DO SOCIO-CULTURAL TRAITS AND OTHER DEMOGRAPHICS

AFFECT OUTDOOR RECREATION CONSTRAINTS?

THE CASE FOR MESA COUNTY, COLORADO

Authors: Nathan Perry, Ph.D., Colorado Mesa University, T. Timothy Casey, Ph. D., Colorado Mesa University, and Steven Ross Murray, D. A., University of California, Berkeley

ABSTRACT

Physical activity from outdoor recreational pursuits is well known to be beneficial and has been encouraged by the US Surgeon General. Specific constraints, often reflected in sociocultural issues, can be impediments to individuals participating in outdoor recreational activities. Data were collected in conjunction with the Bureau of Land Management and the Mesa County Health Department in Colorado via surveys involving 580 residents of Mesa County regarding access to outdoor recreation. Ordinal logistic regression was used to test whether ten different constraints to outdoor recreation were important to the respondents, given their age, sex, education, race/ethnicity, native language, and residential density. The results indicated that the probability of experiencing outdoor constraints increased with lower education, Hispanic ethnicity, native Spanish speaking, and young age. A second model, albeit with fewer respondents, was tested to include income. The results of the second model were similar to the first model and also showed that lower incomes were associated with a higher probability of outdoor constraints. The results of the two models indicated that socioeconomic status was a driving factor behind constraints to outdoor recreation.

INTRODUCTION

Physical activity is known to be beneficial for one’s health and wellness and has been encouraged by the US Surgeon General (US Department of Health and Human Services, 1996). One area where governmental agencies continue to strive to provide access for a safe and acceptable environment for physical activity is with outdoor recreation in parks and other governmentally controlled areas. In addition to physical activity, outdoor recreation is viewed as a quality-of-life factor (Frick, Degenhardt, & Buchecker, 2007; “Improving the Measurement...,” 2012; Muller, 1999) and as a means for individuals “to regulate their physical and psychological resources” (Frick, Degenhardt, & Duchecker, 2007, p. 31), to improve their overall wellness. However, various constraints to leisure have been noted in the literature (Frick, Degenhardt, & Buchecker,

2007; Kaczynski et al., 2015; Jackson, 2005; Shores, Scott, & Floyd, 2007), and these constraints can often be predicted based on socioeconomic status. We embarked to investigate how the combinations of various sociocultural factors affected individuals’ constraints to outdoor recreation. This paper contributes to the literature by furthering the research on the effect of socioeconomic status on the constraints to outdoor recreation. Further, an improved empirical methodology, using an ordinal logistic regression instead of a logistic regression which allows for a more specific interpretation of the exact probability of socioeconomic characteristics on constraints, was employed.

REVIEW OF RELATED LITERATURE

Much of the literature regarding outdoor recreation constraints focuses on the role of socioeconomic status in perceived constraints. The American Psychological Association (n.d.) defines socioeconomic status as “the social standing or class of an individual or group... often measured as a combination of education, income and occupation.” Lee, Scott, and Floyd (2001) used a logistic regression to test the multiple hierarchy stratification perspective - a theoretical view operating on the premise that each person has a specific status or position within society - with a large dataset from a Texas outdoor recreation survey. The results illustrated strong evidence for a multiple hierarchy stratification perspective where older, female minorities, without a college degree and who earn under $20,000 per year rank the lowest in outdoor recreation participation; conversely, young, white men with a college degree who make an income of more than $20,000 per year represent the highest rank of the hierarchy. One of the key ideas in multiple hierarchy stratification theory is that there should be different levels of socioeconomic status, and each of these levels will have different challenges with leisure constraints (Crompton, 1998).

Shores, Scott, and Floyd (2007) used a logistic regression to determine the importance of socioeconomic characteristics on constraints to outdoor recreation. They included nine constraints as dependent variables, with education, income, age, sex, and race as independent variables, and their results indicated that lower socioeconomic characteristics had higher probabilities of having constraints to outdoor activities. The probability was enhanced when an individual had more than one characteristic that may imply a higher probability of constraints.

