A Structural Model of Leisure Constraints Negotiation in Outdoor Recreation

Abstract

A conceptual model tested the leisure constraints negotiation process of outdoor recreation: motivation and the constraints to participate likely influenced by negotiation efforts. Higher motivation to participate encourages using negotiation strategies and resources to overcome constraints. Experiencing constraints was thought to trigger negotiation efforts. Drawing from social cognitive theory, negotiation-efficacy was proposed to encourage motivation, diminish the perception of constraints and promote negotiation efforts, which indirectly influenced positive participation. The model tested used data collected from a random sample of Arizona residents through hierarchical confirmatory factor analysis and structural equation modeling. Results support the conceptual model and suggest the constraints negotiation process is a dynamic interaction of influences promoting outdoor recreation participation.

Keywords:

• motivation
• self-efficacy
• social cognitive theory
• structural equation modeling

Constraints negotiation research has shed new light on social-psychological and behavioral processes fundamental to leisure such as motivation, desired experiences, constraints, negotiation and activity participation. The negotiation concept was introduced to explain how leisure constraints may be overcome or navigated (Jackson, Crawford, & Godbey, 1993; Kay & Jackson, 1991; Scott, 1991). Initial empirical research was mostly descriptive as researchers identified strategies and resources used by people to negotiate constraints (Jackson & Rucks, 1995; Scott, 1991) and categorized respondents based on the use of negotiation resources (Henderson et al., 1995).

Recent research has made advances in theory and method by placing negotiation in the broader context of leisure behavior and by exploring relations between constraints and other concepts using multivariate analysis. Notably, Hubbard and Mannell (2001) used confirmatory factor analysis and structural equation modeling to examine four competing models of the negotiation process. They found support for a constraint-effects-mitigation model in which encountering constraints activates the use of negotiation strategies that limit the negative impact of constraints on activity participation. They also found that higher motivation presages greater negotiation efforts. Loucks-Atkinson and Mannell (2007) extended this research by drawing on social cognitive theory and incorporating a negotiation-efficacy construct. In his overview of the epistemology of leisure constraints, Jackson (2005a) stated that such empirical tests of theoretically derived models through confirmatory factor analysis and structurally equation modeling is one of the most “innovative, and potentially fruitful directions” (p. 8) for research.

Despite the significant contributions of the work by Mannell and colleagues (i.e., Hubbard & Mannell, 2001; Loucks-Atkinson & Mannell, 2007), certain admitted limitations of these studies require additional research be conducted to clarify the conceptual model and demonstrate generalizability of the findings in additional populations. Hubbard and Mannell's study was based on a limited and homogenous sample of employees from companies that offered work site recreation programs and the authors admitted: “Caution needs to be exercised in generalizing the findings of the present study. Further research with other populations and activities will be needed to determine the universality of the processes identified” (p. 158). Their study relied on a global measure of motivation comprised of two items, which the authors recognized as a limitation as did others (e.g., Alexandris, Tsorbatzoudis, & Grouios, 2002). The moderate sample size in Loucks-Atkinson and Mannell's study eliminated the possibility of using structural equation modeling, a multivariate technique that accounts for measurement error in the model and establishes relationships among constructs while holding constant the effects of the other constructs. Thus, Loucks-Atkinson and Mannell's analysis and interpretation relied on the assumption of perfect measurement of the model constructs. Further, their conclusions were based on correlational and not causal analysis. The sample in Loucks-Atkinson and Mannell's study was a purposively selected group of individuals with a specific physical disorder, and thus, the generalizability of the findings was limited.

My study extends prior research by conducting an empirical test of a conceptual model of constraints negotiation in the context of outdoor recreation (see Figure 1). The model proposes that motivation is directly and positively related to outdoor recreation participation. The positive influence of motivation is counteracted by the negative influence of constraints. The model also proposes that these relationships are likely to be influenced by efforts to negotiate: Higher motivation to participate in outdoor recreation is likely to encourage the use of negotiation strategies and resources to overcome constraints. The experience of constraints, however, also has an indirect positive impact on participation by triggering or activating negotiation efforts. Negotiation-efficacy encourages motivation, diminishes the perception of constraints, and encourages the use of negotiation efforts, thus having an indirect positive influence on participation.

FIGURE 1 Structural model of leisure constraints negotiation with hypotheses and parameters to be estimated.

This extension is particularly important for several reasons. First, my study is based on a statewide random sample of Arizona residents, which provided the opportunity to assess the generalizability of previous findings in a diverse population. Second, the relatively large sample size (n = 439) permitted the use of multivariate analyses to assess construct validity, account for measurement error and test for mediation. Hierarchical confirmatory factor analysis (CFA) was used to evaluate the factor structures of the constructs and test the measurement model. Third, structural equation modeling (SEM) was used to specify, estimate and assess the theoretical model to test hypothesized relationships. Further, recommendations from several scholars (Henderson et al., 1995; Hubbard & Mannell, 2001; Loucks-Atkinson & Mannell, 2007) were followed to draw from social cognitive theory and specifically self-efficacy to develop the theoretical specification of the negotiation process. In addition to theory development, this study may provide recreation planners and managers greater appreciation of the factors related to outdoor recreation participation. Such improved understanding can give leisure service providers new insights into strategies to limit constraints, encourage negotiation, and maximize the benefits of outdoor recreation to individuals and society.

