Literature Review

Despite recent rapid growth, research on the economics of agritourism is limited. One vein of research examines what motivates producers to diversity their agricultural business into agritourism. Another vein of complementary research explores consumer behavior and experiences with agritourism, to better inform producers of appropriate agritourism marketing strategies. No research of which we are aware investigates the spatial drivers of agritourism, despite Cole (Reference Cole2007) calling for the use of regional science methods, such as exploratory spatial data analysis, to better address tourism and the impact of tourism promotion.

The relatively fast-paced growth of agritourism is noteworthy, but more interesting is why and how agritourism is developing differentially across the United States. This spatial variation may be partly due to the variation in farm and ranch operators’ motivation to participate in agritourism. Tew and Barbieri (Reference Tew and Barbieri2012) identify some of these motivations as “increase[ing] farm revenues, market opportunities and social bonding, opportunities to keep family together, and personal pursuits.” Additional motivations are local economic development, or positive spillovers from agritourism to the community.

Farmers and ranchers may pursue agritourism to diversify their business portfolio and make their enterprises more resilient. Increasingly, smaller farms and ranches must diversify their agricultural enterprises to stay viable in competitive crop and livestock commodity markets driven by economies of scale (Ilbery et al. Reference Ilbery, Bowler, Clark, Crockett and Shaw1998, Sharpley and Vass Reference Sharpley and Vass2006, Ollenburg and Buckley Reference Ollenburg and Buckley2007, Tew and Barbieri Reference Tew and Barbieri2012). Gains in productivity from large-scale agriculture and exposure to global market forces are widely cited reasons smaller scale (or beginning) farms have trouble staying in business. Thus, alternative business models that integrate direct sales, local markets, and agritourism are important to such farms (Veeck, Che, and Veeck Reference Veeck, Che and Veeck2006, Thilmany and Ahearn Reference Thilmany and Ahearn2013, Hardesty et al. Reference Hardesty, Feenstra, Visher, Lerman, Thilmany-McFadden, Bauman, Gillpatrick and Rainbolt2014).

In some areas of the United States, urban sprawl pressures have increased the costs of operating an agricultural business in or near more densely populated areas (Nickerson, Black, and McCool Reference Nickerson, Black and McCool2001, Veeck, Che, and Veeck Reference Veeck, Che and Veeck2006). Agritourism may take advantage of this urban proximity.

Economic motivators related to employment for family members and increased and/or more stabilized revenue are the most frequently noted agritourism goals by farm/ranch principal operators (Nickerson, Black, and McCool Reference Nickerson, Black and McCool2001, Barbieri and Mahoney Reference Barbieri and Mahoney2009, Barbieri Reference Barbieri2013). Other goals noted in the literature include personal interests, educating the public, and companionship with guests (McGehee and Kim Reference McGehee and Kim2004, Tew and Barbieri Reference Tew and Barbieri2012). Barbieri (Reference Barbieri2013) found that, relative to other farm diversification strategies, agritourism was the most successful at generating spillovers in line with social goals such as creating jobs for non-family members, preserving cultural and historical heritage, and adopting sustainable practices.

Perhaps most intriguing to economic development practitioners and researchers are the positive spillovers to surrounding areas attributed to agritourism. Research supports strong potential for agritourism sites to attract tourists from out of town, who then spend money at local establishments such as restaurants, gas stations, and hotels in the rural area (Veeck, Che, and Veeck Reference Veeck, Che and Veeck2006, Saxena et al. Reference Saxena, Clark, Oliver and Ilbery2007, Barbieri Reference Barbieri2009, Tew and Barbieri Reference Tew and Barbieri2012). Agritourism may broaden the local tax base, making communities more resilient in recessionary periods because agritourism is a relatively low-cost substitute for other types of travel (Veeck, Che, and Veeck Reference Veeck, Che and Veeck2006, Sullins, Moxon, and McFadden Reference Sullins, Moxon and McFadden2010). However, Brown et al. (Reference Brown, Goetz, Ahearn and Liang2014) find that agritourism sales do not significantly increase per capita income or per capita farm sales at a national level but may still have positive effects at regional levels. While many rural communities have recognized potential societal benefits from agritourism and have developed programs supporting agritourism, other regions have been slower to offer support (Veeck, Che, and Veeck Reference Veeck, Che and Veeck2006). There appear to be clear differences across regions, perhaps due to policy.

