Lack of good broadband access is a strong predictor of childhood poverty. That’s the finding of Broadband Communities’ recent analysis combining county-level broadband data it has collected since 2010 with comprehensive, county-level poverty data compiled by the nonprofit organization Save the Children. We looked at overall poverty rankings, and, with sensitivities heightened because of the current need for distance learning, we also analyzed high school graduation patterns.


As Table 1 shows, childhood poverty almost inevitably increases as broadband access worsens. The top half of counties, ranked by broadband access relative to other counties in their states, have a child poverty rate of about 18 percent, and the bottom half have a child poverty rate of more than 21 percent. The gap grows in states that restrict communities from building their own broadband systems where commercial carriers cannot or will not do so (Table 2). In the bottom half of all counties in those states, child poverty related to broadband grows to more than 24 percent. The lowest-ranked counties are at 31 percent.


Rural counties bear the brunt of the bad news, especially when they are at the bottom of the broadband rankings. The top rural counties in “restriction” states have a child poverty rate of 13 percent. The bottom counties have a rate of 30 percent (Table 4).
All data in this article refers to a pre-COVID United States, but broadband disparities now are even worse than they were at the start of 2020 given work-from-home and distance-learning demands brought on by the pandemic. In urban areas where access is available but not always affordable, providers usually have made it available free or at low cost during the COVID-19 lockdown. Almost half of all rural homes have no broadband at any price.
At the time this goes to press, there are numerous new and revitalized efforts at the local, state and federal levels of government to close the broadband gap. The efforts cannot come soon enough, even though beginning in 2018, the federal government alone earmarked about $3 billion in new federal grants and loans for broadband deployments (up from only $40 million available in 2017). About a third of those funds already have been awarded this year by the USDA or soon will be. Later this year, the FCC will begin awarding another $2 billion per year for 10 years to subsidize operations in least-served areas.


Does Lack of Broadband Increase Child Poverty?
Showing a link between child poverty and lack of broadband is relatively simple, but determining which causes which is more difficult. Broadband Communities has studied the broadband gap since 2014. Almost all of our analyses reveal that restricting governmental entities that build or threaten to build their own networks when other entities cannot or will not do so heightens disparities.
For instance, Broadband Communities data shows that rural counties in restriction states have lost population at three times the annual rate of non-restriction states since 2010, even though the restriction states have higher population growth overall. That means rural restriction-state counties have experienced population losses of about 1 percent versus 0.3 percent in non-restriction states. Those rates of population loss, which did slow starting in 2017, have helped determine the direction of causality. At least a quarter, and as much as half, of all rural population loss since 2010 was caused by lack of adequate – or any – broadband access.
Broadband Communities has reported that median family income also declines with lack of good broadband, but the direction of causality is not as clear as it is with population loss, largely because the families that are most poverty-stricken and most in need of employment tend to migrate to urban areas, effectively raising the median income for the areas they left.
Because of the difficulty in showing that inadequate broadband depresses median income, I expected to have the same difficulty showing that it increases child poverty. As Tables 2 and 3 show, however, there is clear indication of a poverty disparity between restriction and non-restriction states, especially in the most underserved 10 percent of counties in each state. In other words, child poverty is even worse in unserved counties of the states where municipalities cannot build their own broadband networks. This suggests that inadequate broadband is the causal factor.
This evidence exists despite the fact that Save the Children gathered only enough data to fully rank 2,617 of the 3,143 counties in the United States. Broadband Communities can and does rank data from all 3,143 counties.
Save the Children graciously supplied all data it was able to gather for the 3,143 counties anyway, even though it could not make the information available on its website’s interactive mapping tool.
Methodology
Because I work with full national datasets, the results using conventional statistical methods are almost always, by definition, statistically significant. But they still are subject to systematic statistical bias I might not detect. As a diligent journalist, I double-check by contacting affected stakeholders.