Review of the Literature

Constraints emerged as a central theme in leisure research over the past 25 years as demonstrated by the proliferation of journal articles and the recent publication of a comprehensive text on the subject (Jackson, 2005b). Early leisure constraints researchers framed the issue in terms of barriers to recreation activity participation (e.g., Buchanan & Allen, 1988; Searle & Jackson, 1985a, 1985b) with the implicit assumption that encountering barriers necessarily resulted in nonparticipation. Constraints were defined as factors that may inhibit activity participation or limit satisfaction (Jackson, 1988). Crawford and Godbey (1987) argued that constraints affect not only participation but also acquisition of leisure preferences. They organized constraints into three categories: intrapersonal constraints defined as individual psychological qualities that affect the development of leisure preferences (e.g., shyness), interpersonal constraints defined as social factors that affect development of leisure preferences (e.g., lack of companions), and structural constraints comprised of factors that intervene between development of leisure preferences and participation (e.g., financial resources). Crawford, Jackson, and Godbey (1991) extended this line of thinking and presented their hierarchical model of leisure constraints, which posited that intrapersonal and interpersonal constraints affect leisure preferences whereas structural constraints intervene between preferences and participation.

The origins of constraints research has been traced at least as far back as the U.S. Outdoor Recreation Resources Review Commission (ORRRC) reports of the 1960s (Jackson & Scott, 1999). As Walker and Virden (2005) noted, however, research on constraints to outdoor recreation has been descriptive and not explanatory with little theoretical development. Walker and Virden pointed out several constraints that scored high on scales across studies, including lack of time, distance to recreation resources, crowding, lack of information, costs, and family commitments. They also noted that constraints to outdoor recreation were similar to other leisure activities but some, like lack of time, may have a stronger impact in the outdoor recreation context because of the travel commitments sometimes necessary to reach undeveloped natural areas. The conceptualization of leisure constraints and empirical studies from outdoor recreation literature led to the first study hypothesis:

H1: Constraints have a direct negative effect on outdoor recreation participation.

It has become increasingly clear that constraints are not necessarily fixed barriers that result in nonparticipation; rather, constraints once encountered might be overcome or negotiated (Crawford et al., 1991; Jackson et al., 1993; Kay & Jackson, 1991; Scott, 1991). Researchers identified strategies and resources used by people to negotiate constraints. For instance, Scott found that bridge players acquired information about opportunities, managed time and schedules and developed new skills. Kay and Jackson discovered that when faced with financial constraints, respondents saved money and identified less expensive opportunities. When faced with time constraints, people reduced time spent on household chores and reduced work time. Jackson et al. synthesized existing research and presented a series of propositions that outlined the constraints negotiation process. Their first and most central proposition stated that “participation is dependent not on the absence of constraints… but on negotiation through them” (p. 4). Jackson and Rucks (1995) were the first to design a study explicitly to investigate negotiation. The descriptive study classified negotiation strategies as cognitive or behavioral with the latter category subdivided into time management, skill acquisition, changing interpersonal relations, improving finances, physical therapy and changing leisure aspirations. Based on existing negotiation research, the second hypothesis in this study was:

H2: Negotiation has a direct positive effect on outdoor recreation participation.

The role of motivation in the negotiation process was addressed explicitly for the first time by Jackson et al. (1993) in their sixth or balance proposition, which stated that “both the initiation and outcome of the negotiation process are dependent upon the relative strength of, and interactions between, constraints on participating in activity and motivation for such participation” (p. 9). In the outdoor recreation literature, motivation is often conceptualized as desire for satisfying recreation experiences and operationalized through the use of the Recreation Experience Preference (REP) scales (Driver, Tinsley, & Manfredo, 1991; Manfredo, Driver, & Tarrant, 1996; Moore & Driver, 2005). The full REP scales include 21 factors each with 2–7 dimensions (Moore & Driver). The REP scales were not based explicitly in motivation, human need or self-determination theory but were developed inductively over 20 years through exploratory factor analyses. Although the REP scales may be criticized for poor theoretical specification, Manfredo et al. (1996) concluded that the scales “can be usefully applied when attempting to determine motivations for or the psychological outcomes desired from leisure” (p. 204). Manfredo et al. determined that the REP demonstrated overall consistency, construct validity and acceptable reliability in a meta-analysis of 36 studies using them to measure leisure motivations. Studies of leisure motivations in outdoor recreation led to the third hypothesis:

H3: Motivation, in the form of desire for satisfying recreation experiences, has a direct positive effect on outdoor recreation participation.

The effect of motivation on negotiation efforts has received more attention in recent years. In their studies of sport and physical activity participation in Greece, Alexandris and colleagues (i.e., Alexandris, Barkoukis, Tsorbatzoudis, & Grouios, 2003; Carroll & Alexandris, 1997) addressed the relations among motivation, constraint, and participation. Carroll and Alexandris found a negative bivariate correlation between motivation and constraints, suggesting that people who are less motivated to participate experience higher levels of constraint. The significance of this finding was limited, however, because it was based on correlational and not causal analysis. Additional research provided some contradictory evidence.

