Content uploaded by Hiroki Uematsu
Author content
All content in this area was uploaded by Hiroki Uematsu on Dec 17, 2013
Content may be subject to copyright.
Agricultural and Resource Economics Review 40/1 (April 2011) 1–19
Copyright 2011 Northeastern Agricultural and Resource Economics Association
Use of Direct Marketing Strategies by
Farmers and Their Impact on Farm
Business Income
Hiroki Uematsu and Ashok K. Mishra
Direct marketing strategies increasingly have been recognized as a viable business option in
U.S. agriculture as they allow producers to receive a better price by selling products directly to
consumers. The objective of this study is twofold. Using a national survey, we first estimated
a zero-inflated negative binomial model to identify factors affecting the total number of direct
marketing strategies adopted by farmers. Then we estimated a quantile regression model to as-
sess the impact of the intensity of adoption of direct marketing strategies on gross cash farm
income. The results show that the intensity of adoption has no significant impact on gross cash
farm income and that participation in farmers markets is negatively correlated with gross cash
farm income at all five quantiles estimated.
Key Words: direct marketing strategies, count data, gross cash income, quantile regression
Direct marketing strategies increasingly have been
recognized as a viable business option in U.S.
agriculture. According to the 2007 Census of Agri-
culture, the number of farms that reported sales of
agricultural products directly to individuals was
136,817, a 17 percent increase since 2002 (USDA
2007). The value of direct marketing sales has in-
creased by about 50 percent over the same period
(USDA 2009). Although the Census of Agricul-
ture accounted only for direct sales to individuals,
a direct marketing strategy generally includes a
wide spectrum of marketing channels such as
farmers markets, you-pick operations, consumer
cooperatives, sales to local restaurants and gro-
cery stores, and locally branded meats (Buhr 2004,
Kohls and Uhl 1998).
The economic incentives of both producers and
consumers have contributed to the recent trend in
increasing use of direct marketing strategies by
U.S. farmers. Direct marketing strategies allow
producers to receive a better price by selling prod-
ucts directly to consumers, who increasingly de-
mand fresh and “local” food due to the growing
concern for a healthier diet (Govindasamy, Hos-
sain, and Adelaja 1999, Morgan and Alipoe 2001,
Uva 2002). Although there is no clear-cut defini-
tion of “local,” and although what constitutes “lo-
calness” is another ongoing debate in the litera-
ture (Hand and Martinez 2010, Martinez et al.
2010), consumers are willing to pay more for lo-
cally grown products even after controlling for
freshness (Darby et al. 2008). The growing initia-
tive to create a sustainable food supply chain is
another important driving force in the implemen-
tation of a direct marketing strategy by farm op-
erators (Ilbery and Maye 2005). Since the major-
ity of the food products sold through direct mar-
keting strategies are typically sourced locally in-
stead of transported from national or international
sources, direct marketing potentially mitigates the
environmental impact of food production by re-
ducing the “food miles” in the food supply chain.
_
________________________________________
Hiroki Uematsu is a graduate research assistant and Ashok K. Mishra
is Professor in the Department of Agricultural Economics and Agri-
business at the Louisiana State University AgCenter in Baton Rouge,
Louisiana.
This paper was presented as a selected paper at the workshop “The
Economics of Local Food Markets,” organized by the Northeastern
Agricultural and Resource Economics Association (NAREA), in At-
lantic City, New Jersey, June 15–16, 2010. The workshop received fi-
nancial support from the U.S. Department of Agriculture’s Economic Re-
search Service, the Farm Foundation, and the Northeast Regional Cente
r
for Rural Development. The views expressed in this paper are the
authors’ and do not necessarily represent the policies or views of the
sponsoring agencies.
The authors wish to thank the participants of the NAREA 2010 work-
shop mentioned above for their useful comments and questions. Mish-
ra’s time on this project was supported by the USDA Cooperative State
Research Education and Extension Service, Hatch Project No. 0212495,
and Louisiana State University Experiment Station, Project No. LAB
93872.
2 April 2011 Agricultural and Resource Economics Review
A broad motivation of this study is to provide a
comprehensive picture of the degree to which
direct marketing strategies are disseminated in the
U.S. farm sector and their impact on the eco-
nomic well-being of U.S. farmers, using a na-
tional survey. Specifically, we first estimate a
zero-inflated negative binomial model to identify
factors affecting the total number of direct mar-
keting strategies adopted by U.S. farmers. Then
we estimate a quantile regression model to assess
the impact of the intensity of adoption of direct
marketing strategies on gross cash farm income.
The rest of the paper is organized as follows.
First, we review existing literature on direct mar-
keting strategies in U.S. agriculture. Then we de-
scribe the data used in this study. We then pro-
vide an empirical model and estimation strategies,
as well as variable descriptions. We follow with
estimation results, before making some conclud-
ing remarks.
Literature Review
The existing literature on direct marketing strate-
gies has mainly focused on the consumer side
from two different perspectives (Brown, Gandee,
and D’Souza 2006, Monson, Mainville, and
Kuminoff 2008). One is consumer preferences for
locally sourced food (Ladzinski and Toensmeyer
1983, Gallons et al. 1997, Lehman et al. 1998,
Kuches et al. 1999, Thilmany and Watson 2004),
and the other is characteristics of consumers pur-
chasing agricultural products through direct mar-
keting strategies (Eastwood, Brooker, and Orr
1987, Schatzer, Tilley, and Moesel 1989, Govin-
dasamy and Nayga 1997, Wolf 1997, Kezis et al.
1998). In contrast, there are relatively fewer
studies on the production side of direct marketing
strategies (Govindasamy, Hossain, and Adelaja
1999, Brown, Gandee, and D’Souza 2006, Mon-
son, Mainville, and Kuminoff 2008). This section
reviews a limited number of such studies.
Brown, Gandee, and D’Souza (2006) identified
demographic and economic factors that influence
direct marketing strategy sales in West Virginia
counties. Factors such as median housing value,
population density, proximity to Washington, D.C.,
and diverse fruit and vegetable production are
found to have a positive impact on county-level
direct marketing strategy sales. Brown et al.
(2007) surveyed vendors at farmers markets in
West Virginia to identify factors affecting the to-
tal sales at those markets, among other things.
The authors found that retired, part-time, or lim-
ited resource farmers generated lower income
from farmers markets. Using data from a mail
survey of Virginia farmers, Monson, Mainville,
and Kuminoff (2008) employed an ordered logit
model to explain farmers’ reliance on direct mar-
keting strategy sales in terms of the share of those
sales in the total farm sales. The authors con-
cluded that smaller farms, farms that typically do
not produce many small fruits, farms using or-
ganic production methods without USDA certifi-
cation, and farms with small households are the
ones most likely to engage in direct marketing.
Factors such as farm size, household size, high-
value crop enterprises, and the use of organic pro-
duction methods without USDA certification are
positively correlated with the higher share of
direct marketing strategy sales to the total farm
sales. An interesting feature of the Monson, Main-
ville, and Kuminoff (2008) study is that the de-
pendent variable is a proxy for the intensity of
adoption of direct marketing strategies, although
the authors could not differentiate between the
direct marketing strategies that contribute to the
share of direct marketing strategy sales in the
total farm sales.
In contrast, using a survey from New Jersey
farmers, Govindasamy, Hossain, and Adelaja
(1999) estimated a binary logit model to examine
the impact of adopting a series of what they
termed “non-traditional agricultural activities,”
including direct marketing strategies, on the prob-
ability of earning “higher” income per acre.
1
They
identified use of agrotourism and direct sales to
consumers as factors contributing to higher in-
come per acre. Although this study does not ac-
count for the intensity of adoption of direct mar-
keting strategies, it could capture the heterogene-
ous effects of non-traditional agricultural acti-
vities on income.
Finally, Goodsell, Stanton, and McLaughlin
(2007) provide a detailed listing of direct mar-
keting opportunities available to livestock and
poultry producers, including but not limited to the
following: classic farm stands, farm-to-retail,
1
Govindasamy, Hossain, and Adelaja (1999) estimated two models
and set a cut-off point between higher and lower income at the median
in one model and at the 75th percentile in another.
Uematsu and Mishra Use of Direct Marketing Strategies by Farmers and Their Impact on Farm Business Income 3
farmers markets, farm-to-school, farm-to-restau-
rant, fundraising dinners, fairs and festivals, and
mail orders. They indicate that the process of es-
tablishing a direct marketing strategy for a live-
stock producer can be complex because of regu-
lations, but that it is one of the best methods for
livestock producers to capture more of the food
dollar.
There are two important aspects missing in the
existing literature on the production side of direct
marketing strategies. First, most studies are lim-
ited to a regional or state-level analysis. An ex-
ception is Payne (2002), who reports the sum-
mary of national survey on farmers markets con-
ducted by the U.S. Department of Agriculture,
but the report does not provide any econometric
analysis. Second, few studies examined the inten-
sity at which farms incorporate direct marketing
strategies into their businesses and their impact
on the farm’s economic well-being, while control-
ling for the impact of adopting different direct
marketing strategies. We address these two issues
in this study. Our analysis uses a large national
survey of U.S. farmers spanning multiple regions
and farm sizes. By examining the influence of the
intensity of adoption of direct marketing strate-
gies on gross cash farm income, this study can
provide significant information to U.S. farmers
on whether direct marketing strategies should be
part of their farm business management plan, con-
tingent on the type and location of the operation.
