This item is only available as the following downloads:
1 PERCEPTIONS OF SOIL TESTING AMONG HORTICULTURAL FARMERS IN THE UNITED STATES By COREY HANLON A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2013
2 2013 Corey Hanlon
3 To my family and loved ones
4 ACKNOWLEDGMENTS I would like to thank my family and friends for their extensive support throughout t his entire process. Your loving support helped me to get where I am today. Thank you for believing in me. All of you have given me confidence to get everything that I want out of life. Dr. Koenig, and Dr. Clark. The experience and advice they gave me was exceptional, both academically and professionally. I would especially like to thank my committee chair, Mickie, for teaching me to think for myself.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 ABSTRACT ................................ ................................ ................................ ..................... 9 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 11 Agricultural Intensification ................................ ................................ ....................... 11 Sustainable Agriculture ................................ ................................ ........................... 15 Research Question ................................ ................................ ................................ 17 Definitions ................................ ................................ ................................ ............... 18 2 LITERATURE REVIEW ................................ ................................ .......................... 19 Theoretical Perspective ................................ ................................ .......................... 19 Precursors to Diffusion of Innovations ................................ .............................. 19 Diffusion of Innovations ................................ ................................ .................... 20 Attribute studies ................................ ................................ ......................... 22 Limitations to diffusion studies ................................ ................................ ... 28 Size of Agricultural Operations ................................ ................................ ............... 33 Soi l Testing and Nutrient Management ................................ ................................ ... 34 3 METHODOLOGY ................................ ................................ ................................ ... 37 Research Design ................................ ................................ ................................ .... 37 Sample Selection ................................ ................................ ................................ .... 37 Instrumentation ................................ ................................ ................................ ....... 40 Comparison Groups ................................ ................................ ......................... 40 Variable Development ................................ ................................ ...................... 40 Cognitive Testing ................................ ................................ .............................. 41 Test of Revised Instrument with Participants ................................ .................... 45 4 RESULTS ................................ ................................ ................................ ............... 51 Sample Statistics ................................ ................................ ................................ .... 51 Demographics ................................ ................................ ................................ .. 51 Post hoc Assignment to Groups ................................ ................................ ....... 52 Test for Normal Distribution ................................ ................................ .................... 52 Test for Central Tendency ................................ ................................ ...................... 53 Spearman Rank Order Correlation ................................ ................................ ......... 53 Research Validity and Explanatory Power ................................ .............................. 56
6 Hypothesis One ................................ ................................ ................................ ...... 57 Hypothesis Two ................................ ................................ ................................ ...... 58 5 DISCUSSION ................................ ................................ ................................ ......... 68 Research Que stion ................................ ................................ ................................ 68 Comparison to Attribute Literature ................................ ................................ .......... 69 Theoretical Contributions of this Study ................................ ................................ ... 72 Pro Innovation Bias ................................ ................................ .......................... 73 Innovation Price ................................ ................................ ................................ 74 Conservation Technology ................................ ................................ ................. 74 Comparison to Attribute Correlations Literature ................................ ...................... 75 Farm Size ................................ ................................ ................................ ............... 77 Demographic Covariance ................................ ................................ ....................... 80 Limitations ................................ ................................ ................................ ............... 85 6 CONCLUSIONS ................................ ................................ ................................ ..... 90 Implications for Future Research ................................ ................................ ............ 91 Theoretical Contributions ................................ ................................ ........................ 92 Policy Implications ................................ ................................ ................................ .. 93 APPENDIX: PERC EPTIONS OF SOIL TESTING INDEX ................................ ............. 94 LIST OF REFERENCES ................................ ................................ ............................. 104 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 115
7 LIST OF TABLES Table page 3 1 attributes that may influence adoption of technologies by large and small farmers, before testing for final instrument use ................................ .................. 50 3 2 attributes that may influence adoption of technologies by large and smal l farmers, after testing for final instrument use ................................ ..................... 50 4 1 Gender and age of respondents in a study of factors affecting adoption of soil testing by large and small farmers, 2012 ................................ ............................ 60 4 2 Gender and years of experience for respondents in a study of factors affecting adoption of soil testing by large and small farmers, 2012 .................... 60 4 3 Farm size and location in watershed with Basin Management Action Plan of participants for respondents in a study of factors affecting adoption of soil testing by large and small farmers, 2012 ................................ ............................ 60 4 4 Farm location and participant ethnicity for respondents in a study of factors affecting adoption of soil testing by large and small farmers, 2012 .................... 61 4 5 Results of the Shapiro W ilks test for normality for five theoretically defined attribute variables separated into post hoc groups based on farm size, in a study of factors affecting adoption of soil testing by small and large farmers ..... 62 4 6 Results of the Shapiro Wilks test for normality for five theoretically defined attribute variables, separated into post hoc groups based on farm size, after data had been through logistic transformation ................................ .................... 63 4 7 Results of the Mann Whitney U test for central distribution for each of five theoretically defined attribute variables, in a study of factors affecting adoption of soil testing by small and large farmers ................................ ............. 63 4 8 Results of the Spearman rank order correlation for five theoretically defined attributes, for small and large farms combined, in a study of factors affecting adoption of soil testing by small an d large farmers. ................................ ............ 64 4 9 Results of the Spearman rank order correlation for five theoretically defined attributes, separated into comparison groups of small and large farmers, in a study of fac tors affecting adoption of soil testing by small and large farmers. .... 64 4 10 Spearman rank order correlations for demographic covariates, for small and large farms combined, in a study of facto rs affecting adoption of soil testing by small and large farmers. ................................ ................................ ................ 64
8 4 11 Spearman rank order correlations for demographic covariates, separated into comparison groups of small and large farms in a study of factors affecting adoption of soil testing by small and large farmers. ................................ ............ 65 4 12 Spearman rank order correlations for theoretically defined attribute variables and demographic va riables, for both small and large farms, in a study of factors affecting adoption of soil testing by small and large farmers. .................. 66 4 13 Spearman rank order correlations for theoretically defined attribute variables and demographic variables, separated into comparison groups of small and large farms ................................ ................................ ................................ .......... 67
9 Abstract of Thesis Presented to the Graduate School of the University of Florida in P artial Fulfil lment of the Requirements for the Degree of Master of Science PERCEPTIONS OF SOIL TESTING AMONG HORTICULTURAL FARMERS IN THE UNITED STATES By Corey Hanlon May 2013 Chair: Marilyn E. Swisher Major: Interdisciplinary Ecology Concerns about water quality and other environmental impacts have increased interest in sustainable agriculture in recent years. Sustainable agriculture techniques like soil testing are simple, inexpensive ways to protect water quality. However, many farmers choose not to use soil te sting. This study examined whether there are differences in how large and small farmers perceive soil testing. It also look ed into some general demographic attributes to see if they affect the soil testing. The theory of d iffusion of i nnovations provided the framework for this research. The sample consist ed of 277 horticultural farmers throughout the United States. An online index measured farmer perceptions of relative advantage, compatibility, complexity, trialability and observab ility of soil testing as well as the age, years of experience, gender and acres in production. Data w ere analyzed using a Mann Whitney U test and Spearman rank order correlations. Results show that perceptions of complexity and observability are significa ntly different between small and large farmers. Females report increased perceptions of relative advantage, compatibility and observability. Years of experience and acres in
10 production are positively correlated with observability. Age had no significant co rrelations with any of the soil testing attribute s Findings suggest that future research should explore which domain to which the theory of diffusion of innovations applies, and which other demographic attributes might affect perceptions of soil testing.
11 CHAPTER 1 INTRODUCTION Agricultural Intensification The estimated world population was 7,021,836,029 people (as cited in Central works in agriculture, but less than 0.7% of the U.S. population works in agriculture (as cited in Central Intelligence Agency, 2012). The land in agricultural use has increased steadily for the past two centuries (Goldewijk, 2001). Cropland increased by 27% from 960 to 1208 million hectares from 1961 to 2005 (Burney, Davis & Lobell, 2010, p. 12053). The current estimate for total agricultural land is about 30% +/ 10% of the includes both croplands (12%) and pastures (28%) (Ramankutty, Evan, Monfreda & Foley, 2008). Global demand for food is increasing because of population growth and dietary changes. Nine hundred and twenty million people were undernourished in 2010, up from the 840 million undernourished in 1990 (Fo od and Agriculture Organization, 2008, p. 6). People want more to eat, and they desire foods that are higher in caloric demand, which typically require more land to produce (Tilman, Balzer, Hill & Befort, 2011). For instance, as people become wealthier, th eir demand for animal protein and foods dense in nutrients increases. Raising livestock requires more land than grain production. Both of these demands call for an increase in agriculture. There are two main ways to increase the amount of crops or animal products available for harvest. We can claim more land for agricultural use, or we can increase 2012, a 123% increase. Global crop production rose by 162% during the same period,
12 indicate a combination of intensification of current agricultural production on land currently in use reducing post harvest loss and increasing land in use could meet worldwide food demand in 2050 (Tilman, Balzer, Hill & Befort, 2011). While the model takes soil fertility, climate, and water availability into account, the authors admit that intensification or expansion of agriculture may not be practically fe asible. Expansion of agriculture implies the increasing use of marginal lands. T he ideal arable land is already developed so expanding agriculture will be moving into lands that might require increased inputs or management for the same amount of yield A gricultural expansion into natural ecosystems has negative impacts on biodiversity. A gricultural processes on marginal land lead to a greater chance of ca using environmental degradation, including soil erosion and impacts on water quality. This potential f or damage occurs because marginal lands require a greater input of fertilizers, water management is often more difficult due to distance from a reliable water source, and increased slope can cause a variety of problems. Agricultural intensification requir es increased inputs for crop and animal production. Water and nutrients are among the main inputs that will potentially become expensive and/or hard to come by for farmers in the future Pesticides, human labor and machinery are also likely to increase in cost. Agricultural water use was responsible for about 85% of the global freshwater consumption at the beginning of the 21st century (Shiklomanov & Rodda, p 375). Withdrawals exceed recharge in many groundwater sources, such as the Ogallala Aquifer in the U.S., which ultimately reduces the volume of water availability in the aquifer and leads to increased energy requirements for
13 pumping. Global agricultural water use increases ann ually, although not in the U.S. Agriculture that is currently in semi arid or arid lands may encounter water prices that force management adaptations to drought tolerant crops (Bower, 1994). Having enough water for irrigation is already an issue in many arid lands, such as the Middle East and Southwest US. In other areas, such as wi th the Floridan aquifer, salt water intrusion due to overdrafting is an issue related to water quantity. Downstream ecosystems may become water stressed or water limited due to groundwater drop or reduced water availability (Maxwell & Kollet, 2008). Nitrat e levels in bodies of water, both surface and subsurface, have been steadily increasing since the middle of the twentieth century (Falkenmark, Finlayson & Gordon, 2007). The main fertilizer inputs for crops are nitrogen, phosphorus and potassium. Humans ha ve caused increased rates of nitrogen input into the terrestrial nitrogen cycle, increased atmospheric concentrations of nitrous oxide increased acidification of soils and bodies of waters in many regions of the globe and increased amounts of nitrogen ent ering marine systems through surface water flow (Vitousek, et al., 1997). Nutrient application to cropland can affect water quality. For example, addition of limiting nutrients to some aquatic ecosystems can result in eutrophication. M ost often nitrogen o r phosphorus leaches into groundwater or otherwise becomes susceptible to runoff into water bodies. Those waters can then become impaired and inappropriate for drinking, recreation or fishing. Pesticides and other agricultural chemicals entering water sour ces are also cause for concern (USGS, 2011). Approximately 10 12% (1.4 1.7 gigatons of carbon equivalent) of total anthropogenic greenhouse gas emissions came from agricultural production in 2005
14 (Smith, et al., 2007). These emissions were mainly in the d irect forms of nitrous oxide and methane from fertilizer use (38%), manure management and ruminant methane emissions (38%), cultivation of wetland crops such as rice (11%) and burning of agricultural lands ( 13%) (US EPA, 2006). There are also indirect emis sions due to use of industrially produced fertilizers and chemicals, as well as the use of homes and machinery that burn fossil fuels. Land use change, such as clearing forest for agriculture or harvesting of forest products accounted for additional emissi ons of 1.5 gigatons of carbon equivalent (Canadell, et al., 2007). Soil erosion causing sedimentation of waters and losses in soil fertility are also concerns associated with increased intensification of agriculture. Agriculture is the leading cause of gl obal soil erosion (McCune, et al., 2011). Soil fertility is affected mostly through the cycle of crop growth and harvest removing nutrients and then fertilizer applications and nitrogen fixing plants adding nutrients. The struggle for farmers is to keep th e soil fertile despite the removal of nutrients with every harvest, especially on lands in poorer nations (United States Department of Agriculture, 2007a; Tilman, Balzer, Hill & Befort, 2011). Improper water management can lead to increased soil salinity o r even desertification (Jacobsen & Adams, 1958; Pereira, Oweis & Zairi, 2002). Threats to biodiversity due to land clearing and fragmentation (Dirzo & Raven, 2003; Kollner, 2000), habitat loss including deforestation and loss of ecosystem services (Rudel, et al., 2009), pesticide resistance (United States Department of Agriculture, 2007a) and effects on wildlife (Le Feon, et al., 2010) are also all problems that can be associated with intensification of agriculture.
15 Sustainable Agriculture Congress defined system of plant and animal production practices having a site specific application that will, over the long term satisfy human food and fiber needs; enhance environmental quality and the natu ral resource base upon which the agricultural economy depends; make the most efficient use of nonrenewable resources and on farm resources and integrate, where appropriate, natural biological cycles and controls; sustain the economic viability of farm oper ations; and enhance the quality of life for farmers and definition is vague by necessity, because sustainable producers to think about the long t erm implications of practices and the broad ng term, stewardship of land, air and water, as well as quality of life for farmers, ranchers and their communities (SARE, 1997), or the triple Thebaud, 2012). It is i mportant to note that sustainable agriculture is not the same as certified organic agriculture. Some commonly accepted sustainable farming practices include crop rotations and diversifications (Giller, et al., 2011), integrated pest management techniques, increased mechanical and biological weed control, use of soil and water conservation practices, soil testing, uses of animal and green manures, conservation tillage, modified rice drainage, precision agriculture, drip irrigation, sequestration of soil org anic carbon and many more. These techniques are designed to produce crops while protecting the
16 environment and protecting quality of life for the workers on the fields, as well as the community surrounding the farm and the final consumer of the product. Th ere are extensive training programs and educational services to teach farmers about these techniques. Conferences, trade association meetings and other venues disseminate information about new and old sustainable agriculture techniques. For instance, the S ustainable Agriculture Research and Education (SARE) program conducts in person trainings, online trainings, sells books and other educational tools and has an extensive website to direct interested people to other resources. The United States Department o f Agriculture (USDA) also has programs devoted to educating farmers about sustainable techniques. There are various federal, state and private grants available that help to offset the cost of participating in these trainings, as well as grants and loans th at help farmers purchase the initial cost for some of the new technologies, such as drip irrigation. Extension agents and farmers often take advantage of the training that is available to them. They learn about the new technologies or new techniques and h ow they will improve their profits, the environment and their quality of life. Extension agents are responsible for working directly with farmers to assist in implementing these techniques and technologies. Farmers still fail to implement certain sustaina ble agriculture techniques, even though they have knowledge of how to use techniques such as conservation tillage and despite the funds available to help with start up costs for expensive technologies. If farmers fail to utilize sustainable practices, the research into what techniques are the best for the triple bottom line is useless. Extension agents and scientists can educate
17 and communicate their ideas extensively, but if farmers choose not to adopt improved practices, then agriculture will not be susta inable. Researchers must study why farmers fail to use the techniques in order to develop effective trainings or programs. Therefore, it is up to scientists to learn what affects adoption to increase adoption in the future. Research Question There are a va riety of techniques and technologies used in sustainable agriculture. The innovation that is the focus of this study is the use of soil testing. Soil testing permits farmers to determine nutrient levels in the soil so that they do not need to apply unneede d fertilizer. Reducing nutrient input is economically advantageous and prevents leaching and runoff of excess nutrients. Soil tests are available through County Extension offices for generally low prices (less than $10 a sample) or through private companie s. They are also easy to collect, but many farmers still do not use any form of soil testing. This study examines why farmers choose not to use soil tests to determine nutrient levels in their fields. I addressed the following specific research question in this study: Is there a difference in the perceived innovation attributes of soil testing between small and large farmers? The study is confined to the use of soil testing by farmers growing horticultural crops for fresh market sale. I will test two hypoth eses in this study to address the research question: H1 : There is a difference in how small farmers and large farmers perceive the innovation attributes. H2 : There are covariate demographic factors that affect the overall perception of innovation attribut es.
