Community Psychology, Political Efficacy, and Trust

1 downloads 0 Views 259KB Size Report
ments (Pretty, 1990), union participation (Catano, Pretty, Southwell, & Cole, 1993), and support systems for adolescents ...... Simon and Schuster. Rosenfeld, R.
Political Psychology, Vol. 31, No. 1, 2010 doi: 10.1111/j.1467-9221.2009.00734.x

Community Psychology, Political Efficacy, and Trust Mary R. Anderson University of Tampa

Much attention has been paid to the effects of political efficacy and trust on political participation. Most studies tend to use efficacy or trust as an independent variable to explain political actions such as voting, campaign involvement, and the like. Despite their importance in explanations of political behavior, relatively little is known regarding mechanisms through which social involvement may influence trust and efficacy. If efficacy and trust are of value, then it is important that we determine how their development can be fostered, and especially whether their development can be promoted through social interaction—such as a sense of community. Borrowing from the field of community psychology, I employ the Sense of Community Index to provide a more nuanced measure of community based on individual perceptions of their community that previous studies were unable to capture. Analyzing original survey data, this paper examines to what extent, if any, a sense of community matters for trust and efficacy. The results demonstrate that social forces, such as community, exert positive and significant effects on internal and external efficacy and personal and political trust, independently of individual traits such as income, age, gender, and education. KEY WORDS: Community psychology, Trust, Efficacy, Political behavior

Much attention has been paid to the effects of political efficacy and trust on political participation (Abramson, 1983; Bennett, 1986; Brehm & Rahn, 1997; Hetherington, 1998; Rosenstone & Hansen, 1993; Uslaner, 2002; and Verba, Schlozman, & Brady, 1995, for example). Most studies tend to use efficacy or trust as an independent variable to explain political actions such as voting, campaign involvement, and the like. The logic underlying these effects is that individuals are empowered and motivated when they believe that their involvement in politics will be consequential and that they can have confidence that the behavior of others will be honorable. A current theoretical perspective among those who study urban social organization is collective efficacy theory. Broadly speaking, the theory suggests (and 59 0162-895X © 2009 International Society of Political Psychology Published by Wiley Periodicals, Inc., 350 Main Street, Malden, MA 02148, USA, 9600 Garsington Road, Oxford, OX4 2DQ, and PO Box 378 Carlton South, 3053 Victoria Australia

60

Anderson

empirical evidence substantiates the claim) that “the prevalence and density of kinship, friendship, and acquaintanceship networks and the level of participation in community based organizations fosters the emergence of collective efficacy, or solidarity and mutual trust (social cohesion) among community residents combined with shared expectations for social control-related action” (Browning, Feinberg, & Dietz, 2004, pp. 506–507). While these studies have been directed at issues of urban policy (Morenoff, Sampson, & Raudenbush, 2001; Rosenfeld, Messner, & Baumer, 2001; Sampson, Morenoff, & Earls, 1999; Sampson, Raudenbush, & Earls, 1997) and health behavior (Browning & Cagney, 2002), there is no reason to believe that collective efficacy theory would not be applicable to studies of political behavior. The implication therefore is that the causal direction specified here flows from sense of community to trust and efficacy. Previous research demonstrates a strong correlation between both efficacy (external and internal) and trust (personal and political) and political behavior. However, despite the fact that these variables contribute to explaining things like voting and campaign involvement (for a discussion of efficacy, see Abramson, 1983; Bennett, 1986; Craig, Niemi, & Silver, 1990; Niemi, Craig, & Mattei, 1991; for a discussion of trust, see Brehm & Rahn, 1997; Hetherington, 1998; Uslaner, 2004, 2002), relatively little is known about the social forces that influence them. Over the last decade, political science research has returned to the Columbia idea that individuals should not be studied in isolation, rather we should pay attention to social forces potentially operating on the individual. Social capital research, for example, demonstrates that social interaction offers an array of resources that can be of benefit at both the individual and collective level (Claibourne & Martin, 2000; La Due Lake & Huckfeldt, 1998; Putnam, 1993, 1995; Stolle, 1998, for example). Context research demonstrates that the physical contexts in which we are embedded impose significant parameters in terms of the type(s) of information we receive (Huckfeldt, Johnson, & Sprague, 2004; Huckfeldt, Plutzer, & Sprague, 1993; Huckfeldt & Sprague, 1995, for example). Despite the importance of efficacy and trust in explanations of political behavior, relatively little is known regarding mechanisms through which social involvement may influence them. If efficacy and trust are of value, then it is important that we determine how their development can be fostered, and especially whether their development can be promoted through social interaction—such as a sense of community. Social forces—where and with whom we work, live, socialize, and worship—play a crucial role in determining many of the choices we make, including our level of political involvement and political attitudes. Previous research has largely neglected to consider whether social forces, such as community, constitute antecedents of trust and efficacy (a notable exception is Uslaner [2002], who devotes considerable time discussing the roots of trust). This omission is unfortunate. Efficacy and trust are important constructs in themselves, and thus evidence that sense of community influences trust and/or efficacy

