Trends and Disparities in Broadband Internet Access in the United States,2013 to 2023

Abstract

Since the turn of the century, sociologists and other scholars concerned about digital inequality have most often been concerned about disparities in the quality of Internet use, not necessarily the availability of Internet access itself. However, the coronavirus disease 2019 pandemic laid bare the fact that the Internet’s potential benefits to mitigating the spread of the virus were available only to those with Internet access. In this visualization, the author uses household-level data from the American Community Survey from 2013 to 2023 (n = 10,713,204 households) to estimate a linear probability model predicting Internet access by race/ethnicity, household educational attainment, and poverty status. The results suggest that household Internet access has increased over the past decade, but disparities still exist on all three dimensions.

Since the turn of the century, digital inequality scholars have paid a great deal of attention to the different ways that individuals use the Internet across sociodemographic groups (DiMaggio et al. 2004; Hargittai and Hsieh 2013). However, disparities in Internet access have received far less attention in this literature. The coronavirus disease 2019 pandemic and its aftermath laid bare the importance of Internet access for individuals’ daily lives as pandemic stay-at-home orders closed workplaces, schools, and businesses and encouraged Americans to take their daily activities online to avoid the virus (McClain et al. 2021). Individuals without broadband Internet access could not access these benefits, and offline individuals disproportionately belonged to the same sociodemographic groups that were hit hardest by the virus (Fielding-Miller, Sundaram, and Brouwer 2020; Ndugga, Hill, and Artiga 2022; Pew Research Center 2024; Zhuo and Harrigan 2023).

In this visualization, I estimate a linear probability model using data from the 2013 to 2023 American Community Survey (Ruggles et al. 2024) to predict the odds of Internet access in non–group quarters households across years by several sociodemographic characteristics. Compared with bivariate estimates (see Pew Research Center 2024), results from multivariate models ascertain the strength and importance of one variable after controlling for other model characteristics. I omit data from 2020 because of pandemic-related data quality concerns (Rothbaum et al. 2021) and missingness, leaving a final analytic sample of n = 10,713,204 households. I present results from the model across four panels in Figure 1. Figure 1A shows the trend in household Internet access by year. Figures 1B, 1C, and 1D depict the predicted margins of householder race/ethnicity, highest educational attainment, and family poverty status of the householder, respectively. (See the Supplemental Materials for full details on variable and model construction, descriptive statistics, and additional visualizations of Internet access disparities by rurality, minors attending K–12 school, and adults aged ≥65 years in a household, as well as an expanded version of Figure 1B that compares rates of Internet access across all major racial/ethnic categories used by the U.S. Census Bureau.)

Figure 1.

Trends and disparities in household Internet access, American Community Survey, 2013 to 2023 (n = 10,713,204 households). All panels are drawn from the results of a linear probability model predicting the odds of household internet access. (A) Point estimates of the trend in the percentage of households with Internet access across the United States. (B) Predicted margins of selected racial/ethnic groups, with non-Hispanic Asian households as the reference. (C) Predicted margins of highest household educational attainment, with at least a bachelor’s degree as the reference. (D) Predicted margins of poverty status, with 501 percent or more of the poverty line as the reference. Horizontal grey dashed lines in (B), (C), and (D) mark parity with the reference category. Ninety-five percent confidence intervals are shown as colored dashed lines. Because of the sample size, estimates are precisely estimated, and most confidence intervals are virtually indistinguishable from trendlines. AIAN = American Indian and Alaska Native; AME = average marginal effect; Assoc. = associate’s degree; Deg. = degree; HS = high school; LTHS = less than high school; NH = non-Hispanic.

Several findings are of particular note in Figure 1. First, net of observed characteristics, rates of household Internet access increased from 2013 to 2023, in line with existing bivariate estimates (Pew Research Center 2024). Second, race/ethnicity, educational attainment, and poverty status predict weaker disparities in 2023 than in 2013. However, this does not mean that disparities have disappeared. Third, the average marginal effect of being a non-Hispanic American Indian or Alaska Native household compared with a non-Hispanic Asian household remained stable between about 2017 and 2023, despite the narrowing of other racial/ethnic disparities. Fourth, the least educated and lowest earning households continue to report far lower Internet access relative to all-else-equal households with higher educational attainment and higher incomes, respectively.

This visualization adds to the current literature on digital inequality by employing a linear probability model to better understand trends and disparities in household Internet access. These results suggest that, though Internet access has improved and disparities appear to have narrowed in the past decade, some groups remain less likely to report Internet access than others. These findings suggest that policy efforts aimed at improving Internet access should continue, and special attention should be paid to non-Hispanic American Indian or Alaska Native, less educated, and low-income households.

Supplemental Material

sj-docx-1-srd-10.1177_23780231251363238 – Supplemental material for Trends and Disparities in Broadband Internet Access in the United States, 2013 to 2023

Supplemental material, sj-docx-1-srd-10.1177_23780231251363238 for Trends and Disparities in Broadband Internet Access in the United States, 2013 to 2023 by Spencer Allen in Socius