Data
The study employs data obtained from the na-
tionwide 2008 Agricultural Resource Management
Survey (
ARMS) conducted by the Economic Re-
search Service (
ERS) and the National Agricul-
tural Statistics Service (
NASS). The ARMS pro-
vides information about the relationships between
agricultural production, resources, and the envi-
ronment, as well as about the characteristics and
financial conditions of farm households, manage-
ment strategies, and off-farm income. Operators
associated with farm businesses representing agri-
cultural production in the 48 contiguous states
make up the target population of the survey. Data
are collected from one operator per farm: the
senior farm operator, who makes most of the day-
to-day management decisions.
For statistical purposes, the U.S. Department of
Agriculture currently defines a farm as an estab-
lishment that sold or normally would have sold at
least $1,000 of agricultural products during the
year (USDA 2005). For the purpose of this study,
our sample includes only farms that are classified
as family farms that are organized as sole pro-
prietorships, partnerships, or family corporations
because they are closely controlled by their opera-
tor and the operator’s household (USDA 2005).
Any operator households organized as nonfamily
corporations or cooperatives and farms run by
hired managers are excluded from this study be-
cause we are interested in farm business decisions
made by individual farmers and their family, not
by hired managers.
Finally, the fact that the
ARMS data has a com-
plex survey design and is cross-sectional raises
the possibility that the error terms are heterosce-
dastic. Using the Breusch-Pagan/Cook-Weisberg
test for heteroscedasticity (Judge et al. 1985, p.
446), we rejected the null hypothesis of constant
variance of residuals based on
2
1df =
χ
= 642,263 (p-
value = 0.00). Accordingly, all standard errors
were adjusted for heteroscedasticity using the
Huber-White sandwich robust variance estimator
based on algorithms contained in Stata (Huber
1967, White 1980). The robust standard errors
were used in all the regression models in lieu of
the jackknife variance estimation method, which
is a method suitable for estimation of standard
errors when the dataset has complex survey de-
sign [for further detail in the context of the
ARMS, see Kott (1997) and Dubman (2000)], but
also for when the dataset is used as a subset rather
than in full.
Empirical Model, Estimation Strategy, and
Description of Variables
Our econometric analysis consists of two stages.
In the first stage, we conduct a count data analy-
sis to estimate the number of direct marketing
strategies adopted by farmers. The predicted counts
of direct marketing strategies adopted is then used
as an instrument in the second stage, to estimate
the impact of the intensity of adoption of direct
marketing strategies on farm business income us-
ing a quantile regression approach.
4 April 2011 Agricultural and Resource Economics Review
Factors Affecting the Number of Direct
Marketing Strategies Adopted: A Count Data
Approach
The count variable is obtained by summing the
seven binary variables, each of which represents
whether or not a respondent adopts a particular
direct marketing strategy. The 2008
ARMS con-
tains specific questions pertaining to the use of
direct marketing strategies by farmers. Specifi-
cally, the survey queried farmers about whether
they have used the following direct marketing
outlets or approaches: (i) roadside stores, (ii) on-
farm stores, (iii) farmers markets, (iv) regional
distributors, (v) state branding programs, (vi) di-
rect sales to local grocery stores, restaurants, or
other retailers, and (vii) community-supported
agriculture (
CSA). Each of the above direct mar-
keting strategies is coded as a binary response
variable that takes a value of one when a respon-
dent uses the direct marketing strategy and zero
otherwise. Table 1 summarizes these binary re-
sponse variables. The most frequently used direct
marketing strategy in our sample is roadside
stores (161 farms), followed by direct sales to
local grocery stores, restaurants, or other retailers
(153 farms). Although it is often called the most
popular direct marketing strategy, only 118 farms
reported using farmers markets. Regional distri-
butors, state branding programs, and
CSA were
used by 57, 27, and 12 farms, respectively.
Table 1. Frequency of Individual Direct
Marketing Strategies
Direct Marketing Strategy Frequency Percentage
Roadside stores 161 25.39
On-farm stores 106 16.72
Farmers market 118 18.61
Regional distributors 57 8.99
State branding programs 27 4.26
Direct sales to local grocery
stores, restaurants, or other
retailers
153 24.13
Community-supported agriculture
(CSA) 12 1.89
Total 634 100.00
Source: USDA (2008).
In order to construct a variable that represents
the intensity of adoption of direct marketing
strategies, we count the total number of direct
marketing strategies a family farm used in 2008.
Table 2 provides descriptive statistics of this vari-
able. Approximately 92 percent of the farms in
the sample did not use any direct marketing
strategies. Given the fact that 6.2 percent of the
total farms reported sales of agricultural products
directly to individuals in the 2007 Census of Ag-
riculture (USDA 2007), in terms of use of direct
marketing strategies our sample appears to be a
good representation of the U.S. farm sector. See
Table 2 for the total number of direct marketing
strategies adopted.
Table 2. Total Number of Direct Marketing
Strategies Adopted
Count Frequency Percentage
Cumulative
Percentage
0 4,251 91.83 91.83
1 221 4.77 96.61
2 88 1.90 98.51
3 49 1.06 99.57
4 13 0.28 99.85
5 5 0.11 99.96
6 1 0.02 99.98
7 1 0.02 100.00
Total 4,629 100.00
Source: USDA (2008).
Model selection procedures for a count data
analysis involve two issues. The first is to test the
distributional assumption of the dependent vari-
able. The basic model for a count data analysis
assumes that the count variable has a Poisson
distribution in the population. A Poisson distribu-
tion assumes that the mean and the variance are
equal, but this assumption has to be tested be-
cause it does not hold in many empirical applica-
tions (Cameron and Trivedi 2005). If this as-
sumption is violated, a common approach is to
assume that the dependent variable follows a
more flexible negative binomial distribution.
Another issue is the number of observations
with zero count. A Poisson or a negative binomial
model with a count variable that has a consider-
able number of observations with zero count re-
Uematsu and Mishra Use of Direct Marketing Strategies by Farmers and Their Impact on Farm Business Income 5
sults in under-prediction of zero count. If this is a
concern, a zero-inflated Poisson (ZIP) or zero-
inflated negative binomial (ZINB) model should
be used. A zero-inflated counterpart (both ZIP and
ZINB) first estimates a binary logit model to de-
termine whether the count is zero or not, after
which it conducts a count data analysis (Cameron
and Trivedi 2005). The Vuong test can be used to
compare a Poisson or a negative binomial model
against its zero-inflated counterpart (Cameron
and Trivedi 2009).
An important underlying assumption in a zero-
inflated model is that it is possible to have a zero
count in the second stage. In the context of adop-
tion of direct marketing strategies, the first stage
estimates whether or not a particular farm is will-
ing to and capable of adopting any direct market-
ing strategy, while the second stage estimates the
number of direct marketing strategies adopted. A
zero count in the second stage indicates that the
farmer is willing to adopt and capable of adopting
a direct marketing strategy, but the respondent
chose not to do so in the 2008 survey period. The
farm operator might have adopted direct market-
ing strategies in the past or may adopt direct mar-
keting strategies in the future, but did not do so at
the time of the 2008 survey period. This is an
important clarification because, if the possibility
of zero count in the second stage is eliminated a
priori, one would have to estimate a truncated re-
gression model instead of a zero-inflated model.
Our count variable, the total number of direct
marketing strategies adopted, has a considerable
number of zero counts (92 percent of the sample).
Due to the large number of zero counts, the mean
number of the total direct marketing strategies
adopted is 0.13 with variance equal to 0.28. Be-
cause our dependent variable is subject to the two
concerns mentioned above, we estimated two
models: a negative binomial model and a
ZINB
model. Results from the negative binomial model
2
show that the test for overdispersion is significant
(LR statistic = 135.71, p-value ≤ 0.000), indicat-
ing strong evidence in favor of a negative bino-
mial model over a Poisson model.
Next, we estimated the
ZINB model to yield the
Vuong test statistic that compares the
ZINB and
negative binomial models. The Vuong test statis-
2
Results from the negative binomial model are not provided here, but
are available upon request.
tic (Z-score of 5.23, p-value ≤ 0.000) suggests
significant evidence in favor of the
ZINB model.
With this result, we decided to maintain the
ZINB
model to estimate the predicted counts of the total
number of direct marketing strategies adopted by
farm operators.
The zero-inflated negative binomial (
ZINB)
model is estimated in two steps. For the first-step
logit estimation of whether a farmer is willing to
adopt and capable of adopting direct marketing
strategies, the independent variables included:
direct payments received ($) by the farm, whether
the farm received Conservation Reserve Program
payments, farm type dummy variables (high-
value crop farms and other field-crop farms), dis-
tance from the farm to the closest city with a
population of at least 10,000, and whether there is
an animal product processing facility within 50
miles of the farm.