Trust and Efficacy

61

would be of considerable substantive significance. This significance is magnified, however, when we recall the critical roles efficacy and trust play as forces affecting a broad array of political behaviors. Any impact of sense of community identified here would suggest an indirect effect on the many factors known to be influenced by efficacy and trust. Theoretically it is quite reasonable to assume that social forces should play a role in influencing levels of efficacy and trust. Context research and social capital research highlights the potential importance of social contexts (i.e., the community) on political behavior and attitudes—those such as trust and efficacy for example. It is entirely rational, therefore, to hypothesize those social forces—such as community—will affect efficacy (internal and external) and trust (social and political). Sense of community at its very core suggests collaboration. Central to the completion of any collective effort—the likes of which church groups, service organizations, and workplace environments undertake—is cooperation. One of the cornerstones of building trust is cooperation (Putnam, 2000), thus the greater one’s sense of community the more likely they are to be trusting. Second, sense of community is built on relationships. Those who are successful in building relationships with others and in influencing the opinions of fellow members, coworkers, or neighbors might be encouraged to believe that they can also be influential in the political arena. In other words, those with higher levels of sense of community may be more likely to have strong feelings of efficacy. Although the theoretical framework outlined here posits that sense of community influences efficacy and trust, alternate causal connections are possible. The most likely alternate scenario is that any identified relationship is spurious, tracing not to the influence of sense of community on trust and efficacy, but rather to some common underlying force that influences all three of these variables. Given the construction of the sense of community index, which I describe in detail below, a less likely scenario is that efficacy and trust influence sense of community. As is always the case in cross-sectional analyses, there is no definitive means to exclude these possibilities. However, cognizant of these concerns, I included in the survey several items designed to capture an important array of individual-level traits; the Big Five personality characteristics. Inclusion of the Big Five as control variables does much to allay concerns about spuriousness because the variables account for the most obvious forces—extroversion, warmth and agreeableness, conscientiousness, and so on—that might incline an individual both to feel efficacious and trustworthy and to develop a strong sense of community. Personality is important as a control as it is likely to play a role in how an individual views his/her community.1 1

The Big Five personality traits are openness to experience, conscientiousness, extroversion, emotional stability, and agreeableness. These are included mainly as control variables to account for personality affects that may be related to various types of political behaviors. It is entirely plausible, even likely, that personality plays a large role in many types of political behaviors and attitudes, including levels of trust and efficacy. For a detailed discussion of personality and political behavior, see Mondak and Halperin (2008).