Farmers and ranchers incorporating agritourism into their businesses are often viewed as agricultural entrepreneurs, and regional variation in entrepreneurship rates has been well documented (Low, Henderson and Weiler Reference Low, Henderson and Weiler2005, Mack and Qian Reference Mack and Qian2016). Regional variation in farm entrepreneurship rates is also expected. In a nationally representative survey, 20 percent of producers participating in agritourism also participated in additional on-farm enterprises that some consider entrepreneurial, such as local food sales and production of value-added goods (Vogel Reference Vogel2012, Liang and Dunn Reference Liang and Dunn2014). The notion of agritourism operators as entrepreneurial is consistent with past work on tourism (Marcouiller Reference Marcouiller2007, Liang and Dunn Reference Liang and Dunn2014) that concludes agricultural entrepreneurs use business skills to differentiate their business as a means to mitigate the volatility of commodity markets (Nickerson, Black, and McCool Reference Nickerson, Black and McCool2001, Barbieri and Mahoney Reference Barbieri and Mahoney2009, Tew and Barbieri Reference Tew and Barbieri2012).

Despite agritourism's rapid growth, research on the spatial distribution of agritourism and its spatial drivers is limited. Bagi and Reeder (Reference Bagi and Reeder2012) found agritourism rates were highest in the Southern Plains and Inter-Mountain West using data from a nationally representative survey, but more spatially detailed estimates were impossible. Operator motivations are not the only factors associated with differences in agritourism rates across space; site characteristics, such as surrounding scenic beauty and proximity to urban areas, affect the success of an agritourism site (Veeck, Che, and Veeck Reference Veeck, Che and Veeck2006, Che Reference Che2007, Bagi and Reeder Reference Bagi and Reeder2012). Research shows that areas rich in natural resources are more likely to be conducive to recreation and tourism (Gartner Reference Gartner2005), but there is at least one contradiction to this theory in the literature. Hill et al. (Reference Hill, Loomis, Thilmany and Sullins2014), in their travel cost model, hypothesized high natural amenities would lead to more agritourism trips, but they found a negative and significant association with natural amenities. Hill et al. suggest the negative relationship could be due to the study being limited to Colorado, where topography plays a role in the location of agricultural operations. We hope that by studying the entire contiguous United States, this sort of limitation will be overcome.

A major contribution of this paper is to expand on initial attempts to add place-based factors into modeling agritourism using regional science methods. We believe these methods are appropriate, given the spatial heterogeneity in the agritourism sector and the presumed dependence of these operations on place-based factors (e.g., scenic highways, present in some places and absent in others, being part of rural tourism systems, according to Gartner (Reference Gartner2005)). We focus on regions with high shares of farms involved in agritourism, or agritourism hot spots, as a first attempt to understand the impacts of place-based factors on agritourism.

Methods and Data

To understand the determinants of agritourism clusters, we must first define the clusters and identify their location. For the former, we conducted a local indicators of spatial autocorrelation (LISA) analysis to identify the agritourism hot spots. LISA results were subsequently used as the limited dependent variable in a probit model to confirm whether or not factors reported as significant in previous business location and tourism literature are relevant in these data.Footnote ¹

The LISA analysis was first proposed by Anselin (Reference Anselin1995) as a way of decomposing a global indicator of spatial autocorrelation, e.g., Moran's I, into individual units. The purpose of this decomposition is to explore the spatial autocorrelation of each observation, and identify patterns or clusters of significant spatial autocorrelation across space (Anselin Reference Anselin1995). Using the results of the LISA analysis, one can produce choropleth maps that identify these areas of significantly high positive autocorrelation (hot spots) and areas of significantly low positive autocorrelation (cold spots). These maps provide the researcher with an intuitive view of the activity in question, allowing them to determine patterns and hypothesize which factors may influence these patterns over the space in question. Despite a call to marry the study of tourism with regional science methods (Cole Reference Cole2007), we do not know of any other research examining spatial autocorrelation of tourism or agritourism.

The LISA analysis serves as an exploratory study highlighting the importance of incorporating the dimensions of space into tourism industry assessments. In fact, the nature of our discrete dependent variable being an indicator of significant positive spatial autocorrelation from neighboring counties could perceivably lead to econometric issues, such as endogeneity arising from describing a significant spatial autocorrelation using a spatial auto regressive parameter, if the model were estimated with a spatial probit model.