This empirical analysis of the multiple stratification hierarchy approach has been applied to more specific outdoor activities such as wildlife watching. Lee and Scott (2011) found that race/ethnicity is the best predictor of wildlife watching activities, with white/non-Hispanics being 2.4 times more likely to engage in wildlife watching than other demographics. Being more educated, living in rural areas, having higher household incomes, being older, and being female were all associated with more outdoor wildlife watching. The results indicate support for the multiple hierarchy stratification perspective, or that socioeconomic status matters for outdoor constraints. Other studies find evidence that race, sex, income, and other variables used to analyze multiple stratification hierarchy perspective are primary determinants to outdoor recreation (Johnson, Bowker, & Cordell, 2001; Scott & Munson, 1994).

METHODS

The data for this study were collected via online and in-person surveys by the staff of the Mesa County Health Department during the period of October 2015 through July of 2016. Several staff members visited various public facilities including governmental agencies (e.g., schools, workforce centers) and community groups (e.g., churches, social organizations) throughout Mesa County, Colorado, to collect data. The survey was constructed from one used previously for the residents of Larimer County, Colorado (“Plug in to nature,” 2012), and it employed a variety of questions to assess demographic information such as age, race/ethnicity, residential density, education, income, native language, and sex. The survey was available in both English and Spanish to ensure an adequate response rate for all area residents. Additionally, lower-income residents were targeted to have a sample size that reflected the true demographics of Mesa County, Colorado (See Table 1). The following 10 constraints to leisure were analyzed:

1. Being in nature makes me nervous. (Nervous)

2. The risks to a child while in nature exceed the benefits. (Risks)

3. My child would rather play with technology. (Technology)

4. Time. (Time)

5. Cost. (Cost)

6. Transportation. (Transportation)

7. Location. (Location)

8. I am aware of the programs we can attend in Mesa County. (Awareness)

9. I am uncomfortable with my family being in nature/outdoors because we don’t see other people who share our culture. (Culture)

10. We would not attend a program if it was offered in a language different from the language we speak at home. (Language)

Table 1. Community survey demographics compared to census demographics.

	Total Surveys	Survey Respondents	Mesa County Census (2014)
White	480	82.76%	82%
Latino	81	13.97%	14.10%
Black or African American	3	0.52%	1%
American Indian or Alaska Native	7	1.21%	1.50%
Asian	4	0.69%	0.90%
Other	5	0.86%	N/A
Total	580	100%	100%

The response categories were “Strongly Disagree,” “Disagree,” “Neutral,” “Agree,” and “Strongly Agree.” For time, cost, transportation, and location, the response variables were slightly different using “No Influence,” “Minor Influence,” “Some Influence,” “A Lot of Influence,” and “Very Strong Influence. ” Each of these constraints served as a dependent variable that was regressed using an ordinal logit regression to determine the effect of the demographic and personal characteristics on the constraints to outdoor recreation. Of the 677 surveys completed, 580 (85.7%) were complete and usable, and 431 (63%) were submitted online.

The data were collected by the Mesa County Health Department in two phases. The first phase did not ask about income on the survey, but did have a Spanish version, which made it possible to survey native Spanish-speakers about constraints to outdoor recreation. The second phase of data collection included income in the survey, but did not capture any Spanish-speakers. Because of this, two models were tested in order to test the effect of being a native Spanish-speaker on constraints, and to test separately the effect of income on constraints. The two models are listed below:

Model 1: Ordinal Logit (π) = α + Σ (βrural, βedu, βage, βsex, βethnicity, βspanishspeaker)

Model 2: Ordinal Logit (π) = α + Σ (βrural, βedu, βage, βsex, βethnicity, βincome)

This study follows the general methodology of Shores, Scott, and Floyd (2007) and Lee, Scott, and Floyd (2001). Shores et. al. (2007) and Lee et. al. (2001) each use logistic regression. In the case of Shores et. al. (2007), the dependent variables had three response choices, and the authors chose to dichotomize the responses in order to perform a logistic regression, which requires two possible responses. We improved the empirical methodology of these two aforementioned studies by employing an ordinal logistic regression, which allows an infinite number of responses (in the case of this study, five). In the case where the choices of responses were value oriented and not perfect numerical scores, an ordinal logit provided the most efficient results. Ordinal logistic regression was a better fit than ordinary least squares (OLS) because in the case of an ordinally ranked dependent variable OLS does not necessarily minimize the variance, the assumption of homoscedasticity is likely to be violated, and distributional assumptions about the dependent variables are likely to be incorrect (Menard, 2010). It is a better model than the standard logit model because it does not force the response variables into two categories, hence allowing the researcher to view the exact probability of the respondent answering any of the response categories from the dependent variable.

