Social Interactions and the 'Digital Divide' : Explaining Regional Variations in Internet Use

The presence of extraordinary geographical variation in Internet use in the U.S. is widely acknowledged. Prior research suggests that individual, household, and regional differences are responsible for this disparity. We argue for an alternative explanation: that individual choice is subject to social influence, and that such peer effects are the cause of the excess variation. We test this assertion with empirical analysis of a matched data set compiled from two sources. The first is a dataset collected by the Pew Charitable Trust in June 2003 comprising of a nationwide random sample where a questionnaire is used to collect details on Internet use as well as personal characteristics such as age, income, and education of respondents. The geographic location of individuals is identified by the FIP code that represents the county where the respondent resides. From the 2000 Census we then extract a number of regional characteristics for this location (such as median household income, median age, etc.). We begin with a linear probability model of choice, and then use an instrumental variables (IV) approach to address simultaneity and other problems. In addition, we present the results from a logistic regression model. Our analysis of the data provides strong evidence of peer effects, suggesting that individual Internet use is influenced by local patterns of usage. In fact, regional differences (such as between urban and rural areas) as well as individual differences appear to matter less when peer group choices are taken into account. Our IV results suggest that the correlation between individual and group choice is not a consequence of unobserved individual or regional characteristics