18 Definitions Some of the terms used in this study have vague or multiple definitions. For clarity, the definitions used are listed below: or other unit of adoption" (Rog ers, 1962). Innovation Attributes: The five qualities that Rogers (1962) uses to describe an innovation are relative advantage, complexity, compatibility, trialability and observability. Soil Testing: A test requiring a collection and compilation of sample s from a field or part of a field that is then tested for percentages of nitrogen, phosphorus and potassium, as well as pH. The tests are available through County Extension Offices or through private companies.
19 CHAPTER 2 LITERATURE REVIEW I will address soil testing from a diffusion of innovations theoretical perspective. There are many theories that address how people decide when and why to adopt a new technology, as well as theories that address why certain technologies are preferred over others. Howeve r, theory of diffusion of innovations combines innovation attributes with suggested reasons for why people choose to adopt or not adopt an innovation, including characteristics of the adopters themselves. This study focuses on how the differences between p relate to their adoption of that innovation (soil testing). In this chapter, I will discuss the theoretical background related to diffusion of innovations, the history of sustain able agriculture movements and delve into soil testing, farm sizes and horticultural crops in the United States. Theoretical Perspective Precursors to Diffusion of Innovations Gabriel Tarde published the predecessor to theory of diffusion of innovations i n 40). His observation that innovations can be adopted or rejected was crucial for the future of diffusion research. He also observed that the rate of adoptions for innovations varies over an S shaped curve, with few people adopting at the beginning of the i availability, then more and more until it was nearly exponential adoption rates, slowing to perceptibly slower rates after a while.
20 conducted by Ryan and Gross (1943) studying the adoption rates for hybrid corn seed in Iowa. Few farmers adopted the hybrid seed for multiple years, despite administrators and researchers touting the effectiveness of it. The adoption rate of the hybrid seed followed the S shaped curve that Tarde had observed (1903). Five years after the introduction of the seed there were practically zero users. Thirteen years after the release of the hybrid seed, 99% of farmers were utilizing it (Ryan & Gross, 1943). This study was the first time that diff erent classes of adopters were described ; early adopters and Gross also looked into how information about the innovation spread and concluded that interpersonal networks were critical to diffusion, and that diffusion was essentially a social process (1943). This study became known as the classical diffusion paradigm (Rogers, 1995; Severin and Tankard, 1992). Diffusion of Innovations that he created through his personal observations, literature review on research on diffusion of innovations, as well as through focusing on the two previously mentioned studies. This theory attempts to explain ho w, why and at what speed new ideas or technology spread throughout societies. Rogers uses four main elements to determine the rate of diffusion: innovation attributes, communication channels, time and social moves from knowledge of the innovation to persuasion to use it, to a decision about whether to use it, to implementation and finally confirmation of the decision to adopt the innovation (Rogers, 1962).
21 In this theory, an innovation is any idea, product, process, concept, system, behavior or combination thereof, which is new to the individual, organization, industry, which an innovation i s communicated through certain channels over time among the defined by Rogers with the ex excepting complexity, are positively related to adoption rates. Complexity is negatively related to adoption rates of an innovation. Communication channels are the next of the four elements Rogers identified 1995). Rogers then differentiates them into mass media channels and interpersonal channels. While mass media channels are more efficient and quick, interpersonal channels are more effective at encouraging adoption. Communication about innovations through interpersonal channels seems enhanced when individuals, organizations, or groups are similar to one another (Smith, 2004).
22 Time, as the next element that affects rate of adoption, is self explanatory. Rogers theorizes a positive correlation between time a nd rate of adoption. Social systems are the final of the four elements that affect adoption rates. A social system and its inherent norms may determine the rate at which an innovation diffuses (Rogers, 2003). Factors in social systems that may affect adopt ion rates include education levels (Waller, et.al, 1998), policies in place (Jensen, Halvorsen, & Shonnard, 2011), cultural practices (Waugh, Hildebrand & Andrew, 1989), age (Rogers, 2003) and more. T he idea that targeting innovative businesspersons to ado pt innovations would lead to less innovative people follow ing suit led to the initiation of the United States Extension Service This agency was expected to help less innovative farmers adopt new technologies and techniques (Stephenson, 2003). The idea was that the social system element was the most influential component of innovation adoption. When the program did not work as expected, Rogers (2003) and Waugh, Hildebrand and Andrew (1989) criticized the rationale because it did not take into account the in novation attributes and because the diffusion tended to stop with the initial farmers contacted by the Extension Service. Attribute studies The study of how characteristics of adopters affect the rates of innovation adoption has been widely researched thro ughout many different social science fields in the last fifty years. Relatively little research had gone into how the differing attributes of relative lack of research i s odd considering that Rogers himself predicted perceived attributes of innovations may explain rate of adoption from 49 to 87 percent of decision, the nature o f communication channels diffusing the innovation at various
23 stages in the innovation decision process, the nature of the social system in which the 206, 1995). In other words, variance in rates of adoption besides the attributes of the innovation are explained by whether the person is or is not an early adopter, communication channels, social systems and the effort associated with diffusion efforts. In most published innovation studies, the rates of adoption follow the S shaped curve, with variations in the gradation of the curve (Bigoness & Perreault, 1981; Rogers, 2003). M any diffusion of innovation studies conclude that one or more attributes expla in more variance in adoption behavior than do other attributes. For example, Atwell, Schulte and Westphal (2009) stated that compatibility and observability were self identified by farmers in the Corn Belt as more important than the other three attributes. This study utilized interviews and well explained theoretical comparisons to come to their conclusions. This study was well designed and well implemented but had an inadequate sample size They also used purposive sampling, which may have introduced syst ematic biases. Waller et al. (1998) also utilized observability and compatibility, but added complexity in a study on Integrated Pest Management adoption rates, whereas Du (1999) stated that complexity and relative advantage were related to adoption more so than the other attributes. The Waller et al. research looked into potato growers in Ohio and their responses to the encroachment of potato pests. They used surveys and interviews to attempt to contact all of the potato growers in Ohio. They based their methods development on well respected literature in the field of diffusion innovation
24 were able to get information from about 50% of their intended survey participants and f rom six farmers that agreed to do an interview. While the researchers attempted to obtain a census of the Ohio potato farmers, a 50% response rate is acceptable. Results were assessed using simple correlation statistics for the surveys, yet the authors did not indicate exactly how they assessed the interview data. Assuming that the researchers used standard methodologies for thematic analysis, the interview data is reliable. However, the researchers did not identify their approach to data analysis when they should have, which leads to questionable internal validity. However, their methods to study in 1999 was based on the theory of diffusion of innovations. He was lookin g at the diffusion of the Internet in China through assessing the adopter category of respondents as well as the attributes of the innovation. One major difference from most innovation studies is that he never assessed compatibility. There was a flaw in th e design of this study b ecause th e fifth attribute was not included Therefore, the external validity of Du's study is low from a building theory perspective Chakravarty and Dubinsky (2004) agree with Du, but add trialability as an important factor. Thei r study was one of the first that looked into innovations that were disruptive innovations. They used decimalization, a part of trading stocks implemented in the United States in the earl y 21 st century. Using development methods including a literature review and a focus group, they developed an online questionnaire and
25 scores ranging from 0.61 to 0.91, ind icating a varying reliability. They used a multiple regression model to conclude the importance of complexity, observability and trialability. This survey was valid and support ed and built on theory, but their instrument needs some work to become truly rel iable. Grieshop, Zalom and Miyao (1988) and Nowak (1987), on the other hand, identified complexity alone as the most important attribute in relation to adoption rates. Grieshop, Zalom and Miyao sampled California tomato growers to assess rate of integrated pest management adoption, characteristics of growers and perceptions of the innovation. The instrument that they created was acceptable, but not outstanding in terms of reliability or internal validity. They were confident in their results, but the conclu sions of their study as a whole should not be generalized to a population outside of tomato growers, or perhaps horticultural producers, because of their limited internal validity and reliability. They used theory to develop the design and methods of their research, but their conclusions are much more specific to integrated pest management than they are to theory. Their external validity is low, in my opinion, because of their low reliability and internal validity. Nowak (1987) compares diffusion, economic and ecological factors to try to explain adoption of conservation technologies on farms. Nowak has published many papers on diffusion studies, and has researched many different aspects of the theory of diffusion of innovations. The reliability of his resu lts is difficult to judge from this study because he reports them so briefly in this publication. Assuming that Nowak had high internal validity and reliability, as many of his other published research does, than I would suspect that this study has high ex ternal validity. The conclusions of Nowak's
26 study towards innovations in agricultural settings are strong, but also able to generalize to innovations in a variety of other settings. Other studies of his have had reliable results and extensive instrument de velopment. Eastlick (1993) and Carter and Belanger (2005), in contrast, identified compatibility and relative advantage as more reliable for predicting adoption behavior than the other three attributes. Eastlick researched the factors that affected adopti on of videotex, an innovation available in the early 1990s. She surveyed hundreds of shoppers, with a response rate of 27%, to rate their perceptions of the videotex, as well as their intents to adopt the innovation. Her sample size and instrumentation lea ds to high reliability scores for this study. The design of her study included a random large sample and was supported through theory. Overall, her study showed reliable methods and valid results. Carter and Belanger put a lot of effort into designing an index with a strong methodo logical background. However, their design was subpar. They contacted t heir entire samp le set at a single community event, despite a theoretical population of the entire United States. The sample ended up having a high proportion of users that had already adopted the innovation (e government services), which likely affected the perceptions of innovation attributes. Few studies identify trialability as a predictor of adoption behavior. Most innovation attribute studies still tend t o conclude that trialability relates to adoption, but that it is not a major explanation of variance (Agarwal & Prasad, 1997; Moore & Benbaset, 1991). Agarwal and Prasad based their research on an instrument created by Moor e and Benbaset. T hey concluded tha t the same attributes are most important to
27 innovation adoption behavior. However, the fact is that trialability had the lowest expected that the attribute with the least reliable results would also en d up to be the least valid in the study. Revision would benefit that instrument LaRose and Atkin (1992) state that compatibility is a good predictor of adoption behavior. They practiced exceptionally precise sampling techniques to eliminate bias, and the reliability of their study is high. However, the explanatory power of their research is based on redundancy. They state that compatibility is the best predictor of that a re similar to one another. They did not include any innovations that were not in innovation clusters in their study and then concluded that compatibility was most important for predicting adoption. That seems to me to be more of a conclusion that compatibi lity has a positive relationship with adoption behavior, not that compatibility causes adoption. Some studies fail to identify predictors of adoption behavior, but instead conclude that some attributes do not explain significant variance. Lin (1998) notes instead that was strong in both methodology and design. The only drawback to it is that the conclusions of the study focus on innovations that are tangible products, inste ad of multiple types of innovations. A single purchase defined adoption behavior instead of continued use of the innovation, in this study. This definition leads to a limited ability for generalization to the entire theory of diffusion of innovations.