For this analysis, a National Agricultural Statistics Service county shapefile was merged with 2012 Census of Agriculture data on the number of farms in each county with agritourism revenue (akin to the measures used in the business location literature).Footnote ² While not a perfect solution, we treat counties with undisclosed agritourism data as zeros in the LISA analysis. Treating counties with an undisclosed number of agritourism operations should not bias LISA hot spot results because these counties, by definition, have few agritourism farms and ranches. Dropping these undisclosed counties could bias the prevalence and location of cold spots (we likely underestimate agritourism participation in counties with undisclosed data by treating the share as zero).

A global indicator of spatial autocorrelation in the percentage of farms reporting agritourism income, Moran's I, was calculated first. Figure 1 is a scatter plot showing the dispersion of spatial autocorrelation across counties. A clear positive trend can be seen, as indicated by the Moran's I of 0.4505; its p-value is significantly below the 1 percent level supporting the choice to use our preferred methodological approach. The horizontal axis of the Moran's I scatter plot indicates the level of agritourism (measured as the percent of farms with agritourism) of a particular county i. The vertical axis indicates the level of agritourism in county i’s neighborhood. Observations in the first (top right) or third (bottom left) quadrants of the Moran's I scatterplot: (1) a high share of farms/ranches participating in agritourism and neighbors who also have a high share of participation; or, (2) a low share of agritourism operations surrounded by neighbors with similarly lower levels. Counties in the second and fourth quadrants exhibit negative autocorrelation and have higher levels of agritourism participation compared to their neighbors or have lower levels of agritourism participation than their neighbors have.Footnote ³

Figure 1. Moran's I Scatter Plot, County-Level Percent of Farms with Agritourism Revenue, 2012

The LISA analysis results are presented as a map in which counties exhibiting significant (p < 0.05) spatial autocorrelation are highlighted. When looking at this output in Figure 2, it is important to remember that these shaded areas are the “cores” of the spatial clusters. For example, counties bordering the hot spots most likely also have high levels of agritourism because they helped define the hot spot as having significant spatial autocorrelation. One can think of the shaded counties as the cores, with the autocorrelation spreading out and diminishing in intensity from these core areas.

Figure 2. Significant Clusters of Spatial Autocorrelation in Percent of Farms with Agritourism Revenue, 2012

The LISA analysis (Figure 2) shows interesting patterns among the percent of farms using agritourism in the United States but does not give any indication, other than regionally focused hypotheses derived from the literature, of what drives these clusters of positive autocorrelation. While the incidence of hot spots from the LISA will be explored in detail subsequently, we can use this map to drive some of our a priori expectations on spatial influences on where agritourism enterprises exist.

First, and perhaps most apparent, is the difference in agritourism across regions. The small number of hot spot counties limit our ability to analyze hotspots in separate regions of the country, so without having well-defined “agritourism zones” we adopt the four U.S. Census regions to capture some regional heterogeneity. These Census regions include the West, Midwest, South, and Northeast. The Midwest has the greatest number of cold spots and fewest hot spots in the contiguous United States, while the three other regions portray at least some significant clusters of hot spots. The map shows patterns of agritourism in areas that differ in climate, culture, and landscape (locational characteristics that are likely important for the tourism sector), indicating interaction terms might be conducive to explaining how heterogeneity among these explanatory variables influences hot spot patterns.

We use a probit regression to explore exogenous factors related to the incidence of a hot spot.Footnote ⁴ We examine the relationship between the dependent variable (agritourism hot spots or not) and a set of independent variables, X, and assume that the error terms are normally distributed. Equation 1 makes use of this assumption, showing that the probability that county i is a hot spot depends on the set of independent variables for that respective county. One important change in the interpretation of the coefficients is that they now represent changes in the z-score, so a transformation is made to retrieve marginal effects. Both z-score coefficients and marginal effects are included in the results.

(1)$$P_i \equiv Pr\lpar y_i = 1 \vert {\bi x}_{\bi i} \rpar = \; {\rm \Phi \lpar }x^{\prime}_i \beta {\rm \rpar }\quad {\rm where}\quad y_i = \left\{{\matrix{ {1\, {\rm if}\, {\rm county}\, i\, {\rm is}\, {\rm a}\, {\rm hotspot}} \cr {0\, {\rm if}\, {\rm otherwise}} \cr}} \right.$$

Twenty-one independent variables were regressed on agritourism hot spot counties in the probit model via maximum likelihood estimation. Summary statistics and data sources for all independent variables are in Table 1.

