Do farm operators benefit from direct to consumer marketing strategies?

AGRICULTURAL ECONOMICS Agricultural Economics 45 (2014) 213–224

Do farm operators benefit from direct to consumer marketing strategies? Timothy Parka , Ashok K. Mishrab,∗ , Shawn J. Wozniakc b Department

a FED/ ERS/ USDA, 1400 Independence Ave, SW, Mailstop 1800, Washington, DC, 20520-1800, USA of Agricultural Economics and Agribusiness, Louisiana State University AgCenter, Louisiana State University, 128 Martin D. Woodin Hall, Baton Rouge, LA, 70803, USA c USAID/Peru/Alternative Development, Avenida La Encalada cdra. 17 s/n, Surco, Lima 33, Peru

Received 12 September 2012; received in revised form 7 March 2013; accepted 4 April 2012

Abstract Using farm-level data this study investigates factors associated with the choice of three direct marketing strategies (DMSs). Particular attention is given to the role of management and marketing skills in selection of DMSs. Additionally, the study applies a selectivity-based approach for the multinomial logit model to assess the relationship between DMSs and the financial performance of the business. Results suggest that both management and marketing skills significantly affect direct-to-consumer sales. Farmers choosing the strategy of sales only through direct-toconsumer outlets report earnings that are significantly lower than earnings from the other marketing strategies. Marketing skills prove to be beneficial to direct-to-consumer (DTC) earnings. Finally, the selectivity correction terms in the direct sales model are significantly negative in the choice of DTC, indicating the presence of sample selection effects. Accounting for selectivity is essential to ensure unbiased and consistent estimates. JEL classifications: C25, Q12, Q13, Q16 Keywords: Direct marketing outlets; Multinomial logit; Farm sales; Selectivity correction

1. Introduction Direct marketing strategies (DMSs) can help improve the bottom line of farms (Detre et al., 2011). However, a review of the literature reveals three limitations in the current state of research on DMSs. First, none of the previous studies has focused on the impact of management and marketing skills on the choice of DMSs. Second, none of the studies investigate the impact of marketing and management skills on the financial performance of the business. Finally, most studies do not correct for selectivity bias in their analysis. A broad motivation of this study is to: first, investigate the role of marketing and management skills in the choice of DMSs by farmers1 in (1) direct-to-consumer (DTC) outlets (such as roadside stand, or on-farm facility, on-farm store, farmer’s market, community supported agriculture); (2) intermediated retail outlets (IMOs; such as direct sales of local grocery stores, regional distributor, and state branding programs); and (3) both DTCs and IMOs

(this combines category 1 and 2 mentioned above).2 A second objective of this study is to assess the impact of management and marketing skills on the farm financial performance. Finally, the issue of selectivity bias is also frequently overlooked in the literature. We develop the econometric methods that address the impact of selection—marketing outlets—by farmers to assess the impact of DMSs on farming revenues. We account for selectivity bias in the observed earnings from a marketing outlet, recognizing that producers choose from a set of marketing options3 to obtain the highest returns. Producers are aware of their managerial skills and aptitudes, which enhance their ability to sell through specific marketing channels (Bruch and Ernst, 2010). Producers also learn about selling through specific marketing outlets and gravitate towards their preferred channels over time. Purchasing agents, sales personnel, and marketing consultants involved in food supply chains also play a role in developing relationships with producers who have the 2

*Corresponding author. Tel.: 225 578 0262; fax 225 578 2716. E-mail address: amishra@lsu.edu (A.K. Mishra). 1 These categories follow the framework established by Low and Vogel (2011) in the USDA report on direct and intermediated marketing of local foods.

C

2013 International Association of Agricultural Economists

Farms with no direct marketing/sales outlets will be used as the base group. Bruch and Ernst (2010) noted that “producers should choose direct market channels that match both their personal strengths and their farm production experience. Producers should also consider their customers, products, resources and the opportunities and threats affiliated with using a certain market channel.” 3

DOI: 10.1111/agec.12042

214

T. Park et al./ Agricultural Economics 45 (2014) 213–224

required marketing expertise to be productive long-term suppliers. Neglecting selectivity effects has policy relevant implications for identifying the key factors that influence earnings from direct marketing by farmers. Trost and Lee (1984) extended the polychotomous choice model to include selectivity corrections and demonstrated that returns to education are underestimated when selectivity is neglected. The implication in the context of our analysis is that unbiased and accurate estimates of the returns to DMSs should account for the presence of selectivity effects. We apply a selectivity approach for the multinomial logit model from Bourguignon, Fournier, and Gurgand (Bourguignon et al., 2007, hereafter BFG), highlighting its advantages over current methods in the section that develops the econometric model. Findings from this study show that management and marketing skills significantly affect DTC sales. A significant level of management skill—more ways to control input costs— increases the likelihood that a farmer uses IMOs and both DTC and IMOs marketing outlets. Larger share of income from sales of vegetable, fruits, and nursery enterprises is associated with a greater likelihood of participation in DTC, IMOs, and both DTC and IMOs marketing outlets. Beginning farmers are more likely to use DTC, IMOs, and both DTC and IMOs as a choice of direct marketing outlets. Finally, the selectivity correction terms in the direct sales model are significantly negative in the choice of DTC, indicating the presence of sample selection effects, hence accounting for selectivity is essential to ensure unbiased and consistent estimates. By examining the influence of choice of direct sales on earnings, the study can provide significant information to U.S. farmers about whether a particular choice of direct sales technique should be part of their farm business management plan, contingent on the type and location of the operation.

2. Marketing and management skills Management and marketing skills are two issues that have garnered much attention in the marketing literature but have fallen short in the agricultural marketing literature, including that which discusses financial returns and factors affecting financial returns. Management and marketing skills are the focus of this study. Increasingly, top-level management is demanding that marketing view its ultimate purpose as contributing to the enhancement of financial returns (Day and Fahey, 1988). Srivastava et al. (1997) argue that the effectiveness of marketing skills (using various channels of marketing) should be evaluated on the basis of their impact on cash flow enhancement and growth in the long-term value of business. Linkages between marketing skills (or marketing activities) that speed-up product diffusion and lead to acceleration of cash flows are well established in the literature (Jain et al., 1995; Robertson, 1993). In the agricultural industry context Robison and Barry (1987) write that farmers can use various forms of marketing initia-

