1 Introduction

In advanced industrialized economies, virtual activity has become recognized as a universal fact of everyday life for all but a large minority that has remained unseduced by the charms of the Internet. In North America, Europe, Japan, Korea, or other parts of the industrialized world, after the adoption rate for households has reached slightly above two-thirds, the adoption rate has, seemingly, plateaued.

Within counties, however, the rate may vary some. In America, the shortfall in adoption and use has been more acute in rural areas. Since the early 1990s, rural residents in America have been more likely than urban residents not to use the Internet, and at first, it was clearly an issue of service availability. Often it was not available locally for rural residents, so if they wanted to connect to the Internet, they needed to dial-up through toll and long-distance phone lines, an expensive proposition. As new technologies and business models were employed to connect people, Internet service became cheaper and more readily available to more households. Despite the advances over the last two decades, rural areas continue to lag. The national goal of universal Internet service, thus, remains incompletely fulfilled. The matter of choice, rather than service availability, however, has begun to be recognized as a bigger factor in determining the rural–urban subscription rate differential.

Globally, and nationally, physical distance and accessibility remain vital factors behind digital connectivity (Tranos et al. 2013). The shortfall in Internet accessibility and use will have, potentially, fundamental socio-economic consequences for individuals, businesses, governments, and regions and have been recognized by many governments. Government plans have been proffered, such as the European Commissions’ Europe 2020 Strategy’s Digital Agenda and the U.S. National Broadband Plan, in part to address Internet issues of adoption, availability, and use.

In America, federal and state policies have been implemented to increase Internet access across the country including some new programs designed explicitly to increase household Internet participation. Historically, federal and state government Internet programs have mostly leveraged private funds to increase the availability of broadband Internet service. The Rural Utility Service of the U.S. Department of Agriculture (USDA) has been a lead agency for rural Internet policy implementation through three on-going programs, namely the: (1) traditional federal rural telecommunication infrastructure program requiring all facilities to be broadband capable; (2) farm bill rural broadband program (authorized by the five-year farm bills, the Agriculture Act of 2014 is the latest of these); and (3) Community Connect Broadband Grant Program. The U.S. Department of Commerce-National Telecommunications and Information Administration (NTIA) and U. S. Department of Agriculture-Rural Utility Service (RUS) also jointly administered broadband programs resulting from the American Recovery and Reinvestment Act of 2009 that has led to, approximately, a $7 billion investment in broadband infrastructure. Recently, the Federal Communications Commission (FCC) reformed the Universal Service Fund and created the Connect America Fund that provided $300 million in phase I monies for rural broadband system development. In September 2015, the FCC announced a further $9 billion over six years in phase II monies.

The research presented in this chapter explores the factors associated with rural household subscription of the more advanced broadband Internet services. The research investigates the increase experienced in rural household broadband Internet subscriptions, socio-economic demographics of broadband Internet subscription, and the rural–urban dichotomy of broadband Internet subscription, thus showing that the provision of broadband Internet services may no longer be the critical issue in reaching universality.

2 Background

The National Telecommunications and Information Administration (NTIA) studies in the mid-1990s are the first national-level research that documented the world of Internet users. They also launched the term Digital Divide into its now familiar place in the telecommunications policy lexicon (National Telecommunications and Information Administration 1998; Greenstein and Prince 2006). The American and OECD descriptive studies offered selected national demographics of computer and Internet users and have over the years delivered snapshots of the Internet’s rapid evolution from its humble origins in the academic community.

Recent studies describe the current and more static situation or examine the adoption of the largely pre-broadband era of the Internet. Household studies by Choudrie and Dwivedi (2005, 2006), Stenberg (2008) and the U.S. General Accounting Office (2001) tested socio-economic factors distinguishing adopters and non-adopters of computers and the Internet. Choudrie and Dwivedi (2005) found age, gender, and social grade were important when distinguishing between adopters and non-adopters of the Internet in British households. Their 2006 British study found that characteristics such as income and education were important factors. Nearly all studies on Internet adoption have focused on the household at the national aggregate level. The early NTIA American and the Dutton and Blank (2013) and Farrington et al. (2013) British studies, for example, described differences across many demographic and geographic groupings, not only for households but also Internet activity in the workplace. In Great Britain, Farrington et al. (2013) found that a divide manifested itself mainly between what they called “deep rural” and other areas.