RESULTS

Of the 677 surveys completed, 580 (85.7%) were complete and usable, and 431 (63%) were submitted online. Although there were 677 respondents, not all respondents answered each question. Only 580 individuals answered the demographic information questions. The regression analysis used requires that each row of data not have any missing data points, so for the purposes of the regression analysis below, 580 samples were used.

Table 2 presents the results of model 1. 1 Twenty out of the 60 coefficients were statistically significant at the 90% level. Rural was positive and significant for the risks constraint, but not significant for the other barriers, indicating that overall there is little difference in perceived risks between rural and non-rural residents. Education was significant in six of 10 regressions, showing the expected signs on the coefficients. Note that with the absence of income information, education can serve as a proxy for income. In the second model, education and income are both controlled for. Age was significant in five of 10 regressions, and the sign on the coefficient indicated that the higher the age, the less the effect of the constraint on recreation. With respect to the sex of the responders, it was significant in one regression only, indicating little evidence that sex played a role in outdoor recreation constraints. The ethnicity variable Hispanic was significant in three of 10 regressions, illustrating that persons of Hispanic culture have less awareness of outdoor activities, more concerns about cultural barriers, and concern about language as a barrier. This was true despite controlling for people who were native Spanish speakers. The native Spanish speaker variable was significant in four of 10 constraints, those being nervous, risk, awareness, and culture. The chi squared test for the statistical significance of the model is listed as the last row in Table 2 and Table 3. Seven of 10 regressions were statistically significant, with the three insignificant regressions being time, location, and transportation.

1 Note that in Tables 1 and 2, the results are in log-odds units and cannot be interpreted as normal regression coefficients. In order to interpret them probabilities must be estimated (see the next section and Table 4 for predicted probabilities). These odds ratios are used to determine statistical significance, but not for interpretation. Note that the number below in parenthesis is the standard error of the odds ratio.

Table 2. Ordinal logit regression estimates of the individual effects of rurality, Education, age, sex, race/ethnicity, and being a Spanish speaker on outdoor recreation constraints for Model 1.

Nervous

Risks

Technology

Time

Cost

Trans- portation

Location

Awareness

Culture

Language

Residential Density

1=Rural

-.011

(.214)

.382**

(.185)

.151

(.175)

.023

(.172)

-.032

(.171)

.187

(.170)

.159

(.172)

-.056

(.169)

.152

(.195)

.197

(.169)

Education

1=college grad

.402**

(.197)

-.365**

(.172)

.189

(.156)

.216

(.155)

-.525***

(.157)

-.331

(.154)

.086

(.155)

.333**

(.157)

-.355**

(.180)

-.346**

(.155)

Age

-.272***

(.007)

-.019***

(.006)

.014***

(.005)

-.003

(.005)

-.001

(.005)

-.003

(.005)

-.006

(.005)

.031***

(.005)

-.011*

(.006)

.006

(.005)

Sex

1=male

-.655**

(.259)

-.309

(.209)

.200

(.182)

-.104

(.178)

-.184

(.180)

.209

(.176)

.245

(.181)

.180

(.178)

.134

(.213)

.020

(.180)

Race/

Ethnicity

1=Hispanic

.059

(.334)

.049

(.294)

.123

(.285)

-.214

(.288)

.378

(.293)

-.056

(.298)

.240

(.283)

-.525*

(.296)

.924***

(.305)

-.469*

(.279)

1=Native Spanish Speaker

1.55***

(.452)

1.59***

(.406)

-.1767

(.418)