28 The innovation decision process, as previously stated is the decision making process that an individual moves through to adopt an innovation. The following chronological steps are (1) knowledge of the innovation to (2) persuasion to use it, to (3) a decision about whether to use it, to (4) implementation and (5) confirmation of the decision to adopt the innovation. Rogers also classifies adopters as innovators, early adopters, early majority, late majority and laggards. Limitations to d iffusion s tudies Some l imitations of diffusion studies are the pro innovation bias, the lack of studies on the consequences of adoption and the fact that the majority of studies with diffusion have been cross sectional. The pro innovation bias is evident in the fact that innovat ions are usually assumed positive and beneficial for the adopter (Stephenson, 2003) and therefore little research has been done on generally rejected innovations (Haider & Kreps, 2004; Rogers, 2003; Strang & Soule, 1998). Rogers estimates that less than on e percent of the published diffusion studies have looked into the outcomes of innovation adoption (2003). In other words, innovations may have been widely adopted, but whether or not they achieve their aims is another question. Finally, because diffusion s tudies have been largely cross sectional in design, they have only been able to conclude about that exact point in time. Increased longitudinal studies would be able to study the previously mentioned gaps in research more fully. Another limitation to some diffusion studies relates to diffusion of agricultural conservation technologies. There is an on going debate in the field of diffusion studies about whether or not to treat conservation technologies as regular innovations. In other words, do conservation technologies adhere to the basic tenants of the theory of diffusion of innovations? Some researchers argue that the decision to adopt agricultural
29 conservation technology requires more background knowledge than the decision to adopt a typical technological innovation. They continue to say that farmers must be aware of the need for the conservation technology, and must have certain information. This knowledge must include sufficient agronomic and economic information to be able to evaluate risks and benefits effectively the proper knowledge or resources to enable the adaptation of the technology to their specific climate, soil type, cropping system, as well as their specific social conditions (Nowak, 1982, 1984; Nowak & Korsching, 1983; Taylor & Miller, 1978 ). Lacking any of the above requirements may lead a farmer to reject the conservation technology, despite a typical classification of an innovation and potential adopter using theory of diffusion of innovations There is also a school of thought that conse rvation technologies are typically unprofitable for the farmer (Fast, 1983; Fliegel & van Es, 1983; Napier et al.; 1984), perhaps due to the ability to externalize costs of soil erosion and run off (Buttel, 1986). One study dichotomized agricultural techno logy, though not necessarily conservation technology, into profitable and un profitable, then studied the farmers that adopted the technology innovations (Pampel & van Es, 1977). There were significant differences between the adopters of the dichotomized t echnologies, reinforcing the possibility that adopters of conservation technologies have different reasons or motivations for adoption than adopters of typical innovations. The other side of the debate about adoption of conservation technologies states tha t they can be treated as regular innovation practices because they are typically beneficial for the farmer, either economically or socially (Nowak, 1984). The conservation practices often have governmental stimulus behind them, such as laws in Germany that pay farmers who are able to reduce their soil nitrogen levels to protect
30 groundwater quality (Linket al., 2006). Another example is the suggestion in the Kyoto protocol to pay farmers able to prove they are sequestering carbon (Palm, et al., 2004; United Nations Framework Convention on Climate Change, 1998). There have been a few research studies about how the size of a corporation or business can affect adoption rates, but the results are mixed (Rogers, 2003). Even more studies have been published that c an prove a relationship, but are unable to quantify or give a direction to it (Chaves & Riley, 2001). For example, Sahadev and Islam (2005) found that size as defined by the number of rooms, was not a predictor of adoption of information and communication technologies for hotels. This innovation is expensive and time consuming. The methodologies of this study were strong, but the design of the research lacked a theoretical basis The explanatory power of the study is lacking, due to the shortage of a theor etical basis or conclusion. Other studies on agricultural settings also concluded that size measured by annual gross income, was not a sole predictor of adoption behavior (Grieshop, Zalom & Miyao, 1988; Ridgley & Brush, 1992, Swisher & Bastidas, 1994). Ri dgley and Brush studied integrated pest management adoption among pear farmers in California. Their original design did not include size of farm as a variable of study. After iterative rounds of interviews with the farmers, farm size arose as a significant factor in adoption rates. They also concluded that the types of farm management and adopter characteristics like age and education correlate with adoption rates. This study has high internal validity. Their external validity is limited to horticultural gr owers of crops of a similar harvest price to pears. Explanatory power from this study is strong, because of a design including case studies and cross sectional comparisons using the theory of diffusion of innovations. Swisher and
31 Bastidas (1994) looked int o a variety of farmers in Florida to study whether their size affected adoption of pest, water and nutrient management strategies. The methodology of this study included a sampling frame of Florida extension producer lists and subsequent interviews with ma ny of the identified farmers. The study was based on diffusion of innovations theory (M.E. Swisher, personal communication, February 15, 2013). Internal and external validity were high However, Mytinger (1968) found exactly the opposite; they suggest that size of programs. He asked various programs throughout California what innovations they wer e using and when they had adopted them. This methodology has the potential to be biased by self responses from the public health programs. With a flawed methodology, the internal validity is low, leading to a low external validity. He does not make any con clusions about the explanatory power of his study. Others (Blake, et al., 2007; Fernandez Cornejo et al., 1992; Hammond, Luschei, Boerboom & Nowak, 2006; Mahler & Rogers, 1999; McNamara et al., 1991; Napier & Camboni, 1993) also concluded that size of an o rganization is positively related to increased rates of adoption. Nowak (1992) states that this relationship between farm size and adoption rate is due to constrictions of finances or space for smaller institutions, particularly farms using the USDA defini tion of size When there is not enough field space to put in new types of irrigation or certain pesticide applications are cost prohibitive for small farms, adoption of an innovation is less likely For inexpensive innovations, the converse is true: adopti on and farm size were not related (Harper et al., 1990). Harper et al. studied rice farmers in Texas and
32 their adoption behaviors of low cost pest management techniques. They obtained a large sample size and obtained data with high reliability and internal validity. The most interesting part of this study is that it did not use diffusion of innovations theory The study was designed to show relationships between a large number of variables and adoption of inexpensive innovations. The conclusion that there w as no relationship between size and adoption is valid for the sample under study. The external validity of the conclusion is also high. There is no reason to dismiss the ability of their conclusion to generalize to a wider population. However, due to the s diffusion of innovations or any other theory, the conclusion cannot further explain theory. Some studies have hypothesized that larger organizations are able to adopt innovations on a trial basis more easily than small organizatio ns due to their larger production surpluses. Increased surpluses cause increased discretionary income Discretionary income is related to an ability to be less risk adverse (Food & Fertilizer Technology Center, 1985). T he relation between farm size and ad option of conservation technologies or practices is explored in the following studies In 1981, Buttel hypothesized that farm size would have no relationship to adoption of such innovations, but did not test his theories. Both slightly before this hypothes is was put forth and in response to the publication, various studies came to the conclusion that farm size did, in fact, have a relationship with adoption rates of conservation practices (Hoover & Wiitala, 1980; Nowak & Korsching, 1983). The conclusions fr om these studies included the following general reasoning for adoption of conservation practices. Large farms have more flexibility in their decision making, they have greater access to discretionary income and equipment
33 resources, they have a larger oppor tunity to utilize new practices on a trial basis and they have more ability to buffer against risk and uncertainty. All of this reasoning is important when considering the adoption of a technology or practice that may not significantly increase harvest or income. There have been many attempts to determine the relationship between certain demographics of the potential adopter and the decision to adopt, within diffusion of innovations research. Various studies have found negative correlations between age and adoption (Blake et al., 2007; McDonald & Glynn, 1994; Rogers, 1995), or concluded that there is no relationship (Grieshop, Zalom & Miyao, 1988; Samiee, Rezvanfar & Faham, 2009; Waller et al., 1998). Many fewer studies looked into the gender of potential a dopters. I only found one and its potential to apply to this research was low, because it was conducted in a society with different gender norms than those of the United States. The study looked at gender and adoption of organic agriculture in Kenya concl uding that gender was a significant predictor of adoption behaviors (Goldberger, 2008). Many, to differing results (Hamilton et al., 1997; Lambur et al., 1985; Napit et al., 1988; Rogers, 1995; Waller et al., 1998), have studied the education level of pote ntial adopters Others have seen positive (Blake et al., 2007) or negative (Rajotte et al., 1987; Waller et al., 1998) correlations between years of experience and adoption behavior in the field surrounding the innovation. Size of Agricultural Operations In the United States, agricultural production has been shifting to larger farms for the last century or so, although small farms by gross annual income, are still a majority at about 90% of total number of farms in production (Economic Research Service, 2012). However, the average farm size by acres, in the US has dropped over the past
34 10 years, from 431 to 418 acres (Economic Research Service, 2012), in part due to the recent increase in the number of small farms. Small farms are increasing for a variet y of reasons, possibly including the widespread Beginning Farmer and Rancher Program (NIFA, 2012) or the increase in demand for organic and local foods (Howard & Allen, 2010). Additionally, the definition for farm size has not changed for decades and there fore has not considered inflation for many years. The USDA defines small farms as those making between $1,000 and $250,000 in gross annual income and large farms as those making more than that (Economic Research Service, 2010). The average farm size for th e entire US is 446 acres, while Florida farmers typically have an average farm size of 244 acres t (NASS, 2006). While considering this average farm size, keep in mind that there are many more small farms, in both acreage and income, than there are large f arms. The majority (79%) of Florida farms have less than $10,000 in farm income per year (Economic Research Service, 2012), whereas 59% of farmers in the United States are in the same category. This discrepancy in farm income indicates that my accessible p opulation will have farm sizes smaller than the US average. One of the problems with defining farm size based on annual gross income is that primary products can be low income, such as wheat, but still take a lot of land. Producing livestock, however, can be high income and take less land. This income gap is one of the reasons that I will be limiting my sample to growers of horticultural crops for fresh market sale. Utilizing participants that all get similar incomes per acre will lower economic differences that may affect innovation adoption decisions. Soil Testing and Nutrient Management yields. However, decisions must now take into account crop yield, purchase prices for
35 the inputs, as well as potential environmental effects of the nutrients. One way that farmers are able to make educated decisions about nutrient inputs is through soil testing. Soil tests can include information on pH, plant available nutrients including nitr ogen, phosphorus and potassium, percent organic matter, micronutrients and more. The current soil testing procedures and recommendations for many counties or states are mandated to provide information that allows farmers to make economically and environmen tally sustainable management decisions (ESTL, 2012; Mylavarapu, 2010). Farmers must keep in mind that the application of any nutrients that do not cause a yield or quality increase would be an economically poor decision. In addition, they must keep in mind that the addition of surplus nutrients would be an environmentally poor decision. The ability to find, using a soil test, the amount of nutrients that are already available to plants and in the soil help guide future nutrient management decisions. Soil t esting is a widely used technique in certain areas and in certain types of crop production systems. The preferred practice is for farmers to test their soil every production season, but this practice is not always the reality For example, a study of tomat o, corn and potato farmers in Florida showed that seasonal soil testing occurred over 90% of the time for large farmers, but only around 58% of the time for smaller farmers (Swisher & Bastidas, 1994). However, some areas lack available access to testing fa cilities, and other farmers do not use soil testing for a variety of reasons. There may be a lack of knowledge (Asthana & Kumar, 2008) or a conscious decision not to use soil testing to make nutrient management decisions (Contant & Korsching, 1997; Zhen, Z oebisch, Chen & Feng, 2006). This study will be exploring some of the
36 perceptions of soil testing that lead to decisions not to use soil testing to assist in nutrient management decisions.
37 CHAPTER 3 METHODOLOGY Research Design This study used a cross sect ional design. Whereas a longitudinal design would have been ideal to study the changes in adoption rates of soil testing over time, it was not feasible due to time restrictions for the researcher. Cross sectional designs are studies conducted at a single p oint in time in order to examine differences between two or more groups (de Vaus, 2001). This design is appropriate if the data collected requires no time dimension and groups have existing differences relevant to the research. In particular, these designs are able to determine if variation between two or more variables is present for two or more comparison groups (Bryman, 2004). In this study, comparison groups were assigned post their farm size, to determin e the variation in participants' views on soil testing attributes. I designated s mall farms in this study as those identifying themselves as having less than $250,000 in gross annual income, while large farms would have a gross annual income greater than $ 250,000. I chose t his value because the United States Department of Agriculture uses the same value in their reports and studies (Economic Research Service, 2010). Sample Selection The theoretical population for this study was farmers producing horticultu ral crops for fresh market sale in the United States. The accessible population was farmers producing horticultural crops for fresh market sale who attended conferences about farming in Florida, or who were in close contact with their county extension pers onnel. The theoretical population was limited to horticultural growers because farmers
38 producing grains, hydroponics, livestock or other types of crops would have different nutrient management decisions to make. The value of the harvest of a horticultural crop is generally much higher than the value of grain harvests. Nutrient management decisions are therefore reached at different economic thresholds. Therefore, a group of farmers producing a similar crop in terms of management decisions was chosen, in the hopes that biases about economic costs of soil testing would be minimized. The sample was limited to farmers producing horticultural crops for fresh market sale for a variety of reasons. Horticultural crops for fresh market sale generally sell for a high er price per unit of crop. This increased potential profit means that the farmer would be able to spend more time and money managing this crop than they potentially would for the same number of units of a cheaper crop, such as pasture grass. A small differ ence between the theoretical and accessible populations for this study is possible Farmers attending conferences or in close contact with county extension personnel may hypothetically have a slightly higher incidence of the use of soil testing, because of an increase in exposure to academic and other personnel advertising and recommending soil testing. However, because this study is considering how farmers perceive the attributes of soil testing instead of just direct adoption/non adoption, this small diff erence is acceptable. The sample consisted of farmers with a variety of farm sizes, from thirteen total states. More respondents were from Florida than other represented state s While Florida farmers typically have an average farm area of 244 acres, the av erage farm size across the entire US is 446 acres (NASS, 2006). The majority (79%) of Florida farmers make less than $10,000 in gross income per year, whereas 59% of farmers in the
39 United States are in the same category (Economic Research Service, 2012). T he sample for this study was probably biased towards smaller farm sizes, in both acreage and income, than the average farm in the United States. I chose a conference to which I could attend, and contacted extension personnel located throughout the United States The chosen conference was the Small Farms Conference, held July 27th 29 th 2012 in South Florida. Potential participants initially screened to determine that they were farmers and that t hey knew what soil tests were. Next, t hey were given the online link to the index, or were given an electronic device to borrow and fill out the index at that time. A few individuals requested to have the index read to them aloud, with myself recording their answers. The index would repeat the initial screening questi ons and then verify that they were also farmers that produced horticultural crops for fresh market sale. Farmers exclusively growing hydroponic crops were excluded from participation due to their presumed lack of soil testing related decisions. I asked fa rmer s growing multiple crops on their property to answer the index questions while considering only the horticultural crops that they grew and the decisions about nutrient management that they made for those crops. The index also had a question that would separate the respondents into the two comparison groups: small farms and large farms, based on annual gross income. After the conference, I contacted Florida extension personnel followed shortly after by extension personnel in Alabama, California, Georgia Michigan, North Carolina, Oregon and Texas. I chose t hose states because they produced the most horticultural sales in 2009 (United States Department of Agriculture, 2009). They provided lists of contact information for farmers they knew that produced ho rticultural crops for fresh
40 market sale. The index prevented farmer s from answering more than once, so repetition of participants was not a concern. Once the contact lists had been provided, emails were sent and telephone calls were made to farmers to asse ss interest levels in participation. If levels were acceptable and farmers agreed, a paper index was mailed or an online link to the index was emailed to the farmer. The participants were mainly men over the age of forty. The final sample size for completi on of the index was 233 respondents, with 177 in the small farm comparison group and 56 in the large farm comparison group. These numbers were not large enough for the required sample size to perform some of the regression models desired for the study Inst rumentation Comparison Groups Participants self identifying themselves using their gross annual income determined the two comparison groups. If the stated income was less than $250,000 a P articipants were if the income was stated as larger than $250,000 Variable Development There are five attributes of any innovation, as discussed in Chapter 2. The attributes of soil testing for nutrients were t he independent variables and included relative advantage, complexity, compatibility, trialability and observability, for this study. I operationalized e ach of these attributes as one construct that needed to for the index. I conducted a literature review t o search for available and thoroughly tested items and variables. If variables were tested and deemed unreliable, I discarded them V ariables
41 of less than 0.8 (Bland & Altman, 1997; Nunnaly, 1978; Vacha Haase, Henson & Caruso, 2002), were discarded, due to lack of reliability. Variables that were low in nomological validity (Adcock & Collier, 2001) were also discarded. Existing items and variables from Carter & Belanger (2005), Chakravarty & Dubinsky (2004), Flight, were adapted to become contextually specific with this study. I also created and tested items myself, to fit with variables research ed in the above listed studies. I did not weight any of the attributes, because the various studies on this subject have not come to any consensus on whether one or more attributes contributes different to adoption behavior. I scored each item the same s ince there was no weighting to take into account Cognitive Testing I conducted c ognitive interviews with seven people before collecting any data in order to refine the index (Collins, 2003). I performed these cognitive interviews in person. The peers con tacted for these interviews had varying backgrounds in both index development and farming. Cognitive testers were chosen in a specific way to continue to eliminate systematic bias originating from myself (Viswanathan, 2005). I tried to pick people who woul d have different views from me when it comes to soil testing. Three of the individuals had developed indices previously in their professional lives. The rest had either never developed indices, or had never before conducted academic research. Five of them had some level of experience with farming and two of them had no practical experience with farming. I asked the cognitive reviewers first to read the instructions and definitions provided to potential respondents at the beginning of the index. After readin g, the peer reviewers were asked to inform me if there was
42 confusion or lack in clarity, as well as how the instructions and definitions should be modified for improvements. The theory (diffusion of innovations) used in this study (Rogers, 1962) was explai ned to the peer reviewers, as was the relevance of the theory to the problem (insufficient nutrient management by farmers) that is addressed with this research. I explained t he two constructs (relative advantage and compatibility) prepared for this stage o f the research to the interviewees. Finally, I explained the indicators for each construct and their relationships. I then asked t he cognitive reviewers to provide comments and feedback about their thoughts of the theory, the constructs and the chosen indi cators. Those reviewers with experience creating indices were asked to help me format and improve the index as a whole, as well as to check i f the response categories were clear (Fowler, 1995). I instructed them to look for missing response categories or i tems that would provide nonsensical answers, as well as to decide if the response categories were appropriate for continuity throughout the index. Those peer reviewers who had experience farming were asked to check each item for clarity, technical expertis e and relevance to the indicator, construct and theory as a whole (Barry, Chaney, Stellefson & Chaney, 2011; DeVellis, 2003). Finally, I explained that the purpose of their review was to help in improving the index by highlighting errors. I also noted that not all of the questions would be included in the final index. Almost all of the cognitive interview participants were concerned about the repetitiveness of many of the indicators and questions and expressed frustration
43 construction experience expressed these views most often and so I took time to explain why th ere was repetitiveness and how I would fix it The initial set of peer reviewers were examining the index that started with 159 items. The total number of items would be much lower in the final index. The cognitive reviewers had a variety of recommendations, which I address shortly. The majority of the peer reviewers responded that statements, instead of questions, would be si mpler and easier to understand, as well as faster for the respondents to complete However, I chose not to follow this recommendation for a variety of reasons. Statements require a higher cognit ive level of thinking to answer because they require a respond ent to first understand the statement and then decide whether they agree with the statement. Finally, the respondent would have to decide which answer choice is appropriate (Wikman, 2006). Respondents easily can become confused, especially with negative st atements and choose the opposite answer choice than what they mean to pick (Collins, 2003). Based on the recommendation that the questions were confusing because of their length, relative to the length of what statements could be, I s horten ed the longer qu estions. I also made the questions as simple as possible to understand. R eviewers also suggested changing some of the response categories. This suggestion was made for grammatical reasons in certain items and to enhance clarity and flow for other items. I changed the response types to a variety of types as indicated in Fowler (1995). In those cases, I also reworded the questions. One of the concerns of every single one of the respondents was that the overall length of the index and that there were too many indicators for each construct. The
44 length of the index was not a concern for reliability or internal validity. Instead it was a concern that if there were too many items, it would be difficult to persuade a large enough sample to agree to respond to the in dex. Therefore, per their recommendations, I remove d one indicator and combine d o the connotations the word has as recommended that the entire indicator be removed. This recommendation was due to the large number of indicators that the reviewers perceived to be more valid. This indicator had only been used in one study (Holak & Lehmann, 1990). Due in part also to rom the index. Three indicators were combined to form the relative advantage construct: these indicators were similar and measured overlapping ideas. While trying to cut dow n the number of indicators I included in the index I concluded that these three indicators were measuring such similar items that discriminatory power between the three indicators would have been low (Adcock & Collier, 2001). The three indicators were com r the compatibility construct. I made t his change because the reviewers found it difficult to answer or that they had different views of what social reward meant. Other reviewers commented
45 also, even after they were read definitions of social reward, that they believed it to be so si milar to social compatibility that the items for them would be measuring the same thing. I concluded that the discriminatory power between the two indicators was likely to idity increased (Adcock & Collier, 2001). I concluded that the nuances between social reward and social compatibility are too light to be distinguished between in less than ten questions. However, social compatibility is definitely an intrinsic part of com patibility, so I focused on drawing out those details in the index. After conducting multiple rounds of cognitive interviews followed by changing the index, I started receiving a distinct lack of new information or comments from our interviewees. Therefor e, I concluded that conducting more interviews at this point would fail to provide me with useful and new insight into the index and so I stopped the cognitive review phase (Collins, 2003; Willis, 2005). By the time I had finished the cognitive testing sta ge, I cut the total number of questions in the index from 159 to 94 Test of Revised Instrument with Participants I decided to use farmers as the testers of the revised index, because the sample population will be farmers who produce horticultural crops for fresh market sale, (Adcock & Collier, 2001). However, I did not plan to limit the sample to that specific type of farmer. This was because obtaining enough respondents for this stag e of testing was time limited. I obtained c omplete responses from seven ho rticultural product farmers, three ornamental plant farmers, two corn and soybean farmers, one citrus producer and preferable, but I was constrained by time and response rates of a pproximately 60%.