Parker and Hudson (1992) were the first to raise the issue of rural–urban equal access to the Internet though when they wrote their book it was still an era where nearly no one connected to the Internet from the home. The issue that they had preternaturally recognized has continued to this day (Oden and Strover 2002; Stenberg 2013). A study recognizing the importance of rural location, however, found, after controlling for income and education attainment, rural and urban households were almost as likely to use dial-up Internet services (Stenberg 2006).

The issue, however, has further evolved, and it is no longer simply whether a household has Internet service available (Stenberg 2013; Camagni and Capello 2005), whereas at the beginning, it was simply a dial-up connection to the Internet, now the issue has become fuzzier. The technology of the service available for any given household, and hence what may be done on the Internet, has become the issue. The new technology comes under the definitional umbrella of what is called broadband technologies, also known as high-speed technologies.

Like the original dial-up Internet technologies, broadband Internet technologies are not as prevalent in rural areas as in urban areas. Unlike the dial-up technology, the economic challenges of broadband deployment are greater in rural areas. Nonetheless, rural areas have been rapidly receiving the technology. Although rural areas are still catching up to more urbanized areas, another issue has come to the fore—not all households that can acquire a broadband connection actually choose to subscribe. As a consequence, in policy forum discussions, the issue of broadband Internet availability has lessened, while the issue of why some do not choose to subscribe has risen (Malecki 2008; Stenberg et al. 2009; Stenberg and Morehart 2012). We will take a beginning look at these issues in this chapter, showing some of the dimensions of regional variance and basic analysis of the underlying reasons for observing the regional variance that exists.

3 Data Used in the Analysis

In the research presented here, we use data from the Bureau of the Census’ Current Population Survey (CPS). The CPS is a monthly survey of approximately 50,000 households covering various socio-economic characteristics such as family income, employment status, and age. The computer and Internet data that we use, however, are not included in every CPS and have been collected only irregularly over the years as a supplement to one of the monthly surveys.

We use the data aggregated at the household level. The household data are not provided with geospatial coordinates although they are coded with state and metropolitan-status location codes. In other words, we are able to differentiate between households located in urban and rural areas for each state (within limits of any survey and the sampling technique employed). The definitions of metropolitan, aka urban, and nonmetropolitan, aka rural, are those used by the Bureau of the Census as defined by the U.S. Office of Management and Budget (Office of Management and Budget 2013).

Our national estimates are based on statistical analysis of the raw CPS data and a number of CPS weighting protocols. The Bureau of Census over-samples some sub-populations when conducting the CPS. The Economics and Statistics Administration of the Department of Commerce constructs weights for the survey data. The data used in the analysis come from the October 2010 monthly survey.

4 General Trends

Overall Internet subscriptions show the upward trends that other new technologies have shown in the past and resemble perhaps most closely that of cable TV’s subscription growth curve (Stenberg 2006), including the recent plateauing of the subscription rates that many have recognized and can be seen to some extent in Fig. 1. Specifically, rural and urban household Internet subscriptions have increased considerably since 2000 though the rate of increase has slowed down considerably in recent years. And, as is evident, many households still do not subscribe. They do not subscribe either by choice or situation.

Fig. 1
figure 1

Source Stenberg (2013)

Internet access in households, 2000–2010. Note HH means households.

Full size image

Broadband Internet service subscriptions during the same period went from nil to a point where nearly all households that subscribed to the Internet had it through high-speed technologies. Rural–urban spatial differences in overall Internet subscription rates, however, remain with roughly 73% of American urban households subscribing to home Internet connections, while only 62% of rural households do so too.

The technologies for gaining access to the Internet have been changing quite rapidly with a number of alternative broadband technologies becoming available at the same time that dial-up was becoming largely insufficient to all but the most mundane Internet activity. With the increasing sophistication of Web sites and the increasing variety of on-line products and services, accessing the Internet through broadband technologies has largely become viewed as necessary in order to fully utilize what the Web has to offer.

5 Urban–Rural Differences in Subscription Rates

While rural household Internet subscription rates remain low in comparison with urban households, the difference between urban and rural adoption rates is highly variable across the country as can be seen in the following maps, Figs. 2 and 3. The lowest urban Internet subscriptions rates are primarily in the southern part of the country. The lowest rural rates are also in southern states.