62

Anderson

Sense of Community Implicitly it has been suggested that community matters for various types of political behavior, yet why it matters remains a mystery. One possible reason for this mystery may lie in the construct of our measures for community. For the most part, measures of community have often been quite coarse; studies often used indicators such as length of residence and home ownership to capture community; at best these can really only serve as a proxy for community connectedness and fail to capture how community comes to matter. Sense of community as described in the community psychology literature can give us a more nuanced measure of community based on individual perceptions of their community that previous studies were unable to capture. Therefore with a more intricate measure of sense of community, we can begin to understand how and why social forces such as community matter for all sorts of political behaviors and attitudes. A sense of community is akin to a feeling of belonging. Borrowing from the field of community psychology where sense of community has a long and distinguished history, I argue that the concept of sense of community is useful to the study of political behavior and attitudes because it captures individuals’ perceptions of their social contexts. Specifically, sense of community is defined as “a feeling that members have of belonging, a feeling that members matter to one another and to the group, and a shared faith that members’ needs will be met through their commitment together” (McMillan & Chavis, 1986, p. 9). McMillan and Chavis (1986) proposed that sense of community is composed of four elements: (1) membership, (2) influence, (3) integration and fulfillment of needs, and (4) shared emotional connection. The first element, membership, creates a sense of belonging and identification and creates boundaries: there are those who belong and those who do not. For example, individuals who belong to a particular association such as the Rotary Club are known as Rotarians and those who do not belong to the group are not Rotarians. Yet, the “us” and “them” division may not be as clear cut as in the Rotarian example described above. It may be “as subtle as to be recognized by only the members themselves” (McMillan & Chavis, 1986, p. 10). The second element, influence, is a sense of mattering. For instance, when a member of a group believes that she can make a difference to the group, that she matters to the group, and that the group matters to the member—this entails a reciprocal relationship. The third element is integration and fulfillment of needs. This refers to the feeling that members’ needs will be met by their membership in the group and that there are shared values among group members; for example, members of a particular church typically hold similar beliefs and values. The final element is shared emotional connection. This element is based, in part, on a shared history or an identification of shared events; an example would be individuals who belong to a cancer survivors group, all the individuals involved in the group share a connection through a similar experience, namely surviving cancer (for a full discussion

Trust and Efficacy

63

of the elements of sense of community, see McMillan & Chavis, 1986). These elements combine to create a sense of community or community connectedness. Note that community as described here extends beyond a simple geographic locale and can include various other contexts such as the workplace, voluntary associations, churches, and the like. In order to measure an individual’s sense of community, McMillan and Chavis (1986) developed what has come to be referred to as the Sense of Community Index, a 12-item True/False questionnaire that taps into the four elements of sense of community.2 Over time the index has been adapted in format to include a 5-point Likert-type response. Using the same statements as the original True/False battery, individuals are asked to respond based on how much they agree or disagree with the statements, ranging from strongly disagree to strongly agree, “a modification that is likely to result in greater sensitivity in representing ‘true’ perceptions of the social connections, mutual concerns and community values” within the respective contexts (Long & Perkins, 2003, p. 291). Here, a similar version of the SCI will be employed to gauge levels of connectedness in various types of social contexts such as the workplace, place of worship, associations, neighborhood, and informal settings such as a social network (circle of friends). In the following pages I will (a) discuss briefly the literature on efficacy and trust and (b) demonstrate how social factors, such as sense of community, affect efficacy and trust. Efficacy Political efficacy has been widely used to explain various types of political activities such as voting, campaign involvement, signing petitions, and the like. To a large extent, much of the recent research on political efficacy itself has focused on how to correctly measure the concept. Scholars generally agree that it is a concept with two distinct components (see, for example, Craig et al., 1990; Niemi et al., 1991). Today, scholars generally agree that political efficacy includes: (1) internal efficacy—beliefs about one’s own ability to influence the political process—and (2) external efficacy—beliefs about the responsiveness of government officials to the concerns of the citizenry (Balch, 1974; Coleman & Davis, 1976; Converse, 1972; Craig et al., 1990; Niemi et al., 1991). While most of the 2

