Information Economics and Policy 40 (2017) 21–25

Contents lists available at ScienceDirect

Information Economics and Policy

journal homepage: www.elsevier.com/locate/iep

The faster, the better? The impact of internet speed on employment

Yang Bai
College of Communications, Pennsylvania State University, 201 Vairo Blvd, State College, 16803 PA, United States.

a r t i c l e i n f o a b s t r a c t

Article history: Does faster broadband generate more economic benefits than normal-speed broadband? This study ex-
Received 15 May 2016 amines the differential economic impacts of broadband of different speeds. Utilizing the broadband avail-
Revised 18 December 2016
ability and speed data from the National Broadband Map, a first-differenced regression analysis was con-
Accepted 6 June 2017
ducted to investigate if faster broadband more effectively boosts the employment. The analysis shows a
Available online 12 June 2017
positive relationship between broadband availability and county-level employment. However, compared
Keywords: to normal-speed broadband, faster broadband did not generate greater positive effects on employment.
Broadband © 2017 Elsevier B.V. All rights reserved.
Download speed
Employment

From the dial-up connection to Fiber-To-The-Home, the Inter- not significantly different from what normal broadband can pro-
net, as Firth and Mellor (2005) pointed out, has increasingly re- duce, it may be more cost-effective to invest more in the normal-
ferred only to the broadband. Not surprisingly, scholars who in- speed broadband.
vestigate the economic impact of ICTs also shift their focus from This study explores the issue of the economic return of In-
the general Internet to the broadband, as if the broadband, as if ternet speed. The question I seek to answer is: as broadband
it is one type of technology. Their enthusiasm for the promising with increasingly higher speed is deployed, does the economic
future of the broadband is not unfounded. Many empirical studies gain which is supposed to follow rise accordingly? As an ex-
do show a significantly positive causal relationship between broad- ploratory research, this study examines the relationship between
band and economic growth (Gillett et al., 2006; Jayakar and Park, the county-level employment rate and broadband availability. Dif-
2013; Czernich, 2014). However, the treatment of broadband as one ferent from many previous studies which treated broadband in
technology may give us a false optimism about the economic gains general, the current research disaggregates the broadband into
it can generate. In fact, the broadband Internet is an umbrella term three sub-categories: superfast broadband (download speed higher
which includes several types of infrastructures. Currently, FCC has than 1 GB/s), fast broadband (download speed between 100 MB/s
recognized six types of broadband technologies: DSL, cable mo- and 1 GB/s) and normal broadband (download speed between
dem, fiber optics, satellite and broadband over power lines. Given 3 MB/s and 100 MB/s). This treatment enables the test for any dif-
that these technologies differ in many aspects such as the difficulty ference in economic impacts of various types of broadband so that
to maintain and upgrade, costs to deploy and use, and the appli- a more nuanced picture of the role played by the information
cations enabled, their economic impact may also differ from each super-highway can be drawn.
other. The article is structured as follows. In the next section, the re-
However, very few studies have studied the potential existence lated literature on the job- creation effect of the Internet and the
of such non-uniformity in the economic impact of broadband. The return of Internet speed are reviewed. Based on the previous lit-
assumption that faster Internet will yield greater economic gains erature, two hypotheses are framed. Then, I describe the sampling
seems to be taken for granted by many national and international strategy, data source and methodology used in this study. There-
policy and business organizations (UNCTAD, 2012; The Executive after, the analyses and results are presented. A summary of the
Office of the President, 2015; FCC, 2015). On the other hand, ex- main findings and the implications of the study are discussed in
cept for case studies and anecdotal evidence (Ericsson et al., 2013; the last section.
Haller & Lyons, 2015), few empirical studies exist which shows
faster Internet can lead to greater economic gain at the macro Literature review
level. Such a lack of empirical evidence should not be taken lightly.
If the economic benefits produced by faster broadband are in fact Broadband and employment