tives (future, options, and cooperatives) to manage enterprise risk and/or income volatility. Turning our attention to management skill (management of inputs) and its impact on profitability reveals that producers use various sources to acquire inputs. These may include buying directly from vendors or input suppliers, locking in the price of inputs—forward contracting of inputs, participating in buying clubs, etc. Additionally, it has been pointed out in the literature that farmers can mitigate the effects of risk by forward contracting, a form of input management strategy, in input markets (Mishra and Morehart, 2001). Forward contracting4 inputs could aid planning and allow farmers to diversify purchases over time (Haydu et al., 1992). Finally, input contracting was found to significantly affect financial returns to dairy farmers in the United States (Mishra and Morehart, 2001). Hence it can be surmised that both marketing and management skills play a role in the choice of marketing outlets and the choice of marketing outlets have an impact on firm/farm performance. 3. Direct Marketing and Value of Farm Sales According to the 2007 Census of Agriculture, 136,817 farms implemented a form of DMS. Moreover, the number of farm operators incorporating direct marketing into their farm business model increased by 17% from 2002 to 2007 (Detre et al., 2011). Over the same period, farmers saw the value of direct marketing sales increase by 49%. A DMS applies to both crop and livestock products. Examples of DMS employed by farmers included use of farmers markets, you-pick operations, consumer cooperatives, and locally branded meats (Kohls and Uhl, 1998). DMSs allow producers to receive a better price by directly selling the products to the consumers who have increasing demand for fresh and “local” food due to the growing concern for a healthier diet (Govindasamy et al., 1999). However, there is no clear-cut definition of “local” and what constitutes “localness” is another on-going debate in the literature (Martinez et al., 2010). Some consumers are willing to pay more for locally grown products even after controlling for freshness (Darby et al., 2008). The growing initiative to create a sustainable food supply chain is another important driving force in the implementation of a DMS by farm operators (Ilbery and Maye, 2005). Finally, DMSs are promoted as marketing strategies that allow farmers to increase their income. Figures 1 and 2 show violin plots of total value of farm sales5 —the variable of interest in this study—by participation in direct marketing and by 4 Farmers choose forward contracting for two reasons. First, this practice allows them to get price discounts and “lock-in” a certain price for inputs. Second, it ensures quality and timeliness if input deliveries. 5 Total value of gross sales is easy to track and assess whether the activities developed to improve marketing outcomes are having an impact. Tracking gross sales can also help in planning for the future. After several years, trends and cycles will be apparent in the market season. Knowing these trends can help in making decisions about when to open or close the market for the season, when to schedule special events, and when additional advertising to draw customers may


215

and nonparticipants, suggesting some volatility in participation. Farmers participating in DTC only are skewed mostly toward the left in the plot, with the mean around $273,000 in net farm sales, a value that is only 55% of the earnings reported by farmers with no direct sales. There are three nodes, heavy in the lower group, and slight for farms in the middle, and higher net value of sales areas. For intermediated marketing outlets (IMO) only, there are two almost symmetrical nodes, showing almost equal participation by small and large farmers, while those having mixed strategies mostly skew toward farmers with lower net value of farm sales.

Violin Plot of Farm Sales by Direct Sales Strategy

No Direct Sales

Direct Sales

Total Value of Farm Sales in ‘000s of Dollars

4. Literature review 0

500

1000

1500

2000

Source: Calculations from 2008 ARMS, ERS, USDA Fig. 1. Participation in direct sales and total value of farm sales.

Violin Plot of Farm Sales by Marketing Strategy

DTC Only

InterM

Both

Total Value of Farm Sales in ‘000s of Dollars

-500

0

500

1000

1500

2000

DTC Only = Direct-to-Consumer Outlets Only InterM = Intermediated Outlets Only Both = Both Strategies

Source: Calculations from 2008 ARMS, ERS, USDA Fig. 2. Direct marketing strategies and total value of farm sales.

the discrete choice to engage in any direct sales or not. Violin plots combine box plots and density traces in one diagram. The box plots display the center, the spread, asymmetry, and outliers in data while the density traces reveal the distribution of the data, with its valleys, peaks and bumps. The mean for farmers participating in any form of direct sales is just about $423,000 for net value of farm sales, while the mean for those not participating is lowest ($361,000). The violin plot demonstrates that there are a larger number of small farms participating in direct marketing compared to large farms, but large farms participating show higher net value sales overall. There is a large standard deviation for participants be warranted. Farm business managers monitor growth in sales (revenues), so that earnings will have the potential to grow over time. It will also be important to monitor earnings growth in order to ensure that sales growth is having the desired effect on earnings.

Although there is a plethora of literature on DMSs as it pertains to consumer desirability and the attributes of consumers who buy directly from producers, there are relatively few studies that focus on producer behavior regarding DMSs and how participation in DMSs affects farm business income. For example, Brown et al. (2006) identified demographic and economic factors that influence DMS sales in West Virginia counties and found that counties’ fruit and vegetable production have a positive impact on county-level DMS sales. On the other hand, the authors found that share of retired, part-time, or limited resource farmers living in the county generated a lower income from a farmers’ market. Using farm-level data from Virginia, Monson et al. (2008) found that small farm households specializing in vegetables, fruits, and nursery are more likely to engage in direct marketing. An interesting feature was that the dependent variable is a proxy for adoption intensity of DMSs, although the authors could not distinguish between DMSs that contribute to the share of DMS sales in total farm sales. The above studies have several limitations. First, they either use aggregated (county) or small sample survey data to analyze participation in DMSs and/or impact of DMSs on county’s agricultural income. Second, these studies do not control for heterogeneity when estimating the empirical model. There are three studies that have used farm-level data, Agricultural Resource Management Survey (ARMS), to investigate: (1) adoption of DMS and its impact on gross sales (Detre et al., 2011); (2) factors affecting the number of DMSs (described it as “intensity of adoption”) adopted by U.S. farmers (Uematsu and Mishra, 2011); and (3) farmers’ use of both DTC and intermediated marketing channels in selling locally produced foods to consumers (Low and Vogel, 2011). Detre et al. (2010) found that production of organic crops and regional locations of the farm were important factors in adoption of DMSs. Farmers who adopted DMSs were likely to have higher income. The study by Detre et al. (2011) was limited in several ways. For example, the authors did not identify the types of DMSs used by the farmer; second, the share of income from each DMS was not reported or estimated in their model; third, the authors failed to assess the impact of the choice of sales outlets on farm business income separately. Finally, the authors do not correct for sample selection bias in their study.

