Broadband and Civic Engagement in Rural Areas: What Matters? Brian Whitacre Oklahoma State University

Transcription

1 Broadband and Civic Engagement in Rural Areas: What Matters? Brian Whitacre Oklahoma State University Report prepared for the Mississippi State University Extension Service Intelligent Community Institute June 2015

2 Introduction As the Internet began to garner interest among everyday Americans during the mid- 1990s, there was some apprehension about what this new technology might mean for social interaction. Initially, there was concern (and some evidence) that individuals who were heavy users of the Internet would become more depressed, have fewer friends, and lead more stressful lives (Kraut et al., 1998). Others argued that the Internet presented a new array of opportunities for increasing a sense of community and trust via lists, support groups, chat rooms, or community websites (Hauben and Hauben, 1997). Still others argued that the Internet neither destroys nor creates social capital rather, it only acts as an enabler for our predispositions to trust (Uslaner, 2000). As such, it complements, rather than alters, the existing behaviors of people (DiMaggio et al., 2001). One point notably missing from these discussions was what specific aspect of connecting to the Internet is most relevant for social collaborations. There has been some research suggesting that simply having a place to access the Internet in one s community can increase levels of community participation and satisfaction (Dutta-Bergman, 2005). Other studies indicate that Internet usage - as opposed to simply having access - can increase social networks, and thus positively influence both social capital and community participation (Wellman, 2001; Simpson, 2005). More work along these lines has shown that this holds true specifically in rural areas (Stern and Dillman, 2006; Stern and Adams, 2010). Alternatively, it has been suggested that it is simply the presence of community places or institutions (places that give a community a source of pride, such as a library, hospital, or university) that foster civic engagement in rural areas (Allen and Dillman, 1995). Most of the studies that have focused on Internet use and civic engagement, however, are based on either limited / outdated survey data (from only a single 1

3 Western region in 2005, or Census data as of 2000), and none have taken advantage of the recent influx of broadband-related data that is now available. The primary question that this research hopes to answer is, Which broadband-related measure is most relevant to metrics of civic engagement in rural areas? Specifically, the relationship between various civic engagement measures and the three broadband-related metrics below will be explored: 1. Number and type of community anchor institutions (CAIs libraries, hospitals, schools, etc. some of which provide public broadband access) 2. The proportion of the population with access to broadband 3. The proportion of the population that has adopted broadband in their house. Data Four distinct datasets were used for this analysis in order to mesh data on civic engagement with each of the broadband-related measures noted above. The most recent nationally representative data on civic engagement came from the November 2011 Current Population Survey (CPS). This survey was conducted on roughly 50,000 households and asked 19 specific questions related to civic engagement, including whether or not each respondent participated in local elections or community groups, the frequency of their discussions with neighbors, how much they trust people in their neighborhood, and other measures. Unfortunately the downside of this data is that the lowest level of geographic information available is whether or not the household resides in a non-metropolitan part of the state (roughly 10,000 of the respondents lived in these areas). Thus, aggregate measures of civic engagement were created for the non-metropolitan portions of each state for example, the percentage that have boycotted a company, or a ranking of how often they discuss politics with family. Figure 1 2

4 lists each of the 19 civic engagement measures from the CPS and their average value across the non-metropolitan portions of 47 states (since DE, NJ, RI do not have non-metro counties). Table CPS Civic Engagement Question Summary (NM Areas Only) The data in Table 1 shows that some civic engagement activities were much more common than others. In particular, while very few expressed an opinion on a political or community issue over the Internet (1.55 out of 6), interacting with friends was much higher (5.14 out of 6). The range of responses for people who talk with or exchange favors with their neighbors was lower (4.0 and 2.8 out of 6). In terms of confidence in various entities, corporations ranked lowest while public schools ranked highest (2.5 and 3.2 out of 4). To obtain relevant broadband-related data, county-level numbers of community anchor institutions (CAIs) per 10,000 population were constructed from the National Broadband Map in December The CAI data itself is broken into six categories of buildings: schools, libraries, healthcare facilities, public safety institutions, colleges / universities, governmental support facilities, and nongovernmental support facilities. There are nearly 300,000 institutions listed in 3