By 2001 researchers had begun to develop conceptual models of constraint negotiation and test these models with empirical data using multivariate analyses. Specifically, Hubbard and Mannell (2001) used CFA and SEM to test four competing models: the independence model, the negotiation buffer model, the constraint-effects-mitigation model and the perceived-constraint-reduction model. Each model included the constructs of motivation, constraint, negotiation and participation (i.e., as the dependent variable). The constraint-effects-mitigation model hypothesized that motivation and negotiation positively and directly influenced participation while constraint negatively and directly influenced participation. The positive relationship between constraint and participation was hypothesized to be mediated by negotiation. Of the four models tested, the constraint-effects-mitigation model provided the best fit to the data. Hubbard and Mannell concluded this model was superior. The model implied that encountering leisure constraints prompts a person to employ negotiation strategies, which mitigates the negative impact of constraint on leisure participation.

The constraint-effects-mitigation model provided some additional insight into the role of motivation in negotiation and participation. Although a weak bivariate correlation existed between motivation and participation, once the effects of other variables were controlled, the final structural model did not include a significant direct path from motivation to participation. Hubbard and Mannell (2001) interpreted this finding to mean that the relationship between motivation and participation was mediated by negotiation, although whether this mediation was statistically significant was unclear from their analysis. Although Carroll and Alexandris (1997) and other studies suggest that motivation plays a more pivotal role in encouraging participation directly, Hubbard and Mannell indicated that motivation's effect on participation is mediated by negotiation. Though the existing research is somewhat equivocal on these points, hypotheses four and five for this study were:

H4: Constraint has a direct positive effect on negotiation.

H5: Motivation has a direct positive effect on negotiation.

As Hubbard and Mannell (2001) stated, however:

Whether motivation is an immediate antecedent and plays a stronger direct role in countering the effects of constraints, when other types of leisure activities, motives, and circumstances is involved, is unclear and will have to be determined by future research. (p. 159)

Self-efficacy in Leisure Constraints Negotiation

Several authors discussed the potential utility of social cognitive theory and specifically the self-efficacy construct to leisure constraint negotiation (Henderson et al., 1995; Hubbard & Mannell, 2001). The idea that successful negotiation of leisure constraints may lead to increased confidence in the ability to negotiate future constraints is implied by propositions three and five from Jackson et al. (1993). Proposition three stated the “absence of the desire to change current leisure behavior may be partly explained by prior successful negotiation of structural constraints” (p. 6) and proposition six stated that “anticipation consists not simply of the anticipation of the presence or intensity of a constraint but also of anticipation of the ability to negotiate it” (p. 8).

Bandura (1994) defined perceived self-efficacy as “people's beliefs about their capabilities to produce designated levels of performance that exercise influence over events that affect their lives” (p. 71). Self-efficacy involves individual judgment about competence to perform a specific task or group of tasks in a given domain. Belief in self-efficacy in one task or domain, however, may be generalized to enhance efficacious feelings in other tasks and domains (Bandura, 1986). People with higher levels of perceived self-efficacy have greater motivation to persevere in the face of adversity. Bandura (1997) argued that self-efficacy is a powerful determinant of human behavior and changing self-efficacy perceptions is a necessary precursor to changing behavior.

Bandura (1997) identified four sources of self-efficacy: enactive attainment through mastery experience, vicarious experience, social persuasion and physiological and affective states. Mastery experience is the most influential source of self-efficacy and is derived from prior success at completing tasks and is undermined by failures. Self-efficacy is also produced through vicarious experience by observing other people succeed or fail in their efforts to accomplish tasks. The influence of vicarious experience on a subject's self-efficacy is greatest when the subject perceives similarity to the model. The third source of self-efficacy is social persuasion. When people are encouraged to sustain their effort in the face of challenges it promotes skill development, sense of achievement, and personal efficacy. Finally, psychological and affective states play a role in the development of self-efficacy. People interpret stress and tension as indicators of poor performance and positive mood as indicators of success. Thus, self-efficacy can be generated by association of success with positive mood.

Self-efficacy affects human functioning through four major psychological processes: cognition, motivation, affect and selection (Bandura, 1994). First, people's self-efficacy beliefs influence cognitive processes in the way people think about and construct scenarios about likely outcomes of future behaviors. Second, self-efficacy affects motivational processes through causal attributions, outcome expectations and goals. Self-efficacy influence people's attributions of their success or failure individual effort or ability; their expectations about the likelihood of valued outcomes of their behavior; and the goals they set, the amount of effort expended to achieve their goals, and their perseverance and resilience in the face of setbacks. Third, self-efficacy plays a role in regulating affective states such as stress, anxiety arousal and depression through the exercise of control over disturbing thoughts. Fourth, self-efficacy beliefs have an effect on the environments and activities people select, thus influencing the course of their lives.

Self-efficacy has been employed in public health and exercise science studies through constructs such as exercise self-efficacy and barriers self-efficacy (e.g., Bloomquist et al., 2006; Marquez & McAuley, 2006; McAuley, 1992, 1993; Motl et al., 2005; Winters, Petosa, & Charlton, 2003). Although the public health and exercise science literature has retained the barriers framework as opposed to the leisure science conceptualization of constraints, these studies provide further evidence to expect that higher levels of confidence in one's ability to overcome constraints to recreation activity participation (i.e., negotiation-efficacy) could improve models of leisure constraint negotiation.

