Determinants of Households’ Adoption of Organic Pesticides for Lawns and Gardens

This study investigates organic pesticide adoption in residential lawn care management, incorporating determinants based on a synthesis of multiple literatures. A mail survey using the Dillman method was conducted in Missouri in 2014 with a random sample of 2000 single-family households (44.1% effective response rate). The data indicate an adoption rate of 17.7% for organic pesticides. This unique dataset also allows us to differentiate distinct non-adopters by familiarity with the practice as well as non-use of any pesticides. Multinomial logit regressions find environmental concerns, awareness of neighbor’s opinions, and gardening behaviors as significant determinants. The effects on relative probability of being an adopter are large: 18 times more likely for people with serious environmental concerns or 5 times more likely for those spending more than 15 hours per month on lawn care.


Introduction
Lawns and gardens are popular in the United States; they are a source of enjoyment, a hobby and a source of home value. Homeowners in the U.S. spent about $47.8 billion on lawn and garden purchases in 2016, with a record average expenditure of $503 per household [1]. Pesticides are an important component; the home and garden sector accounted for about 6% of total U.S. pesticide usage in 2012, valued at about 24% of conventional pesticide sales, compared to 66% by agriculture [2]. The highest expenditure was for insecticides, approximately 80%

Factors Affecting Household's Adoption of Organic Pesticides
The literature on residential adoption of organic pesticides used in lawn care is complex. Households buy organic pesticides to use as inputs in a home production activity like lawn care and then enjoy their lawn on a regular basis. The adoption of inputs like organic pesticides in home production can be related to organic agriculture or IMP production practices while the purchase of products like organic pesticides by end users may be related to organic consumption perspectives. Hence, the literature herein covers studies of both organic consumption and organic or IPM practices in agriculture. Most studies of household organic purchases examine consumers of food [10] [11] [12], or specific kinds of food like fresh vegetables or fruits (e.g. [13] [14] [15]), and drinks [16], but only few cases examine products for other purposes [17]. On the other hand, studying determinants of adoption of pesticide best management practices, work has primarily focused on farmers or agricultural producers rather than households [8] [ 18]. For a comprehensive literature review of organic pesticide adoption in lawn tential determinants.
In empirical models, some socio-demographic characteristics have impacts on adoption. In the literature on organic purchases, for example, women may be more likely to buy organic products, because they express more concern for communal goals than men [25] or they want to protect their children from harmful effects of chemical products [20]. In the context of organic production, women are also likely to be organic producers for the same reasons [26].
Household income level consistently shows a positive effect on adoption of both organic purchases [11] [17] [27] and BMPs such as organic agriculture and IPM [8] [28] because this helps adopters overcome financial constraints related to organic products/inputs. We leave cost issues of organic pesticides to a discussion of price consciousness at the end of this review.
There are not such clear effects for household size, age, and education. Larger households may be less likely to purchase organic products [17], but it does not affect adoption of organic production. The number of children may have competing effects; it is highly correlated with household size, potentially reducing organic consumption, however having children may increase adoption due to health and safety concerns (as discussed below). Young people are more likely to buy organic food than older people [29] [30] and they might be more likely to adopt organic farming methods than older farmers [26], but the relationship between age and probability of BMP adoption is sometimes insignificant [31].
Similarly, higher educational level of homeowners tends to result in pesticide-free purchases [11] [27] [32] or use of organic practices [33], but the relationship is insignificant in other studies [26].
Environmental knowledge and attitudes usually are identified as critical drivers of organic adoption. Most empirical studies of organic consumption show positive effects on purchase intention of knowledge about organic products, awareness of threats from using synthetic pesticides, concerns about current quality of soil and water, or general attitudes of environmental protection (e.g. [11] [34] [35] [36]). However, the approach based on intended purchase is problematic because of the gap between behavior and attitudes toward organic products [37] or potential biases relating to environmentally friendly purchasing behavior [38]. A few studies show that environmental knowledge and attitudes do not positively affect actual purchase of organic products [11], and consumers might overestimate their organic purchases [39].