5 the December 2011 dataset. Some, but not all, of the CAIs provide public broadband access. Each of the entries can be assigned to a county, which allows for construction of the total number of CAIs per 10,000 population. Similarly, the National Broadband Map provides county-level estimates of the percentage of the population that had access to broadband as of Information on residential broadband adoption was obtained from county-level Federal Communication Commission data, also available for December This data does not provide a point estimate of county-level adoption; rather each county is assigned a value from 1 (representing 0-20% adoption) to 5 (representing % adoption). In each of these three cases, the county-level data were aggregated to state-level non-metropolitan averages by combining information (and weighting by population when necessary) from all non-metro counties within a state. Figure 2 summarizes these state-level measures. Table State-level Broadband Measures (NM Areas Only) Figure 2 demonstrates that for the average non-metro portion of a state in 2011, the broadband adoption rate is in the mid 60%s (recall that a value of 3 = 60-80% adoption). This is consistent with other rural-broadband related research from that time (Whitacre, Strover, and Gallardo, 2015). The mean percentage of the non-metro population with access to broadband is 86%, and ranges from 68% (Alaska) to 99% (Connecticut). The number of CAIs per 10,000 non-metro population averages 22.3 across the states, but ranges from less than 6 (California) to almost 79 (Maryland). 4

6 Methodology Two specific types of analysis are used to help determine which type of broadbandrelated measure is most relevant for civic engagement in rural areas. The first uses scatterplots and simple (univariate) regressions to relate civic engagement metrics to the three types of broadband measures listed above at. The comparisons are made a highly aggregated level (using the non-metropolitan averages for each state discussed above). This paints a general picture of how non-metropolitan regions with varying levels of broadband access / adoption rates / CAIs fare in terms of civic engagement. Of particular interest will be which civic engagement measures seem to be most influenced by broadband infrastructure or adoption. The second technique uses household-level regressions to tease out the determinants of individual civic engagement. This technique, using the household-level CPS data, will have measures of broadband access and adoption as part of the explanatory variables (each of the 19 civic engagement measures from Table 1 will serve as dependent variables). The benefit of this approach is that it is able to control for other factors known to impact civic engagement, such as age, race, and income. Thus, the results will tell us what component of broadband truly matters for specific measures of social engagement after controlling for other variables of influence. More formally, these regressions take the form: y i = β 0 + β 1 BB 1 + β 2 BB 2 + β 3 BB 3 + β 4 X i + ε where y i is a yes / no answer regarding civic engagement for individual i, BB 1 is a vector of percapita CAI measures, BB 2 is a measure of local broadband availability, BB 3 is a measure of local broadband adoption (all relevant to individual i), and X i is a vector of other characteristics that are known to influence civic engagement (race, income, education, age, etc.) These are run on a subset of only households in non-metropolitan counties. Statistically significant results for β 1, β 2 or β 3 suggest that certain aspects of broadband do in fact impact the civic engagement 5

7 decision at the individual level. Ordered logistic regressions are used for the CPS engagement questions that are categorical, while traditional logistic regressions are used for yes / no questions. Note that the sample size is much larger for these household-level regressions. Results The scatterplots demonstrate a wide variety of interesting correlations (the full complement of all 19 plots for each of the three broadband measures is available in the Appendix to this report). Several of the more interesting results are displayed below. In particular, nonmetro broadband adoption rates are positively correlated with (and explain a reasonable amount of variation in) the likelihood of expressing opinions over the Internet, the likelihood of boycotting a company, and how often people discuss politics with their family or friends. However, adoption rates also display some negative correlations, including with how often individuals have dinner with other household members, and the frequency of talking with neighbors. Figure 1. Scatterplots for Broadband Adoption & Selected Civic Engagement Measures 6

8 In terms of broadband access, very few of the scatterplots displayed meaningful correlation. One surprising exception was a negative correlation with the frequency of having dinner with other household members. Figure 2. Scatterplots for Broadband Access & Selected Civic Engagement Measures Alternatively, the number of Community Anchor Institutions per 10,000 population was positively associated with several measures, many of which dealt with neighbors or trust. Figure 3. Scatterplots for CAIs & Selected Civic Engagement Measures Overall, the number of statistically significant relationships (at the p = 0.10 level or better) in these univariate regressions can be summarized as follows (recall that 19 civic engagement measures were used): Broadband Adoption: 12 (9 positive, 3 negative) Broadband Access: 2 (1 positive, 1 negative) Community Anchor Institutions: 7 (7 positive) In an effort to determine which of these broadband factors was most important for a specific engagement measure, multivariate regressions were run. Also included was a measure of 7