-.049

(.408)

-.571

(.403)

.308

(.410)

-.473

(.390)

1.13***

(.423)

1.25***

(.415)

-.047

(.385)

Chi squared p-value

.000***

.025**

.666

.005***

.179

.579

.000***

.089*

*** indicates p < .01; ** indicates p < .05; * indicates p < .10

The abbreviations for the columns’ headings are defined as follows: Being in nature makes me nervous. (Nervous); The risks to a child while in nature exceed the benefits. (Risks); My child would rather play with technology. (Technology); Time. (Time); Cost. (Cost); Transportation. (Transportation); Location. (Location); I am aware of the programs we can attend in Mesa County. (Awareness); I am uncomfortable with my family being in nature/outdoors because we don’t see other people who share our culture. (Culture); We would not attend a program if it was offered in a language different from the language we speak at home. (Language)

Because of the two phases of data collection, Model 2 had a smaller sample size than model 1. Model 1’s sample used the maximum number of data points that were usable (N=580), while model 2 used only the datapoints that had income attached to the survey (N=91). Consequently 13 of 60 coefficients were statistically significant, which was less overall significance than in Model 1. Once income was controlled for, rural had no statistical significance in any of the constraints. Education was significant in the cost and location constraints, and age maintained its linear relationship with constraints and was statistically significant in four of 10 regressions. Sex was not statistically significant, and the ethnicity Hispanic was significant in only one regression. The income variable was significant in five of 10 regressions, including technology, cost, transportation, location, and culture. The coefficient on these variables indicated that as income increased, each statistically significant constraint decreased. The chi-squared variables indicated that six out of 10 of the constraint regressions were statistically significant at the 90% level.

Table 3. Ordinal logit regression estimates of the individual effects of rurality, education, age, sex, race/ethnicity, and being a Spanish speaker on outdoor recreation constraints for Model 2.

Nervous

Risks

Technology

Time

Cost

Trans- portation

Location

Awareness

Culture

Language

Residential Density

1=Rural

.006

(.539)

1.39

(.509)

-.157

(.482)

-.351

(.464)

-.388

(.443)

-.276

(.460)

-.272

(.454)

.650

(.468)

.219

(.509)

.763

(.495)

Education

1=college grad

-.649

(.421)

.207

(.428)

.203

(.387)

-.514

(.396)

-1.01**

(.399)

-.390

(.395)

-1.02**

(.405)

.151

(.399)

.354

(.459)

-.697*

(.404)

Age

-.053*

(.029)

.000**

(.028)

.021

(.023)

-.011

(.025)

-.048*

(.026)

-.000

(.026)

-.012

(.025)

.087***

(.026)

.027

(.027)

.014

(.024)

Sex

1=male

-.321

(.752)

-1.06

(.760)

.466

(.640)

.083

(.620)

.342

(.645)

-.278

(.632)

.641

(.600)

-.247

(.563)

-.363

(.751)

-.073

(.600)

Race/

Ethnicity

1=Hispanic

-.133

(.607)

.354

(.588)

-.388

(.541)

-.928*

(.545)

-.694

(.564)

-.368

(.570)

.055

(.574)

.485

(.597)

.712

(.638)

-.661

(.565)

Income

1=Greater than

$50,000

-.085

(.485)

-.597

(.509)

-1.08**

(.482)

-.657

(.464)

-.933**

(.462)

-1.46***

(.477)

-.856*

(.456)

-.647

(.479)

-1.21**

(.519)

-.330

(.490)

Chi squared p-value

.208

.050**

.389

.296

.002***

.020**

.009

.010*

.229

.299

*** indicates p <.01; ** indicates p <.05; * indicates p <.10

PREDICTING CONSTRAINT PROBABILITIES

Using an ordinal logit allows for the calculation of the probability of a constraint affecting a person given specific demographic characteristics. Tables 4 and 5 (provided at the end of the paper) illustrate the probability of experiencing barriers to outdoor recreation by selected characteristics. Table 4 represents model 1, which has Spanish speaking but does not have income, and Table 5 represents Model 2, which has income but not Spanish speaking. The predicted probabilities in Tables 4 and 5 each has three stratum for both Model 1 and Model 2. Stratum 1 uses selected characteristics that would tend to have a higher probability of constraints based on the results of previous literature. In other words, “lower” socioeconomic status variable was used to calculate stratum 1. Stratum 2 was the probability of constraints for the average response for each characteristic. Stratum 3 used characteristics that would tend to have a lower probability of constraints, or higher socioeconomic status characteristics. The three strata can be compared to understand the distinct differences of perceived constraints to outdoor activities between different socioeconomic characteristics.