Fig. 2
figure 2

Urban households with broadband internet subscriptions by state in 2010

Full size image
Fig. 3
figure 3

Rural households with broadband internet subscriptions by state in 2010

Full size image

Northeastern and western rural households, on the other hand, are, on average, more likely to go on-line than other rural households (Fig. 3). In a number of states, such as Colorado and New Hampshire, the rural household adoption rate exceeds the national urban rate significantly and substantially. The variability in rural rates of adoption suggests that more than rural isolation is at play when it comes to household subscriptions.

Once a household has purchased Internet access, however, they are most likely to have acquired high-speed access; in 2010, 96% of on-line households in urban areas had broadband service, while this penetration rate falls to 92% in rural areas (Fig. 1). This rural–urban difference supports the argument that broadband service is not as readily available in rural areas as compared to urban areas.

Analysis by Stenberg and Morehart (2012) suggests that the conversion to broadband Internet is largely from households that had preexisting, i.e., dial-up, Internet subscriptions and were not likely from households that had no existing Internet service. The analysis held no evidence to suggest that when broadband service became available households dropped their Internet subscription altogether. They either remained with their current service or, as was the most common case, moved up to broadband service. While rural northeastern and western rural households generally have higher broadband subscription rates than other parts of the country, they remain with significantly lower rates than their within region urban counterparts.

Rural broadband subscriptions have become more ubiquitous, but many challenges remain for rural service providers. By the very nature of their low population, rural areas do not exhibit the economies-to-scale that urban areas have. Provision costs, therefore, are higher than in urban areas and so tend to be subscription rates for potential customers. As much as the provider can pass on these costs, the additional costs would make broadband Internet access less affordable for businesses and consumers. Given the relative lack of competition faced by rural service providers and the growing inelasticity of demand for the services, they likely have the ability to pass on the costs (Stenberg 2006). On the other hand, the less provision costs can be passed on to potential customers, the less enticement rural service providers have to provide the service at all. Mountainous terrain and harsh weather present additional challenges in rural areas, driving up the cost of service provision in certain areas. Reliable measures of actual costs faced by rural businesses and consumers, however, are not well known and are the subject of new surveys by a number of researchers.

While broadband has increasingly become available in rural and poor areas, the issue has increasingly become a quality issue. In terms of broadband Internet service, quality means the reliability and speed of data transmission. Rural households rely more often on satellite and wireless connections instead of cable or fiber technologies than households in urban areas. Rural households, when they do subscribe to a land-based hardwired service, also use DSL, a generally slower and, arguably, less reliable technology than cable and fiber, more often than urban households (Stenberg and Morehart 2012).

Market analysis for broadband service provision (as noted by the Federal Communications Commission (FCC) and others): the lower the population density, the poorer the community, or the higher the cost of service delivery cost (due to terrain ruggedness and other challenges), the less quality of service provision. All of the characteristics of population density, relative wealth, and physical topographic challenges commonly hold, but with some notable exceptions, for rural areas. They are also consistent with the existing rural–urban dichotomy in technology provision.

As has been shown in NTIA’s National Broadband Map, Native American reservations, rural poverty counties, and other counties stand out with their lower levels of service. The more densely populated areas, such as the megalopolis stretching from Washington, DC to Boston, have the highest percent of fast (4 Mbps) broadband service for households, while the low population areas, such as the Dakotas, have the lowest percentage. Wilderness areas, such as central Maine and parts of the Rockies, also show the expected low percentage.

6 Self-reported Reasons for Non-subscriptions by the Household

Historically, most of the Internet policy discussion has focused on service availability. With broadband Internet service’s rapid rollout, however, it has been a sufficiently long enough period that most households do have at least one broadband service available, even where a land-based system such as fiber optics or DSL is not available. The FCC has estimated that 95% of all households have broadband service available as of December 2013. As a consequence, policy discussion has begun to shift away from the availability issue.

Policymakers are starting to recognize that not having a home Internet subscription is sometimes by choice whether it be voluntary, in one sense, such as they just do not want it, or involuntary, in another sense, such as when they cannot afford it (Fig. 4). Currently, most households that do not have an Internet subscription do not have it largely by choice.

