Survey design

The data used in this study was obtained from an online survey of beef cattle producers in Southeastern US. The survey questionnaire was designed and administered through Qualtrics using computer assisted telephone interviewing (CATI) method.Footnote ⁵ The survey was pre-tested in January and February 2024 with cattle producers and Extension personnel. After pre-testing was complete, the survey was distributed between March and April 2024 in 11 SE states. The final dataset includes 140 completed responses. For analysis completed in this paper, we kept the responses that had credible answers for the questions used.

The survey instrument had three sections. The first section collected information about farm characteristics (i.e. the number of operations, number of pastures, total area in production, and if the operation had a pre-conditioning or stocker operation). This section also included questions to determine the types of commodities produced to form measures of diversification. A second group of questions examined producer perceptions, production goals, and risk management decisions. These included perceptions and enrollment in risk management programs, adoption of conservation practices, sources of information, operational risks, and their concerns about future events and adaptation strategies. Finally, a third section gathered information on producer demographic characteristics (e.g. marital status, level of education, age, gender, years of experience, etc.) and their risk attitude measured with four Likert scale questions. The survey instrument included attention-checking questions to increase response accuracy.

As discussed previously, respondents were asked to indicate their adoption of ten technologies (Table 1). The selection of technologies was based on previous research (Collins Reference Collins2011; Gillespie et al. Reference Gillespie, Whitt and Davis2023; Pruitt et al. Reference Pruitt, Gillespie, Nehring and Qushim2012; USDA-APHIS-VS 2011). In our empirical analysis, producers were asked to indicate if they adopted and currently utilize each of these practices. This removes the potential problem of counting disadopters or producers who adopted technology but do not currently use the listed technology.

Adoption rates for individual technologies ranged from 7% for blockchain to 72% for smartphones/tablets for farm management. Livestock-specific technologies such as automated environmental controls for animal housing, artificial insemination, estrus synchronization, or growth-promoting technologies showed lower rates of adoption compared to technologies such as smartphones used to manage the farm or computers to track or manage finances (Table 1).

Empirical strategy

To estimate the relationship between the number of technological innovations adopted by producer ( $i)$ and diversification among beef cattle producers in the sample, we utilized an Ordered Probit Estimation (consistent with Wollni et al. (Reference Wollni, Lee and Thies2010) and Teklewold et al. (Reference Teklewold, Kassie and Shiferaw2013)). The variable of interest ( ${y_i}$ ) is, besides diversification, explained by farm characteristics, producer perceptions of risk, conservation, information sources, climate adaptation, and producer demographic characteristics measures ( ${X_i}$ ) based on the following equation:

(1) $${{y}}_{{i}}^{{*}} = {\boldsymbol{X}}_{\boldsymbol{i}}^\prime{{\beta }} + {{{u}}_{{i}}}$$

where $i = 1,2, \cdots, 140$ , $y_i^*$ is the latent variable capturing producer $i$ ’s count of technologies adopted, ${X_i}$ is a vector containing the independent variables including state fixed effects, $\beta $ is a vector of parameters to be estimated, excluding the intercept, and ${u_i}$ is the stochastic error that follows a Normal distribution with mean zero and variance 1, ${u_i}\sim N\left( {0,1} \right)$ . The relationship between the variable of interest ( ${y_i}$ ) and the latent variable ( $y_i^*$ ) is given by:

(2) $$\eqalign{ & {{{y}}_{{i}}} = 1{{\;\;\;if}} -\infty \lt {{y}}_{{i}}^{{*}} \lt {{{\mu }}_1} \cr & {{{y}}_{{i}}} = 2{{\;\;\;if\;\;\;\mu }}\_1 \lt {{y}}_{{i}}^{{*}} \lt {{{\mu }}_2} \cr & \hskip 50pt\vdots \cr & {{{y}}_{{i}}} = 7{{\;\;\;\;if\;\;\;\;}}{{{\mu }}_6} \lt {{y}}_{{i}}^{{*}} \lt + \infty \cr} $$

where ${\mu _1}, \cdots, {\mu _7}$ are the threshold parameters known as “cut-points” estimated jointly with the $\beta $ . The data tabulation and the Ordered Probit estimation were completed in Stata 18 (commands oprobit and margins), and the standard errors were clustered at the date of the survey to control for any potential unobserved characteristics that might affect the estimation. Descriptive statistics of independent variables are displayed in Tables 2, 3, 4, and 5 and discussed in the next subsection.