The impact of Internet or broadband availability on employ-

E-mail address: [email protected] ment has long been recognized in the economic literature. How-

http://dx.doi.org/10.1016/j.infoecopol.2017.06.004
0167-6245/© 2017 Elsevier B.V. All rights reserved.
22 Y. Bai / Information Economics and Policy 40 (2017) 21–25

ever, the increased research on this topic only seems to have deep- Cho et al. (2003) surveyed 2206 Internet users in South Korea. Af-
ened the mystery, since evidence for both a significantly positive ter comparing the activities the users of dial-up, ADSL, and cable
impact and non- significant, or even negative impact has been Internet, the researchers found that, although the high-speed In-
found. ternet users used online media, such as online TV and online mu-
A study analyzing the early years of broadband diffusion, i.e., sic, more often than the low-speed users, they did not differ in
1998–2002, detected a significant and positive relationship be- online shopping behaviors. This asymmetrical impact of Internet
tween broadband deployment and growth in employment and speed is also discovered by Kwak et al. (2004). The researchers ex-
the number of new businesses (Gillett et al., 2006). Utilizing the amined two models explaining the social consequences of Inter-
county-level data provided by National Broadband Map, a more re- net adoption and found that the behavior impact of broadband is
cent study by Jayakar and Park (2013) shows a significantly nega- not a uniform one. A linear model is found to explain effectively
tive relationship between broadband deployment and the unem- the impact of the transition to faster Internet for soft engagement,
ployment rate at the county level. Instead of analyzing the eco- such as casual socializing and informal knowledge seeking. How-
nomic impact of broadband in general, Lapointe (2015) focused ex- ever, hard engagement, such as formal knowledge seeking, political
clusively on the employment impact of the fiber Internet at the discussion and civil engagement, is better explained by a differen-
county level and discovered a positive relationship. tial gain model: while the introduction of the Internet generated
Though many studies have shown a universal, positive effect a significant increase in hard engagement activities, such a strong
of broadband on employment, closer examinations have revealed increase halted after people adopted broadband.
that the effect can be conditional and vary across different re- In addition to enabling various activities, the high-speed Inter-
gions and industries. Forman et al. (2012) found that investment net is also regarded as a productivity enhancer for firms (UNCTAD,
in broadband led to employment growth only in economically de- 2012; Ericsson et al., 2013). However, such a claim is mostly sup-
veloped regions with ICT-skilled, high-income populations. After ported by case studies and ethnographies (Haller & Lyons, 2015) or
analyzing the data on employment in non-urban counties in the studies failing to distinguish different types of broadband access
U.S., Whitacre et al. (2014) found no relationship between broad- (Najarzadeh et al., 2014). The non-representative samples studies
band availability and job creation. Scholars have also researched in ethnographies could potentially lead to biased evaluations, and
the industry-specific impact of broadband and found that the job- examining the broadband as a whole may over-generalize the ben-
creation effect existed mainly in technology-concentrated indus- efits engendered by one type of broadband and the positive effects
tries (Kolko, 2012) and high-end service industries such as fi- on one industry or even one firm. In fact, several studies analyz-
nance and insurance,education services and health care services ing representative samples of firms have shown a more complex
(Crandall et al., 2007). Based on the findings in the previous litera- picture. In a longitudinal econometric study examining 790 Ital-
ture, there is no consensus on the impact of broadband availability ian small and medium enterprises, Colombo et al. (2012) found