216


Uematsu and Mishra (2011) found that participation in farmers’ markets and roadside stores had a significantly negative impact on gross cash farm income. On the other hand, direct sales to local grocery stores, restaurants, and/or other retailers and regional distributors had a significantly positive impact on gross cash income. A drawback of this study is that it does not explicitly estimate the factors affecting farmers’ participation in various direct marketing outlets and does not correct for selection bias. Finally, although Low and Vogel (2011) used ARMS data, the study has several drawbacks. First, authors did not explicitly estimate various direct marketing outlets. Second, DTC analysis was based on county level data. Finally, the authors did not correct for selectivity bias in their analysis. The above-mentioned studies fail to assess the impact of management and marketing skills on the utilization of various DMSs and their impact on the value of farm sales. Management and marketing skills are two important issues that may affect farmers’ bottom line (sales) and the decision to participate in DMSs. Unlike previous studies that mainly focused on county level analysis, this study (1) uses a national farm-level data set with the unique feature of a larger sample than previously reported, comprising farms of different economic sizes and in different regions of the United States; (2) corrects for sample selection bias; (3) assesses the impact of management and marketing skills and on participation decision and value of farm sales; and (4) estimates the impact of marketing skills and management skills of farm operators on participation decision and value of farm sales.

5. Data The study employs data obtained from the nationwide 2008 Agricultural Resource Management Survey (ARMS) collected by the Economic Research Service (ERS) and the National Agricultural Statistics Service (NASS). The ARMS provides information about the relationships between agricultural production, resources, and the environment as well as about the characteristics and financial conditions of farm households, marketing strategies, input management strategies, and off-farm income. Data are collected from one operator per farm, the senior farm operator, who makes most of the day-to-day management decisions. We excluded operator households organized as nonfamily corporations or cooperatives and farms run by hired managers from the sample. The 2008 ARMS queried farm operators on choices of marketing outlets and income earned when producers choose different market outlets to sell commodities. The survey instrument contains specific questions pertaining to the use of DMSs by farmers. Farmers indicated whether they have used the following direct marketing outlets: (1) roadside stand or on-farm facility, (2) on-farm stores, (3) farmers’ markets, (4) community supported agriculture (CSA), (5) regional distributors, (6) state branding programs, and (7) direct sales to local grocery stores, restaurants and other retailers. Based on this informa-

tion, a set of three marketing outlets was identified. The first group—DTC outlets only—includes 10% of the producers. The second group, IMOs only—accounts for 7% of the produces. The third group includes farmer who used both DTC and IMOs outlets, and includes 4% of the producers. Farms with no direct sales outlets were used as the base group and comprise 79% of the farms in the 2008 ARMS data set. Variable description and summary statistics is presented in Table 1. To assess the impact of marketing strategies and management skills we bring in two variables that will be used in our empirical model. First, we created a management skill (mangSKILL) variable to measure input management skills of farm operators. This variable combines a farmer’s use of various input acquisition and management practices and includes: locking in the price of inputs (forward purchases); using farm management or other services for advice on input sources/prices; shopping for the best price from multiple suppliers; negotiating price discounts if available; and participating in buying clubs, community supported agriculture, and/or alliances to purchase inputs. Second, we created a marketing skill (mktgSKILL) variable to measure marketing skills. This variable combines a farmer’s use of various marketing practices. These include: using advisory services; options; futures; on-farm storage; contract shipping; collaborative marketing or networking to sell commodities; and farmer owned cooperatives. Based on the literature cited above it is plausible that both marketing skill and management skill may have an impact on financial performance of farms (gross cash income). The 2008 ARMS queried farm operators regarding two separate issues: (1) marketing skills and (2) management skills. The marketing skills (mktgSKILL) variable is drawn from the ARMS survey section asking farmers about their marketing practices including the use of advisory services, options, futures, on farm storage, contract shipping, collaborative marketing or networking to sell commodities and farmer owned cooperatives. By contrast, the management skills (mangSKILL) variable is based on ARMS survey questions on the farmer’s input acquisition and management practices. The practices include forward purchasing of inputs, use of farm management services, comparative pricing across multiple suppliers, attempting to negotiate price discounts, and participating in buying clubs. For instance, Fig. 3 compares types of management practices used by U.S. farmers and those specializing in fruits, nuts, and nursery crops6 —those using DMSs and their counterparts. In 2008, for farmers using DMSs, more than 40% use negotiated price discounts and multiple suppliers. We include the management skills variable (managSKILL) in the discrete choice model and the marketing skills (mktgSKILL) variable in the earnings specifications. The variables show only a modest degree of correlation at 0.54 and the correlations shown are quite similar across each of the marketing options. 6 Literature (Brown et al., 2006; Govindasamy et al., 1999; Monson et al., 2008) provides evidence that farms specializing in fruits, nuts, and nursery production are more likely to engage in direct marketing strategies.


217

Table 1 Variable definition and summary statistics Variable

Description

connectoptns

Internet connectivity options

intntfrmnews

Internet used for farm-related news (hours/week)

intntcommc

Internet used for farm-related commerce (hours/week)

farminpTWN

Inputs available near farm (count of five inputs)d

hldcongdTWN

Household consumer good available for purchase near farms (= 1 if good available, otherwise)e Portfolio of input acquisition and management practices used by operator (5 practices)f Portfolio of marketing practices used by farm operator (7 practices)g

mangSKILL mktgSKIL DISTherf NUMherf

Distribution component of Herfindahl index for crops sold by the producer Number component of Herfindahl index for crops sold by the producer

begfarmer

Operator began operating this farm after 1997

ACRES

Total acres farmed

hrsOperpd

Hours of paid labor per week by other farm operators

hrsWrkrpd

Hours of paid labor per week by workers

male

Gender of operator (= 1 if male; 0 otherwise)

grthCOMP frutntSHR grainSHR vegSHR gvSALGR reasPRC

Off-farm income grew slower than farm sales (= 1 if yes; 0 otherwise) Share of sales accounted for by fruits and nuts Share of sales accounted for by major grains Share of sales accounted for by vegetables Growth in total value of farm sales from previous year (percentage) Farm inputs for which price was a major reason for NOT purchasing in the nearest town Growth in off-farm income (percentage) Total value of farm sales in 2008, in $1,000 (used in logarithm)

offfrMGR tvalfrmsal Sample size

Direct-to-consumer (DTC) outletsa Mean (Std. Dev)

Intermediated (IMOs) outletsb Mean (Std. Dev)

DTC and IMOsc Mean (Std. Dev)

0.95 (0.68) 3.10 (5.08) 2.27 (5.75) 2.38 (1.81) 0.58 (0.81) 1.14 (1.20) 0.81 (1.18) −0.06 (0.14) 0.14 (0.24) 0.21 (0.41) 345.82 (1128.41) 12.61 (61.49) 1034.95 (4772.41) 0.86 (0.35) 0.32 0.18 0.07 0.20 −0.02 0.57

1.23 (1.03) 4.29 (6.76) 3.57 (7.49) 2.12 (1.68) 0.55 (0.78) 1.80 (1.35) 1.09 (1.43) −0.03 (0.10) 0.05 (0.15) 0.20 (0.40) 1649.81 (6927.03) 45.36 (174.22) 1825.17 (3900.59) 0.93 (0.25) 0.31 0.08 0.11 0.11 0.34 0.92