I also use well-tested but less conventional methods, especially Bayesian statistics, to confirm results. Bayesian techniques require a starting point, aka a “prior.” When users can’t generate one, modern software supplies starting points. The Bayesian checks suggest that the odds my conclusions are correct are quite good – eight and usually nine out of 10. Even when I forced the priors somewhat, the odds did not get worse than five to one. That certainty could affect risk decisions by potential broadband deployers.
The last full census was in 2010. The 2020 census was still in progress in August. Data are not likely to be released until winter 2021 at the earliest. Intermediate surveys are taken to help guide industry, commerce and demographic researchers. But they are not complete, especially in rural areas, and they rely on the most recent full census as a starting point. So data quality, though excellent for many purposes, deteriorates as the decade progresses. The most recent partial national sampling was in 2018.
Starting in 2018, massive tax cuts and trillion-dollar federal deficits combined with big changes in immigration and trade policies created new and unprecedented trends in demographics and the economy. The detailed effects of that won’t be clear until the 2020 full census results are released next year.
Broadband Communities’ population base was 2013 for the data tables. This article uses 2018 population data double-checked with 2010 complete census data and the 2013 survey. The earlier data are slightly more reliable for rural areas. The later data are slightly better for urban areas, mainly because urban growth rates have been higher than rural population declines. The Bayesian calculations take this into account when defining the odds noted above.
Save the Children child poverty indicators such as death, teen pregnancy and hunger – the latest nationally available – are based mainly on data gathered by federal agencies in 2019 for the previous year – 2018. For this article, the data on percentage of children (aged 0–18, that is, until they reach their 19th birthday) living in poverty comes from the U.S. Census Bureau 2018 Small Area Income and Poverty Estimates (SAIPE) Program.
The dropout data is derived from the public high school four-year adjusted cohort graduation rate (ACGR) covering the 2016–17 school year. The ACGR is calculated as the number of students who graduate in four years with a regular high school diploma, divided by the number of students who form the adjusted cohort for the graduating class. From the beginning of high school, students who enter a grade for the first time form a cohort that is adjusted by each school district (there are about 17,000 in the United States) by adding students who subsequently transfer into the cohort and subtracting any students who subsequently transfer out, immigrate to another country or die.


Other Insights From Save the Children Data
Combining the Broadband Communities and Save the Children datasets offers insights in the aggregate, even when the data is incomplete. For instance, Broadband Communities ranks all counties against others in each state with regard to broadband access. Comparing the 50 counties on the bottom of the Broadband Communities state lists with Save the Children’s list shows that the 21 that Save the Children could evaluate were marked by childhood poverty as well as poor broadband. Four were ranked at the bottom of their state by Save the Children, another 10 were in the bottom half, and only seven were in the top half – most just barely (Table 8).
Thus, although Save the Children had enough data to rank 83 percent of all U.S. counties overall, it could rank only 40 percent of the counties at the bottom of Broadband Communities’ state lists. More than 1.3 million people, including more than 300,000 children, live in those 50 counties, according to U.S. census data (Table 1). But Save the Children could be sure that only 70,000 of the children (21 percent) live in poverty. Given the missing data and results from counties that can be tracked, the real proportion is probably 30 percent or more – another 30,000.
Save the Children also reported that 13 percent of children in those counties had not finished high school in the regulation four years. Nationally, most kids who don’t complete high school in four years never earn a standard high school diploma. In these counties, because it is far easier to track graduates than dropouts, it is likely that almost none of the missing 30,000 earned a degree, so closer to 25 percent of all children never earn a high-school diploma.
The shortcomings in the data are more obvious in the aggregate. In this case, Table 1 also shows the data for the best-served broadband counties in each state. In general, of course, they are comparatively rich and urban. They have total populations of more than 42 million, including 9 million children. More than 1.6 million are in poverty (18 percent), and 18 percent fail to graduate from high school on time. In urban areas, it also is easier for those who take more than four years to eventually graduate.