46 This stage of instrument development was conducted online and in person at two farmers markets in the Gainesville, Florida area. Once as many responses were received in a timely fashion as possible, a variety of statistical tests were per formed. I needed to make sure that it was first a reliable instrument, to ensure a valid instrument. I chose alpha and item total correlation for this level of analysis. Because I am trying to develop an index for the purposes of measuring five would provide me with a numerical value to indicate the reliability of the index. Reliability is whether participants answered consistently and accurately throughout the index as a whole (Cro nbach, 1951, Hatcher, 1994). It is essential to know that one group of respondents would answer the index questions in the same way and with similar response patterns as an entirely different group of respondents (Santos, 1999). A r greater for each indicator was indicative of a reliable instrument, for the purposes of this study. If a value of at least a 0.8 was not obtained, then that indicator was removed from the index. Researchers generally accept a 0.8 value as satisfactory, w ith a 0.9 or better perceived as a superior instrument, in terms of reliability (Bland & Altman, 1997; Nunnaly, 1978; Vacha Haase, Henson & Caruso, 2002). I used t alpha. While Cronb whole, item total correlation looks at each item and correlates the responses to each other item, within a single indicator. This test can tell if a specific item seems to be drawing
47 provides are for each item. If the value given is less than 0.2 or 0.3 for a specific item, then that item is probably not measuring the same variable that the rest of the items are measuring within an indicator (Field, 2005). In that case, the item was removed from the index (Churchill, 1979). The specific procedure I used was to run the data through a program called Statistica (Release 8) and to test each indicator separately usin item total correlation. Starting with the first indicator, Time Saved by Adoption, the scores were totaled for each item and run through the Split Reliability Test well as an item total correlation value for each item. Please see Table 3 1 for specific removed, based on which had the lowest item total correlation value. Priority was given to the items that had a negative item total correlation value and removed first, because they were probably measuring something different from what the rest of the indicator was measuring. Every time I removed an item, the Split Reliability test was rerun. The recorded with each item removal. After removing items These values are displayed in Table 3 1. The remaining items for each indicator had high ITC scores, all above 0.2 at the least. I also needed to test for discriminatory power. It was necessary to ensure that the index had the ability to discriminate between high and low scorers. One way to ensure discrimination was to determine if the scores were distinguishable between the middle two quartiles and the top quartile, as well as the middle two quartiles and the bottom
48 quartile. If these scores were statistically different, then the items were demon strating discriminat ory power. I sorted t he scores for each respondent from top to bottom for each indicator. The top quartile scores were separated from the middle two quartile scores and from those of the bottom quartile. A Mann Whitney U test was then used to determine whe ther the index measured discriminatory power. The Mann Whitney U test 1957). The test ranks all the values from low to high and then compares the means ranks. This test is particularly useful for small sample sizes and is simple to run, making it ideal for this purpose. The specific procedure to run this test was through Statistica aga in (Release 8), using the Non Parametric T est Comparing Two Independent Samples. I designated t he item total correlation tests. The independent variables were designated as the highest quartile, the mid range two quartiles, or the lowest quartile. I never compared t he highest quartile against the lowest quartile, but instead compared it against the middle. The same is true of the lowest quartile. The one sided exact p val ue was then examined and any items which had a p value of less than 0.15 were chosen. A good p value for any item to show discriminatory power would have been less than 0.05, but I accepted any that were less than 0.15. Items with these values were the que stions that were able to distinguish successfully between respondents that scored at the ends of the response spectrum for the indicators and respondents who were in the middle of the response spectrum for the indicators.
49 Sixty one items and zero indicator s were removed from the index after these tests. The final index that used to gather data consisted of 33 items, including demographic questions. Fifteen indicators measured the five variables and the final index is attached in Appendix A. After data coll ection, I re for each of the five constructs. I removed four total items, one from each construct except for compatibility, to arrive at the final scores listed in Table 3 2.
50 Table 3 alpha scores for five variables representing theoretically defined attributes that may influence adoption of technologies by large and small farmers, before testing for final instrument use Variable Indicator score Relative Advantage Tim e Saved by Adoption 0.99 Efficiency 0.99 Reliability 0.99 Superiority 0.99 Economic Advantage 0.99 Compatibility Social Compatibility 0.99 Fits with Users Existing System 0.99 Personal Compatibility 0.99 Complexity Difficulty/Ease of Use 0.88 Length of Operating Instructions 0.75 General Level of Knowledge Required for 0.99 Trialability Required Large Purchase 0.99 Availability for Use on a Trial Basis 0.99 Observability Visibility of Innovation 0.64 Amount of Communicatio n 0.99 Table 3 attributes that may influence adoption of technologies by large and small farmers, after testing for final instrument use Variable Relative Advantage 0.99 Compatibility 0.99 Complexity 0.99 Trialability 0.99 Observability 0.99
51 CHAPTER 4 RESULTS Sample Statistics Two hundred and thirty three respondents participated in this study. Not every participant responded to all items in the quest ionnaire. Some did not complete the questionnaire. Between 207 and 227 valid cases were included in the statistical analyses, differing for each variable. Demographics Of the 233 respondents, 146 were male (63%), 61 were female (26%) and 26 did not indica te their gender (11%). My sample reflects the overall gender distribution of farmers in the United States, which is 70% male and 30% female (United States Department of Agriculture, 2007b). Most of the respondents were 40 years of age, with 37% between 40 and 55 and 43% over 55. Seventeen percent of the respondents were between 25 and 40 and only three percent were between the ages of 18 and 25. These numbers reflect the generally aging population of U.S. farmers (United States Department of Agriculture, 2 007c). Fifteen percent of respondents had less than four years of farming experience, while 43% of respondents had more than 20 years of experience. Only seven percent of respondents did not indicate how much experience they had and the rest of the partic ipants had 5 9 years of experience (16%), 10 14 years of experience (9%) or between 15 and 19 years of experience (9%). Farm size, by acre of horticultural crops in production, varied from less than one acre (11%) to greater than 50 acres (29%). Most of t he respondents had between one
52 and ten acres (47%), while the remaining respondents had between ten and 30 acres (8%) or between 30 and 50 acres (6%). The majority of respondents had farms that were not located in watersheds with basin m anagement a ction p lans (57%) or were not sure if their watershed was regulated b asin management a ction p lan. Most respondents were White/Caucasian (91%).Three percent were Black/African American and f our percent were Hispanic/Latino. One respondent was Asian and one was American Indian. See Tables 4 1, 4 2, 4 3 and 4 4. Post hoc Assignment to Groups Respondents reported their gross annual sales, which I used to assign them to one of two comparison grou Department of Agriculture (Economic Research Service, 2010) defines small farms as those with gross annual sales of less than $250,000. I used this definition. There were Test for Normal Distribution I tested the scores for each variable, relative advantage, compatibility, complexity, trialability and observability, for normality of distribu tion, because many statistical tests assume normally distributed data. The sample size for the tests varied due to non response, but none of the tests are sensitive to sample size. I used the Shapiro Wilk test for normality of distribution (Shapiro & Wilk, 1965). The test showed that the scores for the 5).
53 were to reduce the impa ct of outliers as well as to normalize the distribution of the data so that further statistical tests could be run. I chose to use a logarithmic transformation, which is particularly useful when data are slightly positively skewed (Sheskin, 2004). Once the data were transformed, I ran the Shapiro Wilk test again. Only the values for complexity for large farms were normally distributed (Table 4 6). Test for Central Tendency I hypothesized that there would exist a significant difference between small and lar ge farmers with regard to perceptions of the attributes of soil testing. I used the Mann Whitney U test for two independent samples to test this hypothesis for each of the five attributes. This non parametric test compares median values for each comparison group (Mann & Whitney, 1947; Sheskin, 2004). I chose the Mann Whitney U test because it does not require normally distributed data, is not sensitive to differences in sample size and is not particularly sensitive to relatively small sample size. As seen i n Table 4 7, there was a significant difference in the scores for farmers. Since I negatively scored all responses for complexity, the lower rank sum for complexity indic ates that small farmers found the complexity of soil testing to be higher than did large farmers. There were no significant differences between small and large farmers for the remaining attributes. Spearman Rank Order Correlation It would have been ideal t o perform a regression analysis of the data to see the relationships between the attribute variables and the demographic variables. However, I
54 did not have continuous data and there were not enough cases to run a regression containing the five attributes a nd the four demographic variables. I decided not to include four other demographic variables for which I had collected data. These variables were race, level of education, state and county where farm is located and whether the farm is in a watershed with a Basin Management Action Plan. The sample contained more than 90% of respondents that identified themselves as Caucasian. This bias could have caused correlations to arise as significant based on the response of less than five minority respondents. The res ponses for level of education, state and county contained so many categories of response that inclusion in a correlation test would have required many more usable cases. The total number of responses was 211, too few to accurately measure correlations for level of education and geographic location of the farm. Many respondents were unsure about whether their watershed had a Basin Management Action Plan and so their responses could not be included in correlation testing. I included years of experience, acres in production, gender and age in the final correlation tests. Instead of a regression, I used the Spearman rank order correlation test. The purpose of this test was to examine the correlations between the five attributes and the four variables. I first ra n the correlation for the five attributes alone. Table 4 8 shows the results for the combined comparison groups and Table 4 9 shows the results for separated comparison groups. There were seven total significantly positively group test. All of those correlations were positive except between complexity and trialability. The strongest correlations were between the score s for compatibility and
55 relative advantage and between compatibility and observability. When the correlation test was run for the separated comparison groups, most of the significant correlations still appeared. Compatibility and trialability for large far ms, however, no longer had a significant correlation. Next, I ran the Spearman rank order correlation test for the four variables. Table 4 10 shows the results for the combined comparison groups and Table 4 11 shows the results for separated comparison gr oups. For the combined comparison groups, there would be logically expected. Years of experience and acres in production were also positively correlated. Gender was neg atively correlated with years of experience and acres in production. I scored g ender in such a way to designate women as a low score and men as a high score. Therefore, a negative correlation between gender and years of experience indicates that, among res pondents to this study, women h ad fewer years of experience tha n men did The same was true of acres in production. When I performed the test with the separated comparison groups, some differences arose. The same significant correlations arose with three e xceptions. Gender and years of experience no longer had a significant correlation for large farms. This difference may have arisen because there were so few women that identified themselves as large farmers. The correlation between gender and acres in prod uction also became insignificant for large farms. Years of experience and age became a much stronger correlation for large farms. This change is possibly because only one large farm respondent had between 0 4 years of experience
56 Finally, I ran the Spearma n rank order correlation test for the five attributes and four variables together. This test was run for the nine combined variables and 211 total responses. Twenty three responses per variable are acceptable for this type of test, but are not ideal. The c om bined comparison group results are found in Table 4 12 and the separated comparison group results are in Table 4 13. For the combined comparison correlations. The previously st ated significant correlations between attributes and between demographic variables were still in place. This test added the ability to see the correlations between attributes and demographic variables. Relative advantage showed a negative significant corre lation with gender. Compatibility also had a negative correlation with gender, but showed a positive correlation with acres in production. Complexity and trialability had no significant correlations with any of the demographic variables. Observability show ed positive correlations with years of experience and acres in production and a negative correlation with gender. The separated comparison groups showed many differences from the combined comparison groups for correlations between attributes and demographi c variables. Compatibility lost the significance of correlation for gender for large and small farms. It also no longer showed a correlation between compatibility and acres of production for small farms. Complexity and trialability remained showing no corr elations with any demographic variables for either comparison group. Observability lost the significant correlation with any demographic variable for large farms and with gender for small farms. Research Validity and Explanatory Power The design of my stu dy was cross sectional and therefore it lacks the ability to give causality to relationships discovered. For example, I can say that there is a
57 correlation between observability and the gender of a respondent, but not the causality of that relationship. I also lack the ability to tell if the correlations are direct or indirect relationships between variables. These problems affect the internal validity of my study. The external validity of my study is high, in my opinion. My sample was limited to horticultu ral growers in a variety of locations throughout the United States, but I am able to generalize beyond the sample I took. I based my research design in theory and therefore am able to generalize my conclusions to other populations. My conclusions are valid for the entire theoretical population of this study : farmer s growing horticultural products in the United States. The respondents to my sample had fewer average years of experience than the average for a typical farming community in the United States. Thi s difference between theoretical and accessible population causes a potential limit to the external validity of my results. It does not affect the explanatory power of my conclusions. The explanatory power of my research is limited. I was able to include m any different variables due to the cross sectional design, as well as many demographic covariates, but the relationships between all the variables are difficult to tease apart. Therefore, I again can conclude that complexity and observability are correlate d with farm size and that years of experience, acres in production and gender are correlated with innovation attributes. However, I cannot assign causality to any of these relationships. Hypothesis One My first hypothesis was that small and large farmers would have differing perceptions of soil testing. I thought I could explain this hypothesis through the innovation attributes in diffusion of innovations. These attributes (relative advantage,
58 compatibility, complexity, trialability and observability) were not all supported through this study as different between comparison groups. I found statistically significant and observability by the participants in this research. In other words, the total scores for small farmers' perceptions of the complexity of soil testing were statistically different from the scores for large farmers' perceptions of the complexity of soil testing. The same is true for the small farmers' scores fo r perceptions of soil testing observability, as compared to the scores for large farmers in the same category. There was not a significant difference in scores between the two comparison groups for the other three innovation attributes (relative advantage, compatibility and trialability). In fact, the p value for trialability was so high that I suggest that there was almost no difference at all It is important to note that there were correlati ons between some of the attributes. There was not a significant correlation between complexity and observability, so there were not likely to be confounding effects causing bias in the Mann Whitney U test results. Hypothesis Two My second hypothesis was th at I could explain any potential lack of correlation between independent variables through demographic differences. There were only 211 cases that could be used in statistical testing. This number of responses was sufficient for multiple Spearman rank orde r correlations for the five attribute variables and four demographic variables. This test showed if correlations were present between demographics. In addition, acre s in production was correlated with gender. Relative
59 advantage was correlated with gender, as was compatibility. Compatibility also correlated with acres in production. Observability correlated with all demographics except for age. The rest of the variable s did not show significant correlations.