Fig. 4
figure 4

Most important reason households gave for not having internet service, 2010

Full size image

The largest pluralities of households who do not have Internet subscriptions are those who do not want it. Rural residents, however, are slightly more likely to cite availability, or more precisely the lack of available Internet service, in their area as a reason for not subscribing. The majority of rural households say that they either do not want it or they can use it elsewhere. Only 2% stated broadband service availability as the primary reason for not subscribing.

7 Revealed Factors in Household Internet Subscriptions

Service cost still remains a major reason cited by rural residents for not having Internet access although the decrease in the cost of broadband technologies over the last decade has had a significant impact on increasing Internet use. Federal Internet programs also have increased Internet use. Nevertheless, the pattern exhibited in Fig. 5 suggests that household income plays a significant role in household Internet subscriptions. Rural household Internet access, at any given income level, generally falls below the correspondent urban household Internet access rate. This is one indication that broadband service has not been as readily available in rural areas as in urban areas.

Fig. 5
figure 5

Rural and urban with in-home internet access using any technology, by income, 2010

Full size image

As stated already, once a household is purchasing Internet services, they are most likely to have broadband (Fig. 6). The gap between rural and urban households, however, remains remarkably flat, outside some data sampling noise and the off-campus college student effect at the lowest income level, when controlling for income. The result indicates that expense is not much of a factor after controlling for income, irrespective of the rurality of the household. The gap between urban and rural households would mostly, but not necessarily entirely, be a consequence of service availability. Some of the aggregate differences in adoption rate between rural and urban households would likely be as a consequence of the lower incomes found in rural households’ vis-à-vis urban.

Fig. 6
figure 6

Broadband as a share of in-home internet access by income, 2010

Full size image

8 Logistic Regression Analysis

We explore the issue of multivariate causality in household broadband Internet subscription demand further by means of logistic regression analysis. The method allows evaluation in the light of observed behavior, in this case the selection of broadband Internet subscription. Statistical inferences are made on a model of choice behavior from US household sample data, in this case the October 2010 Bureau of the Census’ Current Population Survey. The Internet adoption decision is qualitative and is postulated to be a choice between not subscribing or subscribing, i.e., having in the home either no Internet or broadband Internet service.

We hypothesize that income, age, rural–urban place of residence, and some other factors are determinants in broadband Internet subscription as some of them have already shown their influence in computer use as well as the early dial-up Internet use. Our null hypothesis is that broadband Internet use is a random event with no determinants. Furthermore, as is often the case when the dependent variable is categorical, the logit model is employed to examine factors that influence Internet adoption. The logistic specification is well suited to this type of application and has been used in similar studies. See, for example, Gloy and Akridge (2000).

It should be noted that estimates of goodness-of-fit are given in model estimations are not given here. R-squared estimates are traditionally given for OLS regressions, but they are much more controversial for categorical regressions as R-squared estimates do not exist per se. Sometimes, pseudo R-squares are estimated to proxy the R-squared value with a number of different methods often used to proxy it in categorical regression models. Nevertheless, R-squared estimates used in logistic regressions are highly controversial, with no broad acceptance of any one estimation methodology over another, and, as many statisticians argue, may be misleading and should only very carefully be used to compare models, if used at all.

9 Model Results

The results of logistic regression are shown in Table 1. The model is significant with an F-statistic greater than 500 and shows its power to predict having a household broadband Internet subscription. All of the independent variables are significant and with the expected sign and show that each is associated with broadband Internet subscription.

Table 1 Logistic regression of broadband adoption
Full size table

The model uses five income groups, each coded with whether the observation belongs to the household income group (less than $25,000, $25,000–$50,000, $50,000–$75,000, $75,000–$100,000, and over $100,000). The results show that the greater the income the more likely a broadband subscription will be obtained and that the upper income group is seven times more likely to have a broadband Internet subscription than the lowest level. Whilst prices may be perceived cheap with many able to purchase broadband service for less than 50 U.S. dollars a month, purchasing the personal computer, software, and other equipment can still be prohibitive for households of lesser means.

The greater the educational attainment, the more likely the household would purchase an Internet connection. In the model, we have educational attainment in four groups: no high school diploma, high school diploma, some college attainment, and attainment of a bachelor’s degree or more. College-educated households are five times more likely to have home broadband subscription than those who have not obtained a high school diploma.