Loucks-Atkinson and Mannell (2007) examined the role of self-efficacy in the constraints negotiation process for individuals with fibromyalgia syndrome, a condition characterized by chronic pain and related symptoms such as anxiety and depression. For the study, the authors defined negotiation-efficacy as “people's confidence in their ability to successfully use negotiation strategies to overcome constraints they encounter” (p. 22). To measure negotiation-efficacy, they asked respondents to report their level of confidence in their ability to use 37 specific negotiation strategies on a confidence scale ranging from 0% (Very Uncertain) to 100% (Very Certain). Their measure focused on assessing the level and strength of specific efficacy expectations. They examined four nested models of the role of self-efficacy in constraint negotiation using path analysis. Considering fit indices, model comparisons and parsimony, they settled on a model that indicated:

(1) constraints negatively influenced participation and positively influenced negotiation efforts, (2) negotiation efforts positively affected participation, (3) motivation increased negotiation efforts, (4) negotiation-efficacy positively influenced strength of motivation to participate and negotiation efforts, and (5) motivation positively influenced participation. Finally, these results suggested that negotiation-efficacy did not have a significant negative effect on perceived constraints. (p. 32)

Loucks-Atkinson and Mannell's findings provide support for the inclusion of the negotiation-efficacy construct in models of the constraints negotiation process. Specifically, a direct positive relationship was identified between negotiation-efficacy and negotiation indicating that higher confidence in the ability to successfully negotiate constraints resulted in greater efforts to negotiate. The final three hypotheses study stem from studies of self-efficacy in leisure, recreation, and public health:

H6: Negotiation-efficacy has a direct positive effect on negotiation.

H7: Negotiation-efficacy has a direct positive effect on motivation.

H8: Negotiation-efficacy has a direct negative effect on constraints.

Methods

Data Collection and Sample

The data analyzed for this paper were collected as part of a larger study of leisure in Arizona conducted to inform natural and cultural resource managers at Arizona State Parks. Data were collected through a random digit-dialed (RDD) telephone survey of Arizona households followed by a self-administered mail questionnaire of respondents identified through the telephone survey. The telephone questionnaire included questions about frequency of visitation to parks, recreation areas and cultural sites in the past year, awareness of park and recreation management agencies in Arizona, satisfaction with Arizona State Parks, and respondent demographics. The telephone survey instrument was pretested, translated into Spanish and administered by a professional survey research firm using computer-aided telephone interviewing (CATI). The sample was drawn from a database of telephone area codes and exchanges and an RDD procedure was used to select individual telephone numbers. A total of 1,500 telephone interviews were completed in English (n = 1,419) and Spanish (n = 81). Calculation of final outcome rates for the RDD household survey was based on procedures outlined by the American Association for Public Opinion Research (AAOPR: 2006).

Response Rate 1 (RR1) or minimum response rate and Cooperation Rate 1 (COOP1) or minimum cooperation rate were calculated. Minimum response rate is the total number of complete interviews divided by number of interviews plus noninterviews plus all cases of unknown eligibility. The RR1 for this study was 34%. Cooperation rate is proportion of all complete interviews divided by number of complete interviews plus noninterviews where an eligible respondent was contacted. The COOP1 for this study was 80%. Thus, 80% of all households where an eligible respondent was identified agreed to participate in the study. The margin of sampling error for the telephone survey is +/− 2.5% at the 95% confidence interval.

The demographic profile of telephone survey respondents was similar to U.S. Census data for the state. Specifically, the percentage of respondents in the completed telephone sample was within +/− 3% for gender and for each racial and ethnic category with the exception of Hispanic. Fewer Hispanic respondents were in the telephone sample (15%) than expected based on the Census estimates for the state (29%). At the conclusion of the interview, each respondent was asked to participate in the mail survey and offered an incentive in the form of an Arizona State Parks admission coupon. A mail survey questionnaire was sent to 1,033 telephone survey respondents who agreed to participate, or 69% of those contacted by phone.

The mail survey followed the tailored design method (Dillman, 2000) with five contacts including the initial telephone request, survey mailing with incentive for all potential respondents, postcard reminder, second full mailing and final phone or e-mail contact. The mail survey included questions about visitation to parks, recreation areas, natural areas and historic sites in Arizona; motives for recreation participation; activity and settings preferences; information seeking behavior; leisure constraints and constraints negotiation strategies; satisfaction with Arizona State Parks; and respondent demographics. There were 439 completed and usable mail surveys returned and the RR1 for the mail survey was 45%. Nonresponse bias analysis revealed no significant differences between respondents and nonrespondents in terms of length of Arizona residency. Mail survey respondents were, however, younger than nonrespondents (47 years vs. 52 years), and fewer Hispanic respondents were in the completed mail survey sample (7%) than the telephone sample (15%). Thus, with these limitations, the completed sample was deemed to adequately reflect the state population with the notable exception of underrepresentation of Hispanic residents. The data used for the study were taken from the 439 usable mail questionnaires.

Measurement of the Constructs

The motivations or reasons for outdoor recreation participation were measured by 11 items assessing the importance of desired experiences (see Table 1). Respondents rated each item on a five-point Likert-type scale (1 = Not at all Important, 5 = Extremely Important). The items were drawn from the recreation experience preference (REP) scales (Driver et al., 1991; Manfredo et al., 1996; Moore & Driver, 2005). Manfredo et al. suggest that when using REP scales to measure motivation choosing domains and items of relevance to the specific study site and population is reasonable. A four factor solution was proposed for the measurement model with the 11 observed items serving as indicators on one of four first-order factors. The first-order factors in turn loaded on the higher order motivation factor. The first-order factors are listed in Table 1.

