Abstract
This article uses recent data on both broadband and civic engagement to assess how various aspects of broadband access might be related to community involvement in rural areas of the US. In particular, it seeks to answer whether broadband access, adoption, or infrastructure (as measured by community anchor institutions) has the strongest relationship with specific civic engagement metrics. The data include 19 specific civic engagement metrics from the 2011 Current Population Survey and 2011 data on broadband access and adoption from the National Broadband Map and the Federal Communications Commission. Scatterplots and regression analysis using both aggregate (state-level) data and individual (household-level) data were used to assess overall relationships. The results of both univariate and multivariate analyses suggest that it is the adoption of broadband that is the most influential to civic engagement overall; however, community anchor institutions are important for certain metrics such as interactions with neighbors or confidence in local schools.
Notes
1. CPS household-level survey weights were used to construct the civic engagement metrics. This weight represents the number of households in the general population represented by each household in the sample. Thus, the aggregate, state-level metrics used in this analysis represent the average values of all non-metropolitan households surveyed within each state.
2. Unfortunately, we cannot disentangle the CAIs that do not have broadband – and this is a limitation of our analysis.
3. The NBM data provide access estimates for a variety of speeds and connection types (wired vs. wireless). This paper uses data for wired connections with speeds of 3 megabytes per second (Mbps) download/768 Kbps upload.
4. The FCC broadband adoption data includes only fixed connections (i.e. wired and fixed wireless – but not mobile wireless). The data used in this paper are for at least 768 Kbps download and 200 kbps upload. Note that both the NBM and FCC data thresholds are significantly higher than traditional dial-up access speeds of 56 kbps.
5. Variance Inflation Factors for the household regressions indicate that multicollinearity is not a concern (aside from the expected high correlation between age and age2) – all VIFs are less than 2.5.
6. Because these are logistic regressions, heteroskedasticity is tested by a joint significance test of the interaction of residuals from the initial regression and the initial variables themselves. In the seven cases where heteroskedasticity was deemed problematic, robust standard errors were used.