Table 2.

Summary statistics for farm characteristics for Southeast beef cattle producers

Note: n = 140.

Table 3.

Percentage of beef cattle producers in the sample adopting various numbers of technologies across different commodities produced

Note: The column overall displays the overall share for all 140 Southeast beef cattle producers. The column Only Beef displays the share among beef cattle producers that only produce beef cattle. Columns to the right display the share for beef cattle producers that also produce Fruit, or Vegetables, or Row Crop, or Other Livestock, or Poultry and or Forest Products.

^a Of the ten technologies included in the question, respondents reported adopting between one and seven of the ten technologies listed in the survey, with no respondents indicating the adoption of zero technologies or eight or more.

Table 4.

Respondent perceptions of sources of information, threats to the operation, adaptation, and use of conservation practices

Note: Respondents were asked to indicate the number of conservation practices they adopted as well as the number of adaptation strategies they would consider adopting. Both of these variables are the count of practices/strategies indicated. Another question asked respondents to indicate the perceived accuracy of production information. A variable to indicate sales representatives are most accurate is included as a binary variable. Other questions asked producers to select and/or indicate sources of risk and variation in net cash farm income. Respondents who indicated unexpected expense and or/ cattle prices in these questions were included as binary variables.

Table 5.

Summary statistics for beef cattle producer demographic characteristics

Note: Some College or Higher is a binary variable equal to one if the respondent indicated they had more than a high school diploma. High Extension Interaction is a binary variable equal to one if the respondent indicated they had seven or more interactions with Extension either through agents or attendance at events in the last 12 months.

To capture diversification, the survey included the question “Which of the following do you produce?” with the available options to select being: beef cattle, poultry, other livestock, row crop, vegetables, fruit or orchard, and timber or forest products. If an option was selected, respondents were asked to indicate the acreage for each commodity. In the final model, seven dummy variables were added to describe the operation in terms of producing row crop (Produce Row Crop), vegetable crops (Produce Vegetable), fruit and orchard crops (Produce Fruit), poultry (Produce Poultry), other livestock (Produce Other Livestock), timber or forest products (Produce Forest Products). Since all operations produce beef cattle, a dummy variable was added to characterize producers who only produced beef cattle (Only Produce Beef). Technology adoption might also be explained by farm location, size, and if the operation has beef cattle stockers or a pre-conditioning program. We account for a pre-conditioning program or stocking operation with a binary variable equal to 1 if the producer indicated that they have a pre-conditioning program or stocking operation (Stocker). We also capture the number of pastures (No. Pastures), the total acreage in production (Production Total Acres), and the number of operations (No. Operations) that the owner operates. Respondents were instructed to “Consider an operation as a distinct production location more than 50 miles apart.” To account for unobserved differences that may exist between states or cattle production across the region, we added state-fixed effects (10 dummy variables).

Producer perceptions of operational risk, conservation practices, information sources, and possible future climate adaptations might also affect adoption. Respondents were asked “Over the next 10 years, to what extent are you concerned with the following threats to your farm operation?” This included concerns about environmental and market risks (e.g. increased heat stress in livestock, longer dry periods and drought, increasing input costs, and decreasing cattle prices). A binary variable equal to 1 if the respondent indicated that they were “moderately concerned” or “extremely concerned” about decreasing cattle price being a threat to the farm operation over the next 10 years was added (Variation in Cattle Prices). Survey participants were also asked to evaluate the accuracy of information obtained by different sources with 1 being very low accuracy and 5 being very high accuracyFootnote ⁶. A binary variable equal to 1 if the respondent ranked agricultural sales representatives highest or tied for the highest in terms of accuracy of agricultural production information (Sales Rep Most Accurate). In a subsequent question, producers were asked “In the last 3 years, what was the greatest source of variation on your net cash farm income for your largest beef cattle operation in the Southeast?” and instructed to select one source from a list of market and production sourcesFootnote ⁷. A binary variable equal to 1 if the respondent stated that unexpected expenses were the greatest source of variation on their net cash farm income (Variation in Unexpected Expenses). Finally, respondents were asked, “In the last 3 years, did you enroll or receive payment from any of the following programs?” and required to indicate if they had enrolled and/or received a payment from various USDA programsFootnote ⁸. A binary variable equal to 1 if the producer is enrolled in United States Department of Agriculture Livestock Indemnity Program (Enrolled in Indemnity Program).