The independent variables in the second step of
the
ZINB model include years of formal education
for the operator and the spouse, the operator’s
farming experiences, the primary occupation of
the operator and the spouse, and the total number
of acres in operation. Although females’ human
capital in the farm household is sometimes cited
as a key determinant of adoption of a direct mar-
keting strategy (Monson, Mainville, and Kumi-
noff 2008), and although the authors’ casual ob-
servations at farmers markets seem to support
this, it is not confirmed as such in this model.
3
Also included in the ZINB model are the
dummy variable for farms that seek advice from
the Natural Resource Conservation Service (
NRCS)
agents, farm tenure dummy variables (tenants and
full owners),
4
farm type dummy variables (dairy,
other field crops, high-value crops, and live-
stock),
5
whether there is an animal product proc-
essing facility within 50 miles of the farm, dis-
tance from the farm to the closest city with a
population of at least 10,000, direct payments
received ($) by the farm, whether the farm re-
ceived payments from the Conservation Reserve
Program, whether the farm has access to the
3
In fact, our model initially included the number of female operators
in the first three operators in the farm, but it was highly insignificant,
perhaps because of little variability in the variable. The variable was
dropped from the model accordingly.
4
Tenants and full owners are compared to the base group of part
owners.
5
Farm type dummy variables are compared to the base group of
farms, whose primary enterprise is either cotton or cash grains.
6 April 2011 Agricultural and Resource Economics Review
Internet, and dummy variables for production re-
gions defined in the
ARMS data (Atlantic, South,
Plains, and West regions).
6
See Table 3 for a list
of variables and summary statistics.
The Impact of the Intensity of Adoption of Direct
Marketing Strategies on Gross Cash Farm
Income: A Quantile Regression
In the second stage, we estimate the impact of the
intensity of adoption of direct marketing strate-
gies on gross cash farm income, using quantile
regression. In the
ARMS, gross cash farm income
is defined as a sum of the following items: in-
come from crop and livestock operation, livestock
grazing, land rented to others, and other farm-
related activities such as production and market-
ing contracts. We use the logarithm of gross cash
farm income as the dependent variable. Specifi-
cally, the predicted counts of direct marketing
strategies adopted obtained in the first-stage
ZINB
model are used as a proxy for the intensity of
adoption of direct marketing strategies in the sec-
ond stage.
Quantile regression, originally developed by
Koenker and Bassett (1978), enables us to focus
on the underlying socioeconomic factors influ-
encing extreme values in the conditional distribu-
tions of the dependent variable. Quantiles are to
percentiles what probabilities are to percentages.
For example, the 0.50 quantile is the 50th percen-
tile. Instead of estimating conditional means, E(y|x),
as in OLS, quantile regression can estimate any
point on the conditional distribution by estimating
conditional quantiles, Q (β
q
). That is, the qth quan-
tile regression estimator is the one that minimizes
the following objective function:
(1)
{: }
{: }
()min | | +
(1 ) | | , (0,1),
p
ii
ii
N
qiiq
R
iiyx
N
iiq
iiyx
Qqyx
qy x q
β∈
∈≥β
∈<β
⎡
′
β= −β
⎢
⎣
⎤
′
−−β ∈
⎥
⎦
∑
∑
where q is an arbitrarily chosen quantile, p is the
number of parameters to be estimated, y
i
is the ith
observation of the dependent variable, x
i
is a k ×
6
These four regions are compared to the base group of the Midwest
region. Refer to USDA (2010) for a map of the NASS production re-
gions.
1 vector whose each element is the ith observa-
tion of k independent variables, β
q
is a k × 1 vec-
tor of quantile regression parameters to be esti-
mated, and N is the number of observations
(Koenker and Bassett 1978, Cameron and Trivedi
2009).
While OLS minimizes the sum of squared er-
rors, quantile regression minimizes a weighted
sum of absolute values of errors with different
weights being placed on positive and negative
errors, as in equation (1) (Kennedy 2008). The
major advantage of quantile regression is the ro-
bustness to outliers and heteroskedasticity, as quan-
tile regression estimates conditional quantiles in-
stead of conditional means.
Another advantage is that, while OLS estimates
the marginal effects of independent variables at
the conditional mean of the dependent variable,
quantile regression can estimate the marginal ef-
fects of the independent variables at any quantile
of the conditional distributions of the dependent
variable (Koenker and Hallock 2000). Just as the
arithmetical average of a variable often gives an
incomplete picture of the distribution of the vari-
able, OLS estimates can be misleading when the
conditional distributions of the dependent variable
are different across different values of the inde-
pendent variables (Mosteller and Tukey 1977).
Despite these restrictive and naïve assumptions,
most applied econometric analyses are concerned
with the conditional means (Angrist and Pischke
2008). It is these limitations in OLS that quantile
regression can overcome. The advantages of us-
ing quantile regression over OLS are particularly
important in this study considering the fact that its
objective is to provide a comprehensive picture of
the degree to which direct marketing strategy is
disseminated in the U.S. farm sector, and, to the
best of our knowledge, this is the first attempt to
do so using data from the nationwide survey. For
the sake of comparison, we also estimate the same
model with OLS.
Explanatory variables in the quantile regression
model include the predicted counts of the total
number of direct marketing strategies adopted and
dummy variables that represent adoption of each
of the seven direct marketing strategies (roadside
stores; farm stores; farmers markets; regional
distributors; state branding programs; direct sales
to local grocery stores, restaurants, and other re-
tailers; and
CSA), with the intention to measure
Uematsu and Mishra Use of Direct Marketing Strategies by Farmers and Their Impact on Farm Business Income 7
Table 3. Variable Definitions and Summary Statistics
Variable Mean Std. Dev.
Gross cash farm income ($) 855,357 2,957,948
Total number of direct marketing strategies adopted 0.14 0.54
Roadside stores as a direct marketing outlet (= 1 if used, 0 otherwise) 0.03 0.18
Farm stores as a direct marketing outlet (=1 if used, 0 otherwise) 0.02 0.15
Farmers markets as a direct marketing outlet (= 1 if used, 0 otherwise) 0.03 0.16
Regional distributors as a direct marketing outlet (= 1 if used, 0 otherwise) 0.01 0.11
State branding programs as a direct marketing outlet (= 1 if used, 0 otherwise) 0.01 0.08
Direct sales to local grocery stores, restaurants, or other retailers as a direct marketing outlet
(= 1 if used, 0 otherwise)
0.03 0.18
Community-supported agriculture (CSA) as a direct marketing outlet (= 1 if used, 0 otherwise) 0.00 0.05
Operator’s years of education 13.55 1.89
Spouse’s years of education 13.74 1.83
Operator’s years of farming experience 27.98 14.82
Operator’s primary occupation (= 1 if farming, 0 otherwise) 0.72 0.45
Spouse’s primary occupation (= 1 if farming, 0 otherwise) 0.28 0.45
Farm size (total acres operated) 1,577.04 7,819.98
Advice from Natural Resource Conservation Service (NRCS) personnel (= 1 if used, 0
otherwise)
0.12 0.33
Farm tenure – tenant (= 1 if tenant, 0 otherwise) 0.10 0.30
Farm tenure – part owner (= 1 if part owner, 0 otherwise) 0.46 0.50
Farm tenure – full owner (= 1 if full owner, 0 otherwise) 0.44 0.50
Entropy index of diversification 0.01 0.03
Average interest rate charged on loans 1.21 1.68
Dairy farm (= 1 if farm is classified as dairy farm, 0 otherwise) 0.08 0.27
Other field crops farm (= 1 if farm is classified as other field crops farm, 0 otherwise) 0.14 0.35
High-value crops farm (= 1 if farm is classified as high-value crops farm, 0 otherwise) 0.12 0.32
Livestock farm (= 1 if farm is classified as livestock farm, 0 otherwise) 0.44 0.50
Cotton or cash grain farm (= 1 if farm is classified as either cash grain or cotton farm, 0
otherwise)
0.22 0.42
Animal product processing facility (= 1 if farm is within 50 miles of an animal product
processing facility, 0 otherwise)
0.02 0.15
Distance (miles) from the farm to closest city with 10,000 or more population 23.59 23.80
Government payments (= 1 if farm receives any government payments, 0 otherwise)
0.56 0.50
Conservation Reserve Program (CRP) payments (= 1 if farm receives CRP payments, 0
otherwise)
11,701.85 31,480.29
Direct payments received ($) 0.16 0.36
Internet (= 1 if farm has an Internet connection, 0 otherwise) 0.78 0.41
Atlantic region (= 1 if the farm is located in the Atlantic region, 0 otherwise) 0.20 0.40
South region (= 1 if the farm is located in the South region, 0 otherwise) 0.19 0.39
Plains region (= 1 if the farm is located in the Plains region, 0 otherwise) 0.18 0.39
West region (= 1 if the farm is located in the West region, 0 otherwise) 0.19 0.39
Midwest region (= 1 if the farm is located in the Midwest region, 0 otherwise) 0.24 0.43
Total number of observations 4,629
8 April 2011 Agricultural and Resource Economics Review
the impact of adopting each direct marketing
strategy on gross cash farm income after control-
ling for the intensity of adoption of direct mar-
keting strategies. We expect the predicted counts
of the total number of direct marketing strategies
adopted (as a proxy for the intensity of adoption
of direct marketing strategies) to have a positive
effect on gross cash farm income. The intensity of
adoption of direct marketing strategies is ex-
pected to have a larger impact on gross cash farm
income at lower quantiles. On the other hand, we
do not have a priori expectations about the signs
of individual direct marketing strategy variables,
partly because little empirical evidence exists on
the relationship between the direct marketing
strategies and gross cash farm income.