60 Table 4 1. Gender and age of respondents in a study of factors affecting adoption of soil testing by large and small farmers, 2012 Age U nder 18 18 25 25 40 40 55 Over 55 Did Not Indicate Total Male 0 1 24 54 66 0 145 Female 0 4 10 23 25 0 62 Did Not Indicate 0 1 2 1 1 21 26 Total 0 6 36 78 90 21 233 Table 4 2. Gender and years of experience for respondents in a study of factors affecting adoption of soil testing by large and small farmers, 2012 Years of Ex perience 0 4 5 9 10 14 15 19 20+ Did Not Indicate Total Male 16 21 16 15 77 1 146 Female 16 14 5 6 19 1 61 Did Not Indicate 3 2 1 0 5 15 26 Total 35 37 22 21 101 17 233 Table 4 3. Farm size and location in watershed with Basin Management Action Pl an of participants for respondents in a study of factors affecting adoption of soil testing by large and small farmers, 2012 Farm size (in acres) Location in watershed with BMAP <1 1 10 10 30 30 50 50+ Did Not Indicate Total Yes 2 12 3 3 26 0 46 No 13 6 2 13 6 27 2 123 Not Sure 7 25 2 3 8 1 46 Did Not Indicate 1 1 0 0 0 16 18 Total 23 100 18 12 61 19 233
61 Table 4 4. Farm location and participant ethnicity for respondents in a study of factors affecting adoption of soil testing by large and small farm ers, 2012 Farm Location Ethnicity AL CA FL GA NC TX Other State Did Not Indicate Total White/Caucasian 6 14 118 13 1 23 11 4 190 Black/African American 0 0 2 2 0 2 0 1 7 Hispanic/Latino 0 0 8 0 0 1 0 0 9 Other 0 0 0 0 0 0 1 1 2 Did Not Indicate 0 2 3 1 0 0 1 18 25 Total 6 16 131 16 1 26 13 24 233
62 Table 4 5. Results of the Shapiro Wilks test for normality for five theoretically defined attribute variables separate d into post hoc groups based on farm size, in a study of factors affecting adoption of soil testing by small and large farmers Valid N Mean Variance Std Dev Skewness Kurtosis Shapiro Wilks Relative Advantage L 55 0.60 0.00 0.06 2.06 8.77 p = 0.00 S 177 .60 0.01 0.08 1.41 2.38 p = 0.00 Compatibility L 55 0.62 0.01 0.12 1.85 4.15 p = 0.00 S 174 0.58 0.01 0.10 1.86 5.78 p = 0.00 Complexity L 54 0.54 0.01 0.10 0.96 0.74 p < 0.05 S 172 0.46 0.02 0.14 0.76 0.60 p < 0.05 Trialability L 51 0.48 0.01 0.09 1.24 1.29 p < 0.05 S 164 0.46 0.01 0.11 0.51 0.37 p < 0.05 Observability L 53 0.56 0.02 0.16 1.68 3.33 p = 0.00 S 164 0.46 0.02 0.13 0.79 0.82 p < 0.05
63 Table 4 6. Results of the Shapiro Wilks test for normality for five theoretically defined attribute variables, separated into post hoc groups based on farm size, after data had been through logistic transformation, in a study of factors affecting adoption of soil testing by small and large farmers Table 4 7. Results of the Mann Whitney U test for central distribution for ea ch of five theoretically defined attribute variables, in a study of factors affecting adoption of soil testing by small and large farmers Valid N for L Valid N for S Rank Sum for L Rank Sum for S U Z p level Z Adjusted p level Relative Advantage 55 177 5 855.5 21172.5 4315.5 1.26 0.20 1.27 0.20 Compatibility 55 174 6925.5 19409.5 4184.5 1.40 0.16 1.40 0.16 Complexity 54 172 7292.0 18359.0 3481.0 2.77 0.0 1 0.0 1 0.01 Trialability 51 164 5401.5 17818.5 4075.5 0.27 0.78 0.78 0.78 Observability 53 164 72 03.5 16449.5 2919.5 3.58 0.00 0.00 0.00 Valid N Mean Variance Std. Dev. Skewness Kurtosis Shapiro Wilks Relative Advantage L 55 3.95 0.25 0.50 0.96 3.52 p < 0.05 S 177 4.01 0.46 0.68 0.81 0.54 p = 0.00 Compatibility L 55 4.02 0.79 0.89 1.02 0.70 p < 0.05 S 174 3.89 0.62 0.79 0.78 0.68 p = 0.00 Complexity L 54 3.42 0.55 0.74 0.36 0.11 p < 0.20 S 172 3.05 0.78 0.88 0.02 0.62 p < 0.05 Trialability L 51 2.88 0.26 0.51 0.75 0.36 p < 0.05 S 164 2.94 0.47 0.69 0.21 0.10 p < 0.05 Observability L 53 3.50 1.01 1.00 0.61 0.07 p < 0.05 S 164 3.00 0.71 0.84 0.03 0.46 p < 0.05
64 Table 4 8. Results of the Spearman rank order correlation for five theoretically defined attributes, for small and large farms combined, in a study of factors affecting adoption of soil testing by small and large farmers. Values with asterisk Relative Advantage Compatibility Complexity Trialability Observability Relative Advantage 1.0 0 0.60 0.07 0.31 0.32 Compatibility 1.0 0 0.09 0.37 0.5 5 Complexity 1.0 0 0.23 0.0 4 Trialability 1.00 0.33 Observability 1.00 Table 4 9. Results of the Spearman rank order correlation for five theoretically defined attributes, separated into comparison groups of small and large farmers, in a study of factors affecting adoption of soil testing by small and large farme rs. Values with asterisk Relative Advantage Compatibility Complexity Trialability Observability Relative Advantage L 1.00 0.56 0.0 1 0.29 0.51 S 1.00 0.63 0.06 0.3 2 0.3 3 Compatibility L 1.00 0.1 1 0.1 9 0.5 8 S 1.00 0.1 3 0.4 4 0.52 Complexi ty L 1.00 0.3 6 0.0 1 S 1.00 0.20 0.12 Trialability L 1.00 0.29 S 1.00 0.3 6 Observability L 1.00 S 1.00 Table 4 10. Spearman rank order correlations for demographic covariates, for small and large farms combined, in a stud y of factors affecting adoption of soil testing by small and large farmers. Values with asterisk Years of experience Acres in production Gender Age Years of experience 1.00 0.4 4 0.2 4 0.3 4 Acres in production 1.00 0.3 9 0. 10 Gender 1.00 0.0 8 Age 1.00
65 Table 4 11. Spearman rank order correlations for demographic covariates, separated into comparison groups of small and large farms, in a study of factors affecting adoption of soil testing by small and large farmers. Values with asterisk Years of experience Acres in production Gender Age Yea rs of experience L 1.00 0.39 0.02 0.55 S 1.00 0.32 0. 20 0.33 Acres in production L 1.00 0.0 5 0.11 S 1.00 0.32 0.1 4 Gender L 1.00 0.10 S 1.00 0.12 Age L 1.00 S 1.00
66 Table 4 12. Spearman rank order correlations for theoret ically defined attribute variables and demographic variables, for both small and large farms, in a study of factors affecting adoption of soil testing by small and large farmers. Values with asterisk significant at Rel. Adv. Compat. Complex. Trial. Observ Yrs of exp. Acres in prod. Gender Age Rela tive Advantage 1.00 0.60 0.07 0.31 0.32 0.06 0.08 0.2 2 0.08 Compatibility 1.00 0. 10 0.37 0.5 5 0.1 1 0.22 0.16 0.08 Complexity 1.00 0.23 0.0 4 0.0 3 0.0 2 0.05 0.04 Trialability 1.00 0.33 0.0 8 0.06 0.0 3 0.0 1 Observability 1.00 0.25 0.3 2 0.1 9 0.0 9 Years of experience 1.00 0.4 4 0.2 4 0.3 4 Acres in production 1.00 0.3 9 0. 10 Gender 1.00 0.0 8 Age 1.0 0
67 Table 4 13. Spearman rank order correlations for theoretically defined attribute variables and demographic variables, separated into comparison groups of small and large farms, in a study of factors affecting adoption of soil testing by small and larg e farmers. Values with asterisk Rel. Adv. Compat. Complex. Trial. Observ. Yrs of exp. Acres in prod. Gender Age Relative Advantage L 1.00 0.56 0.0 1 0.29 0.51 0.2 2 0.1 8 0.11 0.2 1 S 1.00 0.63 0.06 0.3 2 0.3 3 0.0 6 0.14 0. 2 8 0.0 7 Compatibility L 1.00 0.1 1 0.1 9 0.5 8 0.1 7 0.4 4 0.19 0. 20 S 1.00 0.1 3 0.4 4 0.52 0.07 0.14 0.13 0.0 3 Complexity L 1.00 0.3 6 0.0 1 0.0 5 0.07 0.10 0.0 5 S 1.00 0.20 0.12 0.11 0.1 2 0.00 0.03 Trialability L 1.00 0.29 0.2 1 0.0 7 0.12 0.15 S 1.00 0.3 6 0.04 0.0 6 0.0 2 0.02 Observability L 1.00 0.20 0.24 0.18 0.17 S 1.00 0.1 7 0.18 0.12 0.0 7 Years of experience L 1.00 0.39 0.02 0.55 S 1.00 0.32 0. 20 0.33 Acres in production L 1.00 0.0 5 0.11 S 1.00 0.32 0.1 4 Gender L 1.00 0.10 S 1.00 0.12 Age L 1.00 S 1.00
68 CHAPTER 5 DISCUSSION Research Question I tested two hypotheses in my research. I found evidence that partially supports one of the hypotheses and that does not support the second hypothesis. Below, I will present both the evidence and potential explanations for my results. I will also compare and contrast my findings with those of other published research. Globally, food dema nd is increasing. To meet that demand, agriculture needs to expand to more land, or intensify. Intensification of agriculture requires increased inputs of nutrients and water, so that more crops or animals can be produced on the same amount of land. Farmer s must estimate the nutrient needs of their crop as well as the nutrient availability of their soil. If they cannot make this estimation, they could end up wasting money on unnecessary fertilizer inputs. In addition, excess fertilizer can leave fields and cause problems with water quality in surface and ground water. To prevent these problems, farmers can take soil tests and plant tissue tests to estimate their nutrient input needs accurately Soil tests are relatively inexpensive and accessible. However, m any farmers fail to use soil tests. This research was designed to understand more fully this failure to adopt soil testing. It is possible that large and small farmers have different perceptions or adoption rates of soil testing. If I discovered differenc es between small and large farmer perceptions of soil testing, then attempts to encourage the spread of soil testing could be more specific to the comparison groups. Complexity and observability arose as different perceptions between small and large farmer s. Educators, policy makers and scientists can then focus on making soil testing less complex and more observable for
69 small farmers. Doing the same for large farmers would have less of an effect on adoption rates. Comparison to Attribute Literature The res ults of this study contribute to further understanding of the theory of diffusion of innovations in a few ways. First, this study indicated that two of the innovation attributes differed between participants, while the other three did not. I have found no studies that support the theory completely in stating that every one of the five attributes contributed to the overall perception of the innovation. Every study I have read has indicated that one or more of the innovation attributes contributes more to the decision to adopt an innovation than the other attributes (Atwell, Schulte & Wesphal, 2009; Carter & Belanger, 2005; Chakravarty & Dubinsky, 2004; Du, 1999; Eastlick, 1993; Waller et al., 1998). There have also been studies that conclude that one or more attribute is not correlated, negatively or positively, with adoption behavior (LaRose & Atkin, 1992; Lin, 1998). I was not studying whether one attribute would contribute towards adoption behavior more or less than any of the other attributes. Instead, I w as studying whether the perception of the innovation could be measured through each of the attributes. My conclusion that complexity and observability are the only two attributes that show differences between comparison groups is similar to the conclusion s drawn from Waller, et.al, in 1998. They conducted a study into the relationship between Integrated Pest Management innovations with potential users and looked specifically at many characteristics of the potential users. They concluded that compatibility, complexity and observability were the most meaningful attributes of an innovation for predicting adoption behavior. While I did not find compatibility to be an important attribute in my
70 study, my results are in agreement with them on complexity and observ ability. One main difference between this study and theirs that may explain why I did not find compatibility to arise as statistically significant could be the differing innovations. While soil testing is something that does not interfere with other practi ces on the farm, integrated pest management could. Therefore, it is conceivable that integrated pest management may not be compatible with other farm practices at some points and therefore it is an important part of the innovation attributes. Soil testing would be difficult to classify as incompatible with other practices on the farm, so my respondents may all have scored it as highly compatible. If all the scores were the same between small and large farmers, I would not be able to discern a difference for the attribute as a whole. Grieshop, Zalom and Miyao (1988) identified complexity as the most important attribute for adoption behavior in their study on California tomato farmers. Their study was similar to mine with the theoretical and accessible popul ation. They also focused on both the attributes of the innovation and the characteristics of the respondents. A main difference between this study and their study was that integrated pest management techniques is a much broader category for an innovation. It was also an innovation that had been introduced to the public much more recently than soil testing, especially at the time of publication. These differences may account for why Grieshop, Zalom and Miyao did not identify observability as an important inn ovation attribute. Nowak also identified complexity as an important attribute in the diffusion of innovations (1987). This study focused on conservation tillage techniques, which is a similar innovation to soil testing in economic benefits. The study was b roader in scope than just innovation attributes; it also compared economic and diffusion models,
71 concluding that the models are complementary to one another. He also states that personal characteristics of the adopter often affect the decision to use the i nnovation. Atwell, Schulte and Westphal (2009) interviewed farmers in the corn belt about compatibility and observability as important factors in a decision about adoption of innovations. Some differences between our studies were the type of farmer being studied (corn and livestock farmers for Atwell, Schulte and Wesphal, horticultural for this study) and the use of interview data versus index scoring data. In addition in t he interview study, farmers were encouraged to state what attributes of innovations they thought were most important to their decisions to adopt. Conversely, farmers were asked multiple questions about each construct and their scores were then used to dete rmine which attribute seemed most important, in this study. It is conceivable that farmers would self identify a certain attribute as strongest and their scores on an index would indicate otherwise. Therefore, I think it is important to note that Atwell, S chulte and Westphal concluded that observability was a strong attribute, as did I, but there were many differences between our studies. There were two studies that I considered well based theoretical research in the diffusion of innovations field that con cluded that complexity and observability (as well as trialability) were the weakest of the five innovation attributes. These studies were conducted by Eastlick (1993) and Carter and Belanger (2005) and neither study was focusing on an innovative practice. Instead, they focused on electronic services, such as videotex (Eastlick, 1993) and the use of e government services (Carter & Belanger, 2005). The innovation I was testing requires adopters to do something at a certain point
72 in time for a future return, w hereas electronic products seem much more tangible and convenient. It has been suggested that innovation characteristics might differ between adopters of new conveniences ( cell phones electronics) and adopters of innovative practices (conservati on tillage, drip irrigation). In addition neither of these studies addressed agricultural innovations. It is possible that farmers have a distinct set of characteristics that set them apart from other adopters in biased ways. Theoretical Contributions of this Study The theory of diffusion of innovations is incomplete, as are most theories. The results of this and many other studies have led researchers to conclude that one or more of the innovation attributes does not explain adoption behavior (Agarwal & P rasad, 1997; Atwell, Schulte & Westphal, 2009; Carter & Belanger, 2005; Chakravarty & Dubinsky, 2004; Du, 1999; LaRose & Atkin, 1992; Lin, 1998; Moore & Benbaset, 1991). These conclusions alone are cause for concern about the overall explanatory power of t he theory. The theory of diffusion of innovations has matured over time. Originally, the theory posited that increased education of farmers would lead to agricultural adoption (Rogers, 1962). Since this basis in agricultural innovations, adopter categories and innovation attributes were incorporated into the theory. The theory continued to work well for specific agricultural situations, but then researchers from other domains started to use the theory. However, identifying the attributes of innovations has remained problematic because research has shown that there may be additional attributes or because some of the posited attributes are not explanatory under all conditions. If the validity of attributes varies between populations or innovations, the theory as a whole is potentially flawed if it is intended to explain adoption of all innovations in all populations. This theory varies in explanatory power between populations, as shown in many studies
73 (Atwell, Schulte & Westphal, 2009; Carter & Belanger, 2005; Nowak, 1992; Nowak & Korsching, 1983). The domain to which this theory can be applied is not well enough defined. The theory is unclear about too many situations. Everett Rogers, or somebody else, could fix this by defining the domain for the theory of dif fusion of innovations. There are two main reasons that I think we continue to rely on the theory of diffusion of innovations despite its flaws. First, there is no competing theory to explain why and how innovations diffuse across communities. A competing t heory would stimulate research and debate that could help bring clarity to the conditions under which the theory can be applied. Second, the theory is too complex. It is trying to explain every possible innovation and every possible adopter in the world. S implification and defining the situations in which the theory applies would make the theory more usable in research. Until the theory is more specified, its premises should be questioned before use in research. Pro Innovation Bias The theory of diffusion o f innovations in its current form does not address three additional problems. First, there is a well known pro innovation bias in diffusion research. Most published studies have examined situations in which the innovation was adopted and have focused on th e factors that led to the decision to adopt. There has been little research about innovations that a majority of potential adopters decided not to adopt. I nnovations that were first adopted and later rejected for unknown reasons also lack significant resea rch For example, innovations such as common childhood vaccinations in the United States have stalled, following years of increasing use, at the 2004 levels of coverage (82%) (National Immunization Program, 2003). The spread of vaccination rates has become impeded by something that is not included in the theory.