The results for model 1 (see Table 4) of the predicted probabilities of constraints based on socioeconomic status (multiple stratification hierarchy analysis) illustrate that for five of the 10 constraints (i.e., nervous, risks, costs, transportation, culture, and language) that Stratum 1 had a higher probability of constraints than Stratum 2 or 3, providing evidence that high socioeconomic characteristics perceive less constraints than individuals with low socioeconomic characteristics. This trend was not present for technology, time, location, and awareness. Table 4 can be interpreted as follows: Provided the characteristics in the left columns, the probability of answering “strongly disagree” for the nervous constraint is 18.7% for low socioeconomic status (stratum 1), compared to 73.1% for medium socioeconomic status (stratum 2) and 90.1% probability for high socioeconomic status (stratum 3). For the same constraint, 17.6% of low socioeconomic status respondents strongly agree that going outdoors makes them nervous, compared to medium socioeconomic status (1.7%) and high socioeconomic status (0.5%).

Table 5 illustrates Model 2, which includes income but not Spanish speaking. Nervous, risks, costs, transportation, location, culture, and language all follow the trend that lower socioeconomic status people have a higher probability of constraints to outdoor recreation. Technology and time do not follow this trend, and awareness seems to increase as a constraint for higher socioeconomic status.

Table 4. Model 1 probability of experiencing restraints to outdoor recreation by selected characteristics.

Variables

Barriers

Rural

Sex

College Grad

Age

His-

panic

Spanish Speaker

Nervous

Risks

Tech-

nology

Time

Cost

Trans- portation

Location

Aware-

ness

Culture

Language

Stratum 1

Rural

Non- College

Yes

SD=.187

SD=.098

SD=.289

NI=.071

NI=.090

NI=.128

NI=.109

SD=.060

SD=.126

SD=.239

D=.250

D=.146

D=.234

MI=.093

MI=.165

MI=.171

MI=146

D=.223

D=.230

D=.183

N=.258

N=.110

N=.334

SI=.305

SI=.338

SI=-.286

SI=.305

N=.374

N=.243

N=.237

A=.127

A=.156

A=.106

AL=.326

AL=.204

AL=.236

AL=.246

A=.245

A=.223

A=.185

SA=.176

SA=.487

SA=.035

VS=.203

VS=.201

VS=.176

VS=.193

SA=.094

SA=.177

SA=.154

Stratum 2

Non-Rural

College Grad

SD=.731

SD=.574

SD=.213

NI=.058

NI=.098

NI=.207

NI=.101

SD=.043

SD=.636

SD=.193

D=.170

D=.225

D=.209

MI=.078

MI=.176

MI=.224

MI=.139

D=.174

D=.233

D=.165

N=.062

N=.071

N=.379

SI=.279

SI=.342

SI=-.283

SI=.300

N=.357

N=.077

N=.239

A=.017

A=.056

A=.146

AL=.342

AL=.196

AL=.176

AL=.258

A=.294

A=.034

A=.209

SA=.017

SA=.071

SA=.052

VS=.241

VS=.186

VS=.108

VS=.205

SA=.130

SA=.017

SA=.193

Stratum 3

Non-Rural

College Grad

SD=.907

SD=.777

SD=.147

NI=.059

NI=.144

NI=.226

NI=.095

SD=.018

SD=.747

SD=.198

D=.063

D=.134

D=.170

MI=.079

MI=.224

MI=.233

MI=.132

D=.084

D=.171

D=.167

N=.019

N=.034

N=.402

SI=.280

SI=.344

SI=.278

SI=.295

N=.255

N=.049

N=.239

A=.005

A=.024

A=.200

AL=.341

AL=.157

AL=.164

AL=.259

A=.375

A=.021

A=.206

SA=.005

SA=.028

SA=.079

VS=.239

VS=.129

VS=.097

VS=.217

SA=.265

SA=.010

SA=.188

Table 5. Model 2 probability of experiencing restraints to outdoor recreation by selected characteristics.

