Sign up for our newsletter
The argument has been made for a while that broadband is critical for not only economic development but also for quality of life in general. While the tide was already turning, in terms of community leaders realizing broadband is critical infrastructure, the current COVID-19 public health emergency has shed a bright light on this same issue that has existed for years: many communities across the country lack access to affordable and adequate internet service.
Multiple research has found that broadband does matter for jobs, income, business relocation, civic engagement, and health, among other things. These studies have used a variety of broadband-related metrics, mostly availability and adoption.
However, there are quite a few different ways to measure broadband access and availability in particular communities. For example, is it more important to look at the percentage of households with broadband available to them, or the average advertised download speed in that area? Should we be more concerned with the percentage of households that don’t have an internet subscription, or the percentage that rely only on smartphone connections?
Therefore, these data are often from different sources, with different ways of framing the issue. For example, the Federal Communications Commission (FCC)’s data suggests that 21.3 million people did not have access to broadband, while Microsoft argues that over 157 million did not actually use the internet at broadband speeds.
In other words, broadband data varies tremendously depending on the source. So, which broadband-related metric performs best when attempting to explain specific economic outcomes? To answer this question, a team of researchers from Purdue University, Oklahoma State University, and the University of Tennessee compiled 10 different broadband-related metrics at the county-level to see their impact on job productivity (defined as Gross Domestic Product per job).
A series of statistical analyses were conducted using each of the 10 broadband-related metrics while also including other variables known to affect job productivity such as population size, distance to metropolitan areas, industrial and ethnic/racial diversity, and educational attainment. The analysis also looked specifically at rural areas.
The 10 broadband-related variables included:
1) Percent of households with no internet access, obtained from the Census American Community Survey;
2) Percent of population not using the internet at 25 Megabits per second (Mbps) download, obtained from Microsoft;
3-5) the digital divide index overall score as well as its infrastructure/adoption and socioeconomic scores; these scores range from 0 to 100 where a higher number shows a higher divide. In total, nine different variables are used to calculate the overall digital divide index, including FCC advertised speeds;
6) If a county is in digital distress (i.e. with a high percentage of households with no internet subscription or only a smartphone-based plan or no computing devices or rely only on mobile devices);
7-10) Percent of the population with access to advertised fixed broadband speeds of at least 25/3 Mbps, 25/25 Mbps, 50/50 Mbps, and 1/1 Gigabits per second (Gbps), obtained from the FCC.
The results indicated that the advertised speed variables had no impact (insignificant) on job productivity. On the other hand, other broadband-related variables did have a statistically significant impact on job productivity. In other words, counties with a high share of people not using the internet at 25 Mbps, a higher digital divide score, with a higher share of their population with socioeconomic characteristics known to impact technology adoption, or that were classified as being in digital distress also had a lower job productivity.
When looking at rural only, some different results appeared. The adoption metric – or percent of homes without an internet subscription – did affect job productivity. The digital divide index as well as its infrastructure/adoption and socioeconomic score also affected job productivity while the percent of people not using the internet at 25 Mbps effect disappeared.
The table below summarizes the findings of the 10 broadband-related metrics.
What does this mean?
First, advertised speeds did not affect productivity in both overall and rural. In other words, the current simple availability metrics did not explain differences in job productivity.
Second, adoption did matter for rural areas only (percent homes with no internet subscription). For this reason, policies targeting adoption in rural areas may result in productivity increases.
Third and most importantly, broadband-related metrics that consider multiple factors had a stronger relationship with job productivity. In other words, metrics that consider multiple broadband-related factors and socioeconomic characteristics—better gauging adoption and inclusiveness—are better at explaining productivity, after considering all other factors (e.g. population size, distance to metro, etc.)
At the end of the day, digital access, affordability and adoption are key steps for productive broadband utilization to take place. As more industries and business digitize and as more communities engage digitally with their citizens, ensuring that residents and businesses have the necessary resources to productively use broadband technology is essential. While data in this respect is not perfect, this study has shown that there are some innovative ways to bundle existing data to better gauge the digital divide and its impact on productivity.
Roberto Gallardo is the director of the Purdue Center for Regional Development and Brian Whitacre is a Sarkeys Distinguished Professor at Oklahoma State University, respectively. Their study, with Indraneel Kumar and Sreedhar Upendram, was recently published in the Annals of Regional Science and is entitled “Broadband Metrics and Job Productivity: A Look at County-level Data.”