We also added two variables to control for producers’ willingness to adopt conservation practices and other management strategies used to lower the possibility of damage from weather events. Adoption of conservation practices was measured as a count variable of the number of conservation practices adopted (No. of Conservation Practices). Possible conservation practices appropriate for beef cattle producers in the Southeast that respondents could indicate they used included: incorporation of legumes, poultry litter, silvopasture, forage stockpiling, biochar and incorporation of cool-season forage. Respondents were asked “How likely are you to implement any of the following practices to lower the possibility of damage from weather events?” Possible adaptation strategies appropriate for beef cattle producers in the Southeast included in the question were: irrigation, change in forage species, change in cattle genetics, take out more/better insurance policies, add additional water sources to pasture, and add additional shade to pasture. Then, a count variable of the number of adaptation strategies that a respondent answered “likely” or “very likely” to adopt was included as No. of Adaptation Strategies.

To control for producer demographic characteristics, we account for the respondent’s Age as a continuous variable, whether she/he is Married as a binary variable (equal to 1 if the producer is currently married), and producer’s education (Some College or Higher) as a binary variable equal to 1 if the producer has completed some college, obtained an associate’s degree, bachelors, masters, PhD or any other professional degree. Last, we added a binary variable Beginning Farmer equal to 1 if the producer reported having 10 years or less experience in farming, to account for experience.

Farm characteristics and diversification

Table 2 reports summary statistics for the farm characteristics and diversification measures used to describe operations in the sample. The average size of the operation is 552 acres, with slightly more than 8 pastures. This is consistent with other studies that showed Southeast cattle operations are small, either measured in terms of acreage or total herd size, i.e., (Asem-Hiablie et al. Reference Asem-Hiablie, Rotz, Stout and Place2018; McBride Reference McBride2011). Previous studies have shown the prevalence of both joint cow-calf and stocker, and stocker-only operations as being the dominant operation structure for beef cattle operations across the Southeast (Asem-Hiablie et al. Reference Asem-Hiablie, Rotz, Stout and Place2018; McBride Reference McBride2011). For our sample, 68% of respondents indicated they have a stocker or pre-conditioning program.

More than 60% of respondents indicated that they produce commodities other than just beef cattle. The most common option as a second commodity group was another type of livestock, followed by row crops. Less common commodities produced included fruit or orchard crops (11%), vegetables (12%), and forest or timber products (9%). Further, approximately one-quarter of the sample was highly diversified, indicating they produce three or more commodities out of the seven available in the survey.

Table 3 displays the overall unconditional distribution of the technology adoption and the share (percentage) of respondents in each level (seven levels) with associated measures of diversification. Table S1 in the supplementary material, also displays the adoption of specific technologies, but by diversification measures. While the average number of technologies adopted was 3.58 technologies, differences exist between the number of technologies adopted based on commodity groups. Note that no one adopted zero technologies or more than seven technologies. Beef cattle producers who also produce fruit or orchard crops adopted an average of 4.47 of the listed technologies. The average number of technologies adopted is higher for this group of producers compared to beef cattle producers who also produce vegetables or beef producers who also produce row crops.

Producer perceptions

About half of the respondents are enrolled in the USDA Livestock Indemnity ProgramFootnote ⁹ and are concerned about cattle prices. We also find that the average producer in our sample adopts more than three of both the conservation practices and the practices implemented to combat damage from weather events. While conservation practice adoption and technology adoption are not synonymous, adoption decisions have been shown to be correlated for certain producer groups (Kolady and Van der Sluis Reference Kolady and Van der Sluis2021). Approximately, one quarter of respondent’s state that the greatest variation in income is accredited to unexpected expenses. Given that technological innovations have been shown to be costly to adopt, the perceived source of income variation could be associated with adoption (Gillespie et al. Reference Gillespie, Kim and Paudel2007; Makinde et al. Reference Makinde, Islam, Wood, Conlin, Williams and Scott2022; Rosa Reference Rosa2021).

Survey demographic characteristics

The demographic characteristics of our sample are presented in Table 5. The average age of the sampled producer is approximately 45 years, which is lower than the average age of US farmers (58.1 years) (USDA-NASS 2022); however, this is expected due to the online nature of the survey. Most of the survey participants are married and have attended at least some college. A small group of the sample (∼10%) indicated they had 7 or more interactions with extension agents or attended Extension meeting, event, or field day in the last 12 months.