In the literature on organic farming, environmental knowledge and attitudes also play an important role in the adoption of organic and IPM practices [8] [40] [41]. However, the impacts on a farmer's adoption differ from a consumer's purchase of pesticide-free or lower pesticide products in terms of sensitivity.  [42]. The effect of knowledge on farmers' environmental behavior may also differ based on personal beliefs and emotions [43] lack of trust in information sources [44], and uncertainty and complexity [45]. Extending the concept of organic agriculture to lawn care management, households with more knowledge of lawn care are more likely to adopt BMPs [31].
In the literature, individually perceived benefits and costs associated with the innovation are the main factors affecting the adoption decision. While environmental knowledge and attitudes express public or general benefits, personal health and safety represent individual benefits. Organic products often are perceived as healthier or safer than conventional ones [35] [36] which leads to higher adoption of organic products to avoid health risks [46] [47]. This is related to why women, especially those with children, often show preferences for organic products, as noted in the demographic section. Many studies show that organic agriculture and IPM techniques help farmers reduce or avoid threats of chronic poisoning or cancer due to exposure to chemicals [48].
In the context of lawn care, health issues become important since household members experience a lawn treated with pesticides [3] [49]. Children are highly exposed to these chemicals because they often play in yards and pesticide residue may touch their skin. Hence, households who have children tend to adopt organic and natural pesticides rather than conventional ones [6] [50]. However, the number of children may negatively affect organic pesticide adoption as discussed above due to a possible interaction effect with average income, which implies more research is needed on this relationship [11] [51].
In addition to safety, expected quality of products can affect adoption. Past studies on organic purchase and organic production often view organic products as having better flavor [52] higher vitamins [36] as well as lower pesticide residues [53] which may lead to consumers buying organic rather than conventional products [54] [55] [56].
Regarding lawns, the concept of quality may include property value and aesthetic value (lawn appearance). In general, households who expect that organic lawn management may improve their property values are more likely to adopt this practice [28]. On the other hand, if the household prioritizes the appearance of the lawn, they might be more likely to purchase chemical pesticides, or hire commercial firms to effectively treat a problem such as grubs [6]. Individual preference that relates to enjoyment of yard work or gardening may affect the adoption of organic lawn management. For example, people who are more interested in gardening and able to do yard work tend to adopt alternative lawn care practices [22] because they may be more aware of, and gain more utility from using BMPs.
While safety and high quality or enjoyment of an organic lawn delivers benefits to households, the cost of organic products or methods presents a barrier to their adoption. Consumers often perceive organic products as expensive compared to conventional ones, which implies price negatively affects organic purchases [35] [36]. For households, commercial lawn management is the most expensive, followed by homeowner use of synthetic pesticides, and IPM or organic lawn management, which may include homemade organic pesticides [6]. Using no products obviously has no out-of-pocket costs. However, cost needs to be considered along with the associated quality or outcome. In the context of agriculture, the total cost of IPM or organic practices can potentially be lower, although farmers may substitute additional labor. Availability of cheap labor relative to pesticide cost may positively affect adoption of organic agriculture [59].
Besides cost issues, convenience has been considered a barrier to organic adoption. The concept of convenience refers to the amount of time involved with new practices. Studies found that consumers did not switch to organic food products due to convenience reasons like availability of organic products [35] or search time [60]. However, the implementation of USDA's national organic labeling standards in 2002 has helped to address these issues by providing quality assurance for organic foods [61]. Additionally, since 2006, the "Walmart effect" on the organic product market alleviated convenience issues [12] [62]. Regarding organic production, convenience may be related to availability of labor [59] or other inputs. Lawn care and pesticide use can be analyzed in the household production-consumption model of Becker (1965) [63]. In this model, the indirect utility of the household decreases with chemical pesticide use, but increases with leisure hours. Hence, time scarcity may negatively affect the adoption of organic lawn products.