Variables

Barriers

Rural

Sex

College Grad

Age

Hispanic

Income

Nervous

Risks

Tech- nology

Time

Cost

Trans-

por tation

Location

Aware- ness

Culture

Language

Stratum 1

Rural

Non-College

Yes

Under 50,000

SD=.293

SD=.174

SD=.262

NI=.039

NI=.028

NI=.146

NI=.036

SD=.082

SD=.412

SD=.159

D=.301

D=.238

D=.284

MI=.093

MI=.092

MI=.203

MI=.058

D=.263

D=.388

D=.090

N=.266

N=.100

N=.332

SI=.353

SI=.256

SI=.253

SI=.246

N=.378

N=.107

N=.207

A=.108

A=.123

A=.105

AL=.335

AL=.287

AL=.230

AL=.347

A=.216

A=.068

A=.310

SA=.029

SA=.362

SA=.015

VS=.177

VS=.335

VS=.166

VS=.311

SA=.058

SA=.023

SA=.231

Stratum 2

Non-Rural

College Grad

Over

50,000

SD=.597

SD=.573

SD=.231

NI=.035

NI=.083

NI=.275

NI=.106

SD=.068

SD=.666

SD=.221

D=.242

D=.243

D=.274

MI=.084

MI=.217

MI=.268

MI=.141

D=.233

D=.253

D=.111

N=.117

N=.052

N=.354

SI=.335

SI=.353

SI=.227

SI=.373

N=.380

N=.046

N=.224

A=.034

A=.047

A=.121

AL=.347

AL=.206

AL=.146

AL=.253

A=.246

A=.025

A=.274

SA=.008

SA=.082

SA=.018

VS=.197

VS=.139

VS=.082

VS=.124

SA=.070

SA=.008

SA=.168

Stratum 3

Non-Rural

College Grad

Over

50,000

SD=.899

SD=.836

SD=.118

NI=.050

NI=.254

NI=.437

NI=.146

SD=.017

SD=.665

SD=.262

D=.070

D=.108

D=.195

MI=.115

MI=.363

MI=.271

MI=.175

D=.076

D=.254

D=.122

N=.023

N=.017

N=.420

SI=.385

SI=.259

SI=.164

SI=.380

N=.246

N=.046

N=.227

A=.006

A=.014

A=.226

AL=.304

AL=.082

AL=.084

AL=.207

A=.418

A=.026

A=.248

SA=.001

SA=.023

SA=.039

VS=.142

VS=.041

VS=.042

VS=.089

SA=.240

SA=.008

SA=.138

CONCLUSION AND IMPLICATIONS

We used an ordinal logistic regression to determine the characteristics of outdoor constraints for individuals in Mesa County, Colorado. This model was then used to predict the probability of having or perceiving an outdoor constraint. The constraints were viewed by demographic characteristics classified into socioeconomic status in order to understand if lower socioeconomic status had more constraints to outdoor recreation than high socioeconomic status. The results show that education, income, age, language, and ethnicity were the most statistically significant variables in determining the probability of outdoor constraints, while sex was statistically insignificant, and rural was significant in only one constraint. Age showed the expected signs and confirmed the authors’ previous research (Perry & Casey, 2016), where a distinct negative linear relationship between age and each constraint had been documented. The language result illustrated that Spanish speakers were much more likely to deal with constraints to outdoor recreation. Spanish speakers may be worried about communication with others for information, directions, or in an emergency situation, and a language barrier possibly could create more risk.