1.24 (0.96) 5.50 (8.53) 6.09 (11.23) 1.94 (1.66) 0.54 (0.80) 1.96 (1.37) 1.25 (1.38) −0.10 (0.16) 0.23 (0.27) 0.17 (0.38) 318.80 (539.29) 59.82 (152.71) 1719.81 (6270.36) 0.91 (0.28) 0.27 0.20 0.06 0.24 0.02 0.71

0.57 273.01 (486.39) 183

0.83 681.41 (643.97) 75

0.68 524.63 (568.77) 82

markets products to roadside stands or on = farm facility, on-farm store, farmer’s market, community supported agriculture. markets products to regional distributor, state branding program, direct sales to grocery stores, restaurants, or other retailers. c Includes both DTC and IMOs. d Includes fuel, fertilizer and chemicals, feed and seed, machinery and implements, farm credit. e Includes groceries, clothing, household supplies, cars, trucks, appliances, furniture, etc. f Includes advisory services, options, futures, on farm storage, contract shipping, collaborative marketing or networking to sell commodities and farmer owned cooperatives. g Includes forward purchasing of inputs, use of farm management services, comparative pricing across multiple suppliers, attempting to negotiate price discounts, and participating in buying clubs. a Farmer

b Farmer

Source: Agricultural and Resource Management Survey, 2008.

Farmers in the lowest quartile of earnings adopt the smallest number of both the management and marketing practices while farmers in the highest quartiles adopt the most of both practices. There are several other factors that affect farmers’ decisions to engage in DMSs. For example, access to informa-

tion through the Internet and the use of the Internet—farm news and commerce—could play an important role in the choice of DMSs. One can argue that connectivity to the Internet (connectoptns) and the use of the Internet to seek farmrelated news (intntfrmnews) and farm-related commerce (intntcommc), have a positive impact on the choices of DMSs

218


Adoption of Management Practices by Farmers

0

.1

% of Adopters .3 .2 .4

.5

by Fruit, Nut, or Nursery, Direct-to-Consumer Outlets

Other

Fruit/nut/nursery farm

No Direct Sales Forward Purchases Inputs Negotiates Price Discounts

Other

Fruit/nut/nursery farm

Direct-to-Consumer Outlets ONLY Uses Multiple Suppliers

Source: Calculations from 2008 ARMS, ERS, USDA Fig. 3. Management practices by fruit, nut, or nursery farmers engaging in direct-to-consumer marketing.

(Mishra and Park, 2005). Further, we also include explanatory variables that represent purchase of farm inputs (farminpTWN) and households’ consumption goods (hldcngdTWN) near the farms. Finally, based on the literature, one can argue that farms that specialize in vegetable, fruit, and nursery production, may likely choose DMSs (Govindasamy et al. 1999; Monson et al., 2008). To test the hypothesis that diversified farms may be more likely to participate in DMSs we include Herfindahl index of diversification as an explanatory variable. A measure of diversification in the crops produced and sold by the farmer is based on 2 the Herfindahl index defined as d = 1 − i si where si for each producer is the share of gross earnings from the crop. We use six groupings of crops from the ARMS survey including major grains, other grains which include oilseeds, dry beans, and dry peas, fruit, tree nuts and berries, a category for vegetables, melons, potatoes and sweet potatoes, another group for nursery, greenhouse, cut Christmas trees, floriculture and sod, and a final group composed of hay and other crops. A decomposition of the Herfindahl index is used to address the properties of the diversification decision following Gollop and Monahan (1991): 1 1 2 + − si , d = 1 − y y2 i where y is a count of the number of crops the farmer is marketing. The first term in brackets is the number component (NUMHerf), reflecting the number of different crops produced for sale. The diversification index increases with the number of different crops (NUMherf). A producer who markets crops in each of the six categories has a number component of 0.83 while a producer who markets one crop has the minimum value for the number component of zero (no diversification). Higher values of the number component measure the degree of di-

versification in the farmers marketing strategy in choosing to produce a variety of crops. The second element in the above equation is the distribution component (DISTherf), representing the share of different crops that are marketed. Consider two different farmers who are each marketing two crops in a given year. One farmer reported equal for the two crops for a distribution index earnings of –0.25 212 − 0.52 + 0.52 = −0.25 . The second farmer reported 90% of the farm earnings from crop 1 and the remaining 10% of the earnings from thesecond farmer’s crop. The second distribution index is –0.57 212 − 0.92 + 0.12 = −0.57 , as larger negative values indicate an increasingly unequal distribution of earnings across the marketed crops. A farmer who reported earnings from only one crop would have distribution component of 0, the minimum value. The distribution component of the index shows that an increasingly unequal distribution of earnings from the marketing crops decreases the Herfindahl index. The distribution component is negative and smaller in absolute value than the number component (Table 1). An increasingly unequal product distribution reduces the negative distribution term in absolute value. Decomposition of the diversification index into its number and distribution components yields values of 0.14 and –0.06, respectively for the producers engaged in selling directly to consumers only, with relatively stable values reported across the three marketing options. When investigating the impact of DTC on total farm sales (both BFG and Lee models) we include several explanatory variables, including marketing skills (mktgSKILL). One can argue that marketing skills have a positive effect on total farm sales (Kohls and Uhl, 1998). Similarly, based on the literature one can argue that farm size (LNACRES) has a positive impact on farm sales. Finally, to assess the impact of how well the nonfarming and farming sector is performing we use factors such as growth in off-farm income (grthCOMP) and growth in value of farm sales from previous year (gvSALGR) as explanatory variables in the second model. One can argue that the relationship between growth in off-farm income and farm sales could be negative (Mishra et al., 2002; Nehring and Fernandez-Cornejo, 2005). Similarly, based on economic theory, one would expect a positive relationship between growth in value of farm sales from previous year and total farm sales in the current year.