Organic pesticides are potential replacements for conventional ones for residential use because of their safer yet ready-to-use characteristics. However, the household adoption rate of organic pesticides is currently low. By merging the literatures of organic consumption, organic farming as well as IPM practices, and lawn care management practices, this paper extracts hypotheses for poten-

Conceptual Framework
The study employs a discrete choice framework to examine determinants of or-  Discrete choice models can be derived from a utility maximization approach.
Depending on whether use or non-use of the organic pesticide delivers greater utility (indirect utility value), households decide to be adopters or non-adopters.
This paper uses a random utility-based discrete choice model (RUM), which includes stochastic or latent attributes of alternatives and individual characteristics [66], and has several advantages. First, the model can cover some components of preferences, which are unobservable to the researcher [67] [68]. Second, the RUM has a firm foundation in economics [69]. Another advantage is that the derived logit model is easily tractable with good empirical performance [66] [70].
The indirect utility of a typical RUM can be approximated by an appropriate linear function of observed characteristics of the alternatives, the individual, and the economic environment, and random error representing unobserved factors [71]. For multiple discrete choices, we have a set of random utilities associated with corresponding alternative choices for an individual. The derived logit model, also known as a multinomial logit model (MNL), is widely used to predict the probability or likelihood of the choice that yields the greatest utility among alternatives given a set of observed characteristics hypothesized to affect perceived utility. Specifically, MNL models require that random errors in the RUM are independently distributed and follow type 1 extreme value or log-Weibull distribution for a closed form likelihood expression of the integration [71]. Another important assumption of MNL models is the axiom of independence from irrelevant alternatives (IIA) [67]. The IIA assumption is to model individual choice probabilities and this can be tested by the Hausman-McFadden test or the Small-Hsiao test. Maximum likelihood estimation (MLE) is considered to be the most common method to estimate MNL models [66]. Estimation of the MNL model provides estimates of parameters or weights of observed characteristics used in the RUM [71]. Based on the results of the model, specific hypotheses extracted from the literature review are tested.
The main interest of this study is to examine the effects of environmental knowledge and attitudes, importance of neighbors, health concerns, and lawn care behaviors on a household's organic pesticide adoption decision. We hypothesize that pro-environmental attitudes positively affect adoption. Specifically, households who care more about the environment are more likely to adopt organic pesticides than conventional ones. This effect may be less for residents not applying any pesticides to their lawns. We also expect that households with more knowledge about the environment, and who are more familiar with organic pesticides will be more likely to be adopters. However, the importance given to the opinions of neighbors may negatively affect adoption; if it is important to them that their neighbors think they have a nice lawn they will prioritize aesthetic values over environmental ones.
Based on the previous literature, we assume that families with young children have more health concerns regarding the safety of pesticides applied to their Gardening behavior is expected to influence households' adoption choices because this relates to personal preferences. In this study, dimensions of lawn care behavior are conceptualized by time spent on lawn care and use of professional services for pest control. People often spend more time on yardwork if they enjoy gardening, and thus are expected to be more likely to be aware of and adopt environmentally friendly practices. On the other hand, households who use professional services for pest control will be less likely to adopt organic pesticides.