The entropy index is included to assess the ef-
fect of diversification across enterprises on gross
cash farm income. The entropy index is a meas-
ure of diversification that ranges from 0 to 100,
with 0 indicating a farm producing only one com-
modity and 100 indicating a completely diversi-
fied farm (Jinkins 1992, Harwood et al. 1999).
Since enterprise diversification is a risk manage-
ment tool, it is ambiguous a priori if a higher
degree of diversification leads to a higher income.
Variables that represent human capital include
operator’s education, spouse’s education, opera-
tor’s farming experience, and farming experience
squared. Highly educated and more experienced
farmers are expected to have higher gross cash
farm income (Mishra, El-Osta, and Johnson 1999).
Dummy variables for the primary occupation of
operators and spouses are also included, with the
expectation that farming as a primary occupation
leads to higher gross cash farm income (Mishra,
El-Osta, and Johnson 1999). As a measure of
farm size, the total operated acres and the acres
squared are used. Following the economies of
scale argument, the acres squared is included to
capture nonlinearity between farm size and gross
cash farm income. Farms with higher acreage are
expected to have higher gross cash farm income,
but perhaps at a decreasing rate.
To represent financial performance of the farm,
the average interest rate charged on loans is in-
cluded in the model. Its impact is ambiguous.
Although a higher average interest rate on loans
may be a sign that the farm is in an undesirable
financial position, it may be those farms with a
solid business plan that are willing to take and
capable of taking on a loan with a higher interest
rate. We include the dummy variable for farmers
seeking advice from the
NRCS agents with the
expectation that it has a positive impact on gross
cash farm income. Farm tenure variables are also
included in the model. Specifically, we include two
dummy variables for tenants and part owners,
leaving full owners as the base category. Com-
pared to the base category of full owners, tenants
and part owners tend to operate large farms and
are likely to declare farming as their main occu-
pation. Thus, dummy variables for both tenants
and part owners are expected to have a positive
impact on gross cash farm income. In order to
assess the impact of Internet access on gross cash
farm income, we include the dummy variable for
the Internet. Following Mishra and Park (2005),
we expect that access to the Internet would yield
higher gross cash farm income. As in the first
stage, dummy variables for dairy, other field crops,
high-value crops, and livestock farms are in-
cluded and tested if certain farm types earn higher
gross cash farm income relative to the base group
of farms specializing in cotton and/or cash grains.
Finally, dummy variables for the Atlantic, South,
Plains, and West regions are included in the
model to assess regional differences in gross cash
farm income relative to the Midwest region. See
Table 3 for a definition of variables used in this
study and summary statistics.
Results and Discussion
Factors Affecting Intensity of Adoption of Direct
Marketing Strategies
Parameter estimates of the zero-inflated negative
binomial (
ZINB) model are presented in Table 4.
The coefficient of operator’s years of formal edu-
cation is positive and significant. Considering the
fact that a direct marketing strategy requires a
special set of skills and abilities (Uva 2002),
some of which may not be directly related to ag-
ricultural operations, the positive coefficient on
the operator’s education is expected. The opera-
tor’s farming experience has a negative and sig-
nificant effect on the intensity of adoption of di-
rect marketing strategies, indicating that experi-
enced farmers are unlikely to adopt direct mar-
keting strategies. Findings here support Uva’s
Uematsu and Mishra Use of Direct Marketing Strategies by Farmers and Their Impact on Farm Business Income 9
Table 4. Parameter Estimates from First-Stage Zero Inflated Negative Binomial Model
Variable Parameter Estimate 95% Confidence Interval
Operator’s education
0.112
0.050 0.174
Spouse’s years of education -0.046 -0.106 0.015
Operator’s years of farming experience
-0.012
-0.019 -0.005
Operator’s primary occupation
0.443
0.182 0.703
Spouse’s primary occupation
0.175
-0.022 0.372
Total acres operated
-0.00001
0.000 0.000
Advice from Natural Resource Conservation Service (NRCS) personnel
0.421
0.099 0.743
Farm tenure – tenant -0.192 -0.576 0.192
Farm tenure – full owner -0.075 -0.290 0.140
Dairy farm 0.538 -0.252 1.327
Other field crops farm
1.109
0.410 1.808
High-value crops farm
1.078
0.413 1.744
Livestock farm
0.605
-0.019 1.229
Animal product processing facility
0.509
0.153 0.866
Miles to closest city with 10,000 or more population 0.001 -0.005 0.007
Direct payment received ($)
-0.000003
0.000 0.000
Conservation Reserve Programs (CRP) payments
0.619
0.027 1.211
Internet
0.242
-0.040 0.524
Atlantic region 0.218 -0.045 0.482
South region
-0.366
-0.699 -0.033
Plains region
-0.376
-0.787 0.036
West region
-0.732
-1.037 -0.428
Intercept
-2.241
-3.377 -1.105
Inflation model = logit
Direct payment received ($)
0.00001
-0.000003 0.00003
Conservation Reserve Program (CRP) payments
1.685
0.891 2.479
High-value crops farm
-3.863
-4.548 -3.178
Other field crops farm
-0.739
-1.284 -0.194
Miles to closest city with 10,000 or more population
0.005
-0.006 0.017
Animal product processing facility
-4.783
-6.339 -3.227
Intercept
2.455
0.020 0.650
alpha 0.113
Log likelihood = -1249.251 LR χ
2
(22) = 126.26
Vuong test of ZINB vs. standard negative binomial: Z = 5.23
Note: Bold indicates significance at the 10 percent level.
10 April 2011 Agricultural and Resource Economics Review
(2002) argument that a direct marketing strategy
requires a set of skills different from those for
agricultural operations.
Results in Table 4 show that the coefficients of
farming as a primary occupation are positive and
significant for both operators and spouses, indi-
cating that farmers and spouses who consider
farming as their main occupation are likely to
adopt more direct marketing strategies. This is
also consistent with the aforementioned skill re-
quirements to adopt a direct marketing strategy.
The effect of farm size in terms of the total oper-
ated acres on adoption of direct marketing strate-
gies is found to be negative and significant. This
is consistent with the general understanding that,
compared to large farms, smaller farms tend to
rely more on direct marketing strategies. Further,
large farms are likely to grow commodity crops
and receive government program payments to
support farm business income. A positive and sig-
nificant coefficient of the dummy variable for
farmers seeking advice from the
NRCS profes-
sionals indicates its positive effect on adoption of
direct marketing strategies.
Three of the four farm-type variables—other
field crops, high-value crops, and livestock farms—
have a significant and positive effect on the inten-
sity of adoption of direct marketing strategies,
compared to the base category of cash grain farms
and cotton farms. This finding is consistent with
Figure 1, which shows a breakdown of direct
marketing strategy sales by farm type. Other field
crop, high-value crop, and livestock farms are
more likely to adopt direct marketing strategies at
a higher intensity than cash grain farms and cot-
ton farms. The availability of an animal product
processing facility within 50 miles of the farm
yields a positive coefficient, suggesting that being
in close proximity to such a facility helps live-
stock farms to adopt more direct marketing strate-
gies. Direct payments received by the farm ($)
had a negative and significant coefficient, while
the dummy variable for CRP payments has a
positive and significant effect on the intensity of
adoption of direct marketing strategies. This is,
however, consistent with the prior discussion
about farm types and skill requirements. Just as
cash grain farms are less likely to adopt direct
marketing strategies, farms that receive more
direct payments (regardless of their farm type) are
also less likely to adopt direct marketing strate-
gies because direct payments are tied to produc-
tion of commodity crops like wheat, cotton, corn,
soybean, and others. A positive correlation be-
tween CRP payments and direct marketing strate-
gies adoption is plausible, as farms with more
land retired from production are expected to have
higher labor availability, of course, after control-
ling for primary occupation. Having access to the
Internet on the farm is positively correlated with
the intensity of adoption of direct marketing
strategies. It is likely that Internet access is neces-
sary to set up a successful direct marketing strat-
egy. It may also help farmers to expand the scope
of direct marketing opportunities.
NASS produc-
tion regions also yielded significant impact on
adoption of direct marketing strategies. Relative
to the Midwest region, farms in the Southern,
Plains, and West regions are all less likely to
adopt direct marketing strategies, whereas farms
in the Atlantic region are not significantly dif-
ferent from the Midwest region.
The Impact of the Intensity of Adoption of Direct
Marketing Strategies on Gross Cash Farm
Income
Results from the second-stage quantile regression
are presented in Table 5. The second column in
Table 5 presents parameter estimates from the
OLS model with robust standard errors. The third
through seventh columns are parameter estimates
from the quantile regression, evaluated at the
0.10, 0.25, 0.50 (median), 0.75, and 0.90 quan-
tiles. The last column shows the Wald F-test sta-
tistics that examine the null hypothesis that all
quantile estimates are not significantly different
from each other.