TABLE 1 Standardized Parameter Coefficients, Means, and Standard Deviations for Motivation, Constraint, Negotiation Strategy, Negotiation-Efficacy, and Participation Items

Second-order Factor First-order Factor Indicators	Coefficient	Mean	SD
Motivation
Achievement
Gain a sense of accomplishment	.873	2.65	1.168
Experience excitement/adventure	.551	3.25	1.164
Gain a sense of self-confidence	.786	2.61	1.213
Develop my skills and abilities	.725	2.52	1.210
Enjoy Nature
Be close to nature	.643	4.13	1.019
Observe the scenic beauty	.833	4.50	.760
Enjoy the sounds and smells of nature	.930	4.41	.867
Escape
Get away from the usual demands of life	.741	4.07	1.063
Experience solitude	.540	3.53	1.269
Socialize
Be with family or friends	.592	3.99	1.079
Be with people who share my values	.744	3.14	1.291
Constraints
Intrapersonal
Afraid of getting hurt by animals	.599	1.7692	.982
Lack of interest	.695	2.1389	1.087
Don't feel welcome	.760	1.9453	.957
Afraid of getting hurt by other people	.612	2.1521	1.109
Lack of information	.511	2.6559	1.113
Don't have the skills or physical ability	.563	2.4036	1.242
Interpersonal
Companions prefer other things	.636	2.7369	1.017
Don't have companions/people to go with	.783	2.5820	1.213
The people I know live or work too far away	.736	2.5291	1.057
Structural
Camping fees are too high	.743	2.7150	1.074
Admission fees are too high	.880	2.7338	1.184
It is too expensive	.800	2.4231	1.133
Negotiation
Changing Interpersonal Relations
Try to find people with similar interests	.660	3.3193	.985
Ask my family to share the chores	.546	3.0628	1.020
Bring other people to make me feel safer	.522	2.8324	1.120
Organize events with my own group	.531	3.1494	1.009
Changing Leisure Aspirations
Go to areas that are less crowded	.576	3.8437	.924
Go to areas where I feel comfortable	.785	4.0365	.823
Improve Finances
Try to budget money	.732	3.2840	1.011
Set aside money to use for recreation	.894	3.2617	.963
Negotiation-efficacy
In the past, I have been successful getting around the barriers to my outdoor recreation	.495	3.46	.929
People I admire find ways around challenges they face when trying to recreate	.737	3.55	.921
My family and friends encourage me to participate in outdoor recreation, even when there are obstacles	.675	3.31	.990
I enjoy overcoming obstacles to my outdoor recreation participation	.598	3.38	.959
Participation
Number of Arizona State Parks Visited in previous 12 months.	727	15.888	7.709
Outdoor recreation activity participation scale.	.481	3.294	3.369

To measure constraints, respondents rated the extent to which they agreed or disagreed with 13 statements describing conditions that may limit their outdoor recreation participation (see Table 1). A five-point scale (1 = Strongly Disagree, 5 = Strongly Agree) was used. The list of items was adapted from previous research (Hubbard & Mannell, 2001; Raymore et al., 1993). A three-factor solution was proposed for the measurement model with the 13 observed items serving as indicators on one of three first-order factors. The first-order factors in turn loaded on the higher order constraints factor. The first-order factors are listed in Table 1.

To measure negotiation, respondents were asked the level to which they agreed with eight strategies to start, continue, or increase participation in outdoor recreation (see Table 1) on a five-point scale (1 = Strongly Disagree, 5 = Strongly Agree). As with the constraints items, an initial list of negotiation statements was developed from prior research (Hubbard & Mannell, 2001; Jackson & Rucks, 1995). A three-factor solution was proposed for the measurement model with the eight observed items serving as indicators on one of three first-order factors. The first-order factors in turn loaded on the higher order negotiation factor. The first-order factors are listed in Table 1.

Four items were developed specifically for this study to measure negotiation self-efficacy (see Table 1). These items were designed to tap the four sources of self-efficacy defined by Bandura (1997): mastery experience, vicarious experience, social persuasion and psychological/emotional response. Respondents were asked their level of agreement with each statement on a five-point scale (1 = Strongly Disagree, 5 = Strongly Agree). This method of measuring negotiation self-efficacy differed from Loucks-Atkinson & Mannell (2007) where respondents were asked to rate their confidence in their ability to employ specific negotiation resources from 0% to 100%. Although that approach clearly has merits for assessing perceived self-efficacy to perform specific tasks, the intent in this study was to develop a unidimensional measure of perceived self-efficacy for the domain of negotiation. The four items served as indicators of the negotiation self-efficacy construct in the model.

