Broadband and Unemployment: Analysis of Cross-Sectional Data for U.S. Counties

Abstract

This paper examines the effect of broadband availability on the unemployment rate at the county level in the United States. Using data on broadband availability from the National Broadband Map and census data on unemployment for all 3000-plus counties in the United States, this paper estimates an econometric model that that isolates the effect of broadband deployment after controlling for other factors. This paper builds on our prior work with a smaller sample that found that increasing broadband availability from the current 93% of homes to 100% will reduce unemployment by about 0.5%.

An extensive literature exists on the relationship between telecommunications, and specifically broadband, on economic activity and growth with the consensus favoring a generally positive impact (see for example, Crandall, Lehr, & Litan, 2007; Kolko, 2010). However, the evidence in favor of a positive impact on employment is much more tenuous and contradictory. On the one hand, a positive effect on job creation may be expected from the increased economic activity resulting from broadband deployment. On the other hand, broadband technologies may contribute productivity gains negating any need for additional labor. Kolko (2010) has also shown that though broadband contributes to employment generation, it may not significantly affect the unemployment rate because workers are mobile, and relocate or commute to places with higher labor demand. Atasoy (2011) on the contrary, found that broadband deployment has a significant effect on the employment rate, but the effect is stronger in counties with a higher percentage of college graduates and in industries that employ a more educated workforce. The effect of broadband availability on employment is an unresolved question, which requires further investigation. This paper seeks to analyze this question.

Past attempts at investigating the relationship between broadband and economic growth were complicated by the lack of adequate data on broadband deployment and availability. Previously accessible data on broadband availability, such as the FCC’s ZIP Code-based broadband availability data have been heavily criticized (Grubesic, 2008; Kolko, 2010; Prieger & Hu, 2008). Beard, Ford and Saba (2010) used the 2007 Computer and Internet Use Supplement of the Census Bureau’s Current Population Survey, but this data too is more than six years old and out-of-date. Thus, National Broadband Map data present an attractive opportunity to test the relationship between broadband availability and employment, despite caveats such as the incomplete information and lack of pricing data pointed out by Grubesic (2012), and the measurement errors and selection bias identified by Ford (2011).

In this paper, we test the hypothesis that better broadband availability will be associated with lower county unemployment rates. The county is chosen as the unit of analysis, since it is small enough to capture local variations but large enough to constitute a reasonable labor market. Data on broadband availability and employment are collected from the National Broadband Map and the Bureau of Labor Statistics’ (BLS) Local Area Unemployment Statistics (LAUS) database, for all 3000-plus counties in the United States. We use multiple regression is then used to analyze the data. To control for the effect of education levels found by Atasoy (2011), we also include the percentages of the county workforce with a high school diploma as a control, along with average age and race. Controls are also included for the effects of changes in construction and real estate spending (a critical factor affecting unemployment rates over the last 6 years), and for changes in the statewide economy, since county unemployment rates may be responding to wider regional trends.

Results obtained for counties drawn from a sample of 8 states (N=430) representing different combinations of economic development (state per capita income above or below the national poverty line) and urbanization (percentage of state population resident in urban areas above or below the national mean) indicate that there is a negative correlation between the county unemployment rate and broadband availability. Despite the care we took to select a representative sample of states, purposive sampling may still introduce biases in the results — including the entire population is expected to remedy this. Full results for the population of 3000 counties will be available in time for TPRC.

Keywords: broadband, employment, National Broadband Map, BTOP, stimulus