on employment. Accordingly, the following hypothesis is proposed: that, unless the firm invests heavily in employee ICT skill training
and adjusts its management and operations, the adoption of ap-
H1: Broadband availability has no significant effect on county em- plications enabled by fast broadband had extremely trivial, and, in
ployment rate. some cases, even negative impact on the firm’s productivity. Simi-
larly, Grimes et al. (2009) compared the productivity of 6060 firms
Noticeably, most of the aforementioned studies examine the
in New Zealand. After controlling for the firm’s idiosyncratic char-
impact of broadband in general without distinguishing different
acteristics, they found that though firms using normal broadband
types of broadband. However, broadband Internet is an umbrella
outperformed the ones using dial-up or no Internet by 7%–10% in
term which covers a variety of technologies. Among a limited num-
terms of labor productivity, no significant increase was discovered
ber of studies investigating the differential impact of broadband
between firms adopting the fast broadband and the normal broad-
technologies, most support an increasing return of Internet speed:
band.
compared to the current-generation broadband, a faster Internet
Though these findings may frustrate fast broadband advocates,
can benefit the economy even more. Singer and West (2013) es-
Howell and Grimes (2010) pointed out several potential reasons
timated that, while 40,0 0 0 jobs per year would be created if all
why the scholars failed to find the expected results. First, as
Americans had access to normal-speed broadband by the end of
Colombo et al. (2012) concluded, the adoption of broadband alone
2015, 250,0 0 0 more jobs per year could be created if the next-
cannot guarantee an increase in productivity. It takes years for the
generation broadband were made accessible to 80% of American
potentials of the technology and proper training to manifest. Sec-
households. Shapiro and Hasset (2012) explored the job-creation
ond, the benefits may be at some other stages along the produc-
effect of mobile broadband. While the transition from 2G to 3G In-
tion value chain outside the firm and, thus, are not measured by
ternet is associated with the creation of 1585,0 0 0 jobs from 20 04
the firm’s productivity. Third, the impact of broadband adoption
to 2011, a 10 percent increase in 3G/4G adoption rate could cre-
is industry-specific. It is very well possible that the broadband in-
ate 231,0 0 0 more jobs in less than a year. Though these figures are
creases the productivity of firms in the knowledge and information
exciting, one should not forget that faster Internet, while creating
intensive sectors while having little impact on others. Such a dif-
new employment in some sectors, can also eliminate job opportu-
ferentiated impact may not be detected in an analysis using a rep-
nities in others, such as company management and art, recreation
resentative sample including all types of firms. Based on the find-
and entertainment (Crandall et al., 2007). Therefore, the net and
ings on the incremental effect of Internet speed on employment,
incremental effects of faster Internet on employment still need fur-
the following hypothesis is formulated:
ther investigation.
H2: The economic benefit generated by fast broadband is not sig-
nificantly different from that produced by lower-speed broadband.
Diminishing return of internet speed at the individual and firm level
Empirical studies on the relationship between broadband dif-
Though broadband is widely believed to facilitate the adop- fusion and employment are abundant; however, few consider the
tion of various new types of uses and services (Ericsson, Arthur D. possible existence of differential effects of different broadband
Little & Chalmers University of Technology, 2013), studies have technologies. Though, at the level of individuals and firms, schol-
shown the adoption and use of different online services are ars have discovered evidence of diminishing return of broadband
not equally boosted by the increasing speed of broadband. speed, the nature of the micro level analysis may blind us to
 Y. Bai / Information Economics and Policy 40 (2017) 21–25 23