Results in Table 5 show that the coefficient of
the intensity of adoption of direct marketing strate-
gies obtained from the first-stage
ZINB model is
not significant at all quantiles, nor in the OLS re-
sults. Contrary to our expectation, the intensity of
adoption of direct marketing strategies is found to
have no significant impact on gross cash farm
income. The fact that there are only 20 observa-
tions with four or more direct marketing strate-
gies adopted is the potential cause for this insig-
nificance. Another possible explanation is that
adopting multiple direct marketing strategies may
be a risk management tool rather than a profit-
maximizing strategy. Also, if each direct market-
Uematsu and Mishra Use of Direct Marketing Strategies by Farmers and Their Impact on Farm Business Income 11
SheepandGoat
Farming,7%
PoutryandEgg
Product i on, 6%
HogandPig
Farming,4%
Da i ryCattl e
andMilk
Product i on, 2%
Ca t t l e
Feedlots,5%
BeefCat tleRa nc hi ng
andFarming,26%
OtherCro p Farming,
8%
FruitandTree Nut
Farming,13%
VegetableandMelon
Farming,13%
Greenhouse,Nursery,
andFloriculture
Production,3%
AnimalAqua c ultu re
andOtherAnimal
Production,
7%
OilseedandGr ain
Farming,5%
Figure 1. Distribution of Direct Marketing in the United States for 2007
Source: USDA (2007).
ing strategy included in this model requires a dif-
ferent set of skills and abilities, as discussed ear-
lier, the farmers may prefer to concentrate on and
expand the direct marketing strategy that is al-
ready in place rather than to implement a new
strategy, because of the additional labor require-
ment, learning cost, and other fixed costs associ-
ated with adoption.
Even though the intensity of adoption does not
seem to have any impact on gross cash farm in-
come, adoption of individual direct marketing
strategies showed some significant effect on gross
cash farm income. Quantile regression estimates
reveal that marketing through roadside stores had
a negative and significant effect on gross cash
farm income at the 0.25, 0.50, and 0.75
quantiles.
However, sales through farm stores had a positive
and significant effect on gross cash farm income
for all but the 0.90 quantiles. Finally, parameter
estimates for farmers markets were unexpectedly
negative and significant at all quantiles.
There are many possible explanations for this
unexpected result. First, in comparison to other
direct marketing strategies, farmers who sell their
products at farmers markets may be exposed to
greater competition within the confines of those
markets. Participation in farmers markets is often
cited in the literature as the most popular direct
marketing strategy and considered “the historical
flagship of local food system” (Brown and Miller
2008, p. 1296). A 91 percent increase in farmers
markets from 1998 to 2009 is reported, as the
number grew from 2,756 to 5,274 (Martinez et al.
2010). Perhaps for this very reason, farmers mar-
ket participants are forced to charge prices that
are lower than what they would have charged at
other direct marketing outlets, such as farm stores
and
CSA, where they are exposed to a relatively
lower degree of competition.
Second, empirical evidence suggests that some
farmers markets are failing, despite their increas-
ing popularity. Stephenson, Lev, and Brewer
(2008) report that, from 1998 to 2005 in Oregon,
62 new farmers markets opened, while 32 exist-
ing farmers markets ceased to operate. They re-
port that unsuccessful farmers markets tend to
12 April 2011 Agricultural and Resource Economics Review
Table 5. Parameter Estimates of Second-Stage Quantile Regression Model
Quantile Regression Parameter Estimates
Estimated Quantiles
Variables OLS
0.10 0.25 0.50 0.75 0.90
Wald
F-Score
Predicted counts of the total number of
direct marketing strategies adopted
-0.143 -0.374 -0.004 -0.185 -0.214 -0.143 0.67
Roadside stores
-0.571
-0.439
-0.927 -0.508 -0.310
-0.098
2.10
Farm stores
0.740 0.979 0.890 0.733 0.362
0.166 1.82
Farmers markets
-0.978 -0.759 -0.882 -1.210 -1.035 -0.954
0.64
Regional distributors
0.553 1.080
0.396 0.255 0.439 0.488 1.55
State branding programs 0.078 0.350 0.299 0.344 0.205 0.392 0.07
Direct sales to local grocery stores,
restaurants, or other retailers
0.447
0.022 0.197 0.281
0.494 0.377
0.54
Community-supported agriculture (CSA) -0.228 -0.062 -0.306 0.235 -0.349
-0.786
0.58
Entropy index of diversification
16.988 12.134 17.486 20.556 36.327 51.237 13.80
Operator’s years of education
0.029
-0.017 0.032
0.032 0.028
0.021 1.14
Operator’s farming experience
0.042 0.038 0.050 0.039 0.034 0.027
1.38
Operator’s farming experience squared
-0.001 -0.001 -0.001 -0.001 -0.001 -0.0004
1.66
Spouse’s years of education
0.029 0.054
0.011
0.028
0.027
0.038
0.84
Operator’s primary occupation
1.755 1.447 1.658 1.947 1.794 1.277 12.23
Spouse’s primary occupation
0.321 0.415 0.354 0.281 0.133
0.062
3.80
Total acres operated
0.001 0.0001 0.0001 0.0001 0.0001 0.0001
1.00
Total acres operated squared
-0.000
-0.000 -0.000 -0.000 -0.000
-0.000
0.47
Average interest rate charged on loans
0.186 0.212 0.194 0.143 0.097 0.061 9.45
Advice from Natural Resource
Conservation Service (NRCS)
personnel
0.118
0.122 0.068 -0.006 0.022 0.051 0.41
Farm tenure – tenant
1.010 1.530 1.177 0.910 0.733 0.461 9.33
Farm tenure – part owner
0.643 1.071 0.873 0.564 0.460 0.360 8.44
Internet
0.472 0.306 0.353 0.435 0.568 0.500
1.74
Government payments
0.644 1.003 0.949 0.652 0.454 0.269 9.83
Dairy farm
0.822 0.650 0.662 0.767 0.803 0.855
0.45
Other field crops farm
-0.970 -1.604 -1.185 -0.803 -0.311
-0.032
19.47
High-value crops farm
0.584 0.487 0.602 0.779 0.908 0.915
0.76
Livestock farm
-0.781 -1.175 -0.985 -0.725 -0.503 -0.338 11.14
Atlantic region -0.113
-0.508
-0.116 -0.125 0.063 -0.048
4.55
South region -0.115
-0.257
-0.163
-0.311
-0.080 -0.063
2.46
Plains region
-0.012 -0.108 0.056 -0.116 -0.041 0.023 1.89
West region 0.054 -0.031 0.055
-0.150 -0.155
-0.050 1.38
Intercept 7.607
6.262 6.778 7.808 8.682 9.759
Note: Bold indicates significance at the 10 percent level. Dependent variable = log(gross cash farm income).
Uematsu and Mishra Use of Direct Marketing Strategies by Farmers and Their Impact on Farm Business Income 13
have a small size, less product variety, and inade-
quate administrative resources, such as low reve-
nue and inexperienced managers because of a
high turnover. Again, perhaps because of their
popularity, there may exist a considerable compe-
tition not only within a farmers market but also
among different farmers markets to attract a suf-
ficient number of vendors to generate revenue to
keep the market financially viable.
Third, the economic cost of farmers market
participation may be substantial, while revenue
from them may not be as high as one might ex-
pect. For example, in Iowa only 30 percent of
participants in farmers markets in 2004 reported
sales greater than $5,000 (Varner and Otto 2008).
On the other hand, in New Jersey, variable cost of
farmers market participation for a 20-week season
is estimated at $6,410, excluding production costs
(Rutgers Food Innovation Center 2009). Although
we cannot simply conclude that farmers markets
are not profitable by comparing these two results
conducted in different parts of the country, profit
margins at farmers markets may be very slim. Be-
cause local food markets have relatively shorter
supply chains, direct marketing strategies often
impose additional labor requirements to pro-
ducers such as storage, packaging, transportation,
and advertising (Martinez et al. 2010). But there
are some costs that may be unique to farmers
markets. For example, vendors may need to clear
the inventory of perishable products even below
the marginal cost at the end of the day as farmers
markets are typically open only a few days a
week.
7
Despite the negative and significant impact on
gross cash farm income, there may be reasons for
farmers to continue participating in farmers mar-
kets and for policymakers to continue to endorse
them as the flagship of local food systems. First,
farmers markets can be a risk management tool
because they provide producers with additional
marketing opportunities (Rutgers Food Innova-
tion Center 2009). Therefore, participation in
farmers markets may increase intertemporal util-
ity of risk-averse farmers even if it decreases
gross cash farm income in a given year. Second,
farmers markets may be used to promote other
7
This may not be the case if there are multiple farmers markets
within a reasonable proximity, allowing a farm to sell at farmers mar-
kets more often. We thank the reviewer for this comment.
direct marketing channels such as CSA and to
socialize with other farmers and consumers in the
community.
