
(0.133) (0.007) (0.108) (0.005)
South Dakota 0.750 0.003 Washington 0.557 0.011
(0.103) (0.007) (0.128) (0.004)
Note: Numbers in parenthesis denote standard deviations and all numbers are rounded to the third decimal.
Empirical Results
Convergence
The empirical results of the convergence
model in (9) are presented in table 1. Based on
the data for 1960–2002, the state with the high-
estrate of return withineachregion was chosen
as the index state (y∗in equation (9)). Follow-
ing this criterion, we use Delaware: Northeast,
Minnesota: Lake States, Iowa: Corn Belt,
Nebraska: Northern Plains, North Carolina:
Appalachia, Florida: Southeast, Arkansas:
1304 Number 5, 2004 Amer. J. Agr. Econ.
Table 2. Regression Results for Farm Income and Farmland Value Equations by Region: Lake
States, Corn Belt, and Southeast (1996–2002)
Lake States Corn Belt Southeast
Variable Estimate t-Value Estimate t-Value Estimate t-Value
Farm income equation
CashFlow 0.6300 10.41∗0.1445 5.36∗0.0935 2.48∗∗
TotalCashExpenses −0.4124 −5.82∗−0.1232 −5.07∗−0.0918 2.22∗∗
Depreciation −0.3212 0.32 −1.0920 −4.83∗−1.1513 −8.98∗
Fixed effects
Firm NS ∗NS
Year NS NS NS
Farmland value equation
CashFlow 0.0961 7.88∗0.0367 4.73∗0.1558 3.35∗∗
CostCapital 0.2414 2.03∗∗ −0.1301 −0.84 0.0730 0.36
Productivity −17.9303 −1.42 −19.4439 −3.12∗∗ −8.2662 −1.45
PopDensity −3.1103 −0.22 4.8009 1.24 0.2555 0.11
Fixed effects
Firm ∗∗∗∗
Year NS NS NS
∗and ∗∗ denote statistical significance at the 0.01 and 0.05 confidence levels. NS denotes “not significant.”
Delta, Texas: Southern Plains, Idaho: Moun-
tain region, and California: Pacific States. In
general, convergence occurs if 1is less than 1,
implying that the difference between the rate
of return for a particular state and the regional
index declines over time. The results indicate
that all the rates of return to agricultural as-
sets converge over time in all regions except
Appalachia. Within the Appalachian region,
we fail to reject 1=1atthe 0.05 confidence
level for Kentucky, Tennessee, and Virginia.
Thus, at least conditional convergence for the
rate of return on agricultural assets is sup-
ported in all regions except Appalachia.
To test unconditional convergence, we next
examine convergence between each of the in-
dex states. North Carolina, the state with the
highest average returns over the period, is used
to normalize the index states for each region.
Again, the estimated autoregression coeffi-
cient for each region is less than 1 at any con-
ventional level of statistical significance. Thus,
we conclude that the rates of return on agri-
cultural assets are converging across regions.
Farm Income, Land Values, Wealth,
and Profits Equations
The regression results were generally “best”
for the Lake States and Corn Belt since coef-
ficients’ signs were generally consistent with
economic theory, and levels of significance
were high, especially for the farm income
equation.
Farm income (equation (2)). The results in
the top section of table 2 show some differ-
ences across regions. CashFlow and TotalCash-
Expenses were significant in all three regions,
but with varied coefficients. Depreciation was
not significant in the Lake States, possibly in-
dicating farm structures with relatively greater
fixed assets. Firm fixed effects were significant
in the Corn Belt, indicating that other firm-
related variables also possibly affecting farm
income (such as commodities produced) are
omitted.
Farmland value (equation (4c)). CashFlow
was significant in all three regions (bottom sec-
tion of table 2). This is consistent with the ex-
pectation that land value is determined pri-
marily by its ability to generate agricultural
revenues. The Productivity variable was only
significant in the Corn Belt. This may be due to
the more heterogeneous nature of operations
in the Southeast and Lake States.
Change in wealth (equation (1)). Wealth
consists of both farm and nonfarm capital, al-
though most farm household wealth is held
in the form of farmland. Both components
were highly significant in the combined three-
region area when examining changes in farm
wealth across farm sizes (top of table 3). In-
come generally was not significant, thus wealth
comes from capital, not income, for all farm
sizes.
