Innovation helps restore normalcy to people’s lives, especially during crises like the COVID-19 pandemic, reducing the damage caused by economic or human health shocks (Carlsson-Szlezak et al., 2020). According to Rogers (2003), innovation diffusion is how a new idea spreads and is adopted by people in a community over time. The Diffusion of Innovations (DoI) theory provides a framework to explain why some innovations become popular while others do not, including the process, elements (e.g., innovation, communication channel, time, and social system), and rate of innovation adoption. In the DoI theory, innovation can be an idea, practice, or object recognized as novel by an individual or other unit of adoption. Five qualities of the innovation that affect its adoption rate and likelihood are proposed: (1) relative advantage, (2) compatibility, (3) complexity, (4) trialability, and (5) observability (Rogers, 2003). Potential adopters’ perceptions of these characteristics affect acceptance and use of the innovation.

Time is another central element when discussing innovation adoption. People go through six stages to decide whether to adopt an innovation, including being aware of the innovations (knowledge stage), proactively seeking the information (persuasion stage), deciding to accept or reject the innovations (decision stage), and figuring out how the innovation works (implementation stage), confirming to use (confirmation stage), and ultimately adopting innovation (adoption stage). Furthermore, Rogers (2003) purported that people can be divided into five categories based on the relative speed of adoption, assuming that the speed at which people accept innovation follows a normal distribution. The categories range from those most likely to adopt innovation quickly to those who may never embrace innovation as innovators (2.5%), early adopters (13.5%), early majority (34%), late majority (34%), and laggards (16%). According to the definition by Valente (1996), early adopters are individuals who adopt a particular concept, technology, or practice more than one standard deviation earlier than the average adoption time. The early or late majorities are individuals whose adoption time falls within one standard deviation earlier or later than the average. Lastly, laggards adopt later than one standard deviation from the mean.

The DoI theory is one of the most commonly used theories for understanding innovation adoption and implementation. Research found some differences in basic demographic variables among people in different rates of innovation adoption groups (Avery et al., 2010; Gulati and Williams, 2013; Clarke et al., 2016; DiGiacomo et al., 2023). More specifically, people who are younger, more educated, and have higher social status or income are more likely to adopt innovations faster (Rogers, 2003). Previous studies did not consider the relationships between people’s innovation adoption and non-sociodemographic differences like political beliefs or residency when focusing on food-related topics as an innovation. Based on the DoI theory, innovative information can be more effectively promoted and communicated if it accounts for measurable individual differences. In particular, it would be more conducive to planning future communication or education strategies for late majorities or laggards.

Many studies have used the DoI theory in the context of food supply chains. For instance, Qader et al. (2023) emphasized that perceived compatibility, relative advantage, and religious beliefs positively influence the adoption of halal meat supply chains, while perceived complexity has a negative impact. Similarly, Yi et al. (2022) identified key factors influencing the adoption of traceability practices, including low awareness of the importance of food recalls, perceived relative advantages, and cultural compatibility issues. These studies demonstrate the feasibility of using the DoI theory to analyze food supply chains. However, most research focuses on producers’ adoption of new technologies or concepts in food production rather than consumer perspectives. To our knowledge, no studies have applied the DoI theory as a framework to investigate consumers’ values and purchasing behaviors toward local food and short food supply chains during the pandemic. Thus, this research seeks to bridge this gap in the existing literature, assessing the applicability of the DoI theory in this context.

In this study, consumers’ initiative to purchase local food through short food supply chains for the first time during the COVID-19 pandemic was considered the innovation in the food acquisition process. Such food supply chain shifts were notable in the United States during the pandemic (DuPuis et al., 2022), when these shorter, more localized supply chains helped alleviate food shortages and promote a stable supply (Butu et al., 2020; Lioutas and Charatsari, 2021), thereby reducing food insecurity. Furthermore, short food supply chains reduced the workload of mainstream food channels (Richards and Rickard, 2020) and solved problems in food distribution and price volatility (Butu et al., 2020; Lioutas and Charatsari, 2021). This study also considered the adoption of, at the time, less traditional logistics like online ordering for food purchases as a response to the health risks associated with more traditional logistics, like in-store purchases, as part of the food acquisition innovation.