8
Third, farmers can develop their
entrepreneurial skills through participating in farm-
ers markets (Feenstra et al. 2003). Finally, it is
important to note that some portion of producers’
lost profit margin induced by competition at
farmers markets is shifted to consumers in the
form of lower prices, possibly resulting in an in-
crease in the total surplus from a social welfare
standpoint.
The coefficient of marketing farm products
through regional distributors is positive and sig-
nificant at the 0.10 quantile. The coefficient of
direct sales to local grocery stores, restaurants,
and other retailers has a positive and significant
effect on gross cash farm income at the 0.75 and
0.90 quantiles. This may be an indication that
sales through regional distributors are more suit-
able for farms with smaller gross cash farm in-
come, while farms with larger gross cash farm
income can profit from direct sales to local gro-
cery stores and restaurants, which tend to be a
higher volume transaction. Finally, the coefficient
of direct marketing strategy through community-
supported agriculture is negative and significant
at the 0.90 quantile, suggesting that at the higher
gross cash farm income farmers are not profiting
from
CSA, perhaps due to commodity specializa-
tion or the fact that farms may not be producing
commodities that are being demanded by con-
sumers through direct sales. It is important to
note, however, that the Wald test statistic for all
but one of the seven direct marketing strategies as
well as the intensity of adoption of direct mar-
keting strategies are insignificant, indicating that
estimated coefficients are not significantly differ-
ent at different quantiles. The exception is road-
side stores; the impact on gross cash farm income
of selling products at roadside stores is different
across different quantiles. Nonetheless, consid-
ering the paucity of empirical research on the
impact of direct marketing strategies on the eco-
nomic well-being of farmers, the finding that the
effect of direct marketing strategies adoption on
gross cash farm income is mostly not statistically
different across different quantiles on a national
scale is an important addition to the existing lit-
8
We thank the reviewer for this comment.
14 April 2011 Agricultural and Resource Economics Review
erature on the use of direct marketing strategies in
the U.S. farm sector.
The entropy index of diversification has a posi-
tive and significant coefficient at all quantiles,
and the magnitude of the coefficient increases
with quantiles. The Wald test statistic (F = 12.83,
p-value = 0.000) confirms that the impact of di-
versification on gross cash farm income differs
across quantiles. The operator’s years of educa-
tion has a positive and significant effect on gross
cash farm income only at the 0.50 and 0.75 quan-
tiles. The spouse’s educational attainment also
has a positive and significant effect on gross cash
farm income at the lowest quantile (0.10), the
median (0.50), and the highest quantile (0.90).
While the operator’s farming experience has a
positive and significant effect on gross cash farm
income at all quantiles, experience squared has a
negative and significant impact on gross cash
farm income at all quantiles, confirming the ex-
pectation that the marginal impact of farming
experience on gross cash farm income is increas-
ing at a decreasing rate. However, the Wald test
statistic shows that this trend is not significantly
different across quantiles. Farming as a primary
occupation is positively correlated with gross
cash farm income for operators and spouses, and
its impacts are different across quantiles for both
operators and spouses.
Farm size in terms of the total number of acres
in operation has a positive and significant effect
on gross cash farm income at all quantiles. The
average interest rate charged on loans has a posi-
tive and significant impact on gross cash farm
income at all quantiles, and the positive impact is
larger at smaller quantiles. This is perhaps due to
the fact that smaller farms may tend to be more
financially constrained, and thus that small farms
that are willing to take on a higher interest rate
are likely to have a solid business plan. Higher
interest payments may reflect the debt-repaying
capacity of the farm. Higher interest rates on bor-
rowed capital may be associated with energetic
and dynamic farmers, or entrepreneurs and inno-
vators (Bowler 1992). Another possible explana-
tion is that higher interest rates might also be in-
dicative of the farm business having borrowed in
order to upgrade the commitment to agriculture
(Goodwin and Mishra 2000, Mishra, El-Osta, and
Sandretto 2006).
Two dummy variables for farm tenure (tenants
and part owners) both have positive and signifi-
cant impact on gross cash farm income at all
quantiles, indicating that, compared to full own-
ers, tenant and part owners have higher gross cash
farm income. This is consistent with the fact that
part owners and tenants tend to operate larger
farms and have larger sales than full owners
(USDA 1998). Further, the Wald test statistics are
significant for both tenants and part owners, indi-
cating that the impact of farm tenure on gross
cash farm income differs across quantiles. The
coefficient of having Internet access is positive
and significant at all quantiles, but the estimates
are not significantly different across quantiles.
The dummy variable for government payments
has a positive impact on gross cash farm income
across all quantiles. The Wald test statistic of 9.83
confirms that the impact of government payments
is statistically different across various quantiles.
Farm type dummy variables yielded mostly sig-
nificant estimates. Again, the base group consists
of a combination of cotton farms and cash grain
farmers. The coefficients of dairy farms and high-
value crop farms are positive and significant at all
quantiles. Quantile regression coefficients are
negative and significant at all quantiles in the case
of livestock farms, and are negative and signifi-
cant at all but the 0.90 quantile for other field-
crop farms. For both variables, the negative mag-
nitude of the coefficient is larger for smaller
quantiles, confirmed by the large Wald statistics
(F = 11.14 for livestock farms and F = 19.47 for
other field-crop farms). This may be evidence of
economies of scale in these enterprises.
Because the dependent variable is in a log
form, coefficient estimates are not marginal ef-
fects. Tables 6a and 6b provide marginal effect
estimates of discrete regressors and elasticity es-
timates of continuous regressors, respectively,
from the quantile regression model. Cameron and
Trivedi (2009) recommend using the average mar-
ginal effects (
AME) by multiplying the estimated
coefficients and the exponentiated linear predic-
tions of the dependent variable at each quantile.
However, we opted to use sample quantiles of the
dependent variable because the exponentiated
linear predictions over-predicted the quantiles by
a large margin with a wide confidence interval.
9
A caveat to these marginal effects and elasticity
estimates is that the confidence intervals of the
9
Confidence interval estimates are available upon request.
Uematsu and Mishra Use of Direct Marketing Strategies by Farmers and Their Impact on Farm Business Income 15
Table 6a. Marginal Effect Estimates of Discrete Regressors from Quantile Regression Model
Marginal Effect Estimates ($)
Estimated Quantiles
Variables
0.10 0.25 0.50 0.75 0.90
Predicted counts of the total number of direct
marketing strategies adopted
-1,172 -67 -30,302 -160,131 -291,594
Roadside stores -1,375
-16,121 -83,187 -232,540
-198,425
Farm stores
3,067 15,473 120,097 271,783
338,140
Farmers markets
-2,379 -15,342 -198,174 -776,282 -1,940,909
Regional distributors
3,385
6,885 41,735 329,244 992,494
State branding programs 1,097 5,197 56,310 153,917 796,523
Direct sales to local grocery stores, restaurants, or
other retailers
68 3,432 45,995
370,637 767,174
Community-supported agriculture (CSA) -196 -5,328 38,431 -261,701
-1,598,375
Operator’s years of education -52 550
5,219 20,672
43,299
Spouse’s years of education
170
188
4,514
20,111
77,111
Operator’s primary occupation
4,534 28,849 318,872 1,345,214 2,597,449
Spouse’s primary occupation
1,302 6,160 46,062 99,754
126,877
Average interest rate charged on loans
665 3,373 23,340 72,994 123,795
Advice from Natural Resource Conservation Service
(NRCS) personnel
384 118 -929 16,380 104,224
Farm tenure – tenant
4,795 20,471 148,948 549,623 938,332
Farm tenure – part owner
3,356 15,181 92,339 344,896 731,465
Internet
960 6,146 71,280 425,750 1,016,701
Government payments
3,143 16,514 106,844 340,216 547,859
Dairy farm
2,038 11,524 125,565 602,210 1,739,033
Other field crops farm
-5,027 -20,619 -131,506 -232,889
-66,036
High-value crops farm
1,525 10,465 127,578 681,016 1,860,784
Livestock farm
-3,683 -17,135 -118,755 -377,381 -688,074
Atlantic region
-1,593
-2,020 -20,498 47,076 -96,807
South region
-805
-2,829
-51,002
-60,091 -127,781
Plains region -339 974 -19,054 -30,754 46,101
West region -99 952
-24,604 -116,515
-101,723
Note: Bold indicates significance at the 10 percent level. Marginal effects are estimated using sample quantiles of regressors.
16 April 2011 Agricultural and Resource Economics Review
Table 6b. Elasticity Estimates of Continuous Regressors from Quantile Regression Model
Estimated Quantiles
Variable
0.10 0.25 0.50 0.75 0.90
Entropy index of diversification
0.12 0.17 0.20 0.36 0.50
Operator’s farming experiences
1.07 1.40 1.09 0.94 0.76
Operator’s farming experiences squared
-0.50 -0.72 -0.61 -0.48 -0.37
Total acres operated
0.21 0.17 0.18 0.21 0.23
Total acres operated squared
-0.06
-0.04 -0.03 -0.03
-0.03
Note: Bold indicates significance at the 10 percent level. Elasticities are estimated at the sample means of regressors.
estimates tend to become very large at the higher
quantiles, especially for direct marketing strategy
variables. This could be due to the fact that only
about 8 percent of farms in the sample adopted at
least one direct marketing strategy, and it is those
farms with lower gross cash farm income that are
more likely to adopt a direct marketing strategy.