Two measures of participation were calculated. To assess frequency and diversity of outdoor recreation participation, respondents reported frequency of engagement in each of 32 activities in the 12 months prior to the survey. Participation frequency was measured on an ordinal scale ranging from 0 = Never to 1 = Occasionally to 2 = Frequently.) A summative index was created for each respondent which could range from a minimum of 0 (i.e., never participated in any activities) to 64 (i.e., frequently participated in all activities). For the second measure of participation, respondents reported whether or not they had visited each of 29 Arizona State Parks in the year prior to the survey. The responses were summed so that each respondent was assigned a score for state park visits ranging from 0 to 29. Due to the truncated ordinal scales used to measure participation, summative scales were used instead of higher order factors.

Analysis

All variables in the model were screened for normality by examining skewness and kurtosis. Although significance tests are available to assess normality, in studies such as this one with samples of greater than 200 cases, variables with statistically significant skewness or kurtosis typically do not deviate enough from normality to affect the analysis and thus:

If the sample is large it is a good idea to look at the shape of the distribution instead of using formal inference tests. Because the standard errors for both skewness and kurtosis decrease with larger N, the null hypothesis is likely to be rejected with large samples when there are only minor deviations from normality” (Tabachnick & Fidell, 2000, p. 74).

The variable distributions were assessed using SPSS 14.0 though frequency histograms with superimposed normal distribution, and expected normal probability plots (P-P plots). In P-P plots, the scores are ranked and sorted and compared to an expected normal value for each case. The expected value is the z score for a case with that rank, which would hold in a normal distribution and the normal value is the z score of the actual distribution. If the actual variable distribution is normal, the cases cluster along a diagonal line in the P-P plots with minor deviations shifting some points above and below the line. Based on these analyses, all variables were deemed to be normally distributed and retained for examination using CFA. The data were screened for missing values, which accounted for less than 4% of the data for the variables used in the analysis. Regression data imputation was employed to replace missing values by an estimate called an imputed value.

The measurement model was constructed using SPSS 14.0 and AMOS 6.0. Hierarchical confirmatory factor analysis, a special case of CFA, was used to specify higher order factor structures for the constraints, motivation and negotiation constructs. For these constructs, first-order and second-order factors solutions were hypothesized to represent hierarchical relations. The hierarchical CFAs represent hypotheses that the first-order factors (e.g., structural, intrapersonal, interpersonal) measure the second-order factor (e.g., constraints).

Hierarchical CFA was not used for the negotiation-efficacy construct for several reasons. First, negotiation-efficacy has not undergone adequate theoretical specification, conceptualization or operationalization to allow for hypotheses about hierarchical relations. Second, in this study the construct was measured using four observed variables and thus did not meet the minimum requirements for hierarchical CFA. Thus negotiation-efficacy was treated as a single factor with four observed indicators. Hierarchical CFA was not used for the participation construct because the observed variables were measured on a truncated ordinal scale. Participation was treated as a single construct with two indicators as described earlier. The hierarchical CFA models for motivation, negotiation, and constraint were each assessed for fit, deemed acceptable, and added to the measurement model along with the single factor models for negotiation-efficacy and participation.

The full measurement model was then assessed for fit. The standardized parameter coefficients were significant for all first-order factors on their respective indicators and for all second-order factors on their respective first-order factors (all p-values < .001). Multiple fit indices were used including minimum sample discrepancy functions (chi-square statistic, relative chi-square), measures based on population discrepancy (RMSEA), baseline comparison measures (CFI), and goodness of fit (GFI) and parsimony adjusted goodness of fit measures (AGFI) (see Table 2). The rules of thumb provided by Arbuckle (2005) and Kline (2005) were used to assess model fit. A relative chi-square in the range of 3 to 1 indicates favorable fit between the hypothetical model and the sample data for large samples (i.e., > 200). An RMSEA value of 0.05 or less indicates a close fit to the data, a value of 0.08 indicates a reasonable fit and a value of 0.10 or greater indicates a poor fit. The CFI measure is based on discrepancy between the hypothesized model and a baseline or comparison model (i.e., the independence model). This measures ranges from 0 to 1 with values close to 1 indicating a perfect fit and values below 0.90 indicating potential for improvement. Values for GFI and AGFI of .9 indicate a good fit.

TABLE 2 Summary of Fit Indices for Measurement and Structural Models

	χ^2	df	p	χ^2/df	GFI	AGFI	CFI	RMSEA
Measurement model	1498.94	679	.001	2.20	.87	.84	.87	.05
Structural model	1517.76	681	.001	2.22	.86	.84	.86	.05

Results

A two-step procedure was used to test the study hypotheses. First, hierarchical confirmatory factor analysis was employed to develop and assess the measurement model. Once an acceptable fitting model was developed, the measurement model was modified to represent the theoretical model and tested through structural equation modeling. Maximum likelihood estimation was used to estimate both the measurement and structural models.

Measurement Model

The measurement model was accepted (see Table 3). The relative chi-square (2.20) indicated a favorable fit to the data. The RMSEA value of .05 and the 90% confidence interval around the RMSEA point estimate (.04 – .06) indicated a close fit. The values for the CFI (.87), GFI (.87) and AGFI (.84) indicated potential for some improvement in model fit. Values for CFI are lower when correlations among observed variables are low, as in the case in these data when so many individual items and constructs are being assessed. With the measurement accepted, the analysis proceeded to the structural equation model.