Empirical Model
We build a multinomial logit model to test our hypotheses of organic pesticide The MNL model uses variables constructed from the survey results. Assuming revealed preferences of decision makers for conservation practices [72], the dependent variables related to the respondent's knowledge/experience of using organic pesticides. The choices included "Not applicable", "Never heard of it and not using it" (Never heard), "Somewhat familiar with it but not using it" (Know somewhat), "Know how to use it but not using it" (Know well), and "Currently use it" (Adopter). We do not eliminate "Not applicable" because this choice accounts for a substantial number of observations, 184 respondents, and it may imply various latent possibilities for not adopting organic pesticides, including: not using any pesticides, no availability of necessary organic pesticides, using professional pest control companies, etc. Of these 184 "Not applicable" cases, there are 61 residents who did not use professional services for pest control, never apply fertilizers to their lawns, and said "not applicable" for reading and following pesticide application instructions for their yards or gardens. Given these responses, we assume that these 61 households did not use any pesticides at all for their lawns. Thus, we put these non-adopters into a separate group called "Not using any pesticides" to distinguish them from other "Not applicable" non-adopters. In this way, the dataset enables us to differentiate five distinct "not a problem", "slight problem", "moderate problem", "severe problem", and "don't know," accounting for 19, 130, 276, 221, and 109 cases, respectively. We combine the first two categories into "not or slight problem" and use it as the base category. Agreement with the statement "It is important that my neighbors think I have a nice lawn" was assessed using a five-point Likert scale, from "strongly disagree" to "strongly agree". There were 61, 159, 251, 257, and 24 respondents for the respective categories. We merge "strongly disagree" and "disagree" into a disagree category, and similarly form an agree category, while keeping "neutral" answers separate. To get at health concerns of pesticide use on young children we created a dummy variable for whether the household has at least one child under age 12.
Time spent gardening, one of two aspects of gardening behavior, was measured by four time-interval categories: "0 -5 hours", "6 -10 hours" (the base category), "11 -15 hours", and "more than 15 hours." We keep this variable as described in the survey to represent household's monthly hours spent on gardening and lawn care during the growing season, which may indicate preferences for gardening as a hobby. We also examined the professional lawn care services used by the household; the options included: "No", "Yes, just for mowing", "Yes, for mowing and fertilizing", "Yes, just for fertilizing and pest control", and "Yes, for mowing, fertilizing, and pest control" accounting for 522, 50, 36, 105, and 40 cases, respectively. We constructed a dummy variable by combining the first three responses into "not using pest control services," while the last two responses became "using a pest control service." To increase the explanatory power of the model, we added specific attributes of the background environment and demographic variables [71] [73]. The need for effective pest control is proxied by weed density in the lawn; we expect low weed density to be associated with organic pesticide adoption and for high weed density to be associated with non-adoption. Alternatively, this variable may represent previous use of effective synthetic pesticides in which case the effect on adoption would be opposite. In the dataset, the weed density measured by the average number of weeds per square yard is represented by five options: "None", "1 to 10", "11 to 40", "More than 40" weeds, and "Don't know" resulting in 50, 259, 151, 37, and 254 cases, respectively. We combine the categories of "None" and "1 to 10" into "Less than 10" and use that as the base category. "Some formal schooling", "High school diploma or GED", "Some college or 2 year college degree", "4 year college degree", and "post-graduate degree" with 8, 72, 154, 254, and 265 cases, respectively. We combined the first two categories into a "High school or less" base category. Annual household income included five options: "less than $24,999", "$25,000 -$49,999" (the base), "$50,000 -$74,999", "$75,000 -$99,999", and "$100,000 or more." To sum up, the empirical MNL model is specified as follows: The model takes a simple form of Luce model (e.g. [67] [71] to predict the probability or the likelihood that choice "y i " made by household "i" occurs over the sample of size "n." "k" represents alternative organic pesticide adoption choices: "adopt", and five non-adopter alternatives: "never heard", "know somewhat", "know well", "not use any pesticides", and "other N/A." We can thus classify households as adopters and distinct types of non-adopters. Since we distinguish among various types of non-adopters, we use "adopters" as the base of the model, which differs from traditional adoption studies.