9 the percentage of residents voting republican in the 2012 presidential election, in order to control for any influence of political ideology. Note that none of the broadband variables are highly correlated, with the correlation between adoption and access only This alleviates concerns about multicollinearity in the simple regression. The results are presented in Table 3. Table 3. Regression Results (NM State Averages) The results clearly suggest that it is broadband adoption which matters the most for many of the civic engagement measures. This is true for many of the questions related to participation in local organizations, but also for contacting public officials to express opinions or expressing opinions over the Internet. It is worth noting that in two cases, talking with neighbors and confidence in the media, the relationship with broadband adoption is negative. Relationships with broadband access and CAIs are less robust. Only one variable (dinner with other household members) is significant (and again negative) at more than a p=0.10 level for access, while each of the five measures that are statistically significant for CAIs only hold at the p=0.10 level. It is 8

10 interesting, however, that most of the measures influenced by CAIs in this regression deal with interaction with neighbors and all are positive. The value-added of the household-level regressions lies in the ability to control for things like income levels, education, race, and number of children; all of which have been shown to impact civic engagement. Once these elements are included, is it still true that people living in areas with high broadband adoption rates (or high levels of CAIs) are more likely to get involved? The household-level results in Table 4 help answer that question and generally confirm the results of the simpler, state-level analysis. Although only the highest level of income, a specific education, and a single race are shown as controls in Table 4 due to space concerns; a full host of variables was used, including a dummy variable for being retired and the age of the respondent. As expected, higher levels of income and education are positively associated with many of the civic engagement measures (the only exception for education is for the variable associated with confidence in the media). Being African-American has some positive and some negative relationships with civic engagement measures. Even after these attributes are included, there are meaningful relationships with the level of broadband activity in the area. Notably, individuals living in areas with high levels of broadband adoption are more likely to contact a public official to express an opinion, boycott a company, join a sports or other organization, and discuss politics with friends / family. They are also less likely to vote in local elections, see or hear from friends each day, or talk with their neighbors. They also tend to demonstrate less confidence in corporations and public schools. Individuals living in areas with high levels of broadband access, meanwhile, are more likely to join a religious organization and to have confidence in the media. Negative relationships are found for the likelihood of contacting a public official to express an opinion, discussing politics with friends or family, and having dinner with other household members. Finally, the results for 9

11 individuals living in areas with high levels of CAIs are striking. These families are more likely to join a neighborhood or civic group, to be an officer in an organization, and also to talk / exchange favors with / trust people in their neighborhood. They also tend to have more confidence in the media and in public schools. Each of these results is statistically significant at either the p = 0.05 or 0.01 levels. Conclusion This report has used the latest data available on civic engagement to assess how specific measures might be related to various aspects of broadband. The findings suggest that broadband adoption is generally more important than broadband access in terms of civic engagement. However, there are both positive and negative aspects to having high levels of broadband adoption. While being highly connected is positively related to expressing opinions and joining organizations, it is also negatively related to seeing / hearing from friends or talking with neighbors. This contrasts with the findings for community anchor institutions, which imply that interactions with neighbors and confidence in local schools increase with the prevalence of such infrastructure. Thus, it appears that broadband does matter for civic engagement in rural areas and that the various aspects of broadband contribute differently. Taken to their extreme, the findings suggest that rural communities could encourage specific forms of civic engagement by emphasizing certain components of broadband i.e. to promote more neighborhood engagement, additional anchor institutions should be supported, while to support more civic participation, adoption should be encouraged. Community organizers should be aware of both the positive and negative aspects of these different components before committing to a plan of action. 10

12 Table 4. Regression Results (Individual Household Level) Civic Engagement Measures Income > $150,000 College Degree Black Adoption Access CAI 1 Vote in local elections (1-Never, 4-Always) 0.78 *** 1.77 *** 0.58 *** *** * 2 Contact public official to express opinion (%) 0.49 ** 1.66 *** *** 0.35 *** *** Boycott a company (%) 0.50 ** 2.17 *** ** 0.77 *** ** 4 Expressed opinion over Internet (1-not at all, 6-every day) 0.48 ** 1.22 *** *** ** Join neighborhood or community group (%) 1.04 *** 1.46 *** 0.33 * *** *** 6 Join civic organization (%) 0.75 *** 1.26 *** *** 0.20 * *** 7 Join sports organization (%) 1.48 *** 1.27 *** *** Join religious organization (%) 0.90 *** 1.05 *** 0.44 *** *** 1.93 *** Join other organization (%) 0.70 ** 1.70 *** *** ** Officer in organization (%) 1.45 *** 1.94 *** ** *** 11 Discuss politics with family or friends (1-Not at all, 6-Every da 1.16 *** 1.28 *** ** *** Dinner with other household members (1-Not at all, 6-Every ** *** ** See or hear from friends (1-Never, 6-Every day) ** *** Talk with neighbors (1-Never, 6-Every day) ** 0.49 *** ** *** 15 Exchange favors with neighbors (1-Never, 6-Every day) ** 0.16 * *** 16 Trust people in neighborhood (1-Not at all, 4-Everyone) 0.83 *** 0.51 *** *** ** 17 Confidence in corporations (1-None, 4-Great Deal) 0.83 *** 0.49 *** *** Confidence in media (1-None, 4-Great Deal) *** 0.01 ** 19 Confidence in public schools (1-None, 4-Great Deal) 0.06 *** 0.38 *** ** *** *, **, and *** indicate statistical significance at the p=0.10, 0.05, and 0.01 levels, respectively. n ranges from 4,587 observations to 7,545 (depending on civic engagement measure) 11