A substantial body of research has delved into consumers’ preferences and values concerning local food within the United States (Costanigro et al., 2011; Godette et al., 2015; Abrams and Soukup, 2017). While studies have scrutinized individuals’ personal attributes about their food choices (Rossi et al., 2017), the prevailing wisdom in the realm of food marketing and ideology advocates was for a focus on consumers’ values and perceptions (Costanigro et al., 2011; Grebitus and Dumortier, 2016; Abrams and Soukup, 2017; Reich et al., 2018; Mehrolia et al., 2021). It has been observed that consumers tend to gauge the worth of local foods based on a multifaceted spectrum of factors, including but not limited to quality attributes like taste and freshness (Spiller and Belogolova, 2017), sustainable and/or organic production practices (Abrams and Soukup, 2017; Ellison et al., 2017), the reduction of transportation distance from farm to table (McIlvain-Newsdad et al., 2004), and the prevailing perception that local food is inherently healthier (Robinson and Smith, 2003; Haws et al., 2017). Marketing messages about local foods may more likely resonate with consumers if they believe these values are socially desirable (Costanigro et al., 2011) and align with their personal beliefs (Rossi et al., 2017). Besides, research has identified causal connections in food preferences, from concerns to attitudes and specific behaviors (Kang et al., 2015; Dang and Tran, 2020).

The surge of interest in local food and short food supply chains underscores the importance of a community-centric and decentralized approach to food production and distribution (Barska and Wojciechowska-Solis, 2020; Lioutas and Charatsari, 2021). Consumers’ values for local food are positive, as are their willingness to pay higher prices (Hempel and Hamm, 2016; Printezis et al., 2019). This shift in attitudes, if translated into behavioral changes, could address immediate challenges in the food supply chain and lay the foundation for a more stable and sustainable food system, especially during times of crisis. Furthermore, it underscores prioritizing food security as a core value during crises, ensuring immediate relief and enduring resilience in a community-driven food network (Hobbs, 2020). It also includes endeavors to reduce food waste and promote the consumption of healthier, more sustainably produced food (Schmutz et al., 2018).

Using a national online panel of people, who identified as the main grocery shoppers for their household, two research objectives were pursued to help design educational plans and marketing strategies for people in different innovation adoption groups based on the Diffusion of Innovations framework (Rogers, 2003). The research objectives were:

We conducted this quantitative study from July 13 to August 18, 2021, collecting data from 1,002 adults (18 years or older) in the United States through an online survey administered by Qualtrics. Our sample was stratified to reflect the 2020 U.S. Census population in terms of race, ethnicity, geographic region, age, and income. Qualtrics employs advanced measures like internet protocol (IP) address checks and digital fingerprinting technology to prevent duplicate responses and ensure the validity of the data (Qualtrics, 2019). To acknowledge the valuable time and input of participants, we provided compensation, which encouraged full engagement with all survey questions. To maintain the integrity of our analysis, we excluded any participants who did not complete the entire survey.

Non-probability sampling was used in this study due to its advantages in accessibility, cost-effectiveness, and higher response rates compared to traditional survey methods (Creswell and Creswell, 2018; Lamm and Lamm, 2019). This approach allows for exploring a broader range of perspectives within the population, which is particularly advantageous in studies focusing on societal behaviors and opinions (Lamm and Lamm, 2019). Though it does not provide a fully representative sample, it was considered suitable for making population estimates in this context, especially given limited resources and time constraints (Baker et al., 2013).

This instrument was part of a larger study designed by the researchers with several constructs developed using the DoI theory from Rogers (2003). Three sections of the questionnaire were used to fulfill the objectives of this study: (1) innovation adoption rate, (2) food acquisitions, and (3) food safety concerns. A panel of experts from the University of Minnesota, University of Florida, Kansas State University, University of California-Irvine, and the University of Wisconsin-Madison reviewed the instrument for face and content validity. Their expertise helped refine the questions to ensure comprehension and pertinence to the research objectives. A pilot test (n = 50) was conducted to identify any issues with the survey design, such as unclear questions, and to estimate the reliability of the constructs. It also used reliability to determine if items should be dropped or potentially added.

The operational definition of short food supply chains in our instrument refers to a distribution system in which the number of intermediaries between the producer and the consumer is minimized, often involving direct sales or limited levels of middlemen. This supply chain could help reduce the distance food travels from farm to table, thereby enhancing transparency, system stabilization, and connection between producers and consumers during the pandemic (Aguiar et al., 2018; Cappelli and Cini, 2020; Paciarotti and Torregiani, 2021).

The first section, innovation adoption rate, was designed to measure the extent to which individuals are quick to try new practices, particularly using short food supply chains and online grocery shopping during the pandemic. Statements were related to local food and the four qualities of the DoI theory (relative advantage, compatibility, complexity, and observability) that may influence the speed and likelihood of adopting short food supply chains in response to crisis (Rogers, 2003). Examples of the statements are “Local/regional foods are overall better for me” (relative advantage), “Local/regional food is compatible with the needs of my family” (compatibility), “Switching to local/regional foods is an easy thing for me to do” (complexity), and “I have observed the process of someone acquiring local/regional foods” (observability). All statements used a 5-point Likert scale ranging from 1 = strongly disagree to 5 = strongly agree. Eight items, two questions on each of the four qualities, were included in the scale, and each respondent got an innovation adoption score ranging from 8 to 40 points (α = 0.83, M = 27.34, SD = 5.62). People with more positive values toward local food receive higher points, which means they hold an open attitude toward changing food acquisition channels and have higher innovation adoption rates in this study. Assigning higher scores to positive values toward local food allows the researchers to correlate these attitudes with a greater likelihood of adopting new food acquisition practices.