There exist other measures of sense of community in the literature. However, the McMillan and Chavis SCI is the most used and broadly validated measure of SOC (Chipuer & Pretty, 1999). Support for the reliability and validity of the Sense of Community Index is well documented in the community psychology literature, and the scale has been used to examine a wide variety of communities. Although its most common use has been in the neighborhood context (e.g., Perkins, Florin, Rich, Wandersman, & Chavis et al., 1990), it has also been employed in studies dealing with workplaces (Pretty & McCarthy, 1991), support and demand characteristics of college students’ social environments (Pretty, 1990), union participation (Catano, Pretty, Southwell, & Cole, 1993), and support systems for adolescents (Pretty, Andrews, & Collett, 1994; Pretty, Conroy, Dugay, Fowler, & Williams, 1996). Because it can be adapted to different types of communities, it suits the purpose of this research quite well.

64

Anderson

research on political efficacy has focused on how to measure the concept and its impact on political participation (Abramson, 1983; Bennett, 1986; Rudolph, Gangl, & Stevens, 2000; Verba, Burns, & Schlozman, 1997; Verba et al., 1995), little research has considered the impact of social forces on the development of political efficacy.3 There are several reasons to expect that sense of community will promote feelings of efficacy. First, experience at functioning meaningfully and effectively within a given community may signal to individuals that their capacity to exert influence extends outside of the context at hand, such as to the larger political environment. Even though such activity may not be explicitly political in nature, any success at the group level may engender in individuals the belief that they have the capacity to be influential. Second, sheer strength in numbers—whether because people in a community are genuinely acting in concert or merely because individuals recognize that others share their views—should fuel efficacy. Third, past political activity within many contexts provides empirical evidence of efficacious behavior. For example, members of voluntary associations and neighborhood groups often tackle political issues. By doing so, their members may gain confidence that they as individuals can be politically efficacious. Verba et al. (1995) suggest that among the various factors that shape participation, resources such as civic skills—those that make it easier for individuals to become involved, such as communication and organizational skills—are critical. Involvement beyond simply membership in the workplace, organizations, or church—something such as a sense of community—helps to build those skills. Those who take on leadership roles in their church or their workplace for matters such as fundraising or party planning are cultivating the skills that Verba et al. (1995) argue are important for participation. Therefore, it is entirely logical to hypothesize that sense of community could influence levels of internal efficacy. An individual who has had success in influencing others in his/her workplace, church, neighborhood, and so forth is also likely to believe s/he has the power to influence government. I also expect that a sense of community will come to matter for external efficacy—the belief that government officials actually listen to and care about what citizens have to say. Individuals who are part of a group or context in which they have a high level of sense of community are likely to believe that the government will listen to their concerns, especially when they present a unified voice to relay the message. Secondly, sense of community should affect external efficacy simply because of the belief that there is strength in numbers, that with more people behind an idea (i.e., a chorus as opposed to a soloist), the more likely that idea is 3

Some efforts (although dated) have been made to examine group consciousness as a factor influencing efficacy (Gamson, 1971; Gurin & Epps, 1975; Gurin, Gurin, & Beattie, 1969; Shingles, 1981). Verba et al. (1995) speak to the idea in terms of individual-level traits influencing “roots in the community” and therefore concern about local issues. Most recently, Rudolph, Gangl, and Stevens (2000) conducted research on emotions, efficacy (particularly anxiety), and campaign involvement.