Without the comparison made possible by combining the two sets of data, it might appear that wealthy urban districts have a higher high school dropout rate (calculated in tables as percent of all children) than poor rural districts. I often see those kinds of mistakes riddling papers by academic economists who use only income data, thereby poisoning the arguments against more broadband aid.
It is unclear what the effect of COVID lockdowns will do to graduation rates going forward. We could see lockdowns and disruptions into spring 2021, in part because vaccines are not being tested on children yet, so a vaccine likely won’t be available to them at least until spring. COVID lockdowns did not heavily impact rural counties this spring; graduation ceremonies were stunted, but kids graduated, according to data I was able to collect. The federal government has not posted 2019 data.
Confounders: Issues That Can Invalidate An Analysis
- Chicken-and-egg. Is child poverty increasing (in part) because of poor broadband, or is poor broadband the result of lousy business cases because of poverty? Or is it the result of “outside” influences, unknown or unexplored?
- States have very different ways of organizing themselves into counties. Texas alone has almost 10 percent of all U.S. counties.
- States vary in other ways. States have different geographies, population densities, distributions and temporary economic advantages (new oil discoveries, retiree attraction, etc.). In addition, states with restrictions on municipal broadband have different kinds of restrictions, enacted at different times. Different states and communities calculate key educational performance data differently. New York City public schools understate the problem because they do not include students as dropouts if they were not in the system by 9th grade.
- 25 Mbps download, 3 Mbps upload speed is the FCC’s threshold for “broadband.” That level is somewhat arbitrary, however.
- Variance on X-Axis. National Broadband Map data is flaky and tends to overestimate actual access.
- Take-rate variance. Even in counties with great 25 Mbps access, less than half of households with supposed access bought it in 2018. Still, this is more than double the 17 percent of 2014. Take rate is far higher and churn far lower in rural areas and in urban neighborhoods with school-age children. Anecdotally, COVID-19 has strengthened that effect.
Exploring Save The Children Data
The complete Save the Children report can be downloaded here: www.savethechildren.org/content/dam/usa/reports/advocacy/us-childhood-report-2020.pdf. It notes that the United States, despite having the world’s largest economy, ranks 43rd of 180 countries on helping children reach their full potential. The United States lagged behind most of Europe even before flubbing the handling of the COVID-19 pandemic.
There are also huge differences within the United States, as noted previously. When it comes to nurturing children, New Jersey and Massachusetts are at the top and Louisiana is at the bottom. There are also startling differences within states. The interactive map and database at www.savethechildren.org/us/about-us/resource-library/us-childhood-report can be searched at the county level for all 2,617 counties that had enough available data to directly rank.
According to Save the Children, rural child poverty rates are worse than urban rates in 41 of the 47 states with at least some areas designated as rural. Across the United States, almost 84 percent of the counties where the most children struggle with hunger are rural and high poverty.
This article uses the overall child poverty rate Save the Children uses. Broadband Communities notes high school performance separately, however, because broadband availability has a direct influence on that number – an influence that is now even greater because of distance-learning needs during the COVID-19 pandemic. Children in the worst-performing counties, even within the same state, are as much as 14 times more likely to drop out of school or repeat grades. For instance, in Madison Parish, Louisiana, 60 percent of children do not finish high school in four years. In Lincoln Parish, Louisiana, only 4 percent of children fail to graduate on time.
Some other results of poverty are truly eye-opening. More than half of counties where the most children die are rural and high-poverty. In fact, children in the most disadvantaged counties die at rates up to five times those of children in the highest-ranked counties. Virginia is normally thought of as being quite wealthy. But York County there has a child death rate of 27 per 100,000, whereas in Petersburg, a poorer independent city in the state, the rate is 128 per 100,000.
Child hunger rates can vary by a factor of three. In Kentucky, 12 percent of children in Oldham County are at risk of hunger. Shocking enough. But that rate is below the national average and low compared with 30 percent of children in Kentucky’s Clay County.
Broadband can bring these counties more and better jobs, more and better schooling and more and better health care – including mental health care.
Comments
Read what others have to say, and share your own thoughts with the community.