Therefore, we limit literal interpretation of the
marginal effects and elasticity estimates of direct
marketing strategy variables at the lower quan-
tiles (0.10 and 0.25), as it may not carry practical
meaning at the higher quantiles because of the
wide confidence intervals.
Given the unexpected finding that participation
in farmers markets has a negative impact on gross
cash farm income, the degree to which it affects
gross cash farm income is of interest. Marginal
effect estimates (Table 6a) show that participation
in farmers markets could decrease gross cash
farm income by $2,379 and $15,342 at 0.10 and
0.25 quantiles, respectively, ceteris paribus. The
marginal effect of selling products at roadside
stores is also found to be negative; at the 0.25
quantile of the conditional distribution of gross
cash farm income, selling products at a roadside
store decreases gross cash farm income by $16,121.
The marginal effect of using farm stores is posi-
tive; it increases gross cash farm income by $3,067
at the 0.10 quantile and by $15,473 at the 0.25
quantile. The marginal effect of selling products
through regional distributors increases gross cash
farm income by $3,385 at the 0.10 quantile, but it
is not significant at the higher quantiles.
An additional year of education increases gross
cash farm income by $5,219 for the operator and
$4,514 for the spouse at the 0.50 quantile, sug-
gesting the relative importance of the operator’s
human capital over that of the spouse’s. The mar-
ginal effect of having Internet access is $960 at
the 0.10 quantile, but about $1 million at the 0.90
quantile. Although these estimates are vastly dif-
ferent, they are both about half of the gross cash
farm income at the respective quantiles.
Elasticity estimates (Table 6b) show that a 1
percent increase in the entropy index leads to a
0.12 percent increase in gross cash farm income
at the 0.10 quantile and about 0.50 percent at the
0.90 quantile, suggesting an increasing positive
impact of enterprise diversification on gross cash
farm income, which is consistent with Mishra, El-
Osta, and Sandretto (2006). The estimated elastic-
ity for the operator’s farming experience is posi-
tive at all the quantiles and ranges from 0.76 (at
the 0.90 quantile) to 1.40 (at the 0.25 quantile).
On the other hand, the operator’s farming experi-
ence squared has a negative elasticity estimate at
all quantiles. The percentage change in gross cash
farm income with respect to a 1 percent increase
in farming experience ranges from 0.39 (at the
0.90 quantile) to 0.68 (at the 0.25 quantile). The
elasticity of gross cash farm income with respect
to farming experience is positive but inelastic at
all the quantiles.
The analogous elasticity estimates for the total
acres in operation and the total acres squared
range from 0.15 percent (at the 0.10 quantile) to
0.20 percent (at the 0.90 quantile). The elasticity
of gross cash farm income with respect to the
total operated acres is also positive, but inelastic
at all the quantiles.
Uematsu and Mishra Use of Direct Marketing Strategies by Farmers and Their Impact on Farm Business Income 17
Conclusions
The objective of this study was to estimate the
relationship between the intensity of adoption of
direct marketing strategies and the economic
well-being of U.S. farmers. We employed quan-
tile regression to estimate the relationship at vari-
ous points in the conditional distributions of the
dependent variable, which is gross cash farm in-
come. In doing so, we first conducted a count
data analysis and estimated a zero-inflated nega-
tive binomial model (
ZINB) to obtain the pre-
dicted counts of the total number of direct mar-
keting strategies adopted. The predicted counts
were used as a proxy for the intensity of adoption
of direct marketing strategies in the second-stage
quantile regression, in which we obtained two
unexpected results. One was that the intensity of
adoption of direct marketing strategies was not
found to be significant at any quantile. We pos-
ited several explanations for this unexpected re-
sult. First and foremost, the small number of ob-
servations with four or more direct marketing
strategies adopted is the likely cause of the insig-
nificant estimate. Second, this may indicate that a
direct marketing strategy is a risk management
tool rather than a profit-maximizing strategy.
Third, it may be due to additional labor require-
ments necessary to implement a new direct mar-
keting strategy, as each direct marketing strategy
may demand a unique set of skills and abilities.
The other unexpected finding was the negative
impact of participation in farmers markets on
gross cash farm income at all quantiles. We pro-
posed that this unexpected result can be attributed
to several factors, such as competition among
producers in a farmers market, competition among
farmers markets, inadequate management re-
sources, a low profit margin, and intermittent op-
eration. We also discussed why farmers may con-
tinue to participate in farmers markets despite
such participation’s negative impact on their eco-
nomic well-being from economic and sociologi-
cal perspectives. An important question that re-
mains but that is beyond the scope of this study
is: What lies ahead for farmers markets if there
are few economic incentives for participation?
Despite these unexpected results, this study ful-
filled our primary motivation to provide a com-
prehensive picture of the degree to which direct
marketing strategies are disseminated in the U.S.
farm sector and their impact on the economic
well-being of the U.S. farmers as of 2008.
Finally, some of the challenges that we experi-
enced in this study are noted here. First and fore-
most, the 2008
ARMS data has a small number of
observations of farms that implemented direct
marketing strategies. In our sample, only 378 out
of 4,629 farms implemented at least one direct
marketing strategy, and only 20 farms implemen-
ted 4 or more (Table 2). It is likely that this sam-
ple reflects the actual status of direct marketing
strategies in U.S. agriculture, as direct marketing
strategy sales account for a growing but small
share of the farm sector sales (Martinez et al.
2010), but it may have caused the wide confi-
dence intervals for coefficients, marginal effects,
and elasticity estimates, especially at the higher
quantiles, possibly causing some estimates to be
statistically insignificant. Another challenge we
faced was that our model could not capture the
intensity of adoption of each direct marketing
strategy and possibly heterogeneous skill require-
ments for different direct marketing strategies.
Delineating the relationship between skill require-
ments and the intensity of adoption of different
direct marketing strategies and their impact on
gross cash farm income would be an exciting
topic for another study. Future research will ad-
dress these challenges and build on our first at-
tempt to explore the impact of the intensity of
adoption of direct marketing strategies on the
economic well-being of U.S. farms, using a na-
tional survey.
References
Angrist, J.D., and J.S. Pischke. 2008. Mostly Harmless Eco-
nometrics: An Empiricist’s Companion. Princeton, NJ:
Princeton University Press.
Bowler, I.R. (ed.). 1992. The Geography of Agriculture in De-
veloped Market Economies (1st edition). Harlow, UK: Long-
man Publishing Group.
Brown, C., J.E. Gandee, and G. D’Souza. 2006. “West Vir-
ginia Farm Direct Marketing: A County Level Analysis.”
Journal of Agricultural and Applied Economics 38(3): 575–
584.
Brown, C., and S. Miller. 2008. “The Impacts of Local Mar-
kets: A Review of Research on Farmers Markets and Com-
munity Supported Agriculture (CSA).” American Journal
of Agricultural Economics 90(5): 1296–1302.
Brown, C., S.M. Miller, D.A. Boone, H.N. Boone, S.A. Gartin,
and T.R. McConnell. 2007 “The Importance of Farmers’
18 April 2011 Agricultural and Resource Economics Review
Markets for West Virginia Direct Marketers.” Renewable
Agriculture and Food Systems 22(1): 20–29.
Buhr, B.L. 2004. “Case Studies of Direct Marketing Value-
Added Pork Products in a Commodity Market.” Review of
Agricultural Economics 26(2): 266–279.
Cameron, A.C., and P.K. Trivedi. 2005. Microeconometrics
Methods and Applications. New York: Cambridge Univer-
sity Press.
____. 2009. Microeconometrics Using Stata. College Station,
TX: Stata Press.
Darby, K., M.T. Batte, S. Ernst, and B. Roe. 2008. “Decom-
posing Local: A Conjoint Analysis of Locally Produced
Foods.” American Journal of Agricultural Economics 90(2):
476–486.
Dubman, R.W. 2000. “Variance Estimation with USDA’s
Farm Costs and Returns Surveys and Agricultural Resource
Management Study Surveys.” Staff Paper No. AGES 00-
01, Economic Research Service, U.S. Department of Agri-
culture, Washington, D.C.
Eastwood, D.B., J.R. Brooker, and R.H. Orr. 1987. “Consumer
Preferences for Local Versus Out-of-State Grown Selected
Fresh Produce: The Case of Knoxville, Tennessee.” South-
ern Journal of Agricultural Economics 19(2): 57–64.
Feenstra, G.W., C.C. Lewis, C. Hinrichs, G.W. Gillespie, and
D. Hilchey. 2003 “Entrepreneurial Outcomes and Enter-
prise Size in U.S. Retail Farmers’ Markets.” American
Journal of Alternative Agriculture 18(1): 46–55.
Gallons, J., U.C. Toensmeyer, J.R. Bacon, and C.L. German.
1997. “An Analysis of Consumer Characteristics Concern-
ing Direct Marketing of Fresh Produce in Delaware: A
Case Study.” Journal of Food Distribution Research 28(1):
98–106.
Goodsell, M., T. Stanton, and J. McLaughlin. 2007. “A Re-
source Guide to Direct Marketing Livestock and Poultry.”