13 References Allen, J and Dillman, D Against all odds: Rural community in the information age. Westview Press, Inc. DiMaggio, P., Hargittai, E., Neuman, W., and Robinson, J Social implications of the Internet. Annual Review of Sociology 27: Dutta-Bergman, M Access to the Internet in the context of community participation and community satisfaction. New Media & Society 7(1): Hauben, M. and Hauben, R Netizens. IEEE Computer Society Press, Los Alamitos, CA. Kraut, R., Patterson, M., Lundmark, V., Kiesler, S., Mukopadhyay, T, and Scherlis, W Internet paradox: A social technology that reduces social involvement and psychological well-being? American Psychologist, 53, 1017: Simpson, L Community informatics and sustainability: Why social capital matters. Journal of Community Informatics I(2), Stern, M. and Adams, A Do rural residents really use the Internet to build social capital? American Behavioral Scientist 53(9): Stern, M and Dillman, D Community participation, social ties, and use of the Internet. City & Community 5(4): Uslaner, E. Social capital and the net Communications of the ACM 43 (12), Wellman, B Computer networks as social networks. Science 293 (14), Whitacre, B., Strover, S., and Gallardo, R How much does broadband infrastructure matter? Decomposing the metro non-metro adoption gap with the help of the National Broadband Map. Government Information Quarterly. doi: /j.giq

14 Broadband and Civic Engagement in Rural Areas: What Matters? Appendices A1. Scatterplots of Civic Engagement / Broadband Adoption A2. Scatterplots of Civic Engagement / Broadband Access A3. Scatterplots of Civic Engagement / Community Anchor Institutions A4. Summary of All Simple Regressions Displayed in Scatterplots A5. Full CPS Household-level Regression Results (19 separate regressions)

15 The Scatterplots in appendices A1, A2, and A3 relate to the following measures as detailed in the report Broadband and Civic Engagement in Rural Areas: What Matters? Summary of 19 Civic Engagement Questions in November 2011 CPS: State-level Civic Engagement Questions Mean S.E. Min Max 1 Vote in local elections (1-Never, 4-Always) Contact public official to express opinion (%) Boycott a company (%) Expressed opinion via Internet on political / community issue (1-4 Not at all, 6 - Every Day) Join neighborhood or community group (%) Join civic organization (%) Join sports organization (%) Join religious organization (%) Join other organization (%) Officer in organization (%) Discuss politics with family or friends (1-Not at all, 6-Every day) Dinner with other household members (1-Not at all, 6-Every day) See or hear from friends (1-Never, 6-Every day) Talk with neighbors (1-Never, 6-Every day) Exchange favors with neighbors (1-Never, 6-Every day) Trust people in neighborhood (1-Not at all, 4-Everyone) Confidence in corporations (1-None, 4-Great Deal) Confidence in media (1-None, 4-Great Deal) Confidence in public schools (1-None, 4-Great Deal) Summary of Broadband Measures from December 2011 National Broadband Map or FCC: State-level State-level Broadband Measures Mean S.E. Min Max 1 BB Adoption (1: <20%, 5: >80%) BB Access (% with access to at least 200 kbps) CAI (Total Number of Institutions per 10,000 population) Note: All measures constructed for non-metropolitan areas of states.