Trust and Efficacy

65

to have a positive outcome in their favor. Thus, based on previous research that suggests community matters for behaviors such as efficacy, sense of community should be strong predictors of both internal efficacy and external efficacy. Political Trust It is generally agreed that when it comes to examining trust as a predictor of political participation, there are two separate concepts: (1) trust in government, often referred to as confidence in government or political trust, and (2) trust in others, conveyed a number of different ways such as interpersonal trust (Brehm & Rahn, 1997), social trust (Putnam, 2000), or generalized trust (Uslaner, 2002). Hetherington (1998) defines political trust as “a basic evaluative orientation toward the government founded on how well the government is operating according to people’s normative expectations” (p. 791). Interpersonal trust on the other hand relies on trusting other people in a way that is very different from trusting government. It requires giving people—even those who we may know very little about— the benefit of the doubt (Putnam, 2000; Uslaner, 2002). Uslaner (2002) and Putnam (2000) strongly encourage that these concepts be kept distinct from one another because “they simply are not the same thing” (p. 137, emphasis in original). Political trust may be a consequence of interpersonal trust or visa versa, and they may be correlated with one another; however, they capture two very distinct concepts and therefore should be treated as such. Aside from Uslaner (2002), who devotes a great deal of time to examining the roots of trust, little else has focused directly on the social factors that may influence interpersonal trust and political trust.4 Like efficacy, there are multiple reasons to believe that sense of community will matter for trust. First, sense of community builds relationships; those who belong to a workplace, organization, church, and so on will typically build relationships with those around them. Even at the most basic level, that of an acquaintance, a relationship probably exists simply because members share a common interest in their job, faith, or pastime. Relationships at this most basic level involve some sort of trust, even if it simply means that you trust that the other person in your group shares your interest. Second, and most importantly, sense of community should matter for trust because sense of community entails cooperation. Brehm and Rahn (1997) and Putnam (2000) both suggest that cooperation leads to trust. Those individuals who have a greater sense of community are likely to be involved in activities that require cooperation. Indeed, collaborative effort is central to the very concept of sense of community. Thus, it is quite reasonable, based both on logic and on previous research, that sense of community should affect trust. 4

Brehm and Rahn (1997) demonstrate that “experience with crime and fear of walking in their neighborhood at night undermine trust in others” (p. 1018) and that age may also play a factor. Stolle (1998) provides mixed results on the relationship between voluntary associations and generalized trust.

66

Anderson

Now, as Putnam (2000), Uslaner (2002), and others argue, it is important to distinguish between interpersonal trust—trust in others—and political trust— confidence in government. I argue that sense of community is likely to have stronger effects on interpersonal trust than political trust because of the social nature by which sense of community is developed. Sense of community emerges when the individual has positive bonding experiences within some context. Hence, it is highly plausible that a by-product of this bond with the context will promote similarly positive feelings toward the other individuals in the context with whom the person interacts. More specifically, I expect that sense of community will contribute to trust in others (i.e., personal trust). Additionally, sense of community is also likely to have effects on political trust, because having a sense of community within an organized context such as a church, the workplace, or an organization may also lead one to believe that in formal organized structures (such as government institutions) a firm set of rules and procedures ultimately leads to the best outcome—just as it possibly does in their own context. Therefore, it is reasonable to hypothesize that those who have a greater sense of community are likely to trust both others and government to do what is right. Data Data used for the following analyses are original survey data collected via telephone interviews during the month of October 2004. There were 820 respondents. The survey took approximately 25 minutes to complete.5 The core of the survey was the battery of questions directly related to sense of community. Additionally, there were numerous other items intended to provide data on a wide array of dependent and independent variables related to political behavior. The survey had a completion rate of 58.8%. The survey selection site provided an excellent setting in which to implement a survey of this type for a number of reasons. Although it had some limitations (as does any selection site), the advantages override those concerns and make this particular city a study site that is comparable to many other cities in the United States. As with Huckfeldt and Sprague’s (1995) choice of South Bend as the focal point of their study, the fact that I resided in this city during the time in which the survey was in the field and know it well were important considerations. But this city also possesses diversity of the sort essential for pursuit of my central research questions. It has many different neighborhoods and a varied collection of churches and associations. The area is also diverse in terms of demographics.6 Of course, comparable claims could be made about virtually all cities in the United States. This is to the good. The diversity present in this city is of analytical importance, 5 6

See Appendix C for survey methodology and dispositions. In terms of the area’s demographic characteristics, the survey site has a population of 284,539. 77% reside in urban communities (U.S. Census Bureau, 2000). The per capita income is $19,990. Roughly 48% of the population is male, and the median age is 30.