TABLE 3 Standardized Direct, Indirect, and Total Effects

	F3			F1			F2			F4
	D	I	T	D	I	T	D	I	T	D	I	T
F5	0	.310	.310	−.305	.021	−.284	.465	.029	.494	.09	0	.09
F1	.433	0	.433	—	—	—	—	—	—	—	—	—
F2	−.249	0	−.249	—	—	—	—	—	—	—	—	—
F4	.354	.097	.451	.291	0	.291	.391	—	.391	—	—	—
Keys: F1 = Constraints; F2 = Motivation; F3 = Negotiation-efficacy; F4 = Negotiation; F5 = Participation. D = Direct effects; I = Indirect effects; T = Total effects.

Structural Model

The structural model developed and tested in this study is presented in Figure 1. Following Lee and Scott (2006), the figure represents the parameters to be estimated by the structural model. In the figure the letter “L” represents standardized coefficients between observed (manifest) variables and unobserved (latent) factors. The letter “B” indicates standardized coefficients between higher order latent factors and first-order latent factors, and between the exogenous latent factor and the endogenous latent factors. The letter “D” represents errors of endogenous latent factors and the letter “e” represents error for each manifest variable.

The fit indices (see Table 3) indicated an acceptable fit for the structural model. The relative chi-square value (2.22) demonstrates a favorable fit to the sample data. The RMSEA value of .05 and the 90% confidence interval around the RMSEA point estimate (.04–.06) indicate a close fit. The values for the CFI (.86), GFI (.86), and AGFI (.84) indicated potential for improvement in model fit. Again, values for CFI are lower when correlations among some observed variables are low.

Hypothesis Tests

Figure 2 shows the standardized coefficients for the paths between higher order latent factors and first-order latent factors and between the exogenous latent factor and the endogenous latent factors. (The standardized coefficients between latent factors and observed indicators are displayed in Table 1). The data supported all the Hypotheses except for Hypothesis 2: the coefficient was in the predicted direction, but the path was not significant (p = 45). The paths for Hypotheses 4 thru 7 were positive and statistically significant (p < .001).

FIGURE 2 Structural model of leisure constraints negotiation with standardized parameters estimates.

In summary, the results of the hypothesis tests provide substantial evidence to conclude that the data supported nearly all aspects of the structural model, with the exception of a path from negotiation to participation that was in the predicted direction but was not significant.

Analysis of Direct, Indirect and Total Effects

Table 3 presents the direct, indirect and total effects for the exogenous factor on endogenous factors and the direct, indirect, and total effects for all mediating endogenous factors on the dependent endogenous factor. Overall, the table shows that most effects identified in this study were direct effects, although there were some indirect effects.

Most notably negotiation-efficacy exhibited a weak to moderate positive influence on participation with a total indirect effect of .310. This indirect effect of negation efficacy on participation was mediated through motivation (indirect effect of .201 = .433 × .465), constraints (indirect effect of .076 = −.249 × −.305), and negotiation (indirect effect of .026 = .354 × .073). Thus, most of the indirect effect of negotiation-efficacy on participation is attributable to the path from negotiation-efficacy through motivation to participation. Whether part of the relationship between negotiation-efficacy and participation is accounted for by the relationship each has with motivation cannot be determined by examining the path coefficients alone. When paths on both sides of the hypothesized mediator are significant, as is the case, Sobel (1986) recommended using the multivariate delta method to determine whether mediation occurred. Tests for significance of the mediation effects provide partial support for one mediated path. The indirect effect of negotiation-efficacy through motivation on participation was not statistically significant using the delta method and t-test (p < .05, t = 1.491). The relationship, however, was significant at p < .10 and thus it may be said to be “trending” toward significance. The other indirect effects of negotiation-efficacy on participation (through constraints and negotiation) were not significant.

Discussion

Research on constraints negotiation provides important insights into people's leisure choices and behavior. Researchers have identified constraints as factors that inhibit leisure activity participation or diminish satisfaction (Jackson, 1988), organized constraints into conceptual categories (Crawford & Godbey, 1987), and developed theoretical propositions to guide empirical research (Jackson et al., 1993). Researchers also identified negotiation strategies and resources people use to overcome constraints (Jackson & Rucks, 1995; Kay & Jackson, 1991; Scott, 1991). Recent studies have placed constraints negotiation in a broader context by examining the relations between leisure constraints, negotiation, motivation, and activity participation (Alexandris et al., 2002; Hubbard & Mannell, 2001). Several researchers proposed that self efficacy may provide additional insight into the constraints negotiation process (Henderson et al., 1995; Hubbard & Mannell, 2001), which led to the introduction of the idea of negotiation-efficacy (Loucks-Atkinson & Mannell, 2007).

This study contributes to this research by examining the interaction of motivation, constraints, negotiation, and negotiation-efficacy and their effects on outdoor recreation participation. Results provided partial support for the proposed theoretical model and showed that constraints negatively influenced outdoor recreation participation and positively affected negotiation. Negotiation exhibited a small positive impact on participation, but contrary to the study's hypothesis the relationship was not significant. Motivation, in the form of desire for satisfying recreation experiences, had a relatively strong positive impact on outdoor recreation participation and was the most important single predictor of participation. Similarly, motivation exhibited a positive effect on negation, suggesting that greater desire for satisfying recreation experiences also triggers the use of negotiation resources and strategies. Negotiation-efficacy positively influenced motivation to participate and negotiation efforts and negatively influenced perceived constraints. Finally the results provide partial support for the notion that negotiation-efficacy exhibits an indirect effect on participation mediated through motivation.