16 A1. Broadband Adoption Scatterplots 1 2

17 A1. Adoption 3 4

18 A1. Adoption 5 6

19 A1. Adoption 7 8

20 A1. Adoption 9 10

21 A1. Adoption 11 12

22 A1. Adoption 13 14

23 A1. Adoption 15 16

24 A1. Adoption 17 18

25 A1. Adoption 19

26 A2. Broadband Access Scatterplots 1 2

27 A2. Access 3 4

28 A2. Access 5 6

29 A2. Access 7 8

30 A2. Access 9 10

31 A2. Access 11 12

32 A2. Access 13 14

33 A2. Access 15 16

34 A2. Access 17 18

35 A2. Access 19

36 A3. Community Anchor Institutions Scatterplots 1 2

37 A3. Community Anchor Institutions 3 4

38 A3. Community Anchor Institutions 5 6

39 A3. Community Anchor Institutions 7 8

40 A3. Community Anchor Institutions 9 10

41 A3. Community Anchor Institutions 11 12

42 A3. Community Anchor Institutions 13 14

43 A3. Community Anchor Institutions 15 16

44 A3. Community Anchor Institutions 17 18

45 A3. Community Anchor Institutions 19

46 A4. Summary of All Simple Regressions Displayed in Scatterplots Adoption Access CAI Coeff Signif R2 Coeff Signif R2 Coeff Signif R2 1 Vote in local elections (1-Never, 4-Always) Contact public official to express opinion (%) 0.07 *** Boycott a company (%) 0.09 *** * Expressed opinion over Internet (1-not at all, 6-every day) 0.21 *** Join neighborhood or community group (%) 0.03 ** * Join civic organization (%) 0.04 ** *** Join sports organization (%) 0.08 *** Join religious organization (%) Join other organization (%) 0.04 *** * Officer in organization (%) 0.05 *** *** Discuss politics with family or friends (1-Not at all, 6-Every day) 0.24 ** Dinner with other household members (1-Not at all, 6-Every day) *** See or hear from friends (1-Never, 6-Every day) Talk with neighbors (1-Never, 6-Every day) ** ** Exchange favors with neighbors (1-Never, 6-Every day) *** Trust people in neighborhood (1-Not at all, 4-Everyone) ** Confidence in corporations (1-None, 4-Great Deal) * Confidence in media (1-None, 4-Great Deal) ** Confidence in public schools (1-None, 4-Great Deal) n=47 (non-metro aggregates for states)

47 A5. Full CPS Household-level Regression Results (19 separate regressions) Vote local Pub. Official Boycott Net Opinion Neigh. Group Civic Org. Sports Org. Relig. Org Other Org. Officer in Org. faminc2 ($10-20K) ** faminc3 ($20-30K) ** *** * * faminc4 ($30-40K) *** ** ** *** ** *** *** *** * *** faminc5 ($40-50K) *** *** *** *** * *** faminc6 ($50-60K) *** ** *** *** *** *** *** ** *** faminc7 ($60-75K) *** *** *** *** *** *** ** *** faminc8 ($75-100K) *** ** ** *** *** *** *** ** *** faminc9 ($ K) *** *** *** *** *** *** * *** faminc10 (>$150K) *** ** ** *** *** *** *** *** ** *** hs *** *** *** *** ** *** ** *** *** *** somecollege *** *** *** *** *** *** *** *** *** *** bach *** *** *** *** *** *** *** *** *** *** graddegree *** *** *** *** *** *** *** *** *** *** black *** *** ** * *** *** asian *** *** ** *** * othrace * * hispanic *** *** ** ** *** ** retired *** *** *** * ** ** peage *** *** *** *** *** *** ** ** *** peage *** *** *** *** ** ** * *** numberkids ** *** *** *** ** bbadop *** *** *** *** *** * *** *** *** ** bbaccess *** ** *** ** caitotal * ** *** *** *** # Obs Psuedo R

48 Discuss Pol. Dinner Friends Talk w/ neigh. Favors w/ neigtrust neigh. Conf. in corp. Conf. in media Conf. in schools faminc2 ($10-20K) ** ** ** faminc3 ($20-30K) *** *** * *** *** *** faminc4 ($30-40K) 0.49 *** *** *** ** ** faminc5 ($40-50K) 0.47 *** *** *** * ** faminc6 ($50-60K) *** E ** *** ** faminc7 ($60-75K) *** ** *** *** *** faminc8 ($75-100K) *** *** *** *** * faminc9 ($ K) *** *** *** *** ** faminc10 (>$150K) *** *** *** hs *** ** *** * ** *** somecollege *** *** 0.19 *** ** *** *** * bach *** ** ** ** ** *** *** *** graddegree *** * * *** *** black *** *** *** asian *** othrace * * * *** hispanic *** *** *** *** *** retired *** 0.71 *** ** 0.25 *** *** ** peage *** *** *** ** *** *** peage * -4E-04 *** 3E-04 *** ##### -4E-04 *** 6E-05 2E-04 ** numberkids -5E-04 *** *** *** *** *** *** bbadop ** *** ** *** ** bbaccess *** ** *** caitotal E-04-8E *** *** ** -9E ** *** # Obs Psuedo R