Table 1. Summary Statistics

The four regional dummy variables are South, West, Northeast, and Midwest, with the Midwest region serving as the reference group, due to the small number of hotspots in the region and relatively low levels of agritourism, making it a logical baseline for comparison. Regional dummy variables were expected to capture cultural, agricultural, land, natural resource, and climate characteristics specific to each region.

Because scenic highways are part of rural tourism systems (Gartner Reference Gartner2005), miles of designated scenic byways (national or state) and interstate highways are expected to increase the probability of the county being a hot spot due to greater access to travel infrastructure and past research finding the importance of this transportation access (Bhat et al. Reference Bhat, Paleti and Singh2014). In terms of past agritourism research, Sullins, Moxon, and McFadden (Reference Sullins, Moxon and McFadden2010) suggest, in a cluster analysis of Colorado agritourists, that many travelers treat agritourism sites as secondary or spontaneous visits on their way to some other primary destination. This indicates that counties with more miles of interstate highway and scenic byways may have more travelers passing through. This increase in the potential demand for agritourism may subsequently incentivize more farms and ranches to adopt the activity. Squared terms of these variables were included in the model to capture the hypothesized diminishing effect for each additional byway or interstate mile within a county.

Similarly, based on Sullins, Moxon, and McFadden (Reference Sullins, Moxon and McFadden2010), the coefficients for travel time to a national park and large city were expected to be negative. As mentioned above, more travelers passing through these counties should increase the probability of a spontaneous or secondary stop, further incentivizing local farmers and ranchers to take advantage of the road traffic and adopt an agritourism enterprise. Given this theory, we expect to see agritourism hot spots near other tourist attractions such as large cities and national parks. Using visual inspection of natural breaks in a histogram of visitor numbers to national parks and monuments in 2012, we established a visitor threshold of 40,000 visitors or more per year, because not all national designations lead to large visitation numbers (National Parks Service 2015).Footnote ⁵ To capture regional differences in national parks, we also include interaction terms between travel time to a national park and Census regions.

Although previous literature has not explored this notion, it seems intuitive that counties with higher rates of entrepreneurship and innovation are more likely to have agritourism businesses. On the supply side, farm operators in these areas may be influenced by their cohorts’ entrepreneurial interests and skills, and on the demand side, agritourism may be more attractive to these innovative citizens. In order to measure a county's entrepreneurial propensity, both the breadth of entrepreneurship (the nonfarm proprietorship rate) and patents per 10,000 people were included in the model. Breadth is measured as the number of nonfarm proprietors as a share of the total employment in the county. Because most farmers and ranch operators are proprietors, we used nonfarm proprietors to control for the entrepreneurial nature of the county because farm proprietorships would be highly correlated with the dominance of production agriculture in the county's economy. While patents provide a slightly different measure of the entrepreneurial spirit of a county, we interpret it as a measure of innovation output, whereas breadth is more a measure of the degree of entrepreneurial activity in a county.

Counties with higher per capita income were assumed to have a higher demand for agritourism, and thus per capita income was included and expected to have a positive result on the probability of county i being a hot spot. Counties with larger populations were also assumed to have a higher demand for agritourism, leading to a similar a priori expectation for the natural log of population in county i.

Lastly, to help capture differences in agricultural industry structure, interaction terms between average farm size and regions were included. According to a study in Montana by Nickerson, Black, and McCool (Reference Nickerson, Black and McCool2001), large farms may be more likely to adopt agritourism than small farms, in order to cover larger asset ownership, property tax, and managerial costs. Other studies have suggested small farmers may have a higher propensity to seek diversification options, however.

Regression Analysis Results

The total number of observations (counties) included in the regression analysis was 2,954, after dropping counties with an undisclosed number of agritourism farms and ranches.Footnote ⁶ Dropping undisclosed counties should not bias the model's parameter estimates because the regression analysis only focuses on hot spots, i.e., high concentrations of agritourism. Regression results are in Table 2.Footnote ⁷

Table 2. Probit Results on Agritourism Hot Spots

Observations = 2,954

Pseudo R² = 0.3069

LR Chi²(21)  =  399.24 p-value = 0.0001 (Standard Errors in parentheses).

*Significant at <10 percent, **Significant at <5 percent, ***Significant at <1 percent.