Blank et al. Agricultural Profits and Farm 1305
Table 3. Regression Results for Change in Wealth and Profits Equations by Farm Size: Lake
States, Corn Belt, and Southeast Regions Combined (1996–2002)
Farm Size 1 Farm Size 2 Farm Size 3
Variable Estimate t-Value Estimate t-Value Estimate t-Value
Change in wealth equation
FarmInc 0.1662 0.48 0.3110 1.63 0.0005 0.00
NonFarmInc −0.1638 −1.47 −0.4293 −1.88∗−1.6816 −0.88
ChngFarmCap 0.9908 118.62∗∗∗ 0.9374 69.83∗∗∗ 0.2568 21.68∗∗∗
ChngNFarmCap 0.8597 30.58∗∗∗ 0.9439 18.94∗∗∗ 0.6524 2.22∗∗
Consumption 0.2698 0.93 −1.0688 −1.96∗∗ 2.3527 0.92
Fixed effects
Year ∗∗∗ ∗∗∗ NS
Profits equation
CashFlow 0.0129 0.51 0.0771 3.48∗∗ 0.0054 5.82∗∗∗
TotalExpenses −0.0663 −2.32∗∗ −0.0237 −3.25∗∗ −0.0038 −3.69∗∗
Depreciation −0.0600 −1.15 −0.1678 −9.35∗∗∗ −0.0485 −5.11∗∗∗
Productivity 1.7527 2.23∗∗ 0.7212 2.09∗∗ 0.0456 0.24
HumanCapitalEd 1.5789 1.37 6.8049 3.49∗∗ 0.0022 0.05
Fixed effects
Firm ∗∗∗∗ ∗∗
Year ∗∗
NS
Note: Farm Size 1 corresponds to limited resource, retirement, and residential farms. Farm Size 2 corresponds to farm/lower sales and farm/higher sales. Farm
Size 3 are large family farms and very large farms.
∗,∗∗,and ∗∗∗ denote statistical significance at the 0.10, 0.05, and 0.01 confidence levels. NS denotes “not significant.”
As shown in table 4, both farm and nonfarm
capital were significant in all regions, but had
differential impacts on wealth. For example, a
$1,000 change in farm capital in the Lake States
would raise wealth by $453, compared to $103
in the Corn Belt and $278 in the Southeast.
Table 4. Regression Results for Change in Wealth and Profits Equations by Region: Lake
States, Corn Belt, and Southeast (1996–2002)
Lake States Corn Belt Southeast
Variable Estimate t-Value Estimate t-Value Estimate t-Value
Change in wealth equation
FarmInc 0.4028 2.48∗∗ 0.3323 1.41 −0.9596 −2.26∗∗
NonFarmInc 1.5043 1.50 0.4800 0.81 −0.2574 −0.20
ChngFarmCap 0.4533 17.14∗∗∗ 0.1028 5.27∗∗∗ 0.2776 23.32∗∗∗
ChngNFarmCap 0.6321 3.85∗∗ 1.3783 7.25∗∗∗ 2.2177 9.27∗∗∗
Consumption −2.5824 −1.25 0.9489 0.88 0.1850 0.07
Fixed effects
Year NS ∗∗
Profits equation
CashFlow 0.0177 4.67∗∗∗ 0.0232 7.27∗∗∗ 0.0070 1.77∗
TotalExpenses −0.0031 −0.66 −0.0161 −5.30∗∗∗ −0.0044 −0.96
Depreciation −0.0566 −3.00∗∗ −0.0910 −3.50∗∗ −0.0789 −4.97∗∗∗
Productivity 5.1918 3.43∗∗ −1.5778 −2.86∗∗ −0.1505 −0.63
HumanCapitalEd −16.1208 −5.26∗∗∗ −3.9766 −1.32 2.0530 2.43∗
Fixed Effects
Firm ∗∗ ∗∗∗ ∗∗∗
Year NS ∗∗ NS
∗,∗∗,and ∗∗∗ denote statistical significance at the 0.10, 0.05, and 0.01 confidence levels. NS denotes “not significant.”
Also, a $1,000 change in nonfarm capital would
raise wealth by $632 in the Lake States, by
$1,378 in the Corn Belt, and by $2,218 in the
Southeast. The different impacts across re-
gions may be partly due to differences in the
opportunities and multiplier effects available
1306 Number 5, 2004 Amer. J. Agr. Econ.
off-farm in the regional economies. In all re-
gions, the higher regression coefficients for
“changes in nonfarm capital” imply that there
are economic incentives for shifting resources
out of agriculture and into nonagricultural
uses.
“Nonfarm income” was significant for mid-
sized farms when farms from all three regions
were combined (table 3). This may be because
off-farm income is more stable over time than
is farm income for small-sized firms, and off-
farm income may be a small part of large farms’
total wealth, thus the lack of statistical signifi-
cance in explaining changes in wealth (Mishra
et al.).
Farm profits (equation (6)). There were di-
verse results across regions for the prof-
its equation (table 4) reflecting different
commodity specializations across regions.
“CashFlow” (gross sales) and “Depreciation”
were significant in all three regions. “Human-
CapitalEd,” which represents the productivity
and investment components of human capital
was significant in the Lake States and South-
east (table 4), and for mid-sized farms (table 3).
The productivity variable was significant in the
Lake States and Corn Belt. Combining the
state-level total factor productivity variable
with the farm-level variable (i.e., gross value
of production divided by total cash expenses)
gives a significant relationship between profits
and productivity for small and mid-sized farms
(table 3). The coefficient for productivity de-
creases as farm size increases.
Implications of the Results
These results generally agree with other stud-
ies of convergence of time-series returns on
farm investments, and with other studies that
have used farm-level data to empirically assess
wealth and income patterns across states, farm
types, and commodity specializations. We sug-
gest three implications of these results.