Trust and Efficacy

67

but the fact that similar diversity, albeit of varying levels, can be found elsewhere means that there is no reason whatsoever to expect the general patterns identified in the following analyses to be unique to the selected survey site. Analysis and Results The social factors that potentially contribute to efficacy and trust have been discussed in the literature, yet few studies have tested empirically to what extent (if any) social forces such as sense of community influence efficacy and trust. As previous scholars have noted, both these concepts contain two very distinct components. In the analyses that follow, I will examine the impact that sense of community has on internal efficacy, external efficacy, personal trust, and political trust. I will discuss how the measures for each of the key variables were constructed and then discuss the results of the analysis. Sense of Community. Respondents in this survey were asked about their sense of community in five separate contexts: workplace, organizations, church, neighborhood, and circle of friends (an informal context). Appendix A contains the survey items used to measure sense of community. A variable defined as total sense of community is a summary measure of the respondent’s sense of community scores across all contexts for which they offered answers and is used in the analyses that follow. Total scores range from 11 to 275.7 A table containing the descriptive statistics for the Sense of Community Index—from here on referred to as the SCI—and other key variables can be found in Appendix B.8 Efficacy. Respondents were asked two efficacy questions on the survey meant to measure both internal and external efficacy. The first question addressing internal efficacy asked: How much of a difference do you believe you can make in [city name]? Do you believe that you can make a big difference, a moderate difference, a small difference, or no difference at all? For external efficacy, respondents were asked: How much do you believe your local representatives (such as county commissioners and city council members) care about what you think is important for [city name]? Do you think they care very much, somewhat, only a little, or not at all? 7

8

Some respondents may have scores in fewer than the five contexts in which they were asked about because some individuals may not belong to a church or an organization, in which case they would have opted out of the battery of questions associated with those contexts. The SCI used in this survey was adapted to contain only 11 items; correlational tests conducted on a pilot study reveals that the 11-item battery is highly correlated with the 12-item battery at .989.

68

Anderson Table 1. Sense of Community and Internal Efficacy: Estimated via Ordered Logit

Variable Total Sense of Community Summary Context (No. of Contexts) Total Sense of Community X No. of Contexts Gender Age Age Squared Education Strength of Party Income Marital Status Employment Status Agreeableness Emotional Stability Conscientiousness Extroversion Openness to Experience # of observations (pseudo R2) Log likelihood

Baseline Model Coefficient (se)

Full Model Coefficient (se)

-.085 (.159) .095 (.026)*** -.001 (.000)*** .045 (.045) .16 (.083)# -.004 (.059) -.132 (.112) -.072 (.191) .016 (.022) -.046 (.018)** .013 (.017) .026 (.014)# .030 (.020) 641 (.033) -772.3

.022 (.011)** -.529 (.405) -.001 (.002) -.027 (.161) .074 (.026)** -.000 (.000)** .038 (.045) .147 (.083)# -.041 (.059) -.085 (.114) .005 (.196) .005 (.022) -.038 (.017)* .009 (.017) .017 (.014) .026 (.020) 641 (.05) -760.5

# = p < .10, * = p < .05, **p < .01, ***p < .001

To explore whether sense of community influences internal and external efficacy, I first model efficacy as a function of a full array of demographic variables and other individual level characteristics such as personality. I then add measures of sense of community to the mix to determine whether doing so furthers our understanding of the sources of efficacy. Internal Efficacy. The results for the internal efficacy model are displayed in Table 1. Several interesting findings emerge in these models. First, in the baseline model age, strength of party, emotional stability, and extroversion are significant predictors of internal efficacy.9 In the full model, when social forces, such as sense of community, are introduced, we see strong, significant effects on internal efficacy, beyond age, strength of party, and emotional stability. Recall that the variable total sense of community is the sum of the respondent’s sense of community scores across multiple contexts, and the summary variable for number of contexts is simply the number of contexts for which an individual had a sense of community score. The introduction of the sense of community variable into the model has strong positive effects on internal efficacy. 9

Two of the items used to measure sense of community, items 7 and 8, speak directly to the perceived capacities of the individual. Because these items arguably capture a facet of efficacy, I removed those two items and ran correlations between the index with and without those items. The indices are correlated at .996. Additionally, I ran a second set of models with sense of community constructed with these items omitted. Precisely the same pattern of results emerged bringing added confidence to the causal account presented here.