The results of the predictive model illustrated that nervous, risks, costs, transportation, culture, and language all followed the trend that showed lower socioeconomic status had higher constraints, and vice versa. Model 2 differed from model 1 in that location was more impactful for low socioeconomic status, and after controlling for income, awareness seemed to be more of a constraint for high socioeconomic status. This predictive model thus provides researchers and policymakers with baseline expectations for predicting outdoor constraints based on demographics and socioeconomic status.

These results from this study largely confirmed the role of socioeconomic status in outdoor activities. The lower the socioeconomic status, the higher the constraints to going outdoors. This result was found in all of the papers covered in the literature review. The implications of this study are far reaching. Low socioeconomic residents have an assortment of constraints that high socioeconomic status residents do not have. Suggestions to bridge the socioeconomic gap would be to improve the offerings of outdoor programs, improve awareness of outdoor sites and activities, improve accessibility, and to help alleviate cost. For a comprehensive list of suggestions based on a much wider range of surveys, see Perry and Casey (2016).

Several limitations to this study must be noted. Although the survey has been used in previous studies (Plug into nature, 2012; Perry & Casey, 2016) to address outdoor constraints regarding public lands, the survey instrument’s validity needs to undergo rigorous psychometric testing. Although the survey was directed toward families, it is possible that individuals without children answered these questions. The first question in the survey asks the question, “How many children do you have under the age of 18 and who live in your home at least 50% of the time?” Although this question helps to understand if the respondent is a primary caretaker, it does not dismiss persons who may not have children. This could potentially bias the family-centric questions (i.e., risks to a child while in nature exceed the benefits and my child would rather play with technology). Note that it is a potential bias, and just because individuals do not have children of their own does not mean they cannot accurately assess the risk to other children in their lives. Another limitation was the lack of income in model 1. Although education can serve as a proxy for income, it is preferable to have the income variable for all survey responses, and not just part of them. Future researchers should consider more specific questions about individual constraints in order to narrow down and to measure the exact nature of these constraints.

REFERENCES

American Psychological Association. Socioeconomic status. Retrieved from http://www. apa. org/topics/socioeconomic-status/.

Crompton, R. (1998). Class and stratification (2nd edition). Cambridge: Polity Press.

Plug in to nature: Finding connections to the outdoors for youth and families in Larimer County. (2012, March 7). Retrieved from https://www. larimer. org/sites/default/files/uploads/2017/plug_in_to_nature_project_report. pdf.

National Center for Education Statistics. (2012). Improving the measurement of socioeconomic status for the national assessment of educational progress: A theoretical foundation. Washington, DC: National Center for Education Statistics.

Johnson, C. , Bowker, J. _M. , & Cordell, K.H. (2001). Outdoor recreation constraints: An examination of race, gender, and rural dwelling. Southern Rural Sociology, 17, 111-133.

Lee, K. J., & Scott, D. (2011). Participation in wildlife watching: A multiple hierarchy stratification perspective. Human Dimensions in Wildlife, 16, 330-344.

Lee, J., Scott, D. , & Floyd, M. (2001). Structural inequalities in outdoor recreation participation: A multiple hierarchy stratification perspective. Journal of Leisure Research, 33(4), 427-449.

Menard, S. (2010). Logistic regression: From introductory to advanced concepts and applications.

Thousand Oaks, CA. Sage Publications.

Perry, N., & Casey, T. (2016). Pathways to nature in Mesa County: Connecting children with the outdoors in Mesa County, Colorado. Retrieved from https://coloradomesa. edu/natural-resource- center/documents/Pathways%20to%20Nature%20in%20Mesa%20County.

Scott, D,, & Munson, W. (1994) Perceived constraints to park usage among individuals with low incomes. Journal of Park and Recreation Administration, 12, (4), 79-96.

Shores, K., Scott, D. & Floyd, M. (2007). Constraints to outdoor recreation: A multiple hierarchy stratification perspective. Leisure Sciences, 29, 227-246.

US Department of Health and Human Services. (1996). Physical activity and health: a report of the Surgeon General. Atlanta, Georgia: US Department of Health and Human Services, Public Health Service, CDC, National Center for Chronic Disease Prevention and Health Promotion. Retrieved from https://profiles. nlm. nih. gov/ps/access/NNBBHC. pdf.