The results of this study provide support and clarification of a conceptual model of leisure constraints negotiation reported elsewhere (Hubbard & Mannell, 2001; Loucks-Atkinson & Mannell, 2007). The results support the notion proposed by the constraints-effects-mitigation model that “encountering constraints appears to set in motion two opposing forces – an inhibitory influence on participation stemming directly from the constraints, and a facilitative influence resulting from the negotiation efforts triggered” (Hubbard & Mannell, p. 158). In this study constraints exhibited a relatively strong direct negative influence on participation and moderately strong direct positive influence on negotiation.

The results suggest a central role for motivation than in the constraints negotiation process. Motivation was both a moderately strong direct antecedent encouraging participation and a positive influence on efforts to negotiate. This finding is consistent with Jackson et al.'s (1993) balance proposition, which stated that the initiation and outcome of the negotiation process depends on the relative strength of and interaction between motivation and constraints. The importance of motivation as a direct antecedent of participation contradicts Hubbard and Mannell's (2001) finding that motivation's primary effect on participation was indirect and mediated by negotiation. The results of this study, along with findings from Loucks-Atkinson and Mannell (2007), provide mounting evidence that motivation exerts influence as an immediate antecedent of participation as well as a potential trigger for encouraging the constraint negotiation process.

The results provide support for the inclusion of a negotiation-efficacy construct in the constraints negotiation process. As predicted, people with higher perceptions of negotiation-efficacy were more motivated to participate in outdoor recreation. This finding is consistent with predictions based on social cognitive theory. Self-efficacy is associated with causal attribution, outcome expectations and goals (Bandura, 1994). Thus, people with higher negotiation-efficacy would be more likely to attribute the causes of their leisure behavior to their own individual efforts to negotiate constraints and ability to persevere in the face of challenges. Thus, this study provides support for Loucks-Atkinson and Mannell's (2007, p. 34) suggestion to add a seventh proposition to Jackson et al.'s (1993) list. In the current study, negotiation-efficacy also exhibited a direct negative influence on constraints and a direct positive influence on negotiation, suggesting that people with higher levels of negotiation-efficacy not only engage in greater efforts to negotiate, but also they perceive fewer constraints. Proposition 7 could be revised to state: “The greater people's confidence in the successful use of negotiation resources to cope with constraints, the greater the motivation, the greater the effort to negotiate, the lesser the perception of constraints, and the higher the level of participation.”

Additional research is necessary to address measurement issues especially for the negotiation-efficacy construct. Due to disciplinary differences and the domain-specific nature of self-efficacy, a variety of related constructs in the literature including self-efficacy to exercise, self-efficacy to recreate, self-efficacy to overcome barriers to physical exercise, and now negotiation-efficacy exist. Future research might compare the items developed here, for instance, with an eight-item scale measuring self-efficacy about physical activity developed by Motl et al. (2000). That scale has been subjected to rigorous tests of factorial validity and invariance, but in the context of more general leisure time physical activity. Further, there remain opportunities for researchers to explore the theoretical linkages between perceived self-efficacy and perceived behavioral control (Motl et al., 2005) and to explore the collective negotiation self-efficacy of groups by gender, race and ethnicity.

Continued theoretical development of the leisure constraint negotiation model presented here will also require additional studies of the factorial validity of negotiation strategies. Despite early findings by Jackson and Rucks (1995) that negotiation strategies tend to be consistent with the types of constraints encountered (e.g., a person whose leisure is constrained by lack of skills would employ skill acquisition as a negotiation strategy), little effort has been focused on developing categories of negotiation strategies that parallel intrapersonal, interpersonal and structural constraints. In a qualitative study Daniels, Drogin Rodgers, and Wiggins (2005) suggested there is not necessarily a clear matchup. A structural constraint was found frequently to be negotiated through interpersonal or intrapersonal means.

One limitation of my study is the use of participation as a dependent variable in the analysis, which has been critiqued by both supporters and critics of constraints research. This limitation is partially mitigated because multiple measures of participation were used including a measure tapping not only frequency of participation but also diversity of participation. Regardless, future researchers might consider alternative dependent variables that tap affective and cognitive dimensions of leisure in addition to the behavioral. A second limitation was the cross sectional survey design. Although this approach is useful for analyzing relationships among variables in the completed sample, sample statistics could be produced that are poor estimates of the true population parameters. Finally, the under-representation of Hispanic residents was a limitation.

This study makes several significant contributions to the literature. First, by using hierarchical confirmatory factor analysis and structural equation modeling this study presented an empirical test of the constraints negotiation that examined the relations among motivation, constraints, negotiation, negotiation-efficacy and leisure activity participation. Second, the study assessed the constraints negotiation process in a statewide random sample of Arizona residents, which gave additional weight to the generalizability of the processes identified here. Finally, the study supports the utility of social cognitive theory and the negotiation-efficacy construct for explaining negotiation. Future research will determine how universal the processes identified in this study are in other contexts and with other populations.

Acknowledgements

This research was supported by grant ISA 03-220 from Arizona State Parks. The author thanks Elizabeth Krug from Arizona State Parks and Randy J. Virden, Amy C. Racki and Rebecca M. B. White from Arizona State University for their assistance.