The educational impact of broadband subsidies for schools under E-rate

ABSTRACT

In 1998, the U.S. began spending about $2 billion annually to help fund computer access in elementary and secondary schools. In 2013, the Federal Communications Commission, citing the experience of a school district in North Carolina, increased these annual ‘E-Rate’ subsidies to $4 billion. Do such expenditures actually improve academic achievement? We estimate a model wherein SAT scores, a proxy for student performance, are a function of explanatory factors including federal broadband funding. Examining data from all North Carolina public high schools, 2000–2013, we find no gain in student test results associated with Internet subsidy levels.

KEYWORDS:

• E-rate
• broadband subsidies
• education policy
• internet access
• universal service

Acknowledgements

The authors thank Katie Holba for outstanding research assistance.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 Education Superhigherway Annual Report, 2016, report that 88% of school districts had an Internet connection of 100 kbps/student in 2016, up from 30% in 2013.

2 For example, see Machin, McNally, and Silva (2007); Barrow, Markman, and Rouse (2009); Li, Atkins, and Stanton (2006); and Silva, Milkman, and Badasyan (2016).

3 For example, Belo, Ferreira, and Telang (2014); Leuven et al. (2007).

4 Goolsbee and Guryan (2006).

5 ConnectED: President Obama’s Plan for Connecting All Schools to the Digital Age, White House (June 6, 2013); https://www.whitehouse.gov/sites/default/files/docs/connected_fact_sheet.pdf.

6 From 1998 through 2010, the program was capped at $2.25 billion annually. It was then bumped up to $2.4 billion to adjust for inflation. See FCC (2014). For an explanation of how the spending levels were originally set, pursuant to the 1996 Telecommunications Act, see Furchtgott-Roth (2006).

7 Libraries also receive subsidies, but these constitute only a small fraction of the total disbursements.

8 The FCC officially refers to the ad valorem tax as a ‘contribution.’ Riordan (2001, fn 20) describes the funding source to be ‘revenue-based universal service “contributions” (i.e. revenue taxes).’ Jerry Hausman explains the semantics: ‘The FCC is only allowed to assess “fees,” not taxes. However the FCC fees have the same economic effect as a tax.’ (Hausman 2000, 733).

9 See the interesting historical review in Angrist and Levy (2002).

10 Cox 2017.

11 Angrist and Levy (2002) were also among the first to address endogeneity by exploiting the random nature the subsidies were distributed among schools and a two-stage regression model.

12 They examined the relationship between E-Rate subsidies and the number of classrooms that were connected to the Internet, and the relationship between connected classrooms and educational achievement increases. The study yielded a positive relationship between the subsidies and connections – that is, E-Rate grants appeared to increase classroom broadband penetration. At the same time, it found that the Internet deployment gains had no visible effect on educational results as measured by the Stanford Achievement Test, which is administered to every public school student in California.

13 Leuven et al. (2007) explain that the program provided technology subsidies to schools in which at least 70% of the students were ‘disadvantaged’ based on the number of students classified as disadvantaged. The program provided a one-time subsidy of $90 per student in the school to purchase computers, software, and language materials. They used a regression discontinuity design to test the effects of the subsidies.

14 The authors used an instrumental variable method that exploits a policy change. The authors argue that the positive effects they identified is the result of certain institutional foundations at the school that allowed schools to make better use of the funding. In particular, they find larger effects on English scores at schools with lower overall spending per student but ‘better educational standards as measured by exam pass rates and truancy rates).’

15 Barrow, Markman, and Rouse (2009), 54.

16 Barrow, Markman, and Rouse (2009) used a randomize study in which some classrooms used the program and others used traditional teaching methods. They ‘hypothesize this effectiveness arises from increased individualized instruction as the effects appear larger for students in larger classes and in classes with high student absentee rate’ (52).

17 More than 85% of our observations report that 100% of their classrooms are connected to the internet.

18 While E-Rate funds are not eligible for computer purchases, schools receiving subsidies may be incentivized to increase computer usage given that bandwidth (with subsidies) is increasing.

19 North Carolina does have an end of year exam for certain subjects that are required for students who have taken particular courses. The number of these classes vary year-to-year through our sample period, however.

20 Some school districts appear to interpret these categories broadly, which implies that the federal restrictions are imperfectly enforced. The Montgomery County (Maryland) Public School system has used E-Rate funds to purchase interactive whiteboards. As the district noted in a 2012 memo,

Promethean Boards are the primary whiteboard components of interactive classroom technology systems which include other elements such as student response units. MCPS began implementing an interactive technology initiative in school year 2008–2009, acquiring 2,600 board systems and installing them in about two-thirds of all secondary classrooms. This initial phase was funded primarily by Federal E-rate reimbursement and financed over several years. MCPS made the final payment on the secondary classroom Promethean Boards in FYI2.

McGuire, Essie (Senior Legislative Analyst), ‘Memo to Education Committee, Subject: Worksession – Special Appropriate to the FY13 Capital Budget … .’ http://www.montgomerycountymd.gov/COUNCIL/Resources/Files/agenda/cm/2012/121105/20121105_ED1.pdf (last accessed March 2, 2016).

21 We observe the districts filing the application. We do not observe which schools are part any particular application for a district.

22 There were 310 schools during the 2001–02 school year, increasing to 435 for 2012–13.

23 The SAT consists of three sections: a math section, a comprehensive reading section, and a written section (the written section was added in 2005). Each section is graded out of a possible 800 points.

24 USAC is a non-profit corporation created by the FCC (with officers appointed by the FCC) to administer USF funds at its direction. The FCC uses USAC data to set the tax rate every quarter https://www.fcc.gov/general/contribution-factor-quarterly-filings-universal-service-fund-usf-management-support.

25 We use the log of SAT scores rather than the change in levels that Goolsbee and Guryan (2006) used. The elasticity of SAT scores with respect to E-Rate spending is then readily interpreted.

26 There are other factors that affect SAT performance at the individual level. Freedle (2003) points out that one needs to correct for ethnic and social class bias. Zwick and Sklar (2005) found similar effects as Freedle, but focused on the impact of first language. Since our level of observation is the school average, this should be accounted for by school/district effects as long as the demographics are relatively consistent over time.

27 We also estimated the model with three lagged terms. In the OLS model, the effects for Year 1 and 2 tend to be marginally statistically significant, negative, and small.

28 The authors are indebted to an anonymous referee for making this suggestion.

29 The idea being that access to the internet may offset other challenges that the school might face based on budgetary constraints.