6. Econometric model of choice of sales outlets and earnings in chosen outlet The empirical approach is based on a discrete choice model where producers select a set of marketing channels for agricultural output. McFadden (1986) developed the economic choice theory underlying the multinomial logit model and highlighted its value in linking discrete choice behavior (choice of market outlet) with continuous decisions (sales revenue in each outlet). Ofek and Srinivasan (2002) demonstrated how market


valuation of improved product attributes that account for competition from other brands, potential market expansion, and heterogeneous consumer preferences can be derived from the multinomial logit framework. Producers choose their marketing plans and assess outside options that are available before participating in any marketing channel. The value of farm sales depends on the farmer’s experience in producing and selling farm products and the farmer’s comparative advantage in bargaining and marketing skills combined with differences in the regional development and accessibility of outlets for farm products. Selectivity bias may be present in the econometric model explaining the choice of marketing outlets used by producers. The 2008 ARMS queried farm operators on choices of sales (marketing) outlets and income earned when producers choose different market outlets to sell commodities. Based on this information, a set of three marketing outlets was identified. The marketing strategies included (1) DTC outlets only, (2) intermediated outlets only (IMOs), and (3) both DTC and IMOs outlets. The producer’s choice of a marketing strategy is based on utility maximization among M alternatives, where utility yj∗ depends on features of the outlets and the producer’s marketing expertise. The marketing strategies include the choice to market through any one outlet, any two outlets, all the outlets, or none of the outlets (no direct sales). The utility of the producer who chooses from M (j = 1, 2, . . . ., M) mutually exclusive marketing plans depends on a set of observable exogenous variables Z, estimated parameters γ , and an unobservable stochastic component ηj: yj∗ = Zγj + ηj ,

j = 1, . . . , M.

(1)

We observe only whether a marketing plan is chosen so that yj = 1 if plan j is chosen and yj = 0 otherwise. Given the choice of marketing plan one (the decision to use a single marketing channel), the local-sales related income earned by the farmer is y1 = Xβ1 + u1 ,

leading to the multinomial logit (MNL) model. The probability that the Mth alternative is preferred is exp(ZγM ) . PM = exp(Zγj )

(4)

j

The MNL model offers a framework for dealing with selectivity effects in discrete choice models and has distinct theoretical and empirical advantages. Basuroy and Nguyen (1998) show that the MNL framework is appropriate for establishing equilibrium in market shares and assessing the impact of optimal firm responses to entry and potential market expansion. Choice models based on the MNL formulation are commonly used in marketing science applications and yield optimal pricing policies, which align with observed sales and pricing strategies of firms (Cattani et al., 2002). The parameters of the MNL model can be estimated by maximum likelihood but the estimation of the equation for income earned requires additional assumptions. Following BFG define standard normal variables, ηj∗ , as (5) ηj∗ = −1 G(ηj ) , where is the standard normal cumulative distribution function and assume that the expected values of u1 and ηj * are linearly related for every j, rj∗ ηj∗ (6) E[u1 |η1 . . . ηM ] = σ j =1,...M

The correlation coefficient between u1 and ηj is represented by rj while σ is the standard deviation of the disturbance term from the earned organic income equation. For the multinomial logit model, BFG derive the conditional expectation of ηj∗ . Given that the first marketing option is chosen (j = 1), the outcome equation for income earned, y1 is ⎡ ⎤ Pj ⎦ y1 = Xβ1 + σ ⎣r1∗ m(P1 ) + rj∗ m(Pj ) (7) Pj − 1 +w1 .

j =2,...M

(2)

where X is the set of exogenous variables affecting income earned from the marketing strategy and β is the set of estimated parameters. The idiosyncratic error term u1 satisfies E(u1 |X) = 0 and Var(u1 |X) = σ 2 . The estimation strategy accounts for correlation between the stochastic components ηj and u1 . Following BFG the Mth marketing strategy is observed only ∗ > max(yj∗ ), where j = M. This condition is equivalent if yM toZγM > εM , where εM = max(yj∗ − ηM ), j = M.

219

(3)

When the ηj elements are independent and identically Gumbel distributed, the cumulative distribution function is G(η) = exp(–e−η ) and the density function is g(η) = exp(–η – e−η ),

In this equation P1 is the probability that the first alternative is preferred, m(P1 ) is the conditional expectation of ηj∗ , m(Pj ) J represents the conditional expectation of ηj∗ and m(PJ ) • PJP−1 is ∗ the expectation of ηj for all j = 1. Each conditional expectation can be computed numerically. The residual error term is w1 and is independent of the regressors. In the first stage, the discrete choice model from Eq. (4) is estimated by maximum likelihood method to obtain γˆ . Given that marketing plan 1 is chosen, the second stage in Eq. (7) is estimated by ordinary least squares (OLS), recognizing that the disturbances are heteroskedastic and correlated across the sample observations. The BFG approach for dealing with selectivity has advantages over current methods. The method identifies not only the direction of the bias related to the choice of marketing plan, but also which marketing plan is the source of the bias. This is

220


Table 2 Parameter estimates for choice of direct marketing outlets by farm in the United States Variable

Constant connectoptns mangSKILL intntfrmnews intntcommc farminpTWN hldcongdTWN DISTherf NUMherf begfarmer hayothrcrpSHR grainSHR vegSHR frntSHR nursySHR

Direct-to-consumer (DTC) outlet

Intermediated (IMOs) outlet

DTC and IMOs outlet

Estimate

T-ratio

Estimate

T-ratio

Estimate

T-ratio

−4.819* −0.138 −0.041 0.016 0.001 0.145 −0.195 −1.472 1.411* 0.547* 1.491* −0.568 4.125* 3.530* 4.126*

−16.87 −1.01 −0.55 0.79 0.04 2.41 −1.46 −1.26 2.19 2.48 4.71 −1.34 12.33 11.50 14.45

−5.567* 0.181 0.290* 0.019 0.002 0.024 −0.049 −1.440 −1.272 0.526* 0.800* −0.424 3.158* 2.158* 3.624*

−14.77 1.14 2.97 0.87 0.11 0.28 −0.25 −0.60 −0.84 1.67 1.67 −0.92 6.46 4.27 10.46

−6.506* −0.058 0.346* 0.024 0.040* 0.044 −0.063 −0.121 3.126* 0.594* 1.192* −1.236 4.118* 3.816* 4.074

−13.21 −0.35 3.55 1.09 2.25 0.48 −0.32 −0.09 4.11 1.75 1.97 −1.66 7.76 7.60 8.59

Note: Asterisk indicates asymptotic t-values with significance at α = 0.10 or higher level.

accomplished by estimating a different selectivity term for each marketing strategy, rather than following Lee’s approach that estimates a single selectivity effect for all strategies together. The selectivity correction accounts for all the correlations between the disturbance terms of the earned income equations and the unobservable stochastic components driving the choice of marketing plan. Restrictive assumptions, which are required to implement commonly used selectivity methods, are relaxed. As Schmertmann (1994) initially noted, Lee’s (1983) approach implies a set of strong restrictions. First, unobservable factors that influence the choice of alternative 1 against any other alternative are correlated in the same direction with unobservable factors influencing the observed outcome y1 . That is, the correlations between ui and (ηj – η1 ) are the same sign for all j. A more stringent restriction results when the selection model is based on the multinomial logit model and the residual terms (ηj – η1 ) are assumed to be identically distributed. In this case, the correlations are restricted to be identical. Lee’s method tends to perform poorly in comparison with the BFG’s approach due to these restrictions. Finally, the choice of marketing outlets DTC, IMOs, and both DTC and IMOs) will be estimated using the BFG method and the selectivity term will then be used in the financial performance equation. 7. Results and discussion We will present the results in two parts. In the first part we will discuss the factors affecting the choice of DMSs (Table 2) followed by the assessment of the impact of DMS on farm sales (Table 3). 7.1. Choice of direct marketing strategy Table 2 reports parameter estimates of the choice of marketing strategy used by farmers. Note that the base group for com-