The only statistically significant (p < 0.01) region dummy in the regression results was the propensity to see a higher share of farms and ranches with agritourism in the Northeast, where a given county is 89 percent more likely to be a hot spot than a given county in the Midwest. The Northeast finding is consistent with Liang and Dunn (Reference Liang and Dunn2014) and Brown et al. (Reference Brown, Goetz, Ahearn and Liang2014), who note the higher propensity of agritourism and other local foods activities in that region. Referring to the location of the hot spots (Figure 2), it was surprising that coefficients on neither the South nor the West were positive and significant, but this may simply imply that the hot spots in these regions can be explained by the broader set of independent locational variables. These regional findings may be partially driven by a bias in the LISA analysis due to the relative size difference in counties between east coast and west coast states (the latter having larger counties).

Miles of scenic byway per hundred square miles and its squared term followed a priori expectations, whereas the analogous interstate variables did not. An additional mile of scenic byway for every 100 square miles in county i leads to a 0.28 percent increase in the probability that the county is an agritourism hot spot. Because the mean is just under 1.9 miles of scenic byway per 100 square miles, our result suggests that doubling the miles of scenic byways would lead to a 0.53 percent increase in the probability that the county is a hot spot, so the statistically significant effect (p < 0.05 percent) has a relatively small magnitude.Footnote ⁸

That the coefficient on miles of interstate per 100 sq. mi. was not significant may imply that agritourists are more likely to travel via scenic byways than interstates on their way to their primary destinations for leisure travel, whereas interstates are more conducive to use by travelers hoping to reach destinations in the shortest time possible. Additionally, interstates have few exits relative to the open entry/exit nature of most scenic byways, which may increase the perceived marginal travel cost of potential agritourists to exit and visit the site from an interstate when compared to a byway.

Neither of the entrepreneurial propensity variables, Breadth and Patents per 10 k People had coefficients significantly different from zero. Although these variables are not significant, it does not necessarily mean that the entrepreneurial nature of a county does not play a role in whether or not the county is a hot spot for agritourism.

The coefficient on Travel Time to City of ≥250 k People was only significant at the 10 percent level (p < 0.10), but positive, suggesting a more remote location is marginally associated with the presence of agritourism hot spots. Travel Time to National Park was only significant when interacted with the South and Northeast regions, but with unexpectedly opposite signs. In the Northeast region, a 1-h decrease in travel time to a national park or monument increased the probability of the county being a hot spot by 3.77 percent relative to a county in the Midwest. The South experienced an opposite effect; results suggest counties closer to national parks and monuments are less likely to be hot spots for agritourism in the South, compared to the Midwest. With respect to the Northeast result, urbanization may create travel opportunities from within-region visitors from urban areas.

Nationally, average farm size was unrelated to the probability of a county being a hot spot for agritourism. Compared to the Midwest, counties in the Northeast with a lower average farm size were more conducive to being a hot spot than counties with a larger average farm size, likely due to the type of agricultural enterprises and operators in the Northeast and urban proximity.

Similar to Bagi and Reeder's (Reference Bagi and Reeder2012) findings, the natural amenities scale coefficient is positive and significant (p < 0.01) indicating a positive relationship between natural amenities and agritourism. On average, a one-unit increase in the natural amenities scale index increases the probability that a county is a hot spot by 0.70 percent. This implies that a county one-half a standard deviation above the average natural amenities index score of 0.0525 is nearly 80 percent more likely to be an agritourism hotspot than counties at the mean, a notable difference in absolute terms. Although there is little a county can do about its natural amenity endowment, understanding how locational attributes influence a place's competitiveness is important for the environmental scan that underlies any planning exercise or policy formulation (McGranahan Reference McGranahan1999). The amenity result also suggests some areas may be better places for public investment/support of agritourism than other places.

The coefficient on Per Capita Income followed a priori expectations, suggesting that a higher per capita income is associated with a higher propensity to be an agritourism hot spot, all else being equal. More disposable income may mean more money to spend locally on agricultural recreation.