In the model, "x" is a vector of explanatory variables from household "i" res-

Data Description
Summary statistics for the data used for the regression model can be found in  The summary statistics in Table 2 show that adopters look different from the sample as a whole in seriousness of environmental problems regarding pesticides, neighbor opinions of their lawn being important, and time spent gardening. There are also several noticeable differences across distinct types of non-adopters. For example, "severe" is the most common response regarding pesticide problems for adopters and for those who use no pesticides, 40%, compared to 29% for the whole sample. Compared to the whole sample (38%), a lower percentage of adopters (30%) disagree that neighbors' opinions are important, but this difference masks the much lower level of disagreement among those not using any pesticides (10%) and the much higher disagreement (over 44%) of non-adopter categories know well and other N/A. This data highlights the importance of distinguishing among types of non-adopters and separating those who use no pesticides from other responses. About 40% of adopters spent more than 15 hours per month gardening while the most common response overall was 6 -10 hours. The most common response for those who don't use pesticides was 0 -5 hours, which was the second most common for never heard and other N/A categories (26% and 30%). The weed question is also interesting; 28% of adopters didn't know how many weeds they had, but for those who had never heard of organic pesticides and those who don't use pesticides, over 45% didn't know. This may indicate that these non-adopters are more generally not interested in gardening and lawn care.
Demographic statistics indicate that adopters are somewhat less likely to be male than all non-adopter categories other than those who use no pesticides.
They are also less likely to have young children. Adopters are older than the sample as a whole, especially compared to those who use no pesticides. Educational levels for adopters and those using no pesticides are fairly similar, but those who use no pesticides are much more likely to be in the lower income categories than adopters. Given the population for this dataset (those living in homes with yards), respondents being older, richer, and more educated than the relevant Census data is reasonable [73]. In Table 2

Results and Discussion
We begin this section by discussing the way the regression results are reported and interpreted. For our empirical MNL model where adopters are the base and distinct types of non-adopters represent alternatives, there are separate outcomes for each type of non-adopter category as if we independently estimated logistic regressions for each group of non-adopters compared to adopters. In other words, the model results include estimated coefficients, significance levels, and goodness-of-fit statistics for non-adopter types of "never heard", "know somewhat", "know well", "not use any pesticides", and "other N/A", compared to the base of adopters. Since all explanatory variables are categorical, a coefficient is interpreted as the difference in the logit or log-odds of being a specific type of non-adopter rather than an adopter, due to the effect of that response category versus the base category, all else equal. The exponential of an estimated coefficient can be interpreted as how many times more/less likely it is to be a specific non-adopter type versus an adopter of organic pesticide practices. A positive coefficient means the odds are greater than 1: the respondent is more likely to be a non-adopter than adopter if they chose that response category rather than the base category for that variable. A negative coefficient implies the respondent is less likely to be a non-adopter than an adopter based on that variable. Again, this differs from typical adoption studies; here, negative coefficients imply that variable positively affects adoption.
Regression results of the preferred model are reported in Table 3. In terms of it is acceptable relative to other empirical studies [73]. Moreover, the likelihood ratio test for the model is highly significant, α-level = 1% confirming that the model significantly predicts the likelihood of organic pesticide non-adoption.
From Table 3, we can see important variables in the model. Only statistically significant results are discussed. As expected, perceived seriousness of environmental problems related to pesticides affects adoption of organic pesticides. The estimated coefficients are negative, implying those who indicated "moderate problem" rather than the base "not or slight problem" are less likely to be non-adopters, or in other words, they tend to be adopters. If they indicated it is a "serious problem" they are less likely to be a non-adopter, except those who use no pesticides, and the effects are generally larger. For example, the estimate for  "never heard" is the largest, at 2.7, followed by "other N/A" at 2.0. An alternative interpretation using these exponents is that if a respondent indicates that pesticides are a serious problem (versus the base), the probability of being non-adopter type "never heard" is 15 times less likely than being an adopter, and being "other N/A" is 7 times less likely. The effects are smaller for "know somewhat" and even smaller and less significant for "know well". These results hint at the role of environmental knowledge, in addition to concern, that is highlighted in the organic literature. Households who are more familiar with organic pesticides could more easily transition to being adopters. The fact that those who indicate it is a serious problem are not significantly less likely to not use pesticides than to adopt organic ones implies that at least some people do not use pesticides due to environmental concerns.