In the model, age is a continuous variable ranging from three years of age to 85. Age has a unique property that differs from income and education. As people get older, they are more likely to have an Internet subscription or at least until they reach retirement age when the likelihood starts to decrease. This is a bell-shaped curve. As a consequence, we model age with two variables: age and age-squared. These two parameters largely capture this nonlinear effect. The results for the age parameters are consistent with the bell-shaped curve and the expected positive effect of age on adoption generally. The results, however, show that this factor may not be as important as it has been noted in past research, such as Choudrie and Dwivedi (2006), with age only slightly increasing the odds if having broadband with the odds diminishing less past retirement age.

School-age children, with their exposure to the Internet in their schools as well as their increasing need to get on-line for school assignments and instruction, have been recognized, at least anecdotally, as major demand drivers for household broadband service subscriptions. The model finds that the presence of children in the household significantly increases the odds of having a broadband subscription in the household.

Stability in home environment has been postulated as a factor in adopting in-home broadband service. In the model, we use residency in a house or apartment as a proxy for this possible factor. The results indicate that living in house or apartment significantly increases the odds of having a broadband subscription.

People accessing the Internet away from home have been suggested as more likely to buy home subscriptions. Often the access is through the workplace and, to a lesser extent, libraries. The argument is as follows: the more people are familiar with what the Internet has to offer, the more they want to have greater access to it. The access away from home variable shows a significant positive relationship with household subscription.

Anecdotally, it has been noted that migrant farmworkers are often not citizens and do not have broadband subscriptions. They do, however, communicate with home through smartphones. We use a non-US citizen as a proxy for this. This is not the best proxy, but the results seem to be consistent with this observation; non-citizens are much more likely not to have a broadband subscription.

We have two spatial factors in the model: (1) rural–urban location and (2) geographic region of the country. In the first, the factor is whether a household is located in a rural location or an urban location, as defined before. The model clearly shows rural households are less likely than urban households to have adopted broadband Internet. As we control for the lower income, educational attainment, and other differences in rural vis-à-vis urban households, the negative sign for rural location a sign of rural households has less broadband service availability.

Regions of the country, the Northeast, Midwest, South, and West Census Regions (U.S. Census Bureau 2012), are different, with different rural–urban spatial patterns. That pattern clearly shows in the model results. The Northeast has the most densely populated rural areas of the four regions and thus, relative to the nation, higher rates of rural household broadband service availability. The West, however, has the most urbanized rural area of the four regions and, as a consequence, has the greatest likelihood of household broadband Internet subscription. The Midwest and South have the most spread out of rural populations. Thus, if you are a Midwestern or Southern rural household, you are likely to have a greater challenge in obtaining a broadband subscription.

10 Conclusions

Obtaining broadband Internet service remains more challenging in rural areas. Rural households still are less likely to have broadband Internet service available to them than their urban peers, but nearly all rural households that have the Internet, like their urban brethren, use broadband technologies from the home.

Service availability of broadband, however, is no longer the primary reason for not having home Internet subscriptions as it has now come to pass that a much larger share of households without the Internet choose not to subscribe rather than cannot subscribe. Much of the rural–urban household variance in broadband adoption rates can be explained by the variance in household characteristics. Rural households tend to have less income and lower educational attainment, on average, than urban households. Broadband availability though greatly increased, however, remains a significant negative factor for rural residents in being able to purchase the services.

The rural underperformance in Internet use will have fundamental socio-economic consequences for individuals, businesses, governments, and regions. Federal and state policies continue to address broadband availability, but policymakers have begun to recognize this, so some new programs diverge from the original Internet infrastructure model and are designed explicitly to increase household Internet participation.

Irrespective of location, be it North America, Europe, Japan, Korea, or other parts of the industrialized world, one fact is universal—after the adoption rate has reached around two-thirds, adoption rate has plateaued. In order to understand why it has plateaued, more research is needed to better understand the dynamics of household choice with respect to broadband Internet services to better illuminate the policy discussion and program development. Ideally, this would be through longitudinal studies, but, given the currently short data time frame, other techniques will continue to offer analysis to better guide policy.