Ln(Population) had a significant (p < 0.01), negative coefficient, suggesting that agritourism hot spots are more likely where population density is lower. The log population result makes sense given the positive coefficient on Travel Time to City of ≥250 k People. It seems agritourism hot spots are more likely to exist in less populated areas, contrary to business location models, which find population and the number of enterprises are positively correlated (Bhat et al. Reference Bhat, Paleti and Singh2014). Thus, rural areas could take advantage of factors that draw visitors to the area (history, natural resources, byways, or national parks) and may consider the presence of such factors in their decision to start up an agritourism operation. The negative coefficient on population may also be a story of resiliency; farms and ranches in less-populated areas are more likely to adopt agritourism due to having few other economic development opportunities.

Residuals Analysis

While the probit model identified interesting factors associated with a county's probability of being an agritourism hot spot, we recognize some regional or local subtleties are lost due to the broad geographic scope of the model. For this reason, we examine the regression residuals, which offer valuable insight into how paucity in model specification may result in residuals that are spatially related. Where the share of farms or ranches with agritourism is higher or lower than the model would lead us to expect, we infer a set of unforeseen opportunities or barriers affect that locale.

Counties with high and low residuals show some degree of clustering (Figure 3). High residual counties (0.95 to 1.0, in black) can be thought of as counties that were predicted not to be hot spots by the probit model, but really were agritourism hot spots. In contrast, low residual counties (−0.5 to −1.0, dark grey) were predicted to be agritourism hot spots, but were not (at least as indicated by the LISA analysis). These counties could represent locations where either something innovative emerged, allowing operators to take advantage of opportunities unaccounted for in our model or, they could represent counties in which the agritourism industry has great potential but is hindered by unforeseen obstacles.Footnote ⁹

Figure 3. Counties with High/Low Residuals when Predicting Presence of Agritourism

The Northeast, Texas, and the Rocky Mountain states provide some visual evidence of residual clusters, but there also seem to be spatial outliers with less agritourism than predicted in Louisiana, Florida, and Northern Texas (Figure 3). While these counties could simply be designated as “unique” due to the model's conservative specification, we believe that the robust LISA and probit results illustrate that the model residuals may hold interesting information on what causes agritourism farms/ranches in these counties to be relatively abundant (or scarce) relative to peer counties. The expanding literature on the potential benefits of agritourism to local communities suggests case studies be conducted on counties with unusual levels of agritourism. We hope future research uses both in-depth case studies and in-depth farm-level microdata to better understand drivers (or lack thereof) of agritourism in various regions. A full list of these hot spots is listed in Figures A1 and A2.

Conclusions and Implications for Rural Development

The evolution of market structure within the U.S. agriculture sector has led to increasing support for enterprise diversification strategies, like agritourism, in sustaining agricultural businesses (Liang and Dunn Reference Liang and Dunn2014). Rural communities’ efforts to find economic development strategies that allow them to remain viable in an era of rapid urbanization has further catalyzed interest in agritourism. The spatial heterogeneity of agritourism and its drivers suggest some communities may be more successful in using agritourism as a strategy to sustain agricultural enterprises and their rural communities.

This study examined the drivers of agritourism clusters to try to understand what place-based features are associated with their creation. Our results suggest that in more densely populated regions such as in the Northeast, farms appear to recognize the value of their urban proximity, taking advantage of travelers’ willingness to visit their operations. In short, easily accessible rural counties with scenic beauty are more likely to have a relatively high percentage of farms or ranches with agritourism—an agritourism hot spot.

Perhaps the most significant contributions this research makes to the existing literature are: (1) its scope (i.e., observing agritourism across all contiguous 48 states), (2) identification of significant agritourism hot spots, and (3) use of location and spatial parameters to describe factors associated with agritourism hot spots.

Beyond informing agritourism managers, our results may provide policy makers and economic development practitioners with a more complete picture of the current extent and potential growth of the industry, with a particular focus on the “appropriate fit” of various agritourism models for different locations. It is likely that public support for agritourism industries will have to take different approaches to developing a successful local industry depending on the unique locational assets in a region. Policy makers and economic developers may find our results useful for addressing opportunities and barriers to growth in their region's agritourism industry.

This research was limited by the available data to examine the prevalence of agritourism enterprises on U.S. farms. The next logical step is to try to determine the factors associated with higher agritourism revenue on participating farms and ranches, but this requires farm-level data that include agritourism revenue and total farm revenue. Future work could also compare the relative success of agritourism operations in “spillover” regions and “resiliency” regions, using farm-level data. Unfortunately, the farm-level data necessary to analyze the spillover effects from place-based factors and industry agglomeration is not publicly available. Thus, we leave these questions for future research.