Trust and Efficacy

69

As the table depicts, positive results emerge for the sense of community variable; however, there is a negative coefficient for the summary variable for the number of contexts. These two variables—total sense of community and the summary variable for number of contexts—must be discussed in tandem because they are interrelated. For instance, it is impossible for an individual to have the maximum sense of community score (which is 275) if she has a sense of community score in fewer than five contexts. Individuals were asked about their sense of community in multiple contexts (a total of five) yet a given respondent may not belong to a church or a local organization, and therefore not be eligible to have a sense of community score in those contexts. Thus, we need to consider the sense of community and sum of contexts variables together when discussing the meaning of the results, hence, the full model also includes an interaction term for total sense of community X number of contexts. My first consideration regarding the insignificant coefficient for the interaction term is that it is likely the result of multicollinearity. I employed joint-F tests to substantiate my suspicions and the results verify a joint significance of the number of contexts and sense of community interaction.10 Because the coefficients are derived from ordered logit models which are typically confusing to interpret substantively, I have generated a few predicted probabilities to help clarify the substantive meaning of the results in column two. Figure 1 is a graphical representation of the predicted probabilities for sense of community by number of contexts. It displays estimates of the joint effects of sense of community and number of contexts on internal efficacy. Similar figures will be presented below for the other three dependent variables. In each figure, all other predictors are held constant at their mean values with only number of contexts, sense of community score, and the interaction term allowed to vary. The horizontal axis indicates the number of contexts in which a person is involved; scale values range from two to five (all respondents answered at least two of five context batteries). The vertical axis is the estimated likelihood of high efficacy; that is, this is the estimate that a person will have selected the top response category of the four available. Each graph will include three lines which capture variance in levels of sense of community. Again, the critical point to keep in mind is that minimum and maximum values on the sense of community scale are partly determined by the number of contexts in which the person is involved. Specifically, the minimum value adds 11 points per context and the maximum value adds 55 points per context. One line in each figure will reflect effects for respondents with minimum levels of sense of community; i.e., 10

Berry, Esarey, and Rubin (2007) argue that a statistically significant interaction term is neither a necessary nor sufficient condition to demonstrate that a substantively meaningful interaction effect exists (p. 4) and that the sign of the coefficient may give a misleading signal about the relationship between the product term and the dependent variable (p. 19). In cases such as these, the authors recommend generating and plotting predicted probabilities such as those that are found here in Figures 1–4.

70

Anderson “I believe I can make a big difference.” 1.0

Predicted Probability

.8

.6

Sense of Community

.4

average .2

maximum minimum

0.0 2.00

3.00

4.00

5.00

Number of Contexts

Figure 1. Internal Efficacy Model—Graphical Representation of Change in Predicted Probability for Sense of Community (SOC) at Three Different Levels.

sense of community scores that range from 11 to 55 as we move across the horizontal axis from two contexts to five. A second line will report estimated efficacy for individuals with average levels of sense of community. The third line will reflect effects for maximum sense of community, or values increasing from 55 to 275 across the horizontal axis. Three features of the depicted results will warrant assessment. First, if the three lines slope roughly parallel to one another, this would indicate a general effect of contextual involvement, controlling for sense of community. For example, if all three lines slope upward from left to right, this would suggest that being involved in more contexts generates more efficacy, irrespective of whether the respondent has a high or a low sense of community. Second, the gaps between the three lines will indicate the general effects of sense of community. For example, if there is a large gap between the lines for high and low sense of community, this would mean that sense of community is associated in a positive manner with efficacy. Third, if the slopes of the three lines change relative to one another across the horizontal axis (i.e., the lines are distinctly not parallel), this would reflect a unique joint effect of sense of community and number of contexts. For example, if the lines for high and low sense of community intersect where number of contexts equals two, but diverge widely where number of contexts equals five, this would suggest that efficacy is produced primarily where people have high levels of social attachment within multiple contexts. Figure 1 highlights two key results. First, the slope of the lines are nonparallel, in fact, they are opposite, suggesting that sense of community affects those who