For the importance given to neighbors' perceptions of their lawn, if people disagree this is important (versus agree) they are less likely to be in the "know somewhat", "know well", or "other N/A" categories than to be adopters. In other words, adopters put less weight on the opinions of their neighbors than these categories. The lack of adoption by those who care what the neighbors think may relate to the perceived lower effectiveness of organic products, but examining the summary statistics, adopters do not have weedier lawns. The magnitude of the effect for the two non-adopter categories with some knowledge is similar; they are about two times (=e 0.8 ) less likely to be adopters. The effect of disagreeing with the statement is the highest for the "other N/A" category. Households who are not using any pesticides are distinct from other non-adopters since the coefficients are positive, and for the neutral response, significant. In other words, if they are neutral, they are more likely to not apply pesticides than to be adopters.
This may indicate that they don't care about either their lawn or their neighbors.
The effect of health concerns represented by whether households have child- concern and the financial effect of having a larger family. The role of children regarding organic purchases is still not clear in the literature [51].
Gardening behaviors, defined by time spent gardening and use of pest control services, have different effects on adoption. Similar to the results regarding perceived seriousness of environmental problems, those indicating they spent "more than 15 hours" gardening versus the base category of "6 -10 hours" were significantly less likely to be in any of the non-adopter categories. In other words, serious gardeners are more likely to use organic pesticides, as expected from the literature. The magnitudes are large: the biggest effect is for "never heard," followed by "no pesticides", "other N/A", "know somewhat", and "know well". Thus, the probability of non-adoption relative to adoption would be from 3 -5 times less for gardeners spending more than 15 hours a month for lawn care. The probability is smallest for those with the most knowledge and largest for those with the least knowledge. Note that time can be also treated as a substitute input for pesticides in the model proposed by Templeton et al. (1999) [63].
Holding other factors constant, the other gardening behavior, use of pest control services, is not significant in the study.
There is only one type of non-adopter category for which weed density is significant. Households whose lawn has 10 -40 weeds per square yard (versus less than 10) or simply don't know are more likely to be non-adopters who don't use pesticides rather than being adopters. The results support the hypothesis that not using any pesticides (or not caring) may have allowed the proliferation of weeds, rather than weeds representing a need for effective herbicides/pesticides.
The estimates are approximately 0.95 implying the probability of not using any pesticides is 3 times more likely than adoption for those response categories. The effect of "greater than 40 weeds" is not significant which may be partly due to the low number of responses in this category.
Demographic variables do not seem to have much impact on organic pesticide adoption. While the signs for male versus female generally align with the literature, there are no significant effects, ceteris paribus. This result is similar to Shin and McCann [73] for the adoption of watershed conservation practices. Regarding age, compared to those over 60, those 18 -30 are more likely to not apply pesticides than to adopt organic ones. They may be new to having a home and yard and starting their careers. The middle-aged category, versus over 60, is more likely to have never heard of organic pesticides or to not apply any pesticides than to adopt. The probability of being these non-adopter types is about 3 times more likely than being an adopter. In other words, there is some evidence that those over 60 are more likely to adopt.
If a respondent had some college versus a high school diploma or less, they are  Table 3, the significant coefficient estimates are −1.09, −0.88, and −1.14 for "know somewhat", "know well", and We tested for multicollearity in two ways. We calculated variance inflation factors (VIFs) for all variables. Since all the VIFs are less than the standard cutoff of 10, there is no evidence of multicollinearity in the full model. Almost all correlation coefficients were smaller than 0.3 except for age and having children under 12 (0.6). Using a Chi-square test for independence, we find that these two variables are significantly correlated (p < 0.0001).
Based on the correlation between age and children, we ran two reduced models. The first model is defined by excluding the children under-12 variable while the second one leaves the age variable out of the set of explanatory variables. The estimation results of these models are reported in Appendix 1 and Appendix 2, respectively. The results are generally robust regarding signs and significance.