Table 3 Parameter estimates for direct marketing outlets and its impact on financial performance of farms in the United States: Lee and BFG model. Dependent variable = total value of farm sales, in 2008 (in logarithm) BFG model

Constant mktgSKILL LNACRES LNhrsOperpd LNhrsWrkrpd male offfrMGR gvSALGR grthCOMP reasPRC M(P1 , DTC only) M(P2 , IMO only) M(P3 , DTC&IMO) M(P4 , neither DTC or IMO)

Lee model

Estimate

T-ratio

Estimate

T-ratio

4.73* 0.21* 0.36* 0.17* 0.39* 0.53 0.27* −0.52* 0.73* 0.04 0.60 −5.78* −2.69* −0.256

5.65 2.27 4.01 1.89 7.89 1.66 1.77 −2.51 2.14 0.38 1.86 −3.37 −2.10 −0.48

7.083 0.298* 0.253* 0.221* 0.473* 0.408 0.244* −0.468* 0.616 0.068 0.125

15.55 2.49 2.93 2.28 11.24 1.20 1.65 −2.17 1.49 0.70 0.63

Note: Asterisk indicates asymptotic t-values with significance at α = 0.10 or higher level.

parison is farmers with no direct marketing sales. The multinomial logit model imposes the restriction of the independence of irrelevant alternatives (IIA) assumption, which means that the choice between any category of marketing choices (say DTC and IMO) is unaffected by the availability of another marketing option. We performed Hausman and McFadden (1984) tests and the IIA assumptions are not rejected. We provide a defense of the MNL model based on its use in marketing science and optimal pricing decisions above. Findings reported in Table 2 show that indeed management skills play an important role in the choice of IMO and both DTC and IMO marketing outlets. Results in Table 2 show that the coefficient of mangSKILL is positive and statistically


significant at the 1% level of significance for IMO and both DTC and IMO, indicating that farm operators with higher number of input acquisition skills are more likely to participate in IMOs and both forms of DMSs, compared to farmers with no direct marketing sales. Farmers in the DTC only category typically adopt the fewest marketing techniques compared to the other operators. Over 55% of the DTC only farmers do not use any of the techniques, the largest percentage across the three marketing options. Practices such as working with a marketing advisor and participating in collaborative marketing are the first choices for these farmers. A plausible explanation in that farming operations with more management skills (number of input acquisition methods) may be small farms, producing various organic farm products, such as fruits and nuts, and nursery crops. Further, as one reviewer noted, when selling through IMO channels, producers have to meet specific quality attributes7 and quantity specifications of the buyer. These farms have specific input requirements and tend to have higher production costs (Uematsu and Mishra, 2011) and using methods to reduce input costs, may help the bottom line of these farms. Additionally, Haydu et al. (1992) found that the reasons that farmers contract for inputs are the reduction of risk and speculation on favorable price moves. An interesting finding in Table 2 is a positive and significant coefficient of Internet used for farm-related commerce (intntcommc) suggesting that farming operations using the Internet for farm-related commerce are more likely to use both forms of DMSs. This result is consistent with the findings of Mishra et al. (2009) who concluded that farmers with Internet connections are more likely to explore additional marketing outlets for their farm products. Farmers engaged in direct sales only tend to market a greater diversity of crops (higher NUMherf) and have a larger variation in the earnings shares of those crops (higher negative DISTherf) compared to farmers with no direct sales. A Wald test rejects the null hypothesis that the coefficients on the number and distribution of components are equal. The calculated χ 2 value was 64.06, exceeding the critical value χ22 of 4.61 at the 95% confidence level. Parameter estimates from the MNL analysis reveals that both NUMherf and DISTherf have a positive significant impact on participation in both forms of DMSs (DTC and IMOs). Beginning farmers (begfarmer)—those who began farming after 1997, are more likely to choose DTC, IMOs, and both forms of DMSs. The findings here suggest that new entrants in farming may be more educated and are likely to engage in offfarm work (Mishra et al. 2002). Furthermore, the new entrants are more likely to operate small and diversified farms, located near metro-areas, where the demand for local food items and fresh produce is greater than for farms located in more sparsely populated areas (Detre et al., 2011).

7 For example, restaurants producers/growers may be require producers to supply carrots/potatoes with certain length and size in order to fit in the specific peeler/cutter the restaurant has for carrots or potatoes.

221

Finally, we have used share of income from hay and other field crops (hayothecrpSHR), cash grains (grainSHR), vegetable (vegSHR), fruits (frntSHR), and nursery (nursySHR) in the MNL regression to assess the impact of farm specialization and participation in DMSs. Results in Table 2 show that farm type has a significant impact on farmers’ participation on the choice of DMSs. For example, participation in DTC, IMOs, and both DTC and IMOs, increases with increasing share of income from vegetables, fruits, and nursery in total farm income. The findings are consistent with literature (Detre et al., 2011; Low and Vogel, 2011; Uematsu and Mishra, 2011) that most of direct marketing activities involve fruits, vegetables, and nursery products.

7.2. Impact of direct marketing outlet on value of farm sales We investigate the factors that influence the value of farm sales conditional on the producer’s choice to use DTC outlets. The dependent variable, value of farm sales, is the logarithm of total value of farm sales in 2008 and we use the BFG method as outlined in the econometric section of this article. The estimated coefficients from the BFG model were used to estimate the value of farm sales equation with the results presented in Table 3. There are three issues addressed in this section. First, the BFG selectivity effects, represented by the M(Pi) terms, are related to the alternative DMSs in the multinomial logit model. The four strategies generate four selectivity terms and we focus on the results from the earnings from DTC outlets. These farmers report earnings that are significantly lower than earnings from the other marketing strategies. The average earnings of $273,011 are about 75% lower than earnings for farmers who do not engage in any direct sales. Second, a comparison of results from the more restrictive Lee approach for dealing with selectivity is also presented. The implications for farmers who are deciding on marketing strategies are addressed. Third, the coefficients from the BFG model are discussed. The selectivity correction terms in the direct sales model are significantly negative in the choice of IMOs and both DTC and IMOs outlets, indicating the presence of sample selection effects. Accounting for selectivity is essential to ensure unbiased and consistent estimates of the coefficients in the value of farm sales equation. A negative (positive) selectivity coefficient in the IMOs marketing choice indicates lower (higher) earnings for the farmer relative to a randomly chosen producer. This finding indicates that farmers with unobserved attributes that enhance sales in IMOs market only have moved to alternative marketing strategies. Note the significant negative selectivity effect for producers using IMOs in the model for DTC farm sales, that is the estimated value of –5.78 for the M(P2) coefficient in the DTC option. This is due to lower than expected farm sales through DTC only due to the movement of farmers with better unobserved characteristics away from DTC only outlets and into