Available at http://www.nyfarms.info/FAIDPaper.pdf (ac-
cessed March 20, 2010).
Goodwin, B.K., and A.K. Mishra. 2000. “An Analysis of Risk
Premia in U.S. Farm-Level Interest Rates.” Agricultural Fi-
nance Review 60(1): 1–16.
Govindasamy, R., F. Hossain, and A. Adelaja. 1999. “Income
of Farmers Who Use Direct Marketing.” Agricultural and
Resource Economics Review 28(1): 76–83.
Govindasamy, R., and R.M. Nayga, Jr. 1997. “Determinants of
Farmer-to-Consumer Direct Market Visits by Type of Fa-
cility: A Logit Analysis.” Agricultural and Resource Eco-
nomics Review 26(1): 31–38.
Hand, M.S., and S. Martinez. 2010. “Just What Does Local
Mean?” Choices 25(1). Available at http://www.choicesmaga
zine.org/magazine/issue.php?issue=19 (accessed March 2010).
Harwood, J.L., R.G. Heifner, K.H. Coble, J.E. Perry, and A.
Somwaru. 1999. “Managing Risk in Farming: Concepts, Re-
search, and Analysis.” Agricultural Economics Report No.
774, Economic Research Service, U.S. Department of Agri-
culture, Washington, D.C.
Huber, P.J. 1967. “The Behavior of Maximum Likelihood Es-
timates under Nonstandard Conditions.” In Proceedings of
the Fifth Berkeley Symposium on Mathematical Statistics
and Probability. Berkeley, CA: University of California
Press.
Ilbery, B., and D. Maye. 2005. “Food Supply Chains and
Sustainability: Evidence from Specialist Food Producers in
the Scottish/English Borders.” Land Use Policy 22(4): 331–
344.
Jinkins, J. 1992. “Measuring Farm and Ranch Business Diver-
sity.” In Agricultural Income and Finance Situation and
Outlook Report No. AFO-45, Economic Research Service,
U.S. Department of Agriculture, Washington, D.C.
Judge, G.G., E.W. Griffiths, R.H. Carter, H. Lütkepohl, and L.
Tsoung-Chao. 1985. The Theory and Practice of Economet-
rics (2nd edition). New York: John Wiley & Sons, Inc.
Kennedy, P. 2008. A Guide to Econometrics (6th edition).
Malden, MA: Blackwell Publishing.
Kezis, A., T. Gwebu, S. Peavey, and H.-T. Cheng. 1998. “A
Study of Consumers at a Small Farmers’ Market in Maine:
Results from a 1995 Survey.” Journal of Food Distribution
Research 29(1): 91–99.
Koenker, R., and G. Bassett, Jr. 1978. “Regression Quantiles.”
Econometrica 46(1): 33–50.
Koenker, R., and K.F. Hallock. 2000. “Quantile Regression:
An Introduction.” Available at http://www.econ.uiuc.edu/~
roger/research/intro/rq.pdf (accessed February 4, 2010.
Kohls, R.L., and J.N. Uhl. 1998. Marketing of Agricultural
Products (8th edition). Englewood Cliffs, NJ: Prentice Hall.
Kott, P.S. 1997. “Using the Delete-a-Group Jackknife Vari-
ance Estimator in NASS Surveys.” National Agricultural
Statistics Service, U.S. Department of Agriculture, Wash-
ington, D.C.
Kuches, K., U.C. Toensmeyer, C.L. German, and J.R. Bacon.
1999. “An Analysis of Consumers’ Views and Preferences
Regarding Farmer to Consumer Direct Markets in Dela-
ware.” Journal of Food Distribution Research 30(1): 124–133.
Ladzinski, K.M., and U.C. Toensmeyer. 1983. “Importance of
Direct Markets for Consumers in Their Fresh Vegetable
and Fruit Purchases.” Journal of Food Distribution Re-
search 14(3): 3–11.
Lehman, J., J.R. Bacon, U. Toensmeyer, J. Pesek, and C. Ger-
man. 1998. “An Analysis of Consumer Preferences for Dela-
ware Farmer Direct Markets.” Journal of Food Distribution
Research 29(1): 84–90.
Martinez, S., M. Hand, M. Da Pra, S. Pollack, K. Ralston, T.
Smith, S. Vogel, C. Shellye, L. Lohr, S. Low, and C. New-
man. 2010. “Local Food Systems: Concepts, Impacts and
Issues.” Economic Research Report No. 97, Economic Re-
search Service, U.S. Department of Agriculture, Washing-
ton, D.C.
Mishra, A.K., H.S. El-Osta, and J.D. Johnson. 1999. “Factors
Contributing to Earnings Success of Cash Grain Farms.”
Journal of Agricultural and Applied Economics 31(3):
623–637.
Mishra, A.K., H.S. El-Osta, and C.L. Sandretto. 2006. “Fac-
tors Affecting Farm Enterprise Diversification.” Agricul-
tural Finance Review 64(2): 151–166.
Uematsu and Mishra Use of Direct Marketing Strategies by Farmers and Their Impact on Farm Business Income 19
Mishra, A.K., and T.A. Park. 2005. “An Empirical Analysis of
Internet Use by U.S. Farmers.” Agricultural and Resource
Economics Review 34(2): 253–264.
Monson, J., D. Mainville, and N. Kuminoff. 2008. “The
Decision to Direct Market: An Analysis of Small Fruit and
Specialty-Product Markets in Virginia.” Journal of Food
Distribution Research 39(2): 1–11.
Morgan, T.K., and D. Alipoe. 2001. “Factors Affecting the
Number and Type of Small-Farm Direct Marketing Outlets
in Mississippi.” Journal of Food Distribution Research
32(1): 125–132.
Mosteller, F., and J. Tukey. 1977. Data Analysis and Regres-
sion: A Second Course in Statistics. Reading, MA: Addison
Wesley.
Payne, T. 2002. “U.S. Farmers’ Markets 2000: A Study of
Emerging Trends.” Journal of Food Distribution Research
33(1): 173–175.
Rutgers Food Innovation Center. 2009. “New Opportunities
for New Jersey Community Farmers Markets.” Rutgers
New Jersey Agricultural Experiment Station, New Bruns-
wick, NJ.
Schatzer, R.J., D.S. Tilley, and D. Moesel. 1989. “Consumer
Expenditures at Direct Produce Markets.” Southern Journal
of Agricultural Economics 21(1): 131–138.
Stephenson, G., L. Lev, and L. Brewer. 2008. “When Things
Don’t Work: Some Insights into Why Farmers’ Markets
Close.” Special Report No. 1073, Oregon State University
Extension Service, Corvallis, OR.
Thilmany, D., and P. Watson. 2004. “The Increasing Role of
Direct Marketing and Farmers Markets for Western U.S.
Producers.” Western Economics Forum 3(2): 19–25.
Uva, W.-F.L. 2002. “An Analysis of Vegetable Farms’ Direct
Marketing Activities in New York State.” Journal of Food
Distribution Research 33(1): 186–189.
U.S. Department of Agriculture. 1998. “Agriculture Fact Book
1998.” Available at http://www.usda.gov/news/pubs/fbook
98/afb98.pdf (accessed October 9, 2010).
____. 2005. “Briefing Rooms—Farm Structure: Glossary.” Eco-
nomic Research Service, U.S. Department of Agriculture,
Washington, D.C. Available at http://www.ers.usda.gov/
briefing/farmstructure/glossary.htm#farm (accessed Octo-
ber 10, 2010).
____. 2007. “2007 Census of Agriculture.” National Agri-
cultural Statistics Service, U.S. Department of Agriculture,
Washington, D.C. Available at http://www.agcensus.usda.
gov/Publications/2007/Full_Report/Volume_1,_Chapter_1
_US/index.asp (accessed October 9, 2010).
____. 2008. “2008 Agricultural Resource Management Sur-
vey.” Economic Research Service, U.S. Department of Ag-
riculture, Washington, D.C. Available at http://www.ers.
usda.gov/data/ (accessed November 1, 2009).
____. 2009. “Alternative Farming Information Center.” Avail-
able at http://afsic.nal.usda.gov/nal_display/index.php?info
_center=2&tax_level=1&tax_subject=299 (accessed Octo-
ber 10, 2010).
____. 2010. “ARMS III Farm Production Regions Map.” Eco-
nomic Research Service, U.S. Department of Agriculture,
Washington, D.C. Available at http://www.nass.usda.gov/
Charts_and_Maps/Farm_Production_Expenditures/reg_ma
p_c.asp (accessed October 10, 2010).
Varner, T., and D. Otto. 2008. “Factors Affecting Sales at
Farmers’ Markets: An Iowa Study.” Review of Agricultural
Economics 30(1): 176–189.
White, H. 1980. “A Heteroskedasticity-Consistent Covariance
Matrix Estimator and a Direct Test for Heteroskedasticity.”
Econometrica 48(4): 817–830.
Wolf, M.M. 1997. “A Target Consumer Profile and Posi-
tioning for Promotion of the Direct Marketing of Fresh
Produce: A Case Study.” Journal of Food Distribution Re-
search 28(3): 11–17.