First, although U.S. farm sector returns are
converging over the 1960–2002 period and
across regions, farm profits still vary widely by
farm type, farm size, location, and by other fac-
tors. Constructing a pseudo panel using pooled
farm-level data and estimating the system of
equations linking wealth, income, profits, and
productivity helps explain the linkages be-
tween the various components. For example,
the finding that both the changes in farm and
nonfarm capital are significant in all three re-
gions suggests that nonfarm capital is a substi-
tute for farm capital. This indicates that farm
households have diversified their portfolios.
Second, changes in farm and nonfarm cap-
ital have differential impacts on farm wealth
by farm location and by farm size. In general,
the fact that changes in nonfarm capital have
larger impacts than do changes in farm cap-
ital across all regions implies that there are
economic incentives to shift resources out of
agriculture. However, this may not happen be-
cause there appears to be incentives for small-
scale farms to increase their capital levels.
Third, we found evidence that farm size af-
fects both farm wealth and profits, and that
the relative impacts of the “farm size” vari-
able vary across these three regions, indicating
differences in profitability across the different
commodities produced in each region. We also
found evidence that a firm’s cumulative in-
vestment in human capital and productive re-
sources is important in the Lake States and the
Southeast, but is not statistically significant in
the Corn Belt. This implies productivity differ-
ences exist across the commodities that domi-
nate production in each region. This is impor-
tant because productivity growth is expected
to be a key to future profitability for each re-
gion’s agricultural sector.
References
Andres, J., J. Bosca, and R. Domenech. “Conver-
gence in the OECD: Transitional Dynamics
or Narrowing Steady-State Differences?” Eco-
nomic Inquiry 42(2004):141–49.
Baltagi, B.H. Econometric Analysis of Panel Data,
2nd ed. Chichester: John Wiley and Sons, 2001.
Ben-David, D. “Equalizing Exchange, Trade Liber-
alization and Income Convergence.” Quarterly
Journal of Economics 106(1993):407–43.
Blank, S.C. “Is Agriculture a ‘Way of Life’ or a Busi-
ness?” Choices 17(2002):26–30.
——. “Where is American Agriculture in Its ‘Life
Cycle’?” Journal of Agricultural and Resource
Economics 28(2003):396–418.
Carriker, G., M. Langemeier, T. Schroeder, and
A. Featherstone. “Propensity to Consume
Farm Family Income from Separate Sources.”
American Journal of Agricultural Economics
75(1993):739–44.
Caselli, F., and W. Coleman II. “The U.S. Structural
Transformation and Regional Convergence: A
Reinterpretation.” Journal of Political Econ-
omy 109(2001):584–616.
Deaton, A. “Panel Data from Time Series
Cross-Sections.” Journal of Econometrics
30(1985):109–26.
Blank et al. Agricultural Profits and Farm 1307
Dodson, C.B. Profitability of Farm Businesses: A Re-
gional, Farm Type, and Size Analysis.Washing-
ton DC: U.S. Department of Agriculture, ERS
Stat. Bull. 884, June 1994.
Drozd, D., and B. Johnson. “Dynamics of a
Rural Land Market Experiencing Farmland
Conversion to Acreages: The Case of Saun-
ders County, Nebraska.” Land Economics
80(2004):294–311.
Greene, W.H. Econometric Analysis, 4th ed. Upper
Saddle River, NJ: Prentice-Hall, 2000.
Gutierrez, L. “Convergence in U.S. and EU Agricul-
ture.” European Review of Agricuiltural Eco-
nomics 27(2000):187–206.
Hoppe, R., and J. MacDonald. America’s Diverse
Family Farms: Assorted Sizes, Types, and Sit-
uations.Washington DC: U.S. Department of
Agriculture, ERS Agr. Info. Bull. 769, May
2001.
Klepper, S. “Entry, Exit, Growth, and Innovation
Over the Product Life Cycle.” American Eco-
nomic Review 86(1996):562–83.
Mishra, A., H. El-Osta, M. Morehart, J. Johnson,
and J. Hopkins. Income, Wealth, and the
Economic Well-Being of Farm Households.
Washington DC: U.S. Department of Agricul-
ture, ERS Agr. Econ. Rep. 812, July 2002.
O’Rourke, K., and J. Williamson. “Late Nineteenth-
Century Anglo-American Factor-Price Con-
vergence: Were Heckscher and Ohlin Right?”
The Journal of Economic History 54(1994):
892–916.
Sala-i-Martin, X. “The Classical Approach to
Convergence Analysis.” Economic Journal
106(1996):1019–36.
Schmitt, G. “Simon Kuznets’ ‘Sectoral Shares in
Labor Force’: A Different Explanation of His
(I +S)/A Ratio.” The American Economic Re-
view 79(1989):1262–76.
Schott, P.K. “One Size Fits All? Hechscher-
Ohlin Specialization in Global Production.”
The American Economic Review 93(2003):686–
708.
U.S. Department of Agriculture (USDA). Eco-
nomic Research Service, Agricultural Re-
source Management Survey (ARMS) Briefing
Room, March 2004.
Verbeek, M., and T. Nijman. “Can Cohort Data Be
Treated as Genuine Panel Data?” Empirical
Economics 17(1992):9–23.