222


using IMOs only. Alternatively, farmers with a set of marketing skills that are actually less suited to selling in DTC outlets have shifted from intermediated markets to pursue the DTC only strategy. A second issue is to demonstrate how the BFG selectivity method provides deeper insight into the factors that influence earnings from the marketing strategies. We also estimated Lee’s model which is the most common approach to account for selectivity bias in the multinomial logit specification and the restrictions implicit in this model are outlined above (Table 3, column 4). Lee’s model estimates a single selectivity effect and does not identify the marketing plans that are the source of bias. This model does not find any selectivity bias in the value of farm sales model for producers using DTC outlets only. Lee’s model does not recognize that observed earnings have been influenced by farmers who do not perform well in a given DMS moving away from this strategy. The implication is that value of farm sales from the single outlet strategy are overestimated (biased upward) when the Lee approach is used. Farmers who are considering pursuing this strategy by looking at summary statistics of sales could be misled by these artificially high returns. This model is misleading in suggesting that selectivity effects are absent and that OLS is an appropriate estimation method for identifying factors influencing value of farm sales from the DTC outlets. We also evaluate the coefficients from the BFG model to identify the important factors that influence earnings. The level of marketing skills used by the farm operator (mktgSKILL) shows a significant positive relationship in the DTC sales model. For producers marketing through DTC outlets, the elasticity for the marketing skills variable indicates that adopting one additional practice increases total farm sales by 0.21%. Activities such as using on farm storage, developing collaborative marketing arrangements, and participating in farmer-owner cooperatives are most frequently reported by the DTC farmers and these activities are relatively low cost to adopt. The findings here indicate a difficult choice uncovered by our approach—higher marketing skills lead farmers away from DTC, but success (that is higher farm sales) in DTC relies on marketing skills. For producers marketing through DTC outlets, the variables for acreage (ACRES), hours of hired labor (hrsWrkrpd), and hours of farm operator’s labor (hrsOperpd) have significant positive impacts on earnings. Farm operators who market through DTC outlets only tend to use significantly less labor than farmers choosing the other DMSs. The positive hired labor coefficient (hrsWrkrpd) indicates the potential to increase earnings by hiring additional labor. In the log linear model for value of farm sales, the coefficients on acreage and labor represent elasticities. The output elasticities measure the change in the producer’s income (sales) as the input changes. For these producers marketing via DTC, the output elasticities indicate that a 1% increase in labor used increases the value of production or sales by 0.56 (0.17 + 0.39)%, while expanding the acreage farmed by 1% increases this farm

sales value by 0.36%. Paid workers provide a larger boost to the value of farm sales than expanded hours provided by farm operators. Higher annual growth rates for off-farm income (offfrMGR) tend to increase sales in DTC outlets. This is consistent with the fact that a majority of the local foods are produced by farmers who have off-farm employment (Low and Vogel, 2011). By contrast, higher growth rates for sales in the farm operation (gvSALGR) are associated with lower sales in DTC outlets. Our results here are surprising because the DTC farmers report the lowest growth rates for both off-farm income and value of farm sales compared to farmers using the other marketing strategies. The mean growth rate of off-farm income was 58% for the DTC farmers, below the rate of 63% recorded by farmers with no direct sales. Overall farm sales for farmers pursuing the DTC only strategy showed a negative growth rate of 2.0% on average and much lower than the 20.5% growth rate for farmers with no direct sales. Finally, results in Table 3 indicate that if the growth in off-farm income was slower than the farm sales then farm operators are more likely to increase sales through DTC outlets—indicating higher returns to labor in farming.

8. Conclusions In the era of a global economy, farmers face increasing pressure in developing a portfolio of various marketing channels and in bargaining competitively with increasingly sophisticated marketing participants in the supply chain of agricultural products in local and regional markets. Many farmers begin by selling directly through farmers’ markets, roadside stands, community supported agriculture, and other intermediated channels like regional distributors, and direct sales to grocery stores, local restaurants, and other retailers. The U.S. Government is also supporting farmers in direct marketing. The “Know Your Farmer, Know Your Food” (KNF2 ) initiative builds on the 2008 Farm Bill to strengthen USDA programs promoting local foods and includes plans to enhance direct marketing and farmers’ promotion programs, to support local farmers and community food groups, to strengthen rural communities and to promote local eating. The KNF2 website (http://www.usda.gov/knowyourfarmer) lists opportunities for farm loan programs such as direct and guaranteed ownership loans for beginning farmers and socially disadvantaged groups, farm storage facility loans, value-added producer grants, beginning farmer and rancher development programs, and technical assistance and marketing services for farmers engaged in local selling. Results from the discrete model (multinomial logit model) highlight variables that may influence the choice of direct marketing outlets by farmers in the United States. Extension agents, crop consultants, and marketing analysts can adapt this information to predict the type of marketing outlet that a given farmer might use and provide better information for farmers as well. For example, increasing management skills—more ways