74 The underlying assumption of most diffusion studies is that innovations should be adopted by users. However, not all innovations are equally beneficial for all users. Increased research that directl y addresses the pro innovation bias could shed light on whether the theory is equally useful in explaining the failure to adopt certain innovations. In this study, I assume that soil testing is positive for the majority of users. Innovation Price S econdly few studies consider the cost of innovations. An inexpensive versus an expensive innovation would be expected to have different characteristics in two regards A difference in innovation cost could cause differences in the theoretical innovation attribut es, as well as in a cost benefit analysis. However, the only innovation attribute from the theory that might address cost is relative advantage. Even with relative advantage however, the cost of the innovation is only considered in comparison to the cost o f the previous practice, technology, or item. The stand alone cost of the innovation is not included in the theoretical attributes. Cost could be a serious constraint for some potential adopters. For example, small farmers often have less discretionary inc ome than large farmers, which could prevent them from adopting costly technologies. Therefore, even if an innovation has many very desirable attributes and the adopter has the characteristics of an innovator or early adopter, cost alone may dissuade adopti on. innovations and low priced innovations could help incorporate this factor in the theory. Conservation Technology Finally, there is on going research into whether the theory o f diffusion of innovations explains conservation technology adoption. The contemporary theory is assumed applicable for all innovations. However, a number of studies have shown that
75 conservation values can affect adoption behavior both positively and negat ively. We need more research to determine whether potential adopters consider the same attributes for conservation technologies as for other innovations. If so, then the theory can stay the same, albeit with a stated domain that includes conservation techn ologies. If not, it may be necessary to change the domain of the theory in certain ways. It may be that the theory is applicable for only certain types of innovations. If this limited applicability is present than this assumption of innovation type needs to be clarified and stated as part of the domain to which the theory applies. added to the theoretical innovation attributes. Values have affected adoption of conservation tec hnologies, for example. They have also affected adoption of other technologies, such as childhood vaccinations in the United States (National Immunization Program, 2003). This research could explain factors that lead to conclusions on whether or not conser vation technologies should be included in the domain to which the theory applies. Comparison to Attribute Correlations Literature with each other, with one exception. Complexity did not correlate with any of the other attributes, except for a negative correlation with trialability (Table 4 8). A corr elation between attributes could indicate several things. Correlation may be a result of items in the index representing multiple attributes. It could also indicate that the attributes are inter related. For example, an adopter might consider the compatibi lity or observability
76 Correlations between attribute scores describe the strength and direction of relationships between variables. A relationship between variables m akes it difficult to pinpoint a specific attribute in explaining innovation adoption. This possible relationship between attributes may be part of why the theory has shown such differing results in the varying attribute studies. Valid research has conclude d that compatibility is the most important attribute, but other valid research has found that relative advantage is the most important attribute. I found a strong positive correlation (R 2 =0.60) between the two attributes (Table 4 8). If previous researcher s tested the relationships between variables, they might have drawn different conclusions. It would be imprecise to consider a specific attribute as explanatory if it is highly correlated to another attribute that is not a significant predictor of adoption behavior. In my study, I identified complexity and observability as the two attributes that showed significantly different scores between comparison groups. Complexity shows no strong correlations to other attributes, but observability does. The scores fo r observability are significantly correlated with the scores for relative advantage, trialability and compatibility. Those attributes did not show significantly different scores between comparison groups, but their scores are somehow correlated to the scor es for observability. The presence of significant correlations between scores suggests a relationship between observability and relative advantage, trialability and compatibility. The correlation between the scores for relative advantage and observability has a much higher R 2 for large farms than they are for small farms (Table 4 9). The relationship between relative advantage and observability is thus stronger for large
77 farmers than for small farmers. This difference suggests that large farms may view the observability of soil testing as an advantage. Observability, as a construct, includes the degree to which the adopter has seen or communicated about the innovation before adoption, as well as the degree to which other people are able to see or communicat e about the innovation after the adopter has begun use. Perhaps farmers think that if the public observes them using soil testing, they will view the farm as more environmentally conscious, which could increase their business. Therefore, my conclusions are not that complexity and observability are the only explanatory attributes in the theory because the other three attributes are all related to observability in some way. It is also significant that complexity and observability did not correlate with each o ther. Had they been highly correlated, it is possible that one of the attributes created statistical differences that were not as pronounced as the scores would suggest. The insignificant R 2 for the correlation between the two attributes indicates that the re is not a relationship between the two attributes. This increases my confidence that complexity and observability differ between small and large farmers. Farm Size This study examined the relationship between farm size and perceptions of soil testing at tributes. Multiple studies have examined the association between the size of an institution and the decision to adopt an innovation. My conclusion that farm size is significantly related to only two out of the five innovation attributes is unusual in diffu sion studies. To my knowledge, other studies have not reached this conclusion, but the literature is extensive and it is possible that others have reached this conclusion. The distinction of this study was the comparison between size of an institution and
78 perceptions of attributes, instead of size and adoption behavior. Extremely positive perceptions of attributes do not equate to adoption in all cases. I agree with the conclusions of other studies that institution size is not a sole predictor of adoption b ehavior (Grieshop, Zalom & Miyao, 1988; Ridgley & Brush, 1992; Swisher & Bastidas, 1994). The theory of diffusion of innovations would also support this conclusion since this size alone does not take into account innovation attributes, innovator categories communication channels, social systems and other potentially relevant factors. Grieshop, Zalom & Miyao were unable to conclude that size can directly affect adoption behavior, but their instrument had major flaws. Both their study and my own used an inde x to measure data, with similar data analysis techniques. However, the reliability of my instrument was much higher. A small sample size and low reliability limits the ability to perform some statistical tests that could predict relationships instead of si mply test the correlation among variables. Ridgley and Brush used a different design and methods than I did and yet reached a similar conclusion. Their sample was of pear farmers in California and their methodologies were a combination of interviews and ca se studies. They did not originally include size of the institution as a variable of study, but it arose in their interviews. Despite these contrasts between my study and theirs, we both concluded that farm size relates to adoption decision making. Swishe r and Bastidas (1994) looked into the relationship between farm size and adoption of environmentally sensitive practices. There was no explicit use of a theory in the study, although the theory of diffusion of innovations was used in the design of the stud y (M. Swisher, personal communication, February 10, 2013). They studied
79 horticultural and agronomic farmers and concluded that size was not a predictor of adoption behavior. However, when they combined size and certain institutional characteristics, the re lationship became clearer. My study was similar to the selected sample, excepting the agronomic producers. It was also similar in conclusion that complexity and observability are related to farm size, but that there is not necessarily a relationship betwee n farm size and adoption of soil testing. Some main differences include that my sample included participants from more than just Florida, and that my study was not focusing on adoption behavior, but on perceptions of innovation attributes. Nowak (1992) co ncluded that farm size is related to adoption rates of innovations. He identified several factors that lead small farmers to decide against adoption, including limited financial, space or other resources and knowledge of how to use the innovation. I reache d similar conclusions about small farmers and their views of innovations including complexity and observability. Small farmers may feel that their ability to afford an innovation affects their decision to adopt even at initial exposure to the innovation. S mall farmers that do believe that they have the needed resources to adopt might then consider the perceived complexity and observability of the innovation. Large farmers generally have more knowledge or access to resources that can simplify the use of an i relates adoption rates to farm size, while I am relating farm size to perceptions of attributes of an innovation. However, my study may help explain his results. Harper et al. (1990) conclud ed that, for inexpensive innovations, farm size is not related to adoption of integrated pest management techniques. They did not conclude
80 that institution size has no relationship with innovation adoption and they did not use the theory of diffusion of in novations. However, their observations about innovation adoption behavior are interesting. They did focus on adoption behavior while I examined attribute perception. Again, I think the conclusions that I reached about complexity and observability can help to explain their results, especially from a theoretical basis. The theoretical framework I used makes it possible to generalize their conclusions beyond their sample of rice farmers in Texas and enhances the explanatory power of their work. Size of an inst itution should be taken into account with the theory of diffusion of innovations when measuring complexity and observability of an innovation. My results make it clear that there is a relationship between farm size and the perceptions of complexity and obs ervability of soil testing (Table 4 7). It is important to consider this relationship for future research based on the theory of diffusion of innovations. Since the theory does not currently consider size of an institution as a characteristic of potential adopters or as a factor that can affect perceptions of attributes, the theory could be strengthened by including size. This study made it clear that there is a relationship between some attribute perceptions and the size of an institution, something that t he theory cannot currently explain. I suspect that the five attributes of an innovation are not as clearly related to adoption behavior as the theory states. Size should be included as one of the characteristics that determine the adopter category of an in stitution. Demographic C ovariance There is a relationship between years of experience and acres in production with gender when the comparison groups are combined into a single data set. The R 2 value for the correlation between years of experience and acre s in production with gender was significant and negative (Table 4 10). Scores for women are lower than for men for
81 both characteristics. However, for large farmers there is not as significant correlation between the scores for gender and any other demograp hic characteristic (Table 4 11). This weaker correlation may be because there were very few women respondents assigned to the large farm comparison group. Women in agriculture work more often on small farms than large farms throughout the United States (Un ited States Department of Agriculture, 2007e). Extension agents or policy makers trying to reach farmers with few years of experience or small acreage focus on working with women as well as men on small farms. I also concluded that small farmers perceive s oil testing as being complex and low in observability. Policies or programs designed to reduce the perceived complexity of soil testing complexity by small farmers could focus on recruiting women with few years of farming experience or women with small acr eage in production. The potential programs or policies would be able to focus their efforts on those audiences, potentially creating more value for the money invested in the programs. One example of a program that is already partially targeting female farm ers with few years of farming experience is the Beginning Farmer and Rancher Development Program, funded through the USDA. The correlation between the scores for age and years of experience for combined comparison groups is positive (Table 4 10). This corr elation suggests that as farmers age, they typically have more years of farming experience. This suggestion makes logical sense, since it would be difficult to have many years of experience at a young age. The correlation between the scores for age and yea rs of experience for large farms is even stronger than the correlation for the combined comparison groups (Table 4 11). This correlation is logical as well because it takes time and experience to
82 build a business. It is therefore probably not fruitful to d evelop policies or programs designed for younger farmers based on concepts that make sense for those with extensive experience. The correlation between the scores for years of experience and acres in production suggests a relationship between the attribute s that reflects the same logic. Buying and managing large amounts of land takes substantial capital, which is typically not accumulated until a business has grown. This growth takes time and experience. Programs targeted towards larger farms could focus on farmers with many years of experience. For example, extension programs provide information about new technologies or innovations that are most applicable to farms with large acreage might recruit through trade associations or other organizations that incl ude high numbers of farmers with many years of experience. When the theoretical attributes were tested against demographic variables, some interesting relationships emerged. For the combined comparison groups, scores for gender showed correlations with sco res for relative advantage, compatibility and observability (Table 4 12). Scores for those variables were higher for men than for women. This variance in scores indicates that programs or policies designed to increase the perceptions of soil testing as adv antageous, compatible with other agricultural techniques, and/or observable should be targeted towards men. Women already perceive soil testing positively with respect to those attributes. Further policies or programs would have more of an affect if they f ocused on parts of the population that did not already accept their conclusions. When separated into farm size, the correlations only remain for relative advantage and gender (Table 4 13). Gender scores no longer show significant correlations with scores f or compatibility or observability. If
83 future policies or programs are targeting specific sizes of farms, then both genders should be included. Scores for years of experience and observability were correlated for combined comparison groups (Table 4 12) and for small farms when the groups were separated, but not for large farms (Table 4 13). A positive relationship between years of experience and observability is logical. The more years of experience a farmer has, the more likely s/he is to have heard of, or seen, soil testing performed. The insignificant correlation between scores for years of experience and observability for large farmers indicates a lack of a measurable relationship between years of experience and observability for large farmers. Large far mers either do not see mention of soil testing as often as they do other agricultural techniques, or that they see soil testing performed much less often than other techniques. It is possible that small farmers discuss soil testing more often among each ot her, enhancing the perception of observability. If extension agents or trade associations were to focus on soil testing for a short period large farmers might increase their perception of observability. Scores for acres in production and compatibility wer e correlated for combined comparison groups (Table 4 12). When separated into comparison groups, the correlation was not significant for small farmers (Table 4 13). Horticultural farmers in general have a relationship between the number of acres that they have in production and their perception of the compatibility of soil testing, whereas small farmers do not show this measurable relationship. Programs that focus on showing how soil testing is compatible with other nutrient management techniques should be targeted to small farmers. The large farmers already perceive compatibility as high. The same is true of
84 observability, which has a relationship with acres in production when the comparison groups are combined (Table 4 12), but not when small and large gro ups are separated (Table 4 13). The scores for observability and acres of production produce a significant R 2 for the combined farm sizes, but not when separated into comparison groups. However, in this case there is no significant correlation between the two variables for large farmers. A trade association or extension program aimed at increasing the observability of soil testing for large farms could improve adoption rates. There are varieties of ways in which the results of this study could be applied to practical on farm decision making. This study suggests that small farmers and large farmers see the complexity of the innovation differently (Table 4 7). Small farmers overall perceive soil testing as more complex than do large farmers. These results su ggest that institutions or personnel working to expand the use of soil testing should focus on reducing the perception that soil testing is complex, particularly among small farmers. If extension personnel were to offer trainings or explanations of how to better use soil testing, it might lead to a greater adoption Farm demonstrations combined with trainings that utilize an experiential learning approach might lower perceptions of complexity. There are a large number of decisions that must be made to take an adequate soil sample. If the institution that processes soil samples would create a web page explaining the basics of collecting a representative sample, small farmers might perceive of soil testing as less complex. If the results include suggestions fo r nutrient management or a way to contact an extension agent with questions, small farmers might consider soil testing less complex.