The data were used further to divide the sample into five levels of different “innovation adopter groups” with the combined index including all of the DoI attributes. Based on Rogers (2003) and Valente (1996), respondents’ innovation adoption scores in this study between the average and above/below one standard deviation can be early (28–32 points) and later majority (22–27 points). Those with scores higher than a standard deviation above the average (33–38 points) were early adopters and higher than two standard deviations (> 38 points) were innovators. Lastly, people scoring lower than a standard deviation below the average (< 22 points) were defined as laggards.

The second section about food acquisitions aimed to identify changes in how respondents acquired food, focusing on new purchasing methods that emerged during the pandemic. The questionnaire included questions about how people obtained food during the pandemic, focusing on both the outlets (e.g., supermarket, farmers’ market, convenience store) and logistics (e.g., in-store/eat-in, online orders for at-home delivery). Specifically, the study focused on outlets related to short food supply chains, such as “farmers’ market (e.g., farm stands and community-supported agriculture (CSA))” and “direct from farmers and/or food makers.” These outlet items were re-coded to represent participants who purchased from these channels for the first time during the pandemic (coded as 1) and those who did not (coded as 0). Regarding logistics, the study analyzed methods for obtaining food during the pandemic that minimized public health risks, which included options like “online order for at-home delivery,” “online ordering for bundled at-home delivery (e.g., DoorDash, Instacart),” and “online order for in-store or curbside pickup.” These logistics were also re-coded to identify participants who utilized these safer methods for the first time during the pandemic = 1 and those who did not = 0. These were matrix-style questions where participants could select different logistics for each outlet that applied to them during the pandemic.

The last section of the questionnaire focused on participants’ concerns about food safety issues. Specifically, two questions were asked to assess their level of concern during the pandemic. The first question inquired about their level of concern regarding “where food comes from,” while the second question explored their concerns regarding the “safety of food we consume (from COVID-19).” Participants were asked to rate their level of attention to food safety issues during the pandemic compared to non-pandemic periods, using a three-point scale where 1 indicated “more attention,” 0 represented “the same amount of attention,” and − 1 indicated “less attention than before.” Previous studies have suggested that people’s values can influence their food selections, and therefore, this study sought to control for participants’ concerns.

We also collected and controlled for participants’ demographic variables, which included age, household income, and highest educational level. Additionally, participants were asked to indicate their political preferences, which were categorized into five options, ranging from very liberal (coded as 1) to very conservative (coded as 5). Participants’ place of residence was coded into five categories: “a farm in a rural area,” “rural area, not a farm,” “urban or suburban area outside of the city limits,” “subdivision in a town or city,” and “downtown area in a city or town.” Their geographic region in the United States, such as the Midwest, Northeast, South, and West, was also collected. These demographic variables were included as control variables in the regression models.

SPSS 29 was used for analysis. Descriptive statistics, one-way analysis of variance (ANOVA) with post hoc Scheffe analyses, and several regressions were used to answer the research objectives. Descriptive statistics were used to analyze sample characteristics, while ANOVA was used to compare differences in innovation adopter groups between different demographic variables.

To conduct multivariable regression analysis, the categorical data types, such as age and annual family income, were transformed into numerical values represented by their respective median values within each group. For example, household income from $15,000 to $24,999 was transformed to $20,000. For the educational level, those with a 4-year college degree and higher degree were re-coded as 1, and others were 0. Other characteristics were converted using dummy variables. Regarding political beliefs and place of residence, the reference category for comparison with others is neutral to simplify interpretation and highlight differences. To elaborate, moderate political beliefs are the basis for comparing with those that hold liberal or conservative political beliefs. This makes it easier to interpret the effect of being more conservative or liberal relative to being moderate. Those living in an urban or suburban area outside of the city limits were compared with those who live in rural and city areas, allowing for a clearer understanding of how location makes an influence. For the geographic region variable, since the majority of participants (41.4%) reside in the South, it was designated as the reference category for comparison with other regions. Regarding the level of innovation adoption, the laggards were selected as the base category for comparison, as it facilitates a clear interpretation of the study’s findings by contrasting the other groups against this reference group.

Before conducting regression analyses, we checked for and ensured that the data met the necessary statistical assumptions for linear models. This included testing for linearity, multicollinearity, and independence of errors.