to control input costs—increases the likelihood that a farmer uses IMOs and both DTC and IMOs marketing outlets. Additionally, a larger share of income coming from sales of vegetable, fruits, and nursery enterprises is associated with a greater likelihood that a farmer uses any combination of DTC and IMO marketing outlets. Beginning farmers are more likely to use DTC, IMOs, and both DTC and IMOs as a choice of direct marketing outlets. The selectivity coefficients in the DTC outlet model confirm that the BFG selectivity model is appropriate for the analysis of marketing outlet choices of U.S. farmers. Total farm sales through DTC are downward biased since farmers who are better suited to market through multiple outlets (both DTC and IMOs) have moved toward this marketing strategy. An accurate evaluation of the projected sales from the DTC outlet must account for selectivity effects. Results from this study show that farm operators with a broader portfolio of marketing skills are more likely to increase farm sales relative to farmers who are using fewer marketing skills. Additionally, we find that farm size, hired labor, and farms with faster growth in farm sales compared to growth in off-farm income are correlated with higher sales, even after accounting for selectivity effects. The results from selectivity corrected multinomial logit model identify factors influencing the sales when producers engage in selling to local outlets. Extension agents, crop consultants, and marketing analysts can gather this information to predict sales in local markets, assess market outlets that a given farmer might use, and provide better information to increase their credibility with farmers. The key variables identified in the model such as the portfolio of key management practices and the hiring and scheduling of labor and the time allocation of the operator is monitored by the farm manager and can be modified over time. Information about the growth in farm sales and earnings from off-farm income sources is readily gathered from farm records to assess the potential sales from DTC outlets. Acknowledgments We would like to thank two anonymous referees for useful comments, and appreciate the comments and advice of Sarah Low, Todd Kuethe and Stephen Vogel in developing the manuscript. Mishra’s time on this project was supported by the Louisiana State University Experiment Station project # LAB 94163. This work was completed while Shawn Wozniak was working at Economic Research Service, USDA. The views expressed here are not necessarily those of the Economic Research Service, the U.S. Department of Agriculture, or the U.S. Government. References Basuroy, S., Nguyen, D. 1998. Multinomial logit market models: Equilibrium characteristics and strategic implications. Manag. Sci. 44(10), 1396–1408 .

223

Bourguignon, F., Fournier, M., Gurgand, M. 2007. Selection bias corrections based on the multinomial logit model: Monte Carlo comparisons. J. Econ. Surv. 21, 174–202 . Brown, C., Gandee, J. E., D’Souza, G. 2006. West Virginia farm direct marketing: A county level analysis. J. Agric. Appl. Econ. 38(3), 575–584. Bruch, M.L., Ernst, M.D. Choosing Direct Marketing Channels for Agricultural Products. Center for Profitable Agriculure, Institute of Agriculture, The University of Tennessee. Available at: https://utextension.tennessee. edu/publications/Documents/PB1796.pdf. accessed January 2013. Cattani, K.D., Dahan, E., Schmidt, G. 2002. Spackling: Smoothing maketo-order production of custom products with make-to-stock production of standard items. Marketing Paper 384, October. Available at: http:// repositories.cdlib.org/anderson/marketing/384. Accessed March 10, 2013. Darby, K., Batte, M.T., Ernst, S., Roe, B. 2008. Decomposing local: A conjoint analysis of locally produced foods. J. Agric. Appl. Econ. 90, 476–486. Day, G., Fahey, L. 1988. Valuign market strategies. J. Marketing 57(July), 45–57. Detre, J., Mark, T., Mishra, A., Adhikari, A. 2011. Lingake between direct marketing and farm income: A double-hurdle approach. Agribus., Int. J. 27(1), 19–33. Gollop, F.M., Monahan, J.L. 1991. A generalized index of diversification: Trends in U.S. manufacturing. Rev. Econ. Stat. 73(2), 318–330. Govindasamy, R., Hossain, F., Adelaja, A. 1999. Income of farmers who use direct marketing. Agric. Resour. Econ. Rev. 28(1), 76–83. Hausman, J.A., McFadden, D.L. 1984. A specification test for the multinomial logit model. Econometrica 52, 1219–1240. Haydu, J.J., Myers, R.J., Thompson, S.R. 1992. Why do farmers forward contract in factor markets? S. J. Agric. Econ. 24, 145. Ilbery, B.,Maye, D. 2005. Food supply chains and sustainability: Evidence from specialist food producers in the Scottish/English borders. Land Use Policy 22(4), 331–244. Jain, D., Mahajan, V., Muller, E. 1995. An approach for determining optimal product sampling for the diffusion of a new product. J. Prod. Innovat. Manag. 12, 29–37. Kohls, R.L., Uhl, J.N. 1998. Marketing of Agricultural Products. 8 ed. Prentice Hall, Englewood Cliffs, NJ. Lee, L-F. 1983. Generalized econometric models with selectivity. Econometrica 51, 507–512. Low, S.A., Vogel, S. 2011. Direct and intermediated marketing of local foods in the United States. Economic Research Report Number 128, November 2011. Martinez, S., Hand, M., Da Pra, M., Pollack, S., Ralston, K., Smith, T., Vogel, S., Clark, S., Lohr, L., Low, S., Newman, C. 2010. Local food systems. Concepts, impacts and issues. Economic Research Report Number 97, May 2010, U.S. Department of Agriculture, Economic Research Service, Washingotn DC. McFadden, D. 1986. The choice theory approach to market research. Market. Sci. 5, 275–297. Mishra, A., Morehart, M. 2001. Factors affecting returns to labor and management on U.S. dairy farms. Agric. Finance Rev. 61, 123–140. Mishra, A., Morehart, M., El-Osta, H., Johnson, J., Hopkins, J. 2002. Income, wealth, and well-being of farm operator households. Agricultural Economics Report # 812, (2002) U.S. Department of Agriculture, Economic Research Service, Washington, DC. Mishra, A., Park, T. 2005. An empirical analysis of Internet use by U.S. farmers. Agric. Resour. Econ. Rev. 34(2), 253–264. Mishra, A., Williams, R., Detre, J. 2009. Internet access and Internet purchasing patterns of farm households. Agric. Resour. Econ. Rev. 38(2), 1–10. Monson, J., Mainville, D., Kuminoff, N. 2008. The decision to direct market: An analysis of small fruit and specialty-product markets in Virginia. J. Food Distrib. Res. 39(2), 1–11. Nehring, R., Fernandez-Cornejo, J. The impacts of off-farm income on farm efficiency, scale, and profitability for corn farms. Paper presented at the 2005 AAEA Meeting in Providence, RI, July 25–27, 2005. Ofek, E., Srinivasan, V. 2002. How much does the market value an improvement in a product attribute? Market. Sci. 21, 398–411.

224


Robertson, T.S. 1993. How to reduce market penetration cycle times. Sloan Manag. Rev. 35, 87–96. Robison, L.J., Barry, P.J. 1987. The Competitive Firm’s Response to Risk. 1987. Macmillan Publishing Company, New York, NY. Schmertmann, C.P. 1994. Selectivity bias correction methods in polychotomous sample selection models. J. Econometr. 60, 101–132. Srivastava, R.K., Shervani, T.A., Fahey, L. 1997. Driving shareholder value:

The role of marketing in reducing vulnerability and volatility of cash flows. J. Market Focused Manag. 2, 49–64. Trost, R.P., Lee, L.F. 1984. Technical training and earnings: a polychotomous choice model with selectivity. Rev. Econ. Stat. 66, 151–156. Uematsu, H., Mishra, A.K. 2011. Use of direct marketing strategies by farmers and its impact on farm business income. Agric. Resour. Econ. Rev. 40, 1–19.