More specifically, the children under-12-variable is still not significant after dropping age. For both alternative models, the neutral response regarding neighbors became significant at the 10% level for the "know somewhat" type of non-respondent, and 0 -5 hours spent gardening became significant at the 10% level for the "other N/A" type. When the variable relating to young children was dropped, there were minor changes to the age variables, the 18 -30 category is Journal of Environmental Protection no longer significant at the 10% level for those using no pesticides, and the 46 -60 age category for "never heard" went from the 10% to 5% level of significance.
We consider McFadden's pseudo R 2 , Nagelkerke's pseudo R 2 , and information measures like the Akaike information criterion (AIC) and Bayesian information criterion (BIC) to assess which model is a better fit given the dataset. We find that the full model has the largest pseudo R-squared values, followed by the model excluding children and the one excluding age. The smallest AIC and BIC are for the model removing age, followed by the one removing children and the full model. We cannot improve AIC and BIC without reducing the pseudo R-squared values. For consistency with other studies, we prefer using pseudo R-squared to evaluate the models. The full model in Table 3 would be the best in this regard, offering more explanatory power than the other models.
Third, we compare the full model to one keeping all "not applicable" respondents lumped together, rather than separating out those who apparently use no pesticides. The results of this model are reported in Appendix 3. In general, the significant coefficients of the other non-adopter groups ("never heard", "know somewhat", and "know well") are the same in sign and magnitude.

Conclusions, Implications, and Limitations
The low adoption rate of residential lawn care BMPs such as organic pesticides means that further environmental improvement is possible if we can identify factors that could lead to improved management practices. In this paper, we examine different characteristics of adopters and non-adopters to identify important drivers of organic pesticide adoption. Given the paucity of research on organic practices in residential lawn care, possible determinants are extracted from multiple literatures. Using unique household data from Missouri, the study also distinguishes five distinct non-adopter groups for a deeper understanding of the factors affecting adoption. A standard multinomial logistic regression is employed to test significance of these determinants and explain household behavior regarding organic pesticides.
Overall, the estimation results support most of the hypotheses of the study. There are also implications that flow from some of our key results. First, there are two groups that are low-hanging fruit, people with pro-environmental attitudes and those whose hobby is gardening. The effects of these characteristics are so dominant that these people may be more likely to adopt organic pesticides and other residential BMPs, regardless of other factors. Environmental organizations that focus on unrelated issues, such as climate change, may be targeted for outreach regarding personal behaviors that affect water quality. People who spent more than 10 hours gardening were more likely to be adopters, and this effect was particularly large for those spending over 15 hours per month. This implies that gardening clubs, magazines, and websites may be a good way to reach people who may be predisposed to environmentally friendly practices that may not be particularly convenient. They may also gain utility from trying new gardening practices.
The fact that 24% of respondents indicate they know the practice well but have not adopted implies that awareness is not the only barrier. This may relate to perceived or actual problems with effectiveness and convenience of current organic pesticides compared to conventional ones or pesticide free solutions. While similar to other adoption studies, the explanatory power of the model is low. Including barriers related to the practice may be helpful; for example, physical limitations may limit the use of some practices by certain individuals. The heterogeneous characteristics of organic pesticides such as price and convenience are not covered in the study. The cost of home-made organic pesticides is often lower while that of commercial products is typically higher than conventional pesticides. The same holds for convenience; many organic pesticides are ready to use but some of them require preparation time or additional equipment. Including these characteristics as well as perceptions could increase the explanatory power of the model. Additional research to examine gardeners who are not using any pesticides and those who adopt environmentally friendly, but not organic, pesticide management practices may be interesting since the water quality impacts are different from other non-adopters. Notes: Superscripts *, ** and *** indicate statistical significance at the 10 percent, 5 percent and 1 percent levels, respectively.