The 1,002 respondents were classified into five different levels of innovation adopter groups, as depicted by Rogers (2003) and shown in Figure 1. The demographic variables of the respondents in different groups are shown in Table 1. For some respondents who chose not to disclose their political beliefs and residency, missing values were replaced with the average values in the analysis.

Comparing demographic data across the innovation adopter groups yielded some significant differences in age, income, educational level, and political belief. Table 2 shows the ANOVA results. The post hoc results indicated no significant differences for age between different innovation adopter groups, F(4, 997) = 2.57, p = 0.04, partial η2 = 0.01. However, certain trends were observed. For instance, respondents with slower adoption rates tended to be older in age. In addition, early adopters tended to have a higher household income, F(4, 997) = 5.63, p < 0.001, partial η2 = 0.02, and a higher educational level, F(4, 997) = 4.81, p < 0.001, partial η2 = 0.03, than the late majority and laggards. Moreover, respondents who held liberal political beliefs typically had a slower speed of innovation adoption than those with conservative beliefs, F(4, 997) = 4.03, p = 0.003, partial η2 = 0.02. On the other hand, participants’ place of residence and geographic region showed no significant differences across the various innovation groups.

To scrutinize the disparities in food acquisition channels, distinguishing between short food supply chain sources and non-short food supply chain sources, this study specifically defined short food supply chain sources as either farmers’ markets or direct purchases from farmers. These two local food outlets emphasize the proximity of food production and acquisition. Figure 2 presents the percentages of respondents, categorized by innovation adopter groups, who indicated their use of these specific food acquisition channels.

The linear regression results are presented in Table 3. The explanatory variables used in the regression model accounted for approximately 25.8 and 29.7% of the variance in the respondents’ choices of acquiring food from farmers’ markets and direct sources, respectively. The results indicated a significant positive association between respondents who demonstrated a higher readiness to adopt innovative food acquisition practices as measured by the DoI innovation adopter groups and their self-reported use of local food outlets during the pandemic. Moreover, the regression analysis identified other factors influencing the first-time use of local food outlets during the pandemic. It was observed that accounting for their innovation tendencies, people with younger ages (β = −0.40, p < 0.001), higher household incomes (β = 0.11, p < 0.001), and those who identified themselves as more conservative in political beliefs (β = 0.07, p = 0.04) were more likely to acquire food from farmers’ markets for the first time during the pandemic, while those with younger ages (β = −0.45, p < 0.001), higher household incomes (β = 0.11, p < 0.001) and educational levels (β = 0.08, p = 0.01), and identified themselves as more conservative in political beliefs (β = 0.08, p = 0.01) were more likely to acquire food directly from farmers. Additionally, participants who expressed higher levels of concern regarding the origin of their food during the COVID-19 pandemic also exhibited a stronger inclination toward acquiring food for the first time from farmers’ markets (β = 0.06, p = 0.04). On the contrary, it is worth highlighting that people who expressed heightened concern about the safety of food (from COVID-19) during the pandemic, compared to before it, were associated with less tendency to purchase food through these local food outlets, all else equal.

Moreover, the COVID-19 pandemic significantly impacted food purchasing logistics, leading to a notable surge in online food ordering as a prominent method of acquiring food. As online ordering represents an innovative approach compared to traditional in-store purchases, it became crucial to assess whether respondents who used local food outlets also adopted online ordering methods. Next, the study considered two specific online ordering methods for acquiring food from local food outlets: “online ordering for at-home delivery” and “online ordering for in-store or curbside pickup.” Figure 3 presents the distribution of respondents, categorized by their innovation adopter groups, who reported using online ordering methods from local food outlets, like farmers’ markets or direct purchases from farmers.

The subsequent multivariate regression analysis revealed a significant positive correlation between early adopters and the preference for purchasing food at local food outlets through online ordering for at-home delivery (Table 4). Several demographic variables were also identified as influential factors impacting this behavior. Participants with younger ages (β = −0.30, p < 0.001) and higher educational levels (β = 0.13, p < 0.001) displayed a stronger inclination toward acquiring food at local food outlets through online ordering for at-home delivery. Additionally, individuals residing in the northeastern region of the United States (β = 0.08, p = 0.01) were more likely to exhibit this behavior.

However, the findings indicated no significant association when examining the relationship between respondent innovation adopter groups and acquiring food at local food outlets through online ordering for in-store or curbside pickup. Similar to online ordering for at-home delivery, people of younger ages (β = −0.35, p < 0.001) and higher educational levels (β = 0.07, p = 0.04) were more likely to purchase food at local food outlets through these logistics during the pandemic, in addition to those earning higher household incomes (β = 0.15, p < 0.001).