Trust and Efficacy

71

are more and less attached to various contexts differently. Second, there is a large gap between the lines for minimum sense of community and maximum sense of community signaling that sense of community is positively associated with internal efficacy. For instance, for an individual involved in two contexts, and who receives the minimum possible sense of community score, 22, the estimated likelihood of a high level of efficacy is only 0.03. In contrast, were this individual to have the highest possible sense of community score given involvement in two contexts, a score of 110, the estimated likelihood of high efficacy is 0.12. The estimated probability of high efficacy for individuals with the lowest and highest possible levels of sense of community are 0.01 and 0.22 given involvement in three contexts, and 0.01 and 0.54 for involvement in five contexts. Regardless of the number of contexts in which an individual is involved, the higher her sense of community the greater the probability of having positive feelings of internal efficacy. But what about the average person, the person who has an average sense of community score? If we look at the observed mean by the number of contexts in which one is involved, a clear positive pattern emerges for internal efficacy. For an individual with average sense of community, the result of joining multiple contexts results in positive net effects, and the predicted probability of high efficacy increases moderately from 0.05 to 0.09. For instance, having an average sense of community in two contexts results in a probability of 0.05 for strong feelings of internal efficacy; it increases to 0.06 for three contexts; and peaks at .09 for four and five contexts. Thus, sense of community matters most for internal efficacy for those individuals with high levels of attachment. Being involved in more contexts is better only when you are at least moderately attached in those contexts and best if your level of attachment is high and beyond three contexts you level off in the likelihood that you will feel strongly efficacious. If you are involved in multiple contexts and have a low sense of community, then the results are actually modestly negative for every context added. External Efficacy. The results for external efficacy are displayed in Table 2. Again, an interesting pattern emerges. While the total sense of community variable fails to reach significance, the culprit is the addition of the interaction term to the model and the issue of multicollinearity. When the model is run without the interaction term, the sense of community variable is positive and significant. Figure 2 displays the predicted probabilities for external efficacy, with focus on the likelihood that respondents believe local officials care “very much” about what they think is important. The figure demonstrates the decreasing shift in external efficacy for those with low, average, and high sense of community scores as the number of contexts increases. This suggests, then, that having a high sense of community in a few contexts matters more than having a high sense of community in multiple contexts. Predicted probabilities decrease from 0.42 to 0.02 when moving from two to five

72

Anderson Table 2. Sense of Community and External Efficacy: Estimated via Ordered Logit

Variable Total Sense of Community Summary Context (No. of Contexts) Total Sense of Community X No. of Contexts Gender Age Age Squared Education Strength of Party Income Marital Status Employment Status Agreeableness Emotional Stability Conscientiousness Extroversion Openness to Experience # of observations (pseudo R2) Log likelihood

Baseline Model Coefficient (se)

Full Model Coefficient (se)

-.242 (.165) .025 (.027) -.000 (.000) .090 (.047)# .059 (.085) .058 (.063) .005 (.117) .185 (.198) .038 (.022)# -.004 (.018) .026 (.017) .004 (.015) -.049 (.021)* 635 (.03) -698.9

.008 (.001) -1.21 (.412)** .003 (.002) -.23 (.167) .004 (.027) .000 (.000) .083 (.048)# .053 (.086) .038 (.064) .031 (.118) .186 (.203) .024 (.022) .000 (.018) .022 (.017) -.005 (.015) -.052 (.022)* 635 (.04) -689.5

# = p < .10, * = p