85 The observability of soil testing was the other attribute for which small and large farmers had differing perceptions. Sma ll farmers perceived the observability of soil testing to be significantly lower than did large farmers (Table 4 7). These results suggest that institutions or personnel working to expand the use of soil testing should focus on improving the observability of soil testing, particularly among small farmers. Extension personnel could try to achieve this in part through on farm demonstrations of soil testing or perhaps an on line video showing a systematic process conducted by a farmer taking the sample and usi ng the results. A site tour of a farm followed by walking through the steps involved in taking a representative sample from the farm might increase observability of soil testing, while also decreasing the perceived complexity. Limitations There were certai n trends apparent in the covariates that had the potential to bias results. Twenty five percent of male respondents and 49% of female respondents had less than 10 years of experience with farming (Table 4 2). This difference is statistically significant an d may have affected the results of this study. Gender, as a covariate, may also be linked with years of experience because so many more women than men had relatively little farming experience. The majority of farmers in the United States have more than 10 years of experience (United States Department of Agriculture, 2007c). This difference between the sample and the theor etical population in this study may affect the internal validity of the results. There were significantly more responses from participant s that self identified themselves as small farmers than from those that self identified as large farmers. In the United States, there are a greater number of small farms than large (United States Department of Agriculture, 2007d) and the ratio of small to large farmers was
86 representative of this number I wanted more responses from large farmers for statistical purposes only. The low response rate placed limitations on the available statistical tests I could perform, and an increased number of responses fro m large farmers would have increased the number of tests I could use. Because I attended a Small Farms Conference, I should have also attended a conference focusing on large farms. When planning my sampling procedure, I focused my efforts on the seven sta tes with the highest levels of horticultural production. I chose to use those states because they were the states listed with the highest total horticultural sales from the 2007 Census of Agriculture (United States Department of Agriculture, 2009). I conta cted extension agents in those states first. Most responded positively, offering to send out the index link to at least three farmers. However, the extension agents in Oregon politely responded that it was against their policies to contact farmers on behal f of data collectors that the agents did not personally know. I then tried to contact individual farms in Oregon with no success. Responses from farmers in Michigan were also a problem. Both farmers and extension agents from Michigan informed me that the t ime of year in which I was contacting them (August) was a busy harvest season and that farmers would be unlikely to respond to my requests for feedback. Despite some further attempts, I received no responses from Michigan. At that point, I reached out to i ndividual farmers in states other than the top horticultural producers and received 13 usable responses. I do not know the response rate to my index. Response rates can be low using electronic requests for information. I asked extension agents to send the link for my
87 index to farmer s they thought would be eligible for and interested in the index. Most of the extension agents did not tell me how many people they contacted about the study. Therefore, I do not know how many total requests to fill out my index were distributed. I do know that the total number of usable responses was two hundred and thirty three. The total number of people that attempted to answer the index but were excluded for reasons involving that they did not fit the requirements of the inde x (under 18, not producing horticultural crops, etc), or for break off during response was two hundred and ninety nine. If a respondent filled out less than a quarter of the index, I eliminated their response as a whole. Rates of break off were approximate ly 11%. There are some demographic characteristics that I wish I could have tested further. Educational background and geographic location would have potentially yielded interesting results. T he total number of usable responses (233) limited my available s tatistical tests and I could not run a general regression for these covariates because there were so many answer choices. It would be particularly revealing to be able to run a correlation test between educational background and farm size. A test for farm size against geographic location could have informed me whether my sample was biased to a different farm size than was typical for the area surveyed. A test for farm size against location in a watershed with a basin m anagement a ction p lan could inform rese arch about whether voluntary or mandatory policies for water quality affect attribute perceptions. All of these demographics were included in the index, but were unable to test due to the limited number of usable responses. An item that was not included in the index but would also have been interesting to study would have been the amount of typical interaction between the respondent and
88 extension personnel in their county. I suspect that my sample was strongly biased towards farms and participants with regu lar contact with extension agents, mostly because my method of sampling required extension personnel to contact farmers for me. M easurable differences between farmers that regularly interact with extension agents and farmers that rarely or never contact Ex tension may exist Farm size could even differ between those populations. It is conceivable that large farmers might not contact extension agents as often because they have the financial ability to contract with private companies to get similar information more quickly or efficiently. Large farmers sometimes have specialists and laboratories as part of their business and therefore have no need to send soil samples to county agents. In addition when large farms come across a problem that they are unable to identify quickly they tend to go directly to the state specialist instead of first contacting their County Extension. I collected responses both at the Small Farms Conference and through personal contact with some farmers. These participants included app roximately 30% of the overall usable responses. The percentage is approximate because responses were anonymous and so I am only able to count the number of responses from the conference and that I know I personally recruited. It would be interesting to tes t for differences between those participants and the participants that extension agents recruited for me, for both farm size, perceptions of innovation attributes and demographics. The main limitation to this study is the low number of responses. I could h ave received more responses if I had done a few different things. First if I had attended at more than one conference in person, I could have asked more people in person to fill
89 out the index. Increased travel would have required time and funding that wer e not immediately available. Secondly, if I had contacted extension personnel in other states, they could have expanded the overall number of farmers that were asked to respond. This increased communication with extension personnel would have reached a sta te of diminishing returns at some point, because I had already contacted the states the produced the most horticultural products. For example, contacting states with little agriculture would have been a potential waste of time and resources. Thirdly, I cou ld have identified and contacted farmers over the Internet, but that would have potentially biased the results in favor of respondents that used e technology. In other words, that numbers had previously contacted to send out the request to farmers in their area, as this action might have increased responses with no associated biases.
90 CHAPTER 6 CONCLUSIONS The population of the world is still increasing and the amount of land available for agriculture is limited Therefore, agriculture must intensify production on the same amount of land, so that enough food is available for the population This increase in production requires increased nutrients and water use. When inputs increase, greater potential for damage to water quality on a variety of geographical scales is present Nutrient loss from farms is negative both for the farmer and for the envi ronment. One way to prevent nutrient loss is by utilizing soil testing, but many farmers do not use soil testing every production season, as is recommended. This research has shown that there are differences between small and large farmers in how they perc eive the complexity and observability of soil testing. It is possible that the gender and years of experience of the respondents affected and/or strengthened those results. Extension agents, educators and people looking to improve water quality can use the results of this study in multiple ways. It is apparent that the complexity of soil testing is important to small farmers. They could try to make soil testing less complex by using on farm demonstrations, discussions of how to take representative samples, or trainings on how to inte rpret results of soil testing. Aiming these attempts at small farmers would be more likely to cause changes in the perceptions of soil testing. They could also try to make soil testing more observable, again focusing on small far mers. Many extension agents send out periodical newsletters or updates and if that communication were to include discussion of soil testing, observability would increase Making available links to news clippings or journal articles about soil testing would also make soil testing more observable. Communication about an innovation is an important part of observability,
91 and so even in person conversations about soil testing could improve the rates of testing done by small farms. The less complex and more obser vable small farmers see soil testing as, then the more likely they are to adopt soil testing as a regular practice. That could lead to decisions about nutrient management on their farm that prevents nutrient loss and protects water quality. Implications f or Future Research This study focused on farmer perceptions of soil testing attributes, as explained through the theory of diffusion of innovations. While this study was cross sectional in design, it would be interesting to set up a longitudinal study. Re contacting the farmers that originally responded to the index over time (perhaps every year for ten years) and tracking their change in perceptions over time, as well as their change in demographics would allow for fascinating results. A dditional studies s hould be pursued to explain the interactions between soil testing as an innovation and farmer characteristics. Future research should explore the connections between perceptions of soil testing and farmer demographics such as educational background, farmer age and the geographic area where the farmer grows. In addition research to look into how farmers of livestock and agronomic crops perceive soil testing attributes should be performed Research into how mandates for soil testing or other conservation tec hnologies would affect perceptions and use of those innovations would inform scientists about whether or not farmers would abide by those laws. There are watersheds and geographic areas that currently have voluntary programs for protection of water quality such as b as in management action plans and best m anagement p ractices. Taking a sample of participants in these voluntary programs and assessing their perceptions of attributes could yield interesting results and conclusions.
92 When a farmer receives soil te sting results, they are able to see the amounts of nutrients that are in their soil. However, not all of those nutrients are in plant available forms. F armers also utilize plant tissue tests to see how many nutrients are in the plant itself to note differe nces between total and available nutrients in the soil If a similar index were given to horticultural producers to measure attributes of a plant tissue test, would results be similar? Organic farmers often take soil samples and receive the results that st ate how many nutrients are present in their soil. However, the decision they must make at that point differs from that of farmers using formulated fertilizers. While formulated fertilizers guarantee certain amounts and ratios of nitrogen, phosphorus and po tassium, fertilizers available for use on certified organic farms are not as obtainable as conventional fertilizers In addition, these available fertilizers often contain nutrients in forms that are less readily available to the crop. Organic farmers typi cally utilize composted plant materials or manures to add nutrients to fields and these materials vary in nutrient ratios and amounts. Therefore, a study could be done that compares the perceptions of soil testing attributes between certified organic farme rs and conventional farmers of the same crop. Theoretical Contributions Besides studying soil testing, this study also helped to explain parts of the theory of diffusion of innovations. While I focused on the attributes of the innovation with this research it would be interesting to be able to look at the interactions between perceptions of innovation attributes and the adopter categories for respondents. This potential research could be used with different innovation types, such as a tangible good, a prac tice and a conservation practice. If the same respondents were able to
93 score their perceptions of the attributes of these different innovation types, as well as their own adopter categories, the field of innovation diffusion research would be able to tell if the theory of diffusion of innovations is appropriate for most innovation types, or if the theory should differentiate between innovation types. Rogers (1962) defined innovations as any idea, product, process, concept, system, behavior or combination th ereof. I suspect that the differences between these types of innovations may contribute to some of the discrepancies in innovation diffusion studies. Policy Implications Soil testing is not currently mandated through any type of national policy. There are some accreditation programs, such as the National Organic Standards, which suggest the use of soil testing. However, even within the National Organic Standards, soil testing is not necessary to retain certification. This study is not suggesting that soil t esting should be mandated, but instead is looking to the future of water quality issues and considering how soil testing could be a part of a multi step solution Currently, many agricultural commodities have b est m anagement p ractices developed for them, b ut none are mandated. If b est m anagement p ractices become mandated in the future, soil testing may also become mandated. To both ensure and encourage compliance, the results and conclusions from this study could be helpful.
94 APPENDIX PERCEPTIONS OF SOIL TE STING IN DEX
104 LIST OF REFERENCES Adcock, R., & Collier, D. (2001). Measurement Validity: A Shared Standard for Qualitative and Quantitative Research. American Political Science Review Vol. 95, No. 3. pp.529 546. Agarwal, R. & Prasad, J. ( 1997). The role of innovation characteristics and perceived voluntariness in the acceptance of information technologies. Decision Sciences 28(3), 557 582. Asthana, A. & Kumar, S. (2008). Soil health and sustainable crop production: Adoption and awareness of soil testing. Journal of Sustainable Agriculture 32(2), 303 320. Atwell, R.C., Schulte, L.A. & Westphal, L.M. (2009). Linking resilience theory and diffusion of innovations theory to understand the potential for perennials in the U.S. Corn Belt. Ecolo gy and Society 14(1), 30 48. Barry, A.E., Chaney, E.H., Stellefson, M.L. & Chaney, J.D. (2011). So you want to develop a survey: Practical recommendations for scale development. American Journal of Health Studies 26(2), 97 105. Bigoness, W.J. & Perreault W.D. (1981). A conceptual paradigm and approach for the study of innovators. The Academy of Management Journal 24(1), 68 82. Blake, G., Sandler, H.A., Coli, W., Pober, D.M. & Coggins, C. (2007). An assessment of grower perceptions and factors influencin g adoption of IPM in commercial cranberry production. Renewable Agriculture and Food Systems. 22, 134 144. BMJ, 314(1), 572. Bower, H. (1994). Irrigation and global water outlook. Ag ricultural Water Management 25(3), 221 231. Bryman, A. (2004). Social Research Methods Oxford: Oxford University Press. Burney, J.A., Davis, S.J. & Lobell, D.B. (2010). Greenhouse gas mitigation by agricultural intensification. Proceedings of the Nationa l Academy of Sciences of the United States of America. 107(26), 12052 12057. Buttell, F., Gillespie, G., Larson, O. & Harris, C. (1986). Soil and water conservation: a farm structural and public policy context. In S. Lovejoy and T. Napier (Ed.), Conservi ng Soil: Insights from Socioeconomic Research (pp. 26 39). Ankeny, Iowa: Soil Conservation Society of America. Marland, G. (2007). Contributions to accelerating atmosph eric CO2 growth from economic activity, carbon intensity, and efficiency of natural sinks. Proceedings
105 of the National Academy of Sciences of the United States of America. 104,18866 18870. Carter, L. and Belanger, F. (2005). The utilization of e government services: citizen trust, innovation and acceptance factors Information Systems Journal, 15 (1): 5 25. Central Intelligence Agency. (2009). The World Factbook. Washington, DC. Accessed Online June 22, 2012 at https :// www cia gov / library / publications / the world fac tbook / geos / xx html Innovation diffusion in fin ancial markets. Journal of Economic Psychology 26, 89 103. Chaves, B. & Riley, J. (2001). Determination of factors influencing integrated pest management adoption in coffee berry borer in Columbian farms. Agriculture, Ecosystems and Environment. 87, 159 1 77. Churchill, G.A., (1979). A paradigm for developing better measures of marketing constructs. Journal of Marketing Research ,16 (1) 64 73. Collins, D. (2003). Pretesting survey instruments: An overview of cognitive methods. Quality of Life Research 12, 2 29 238. spring soil nitrogen test. American Journal of Alternative Agriculture 12(1), 20 29. Cronbach L. J. (1951). Coefficient alpha and the internal structure of tests Psychometrika 16 (1), 297 333. de Vaus, D. A. (2001). Research design in social research London: Sage Publications. DeVellis, R. (2003). Scale Development: Theory and Application Thousand Oaks, California, Sage Publications. Dirzo R. & Raven P. (2003) Global state of biodiversity and loss. The Annual Review of Environment and Resources. 28(1), 137 167. Du, X. (1999). Internet diffusion and usage in China. Prometheus 17(4), 405 420. Eastlick, M. A. (1993). Predictors of videotex adoption. Journal of D irect Marketing 7(3), 66 74. Economic Research Service. (2010). Structure and Finances of U.S. Farms: Family Farm Report 2010 Edition. United States Department of Agriculture. Accessed
106 Online July 4, 2012 at http://www.ers.usda.gov/publications/eib econom ic information bulletin/eib66.aspx. Economic Research Service. (2012). State fact sheet: Florida. United States Department of Agriculture. Accessed Online July 4, 2012 at http://www.ers.usda.gov/data products/state fact sheets/state data.aspx?StateFIPS=12& StateName=Florida#P0213bed734564eda8916eb7db 4ec2f64_2_499iT18C0x0. Extension Soil Testing Laboratory. (2012). Mission statement. University of Florida Analytical Services Laboratories. Accessed Online July 3, 2012 at http :// soilslab ifas ufl edu / ESTL %20 Home asp Falkenmark, M., Finlayson, M. & Gordon, L. (2007) Agricu lture, water and ecosystems: avoiding the costs of going too far. Chapter 6. D. Molden (Ed.), Water for Food, Water for Life: A Comprehensive Assessment of Water Management in Agriculture Water Management Institute, London, Earthscan and Colombo. Fast, S. (1983). Conservation: A tillage by product. Journal of Soil and Water Conservation. 38(6), 443 449. Fernandex Cornejo, J., Beach, E. & Huang, W. (1992). The adoption of integrated pest management technologies by vegetable growers. ERS Staff Report. No. A GES9228. Field, A. (2005). Discovering Statistics Using SPSS 2nd ed. London: Sage. Fliegel, F. & van Es, J. (1983). The diffusion adoption process in agriculture: changes in technology and changing paradigms. In G. Summers (Ed.), Technology and Social Ch ange in Rural Areas. (pp. 13 28). Boulder, Colorado: Westview Press. Flight, R.L., D'Souza, G. & Allaway, A.W. (2011). Characteristics based innovation adoption: scale and model validation. Journal of Product & Brand Management. 20(5), 343 355. Food and A griculture Organization. (2008). The State of Food Insecurity in the World 2008 Food and Agriculture Organization Rome. Food & Fertilizer Technology Center. (1985). Agricultural research to help the small scale farmer in developing countries. Retrieved o n July 5, 2012, from h ttp://www.fftc.agnet.org/library.php?func=view&style=type&id=20110727132617 Fowler, F.J., Jr. (1995). Improving survey questions. Design and evaluation. Sage, Thousand Oaks, pp. 156 165 (Appendix A).
107 Giller, K., Murwira, S., Dhliwayo, D., Mafongoya, P. & Mpepereki, G. (2011). Soyabeans and sustainable agriculture in southern Africa. International Journal of Agricultural Sustainability. 9(1), 5 0 58. Goldberger, J. (2008). Diffusion and adoption of non certified organic agriculture: A case study from semi arid Makueni District, Kenya. Journal of Sustainable Agriculture 32(4), 531 564. Goldewijk, K.K. (2001). Estimating global land use change ove r the past 300 years: The HYDE Database. Global Biogeochemical Cycles 15(2), 417 433. Grieshop, J., Zalom, F. & Miyao, G. (1988). Adoption and diffusion of the integrated pest management innovations in agriculture. Bulletin of the Entomological Society of America 34(2), 72 79. Haider, M. & Kreps, G.L. (2004). Forty years of diffusion of innovations: Utility and value in public health. Journal of Health Communications 9(1), 3 11. Hamilton, G., Robson, M., Ghidiu, G., Samulis, R. & Prostko, E. (1997). 75% adoption of IPM by the year 2000? American Entomologist 43(2), 74 77. Hammond, C., Lushei, E., Boerboom, C. & Nowak, P. (2006). Adoption of integrated pest management tactics by Wisconsin farmers. Weed Technology 20, 756 767. Harper, J.K., Rister, M.E., Mjelde, J.W., Drees, B.M. & Way, M.O. (1990). Factors influencing the adoption of insect management technology. American Journal of Agricultural Economics 72, 997 1005. Hatcher, L. (1994). A step by step approach to using the SAS(R) system for factor ana lysis and structural equation modeling. Cary, NC: SAS Institute. Holak, S. L. and Lehmann, D.R. (1990). Purchase intentions and the dimensions of innovation: An exploratory model. J Product Innov Manag, 7 : 59 73. Hoover, H. & Wiitala, M. (1980). Operator and landlord participation in soil erosion control in the Maple Creek watershed in northeast Nebraska. Washington, D.C.: Economics, Statistics, and Cooperative Service Staff Report, Natural Resources Economics Division 80 4. Hosmer, D.W. & Lemeshow, S. (19 89). Applied Logistic Regression. Wiley series in probability and mathematical statistics New York: John Wiley and Sons. Howard, P.H. & Allen P. (2010). Beyond organic and fair trade? An analysis of ecolabel preferences in the United States. Rural Sociolo gy 75(2), 244 269. Jacobsen, T. & Adams, R.M. (1958). Salt a nd silt in ancient Mesopotamian agriculture. Science 128(1334), 1251 1258.
108 Jensen, J.R., Halvorsen, K.E., & Shonnard, D.R. (2011). Ethanol from lignocellulosics, U.S. Federal energy and agricul tural policy, and the diffusion of innovation. Biomass and Bioenergy 35(4), 1440 1453. Jolly, W., Edwards, W. & Erbach, D. (1983). Economics of conservation tillage in Iowa. J ournal of Soil and Water Conservation 38(3), 78 91. Kollner, T. (2000). Specie s pool effect potentials (SPEP) as a yardstick to evaluate land use impacts on biodiversity. Journal of Cleaner Production 8, 293 311. Kruskal, W. H. (1957). Historical notes on the Wilcoxon unpaired two sample test. Journal of the American Statistical A ssociation, 52 (279), 356 360. Lambur, M., Whalon, M. & Fear, F. (1985). Diffusion theory and integrated pest management: illustrations from the Michigan Fruit IPM Program. Bulletin of the Entomological Society of America 31(3), 40 45. LaRose, R. & Atkin, D. (1992). Audiotext and the re invention of the telephone as a mass medium. Journalism Quarterly 69(2), 413 421. Le Feon, V., Schermann Legionnet, A., Delettre, Y., Aviron, S., Billeter, R., Bugter, R., Hendrickx, F. & Burel, F. (2010). Intensification of agriculture, landscape composition and wild bee communities: A large scale study in four European countries. Agriculture, Ecosystems, and Environment 137(1/2), 143 150. Lin, C.A. (1998). Exploring personal computer adoption dynamics. Journal of Broadc asting & Electronic Media 42(1), 95 112. Link, J., Graeff, S., Batchelor, W.D. & Claupein, W. (2006). Evaluating the economic and environmental impact of environmental compensation payment policy under uniform and variable rate nitrogen management. Agricu ltural Systems 91(2), 135 153. Mahler, A. & Rogers, E.M. (1999). The diffusion of interactive communication innovations and the critical mass: the adoption of telecommunications services by German banks. Telecommunications Policy 23, 719 740. Mann, H. & Whitney, D. (1947). On a test of whether one of two random variables is stochastically larger than the other. Annals of Mathematical Statistics 18, 50 60. Maxwell, R. & Kollet, S. (2008). Interdependence of groundwater dynamics and land energy feedbacks under climate change. Nature Geoscience 1, 665 669. McCune, N.M., Hernandez, F. G., Gonzalez, Y.R., Alcantara, E.A., Martinez, F.O., questions, local answers: Soil management a nd sustainable intensification in diverse socioeconomic contexts of Cuba. Journal of Sustainable Agriculture 35(6), 650 670.
109 McDonald, D. & Glynn, C. (1994). Difficulties in measuring adoption of apple IPM: a case study. Agriculture, Ecosystems, and Envir onment. 48, 219 230. McNamara, K., Wetzstein, M. & Douce, G. (1991). Factors affecting peanut producer adoption of integrated pest management. Review of Agricultural Economics 13(1), 129 139. Moore, G. C. and Benbasat, I. (1991). Development of an instru ment to measure the perceptions of adopting an information technology innovation. Information Systems Research, 2( 3): 192 222. Mylavarapu, R.S. (2010). Diagnostic nutrient testing. Horttechnology 20, 19 22. Mytinger, R. (1968). Innovation in local health services: A study of the adoption of new programs by local health departments with particular reference to newer medical care activities. Arlington: U.S. Dept. of Health, Education, and Welfare, Division of Medical Care Administration. Napier, T.L. & Camb oni, S.M. (1993). Use of conventional and conservation practices among farmers in the Scioto River basin of Ohio. Journal of Soil and Water Conservation 48, 231 237. Napier, T.L., Thraen, Gore, A. & Gore, W. (1984). Factors affecting the adoption of conv entional and conservation tillage practices in Ohio. Journal of Soil and Water Conservation. 39, 205 208. Napit, K., Norton, G.W., Kazmierczak, RF.. & Rajotte, E.G. (1988). Economic impacts of extension integrated pest management programs in several state s. Journal of Economic Entomology 81(1), 251 256. National Agricultural Statistics Service. (2006). Average farm size by state, 2006. Charts and Maps, United States Department of Agriculture. Accessed Online July 4, 2012 at http://www.nass.usda.gov/Charts_and_Maps/Farms_and_Land_in_Farms/fncht6. asp National Immunization Program. (2003). Immunization coverage in the U.S. Results from National Immunization Coverage in the U.S.: Results from National Immunization Survey. Centers for Disease Control and Prevention. [Data File] Retrieved from http://www.cdc.gov/vaccines/stats surv/imz covera ge.htm#nis National Institute of Food and Agriculture. (2012). Beginning farmer and rancher program. Accessed Online July 4, 2012 at http :// www nifa usda gov / fo / beginningfarmerandrancher cfm Nowak, P. (1984). Adoption and diffusion of soil and water conservation practices. In B. English, j. Maetzold, B. Holding and E. Heady (Ed. ), F uture Agricultural
110 Technology and Resource Conservation (pp. 214 237). Ames, Iowa: Iowa State University Press. Nowak, P. (1987). The adoption of agricultural conservation technologies: economic and diffusion explanations. Rural Sociology 52(2), 208 220. Nowak, P. (1992). Why farmers adopt production technology. Journal of Soil and Water Conservation 47(1), 14 16. Nunnaly, J. (1978). Psychometric theory New York: McGraw Hill. Palm, C. A., Machado, P.L., Mahmood, T., Melillo, J., Murrell, S.T., Ny Galloway, J. N. (2004). Societal needs for addressing nitrogen fertilizer needs: Balancing food production and environmental concern. In A. Mosier and J. Syers (Ed.), Agriculture and the Nitrogen Cycle: Assessing the Impacts of Fertilizer U se. (pp. 71 92). Washington D.C.: Island Press. Pampel, F. & van Es, J. (1977). Environmental quality and issues of adoption research. Rural Sociology 42(1), 57 71. Pereau, J.C., Doyen, L., Little, L.R. & Thebaud, O. The triple bottom line: Meeting ecolo gical, economic and social goals with individual transferable quotas. Journal of Environmental Economics and Management 63(3), 419 434. Pereira, L.S., Oweis, T. & Zairi, A. (2002). Irrigation management under water scarcity. Agricultural Water Management 57(3), 175 206. Pfister, S., Bayer, P., Koehler, A. & Hellweg, S. (2011). Projected water consumption in future global agriculture: Scenarios and related impacts. Science of the Total Environment. 409(20), 4206 4216. Rajotte, E., Kazmierczak, R., Norton, G., Lambur, M. & Allen, W. (1987). The national evaluation of extension integrated pest management (IPM) programs. Virginia Cooperative Extension Services/USDA VCES Publication 491 010. Blacksburg, VA, USA. Ramankutty, N., Evan, A.T., Monfreda, C. & Foley J.A. (2008). Farming the planet: Geographic distribution of global agricultural lands in the year 2000. Global Biogeochemical Cycles 22, GB1003 1 19. Ramankutty, N., Hertelb, T. & Leeb, H. (2004). Global land use and land cover data for integrated asses sment modeling. Global Trade Analysis Project. Purdue, Indiana. Accessed Online June 22, 2012 at https :// www gtap agecon purdue edu / resources / download /1990. pdf Ridgley, A.M. & Brush, S.B. (1992). Social factors and selective technology adoption: the case of Integrated Pest Management. Human Organization 51, 367 378.
111 Rogers, E.M. (1962). Diffusion o f innovations (1 ed.). Glencoe: Free Press. Rogers, E.M. (1995). Diffusion of Innovations. (4th ed.). New York: The Free Press. Rudel, T.K., Schneider, L., Uriarte, M., Turner, B., DeFries, R., Lawrence, D., Geoghegan, J., ... & Clark, W. (2009). Agricul tural intensification and changes in cultivated areas, 1970 2005. Proceedings of the National Academy of Sciences of the United States of America. 106(49), 20675 20680. Ryan, R. & Gross, N. (1943). The diffusion of hybrid seed corn in two Iowa communitie s. Rural Sociology 8(1), 15 24. Sahadev, S. & Islam, N. (2005). Why hotels adopt ICTs: a study on the ICT adoption propensity of hotels in Thailand. I nternational Journal of Contemporary Hospitality Management. 17(5), 391 401. Samiee, A., Rezvanfar, A. & Faham, E. (2009). Factors influencing the adoption of integrated pest management (IPM) by wheat growers in Varamin County, Iran. African Journal of Agricultural Research 4(5), 491 497. iability of scales. Journal of Extension, 37 (2). Severin, W. J., & Tankard, J. W. Jr. (1992). Communication Theories: Origins, Methods, and Uses in the Mass Media (3rd ed.), White Plains, NY: Longman. Shapiro, S. & Wilk, M. (1965). An analysis of variance test for normality (complete samples). Biometricka 52, 591 611. Sheskin, D. (2004). Handbook of parametric and nonparametric statistical procedures (3 rd ed.), New York: Chapman & Hall/CRC. Shiklomanov, I.A. & Rodda, J.C. (2003). World Water Resources at the Beginning of the 21st Century Cambridge University Press, Cambridge. 435 pp. Smith P, & Martino, D. (2007). Climate Change 2007: Mitigation Contribution of Working Group III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Ch ange: Agriculture. eds. Metz B, Davidson O, Bosch P, Dave R, Meyer L. Cambridge University Press, Cambridge, UK, pp 497 540. Smith, W. (2004). E v Rogers: Helping to Build a Modern Synthesis of Social Change. Journal of Health Communication 9(0), 139 142. Stephenson, G. (2003). The somewhat flawed theoretical foundation of the extension service. Journal of Extension 41(4). Strang, D. & Soule, S.A. (1998). Diffusion in organizations and social movements: From hybrid corn to poison pills. Annual Review of So ciology. 24(1), 265 290.
112 Sustainable Agriculture Research and Education. (1997). Exploring Sustainability in Agriculture: Ways to Enhance Profits, Protect the Environment and Improve Quality of Life. Accessed Online July 2, 2012 at http :// www sare org / publications / exploring htm Swisher, M.E. & Bastidas, E.P. (1994). Size of grower operation as a predictor of adoption of selected production practices. Proceedings of the Florida State Horticultural Society 107, 142 147. Tarde, G. (1903). The Laws of Imitation Translated by Parsons, E. Henry Holt and Company, New York. Taylor, D. & Miller, W. (1978). The adoption process and environmental innovations: a case study of a government project. Rural Sociology 4 3(4), 634 648. Tilman, D., Balzer, C., Hill, J. & Befort, B. (2011). Global food demand and the sustainable intensification of agriculture. Proceedings of the National Academy of Sciences of the United States of America. 108(50), 20260 20264. United Natio ns Framework Convention on Climate Change. (1998). Kyoto protocol to the United Nations Framework Convention on Climate Change. Accessed Online January 19, 2013 at http://unfccc.int/essential_background/kyoto_protocol/background/items/1351.ph p United States Congress. (1990). Food, Agriculture, Conservation, and Trade Act of 1990 (FACTA), Public Law 101 624, Title XVI, Subtitle A, Section 1603. Government Printi ng Office, Washington, DC, 1990. United States Department of Agriculture. (2007a). Sustainable agriculture: Definitions and terms. Alternative Farming Information Center. Compiled by Mary Gold. Accessed Online July 2, 2012 at http :// www nal usda gov / afsic / pubs / terms / srb 9902. shtml # toc 2. United States Department of Agriculture. (2007b). 2007 Census of Agriculture Women Farmers [Data File]. Retrieved from http://www.agcensus.usda.gov/Publications/2007/Online_Highlights/Fact_Sheets /Demographics/women.pdf United States Department of Agriculture. (2007c). 2007 Census of Agriculture Selected Operator Characteristics for Principal, Se cond, and Third Operator. [Data File]. Retrieved from http://www.agcensus.usda.gov/Publications/2007/Full_Report/Volume_1,_Chapt er_1_US/st99 _1_049_049.pdf United States Department of Agriculture. (2007d). 2007 Census of Agriculture Demographics. [Data File]. Retrieved from
113 http://www.ag census.usda.gov/Publications/2007/Online_Highlights/Fact_Sheets /Demographics/ United States Department of Agriculture. (2007e). 2007 Census of Agriculture Race, Ethnicity, and Gender Profiles. [Data File]. Retrieved from http://www.agcensus.usda.gov/Publications/2007/Online_Highlights/Race,_Ethni city_and_Gender_Profiles/reg99000.pdf United States Department of Agriculture. (20 09). 2007 Census of Agriculture: 2009 Census of Horticultural Specialties. [Data File]. Retrieved from http://www.agcensus.usda.gov/Publ ications/2007/Online_Highlights/Fact_Sheets /Production/horticulture.pdf United States Environmental Protection Agency. (2006). Global Anthropogenic Non CO2 Greenhouse Gas Emissions, 1990 2020. EPA Report 430 R 06 003 (U.S. Environmental Protection Agenc y, Washington, DC). United States Geologic Survey. (2011). Pesticide national synthesis project. National Water Quality Assessment Program. Accessed Online July 2, 2012 at http://water.usgs.gov/nawqa/pnsp/. Vacha Haase, T., Henson, R. & Caruso, J. (2002). Reliability generalization: Moving toward improved understanding and use of score reliability. Educational and Psychological Measurement 62(4), 562 569. Viswanathan, M. (2005). Measurement Error and Research Design Sage Publications, Thousand Oaks, Calif ornia. Vitousek, P., Aber, J., Howarth, R., Likens, G., Matson, Schindler, D., Schlesinger, W. & Tilman, D. (1997). Human alteration of the global nitrogen cycle: Sources and consequences. Ecological Applications 7, 737 750. Waller, B.E., Hoy, C., Henders on, J., Stinner, B. & Welty, C. (1998) Matching innovations with potential users, a case study of potato IPM practices. Agriculture, Ecosystems and Environment 70(3), 203 215. Waugh, R.K., Hildebrand, P.E. & Andrew, C.O. (1989). Farming systems research a nd extension. Research and Development (pp. 207 224). Boulder, CO: Lynne Rienner Publishers. Wikman, A. (2006). Reliability, validity and true values in surveys. Social Indicators Researc h. 78(1), 85 110. Willis, G.B. (2005) Cognitive interviewing: A tool for improving questionnaire design. Sage, Thousand Oaks, pp. 273 298.
114 management practices: A ca se study in the North China Plain. Journal of Environmental Management. 79(4), 409 419.
115 B IOGRAPHICAL SKETCH Corey Hanlon was born in Colorado Springs, but raised in Gainesville, Florida. She graduated from the International Baccalaureate Program at Eastsi de High School in 2007. She attended the University of Florida as an undergraduate student and received a B.S in Environmental Science and a B.S. in Biology in May 2011. She continued at the University of Florida pursuing a Master of Science degree in Int erdisciplinary Ecology. After completing her graduate education, she will pursue a career in resource management.