Table 1 data at the county level. Therefore, this study used the per capita
States Samples.
expenditures of the state government to indicate the level of public
Rich Poor services. Using the expenditures of the state government as such
Urban Illinois (102, $47,643, 88.5%) Oregon (36, $41,220, 81.9%) an indicator is common in county-level analysis (Whitacre et al.,
New York (62, $55,611, 87.9%) Utah (29, $37,664, 90.6%) 2014), since the relevant data for counties is often not available.
Rural Nebraska (93, $47,557, 73.1%) Idaho (44, $36,734, 70.6%) Table 2 gives a summary of the variables and descriptive statistics.
North Dakota (53, $55,802, 59.9%) Oklahoma (77, $43,637, 66.2%)
∗
The figures following each state indicates the number of counties, per capita in- Method
come and percentage of population living in urban areas.

To control for the unobserved time-invariant effects, a first-

the externality and overall social value created by the high-speed differenced model was used. Though the fixed effect estimator
broadband. Therefore, an econometric study was conducted to ex- might be a more popular model for panel data analysis, it is not
plore the nuanced effects on employment of different types of preferable in our case. Fixed effects estimator is more efficient if
broadband characterized by their download speeds. the error term displays no serial correlation, and this assumption
may not be true in many situations (Woodridge, 2009). In order to
Data and method test if a fixed effect model should be used, a Durbin–Watson test
was conducted, and a DW statistics of 0.868 was obtained. This in-
In order to obtain a representative sample of counties, a strat- dicates a positive serial correlation of 0.566 (p < 0.01). This signif-
ified random sampling technique was used. First, all states of the icant, positive serial correlation shows that the first-difference es-
U.S. were assigned an ID number and categorized into one of the timator is better than the fixed effects model. Moreover, no serial
four groups: rural rich, rural poor, urban rich and urban pool. Ac- correlation in the first-differenced errors was detected (DW = 2.41,
cording to U.S. Census Bureau (Census), on average, 80.7% of the p < 0.01), which supports the selection of the first-differenced es-
American population live in urban areas. Therefore, states with timator. Therefore, the primary model can be written as:
higher than 80.7% of the population living in urban areas were D_EMPit = β 0 + β 1 D_Superfastit + β 2 D_Fastit
classified as urban states. In 2014, the average per capita income
was $46,049. Thus, rich states were defined as the ones with per
+β 3 D_Normalit + β 4 D_Xit + uit
capita income above the national average. The ID numbers of the where D_ EMP it is the change in the employment rate for each
states were entered in the online random number generator, Ran- county, D_ Superfast it D_ Fast it and D_ Normal it are the changes
dom.Org. For each category, two random numbers were generated, in the availability of Superfast, Fast and Normal broadband, respec-
and the corresponding states were selected. Then we collected data tively, and D_ X it is a vector of changes in the control variables.
for the counties within each selected state. The detailed informa-
tion for the selected states are summarized in Table 1. Results
A panel was then constructed for the 496 counties sampled
from 2011 to 2014. County annual unemployment rates were Before the analysis, the data was tested for the existence
retrieved from the Local Area Unemployment Statistics (LAUS) of non-normality and outliers. After running the check, non-
database of the Bureau of Labor Statistics (BLS). According to the normality was detected. All IVs are slightly skewed, with skewness
BLS, the employment rate is calculated as the percentage of the and kurtosis statistics smaller than 3. According to Kline (2016),
job-seeking population who are employed. Therefore, the employ- asymmetry and kurtosis between −3 and 3 are acceptable. Also,
ment rate was calculated as 1–unemployment rate. the violation of normality assumption does not lead to biased es-
Data on broadband availability and connection speeds was ob- timates in a first-difference model, and, given a large sample size,
tained from National Broadband Map and State Broadband Initia- the inferential statistics are still valid with the presence of non-
tive. In order to test for the incremental effects of Internet speed, normality (Woodridge, 2009). Therefore, no variables were trans-
the availability data was further disaggregated into four categories. formed. The result of the regression analysis was presented in
Superfast measures the percentage of the population with access Table 3.
to the Internet with 1 GB/s or faster download speed. Fast mea- The analysis shows that increases in the availability of all three
sures the availability of the Internet with 100 MB/s to 1 GB/s down- types of broadband is positively related to the changes in county
load speed. Normal measures the availability of the Internet with employment rate. The more population have access to broadband
3 MB/s to 100 MB/s download speed. NO, as a reference group is Internet, the higher employment rate the county will have. There-
the percentage of the population with no broadband or Internet fore, the first hypothesis is rejected. As for the control variables,
access. increases in state unemployment rate, the percentage of the pop-
A total of 6 control variables were included to control for other ulation with at least high school diplomas and minimum hourly
socioeconomic factors that may impact employment. Several stud- wage are found positively related with changes in the county em-
ies have indicated that percentage of the working-age population, ployment rate, while the percentage of the minority population is
defined as the population aged between 20–59, the percentage negatively associated with the county employment rate. There is
of minorities and education can influence the employment rate also a trend toward significance that higher government expendi-
(Whitacre et al., 2014; Czernich, 2014). Data on these demographic tures are related to more employment.
variables was obtained from the database of Census. County em- The second hypothesis is the main focus of this study. It is hy-
ployment rate and minimum wage were also found to be related pothesized that the effects generated by faster Internet are not dif-
to state-level employment statistics (Jayakar & Park, 2013). To con- ferent from what the lower-speed Internet generates. The analy-
trol for this potential relationship, the annual employment rate for sis shows a more complex picture of the effect of higher Inter-
each state and the state minimum wage were also included in the net speed. While the positive impact of the fast broadband, de-
model. fined as the Internet with 100 MB/s to 1 GB/s download speed, is
The expenditures of local government are also a potential factor greater than that of the normal broadband, the impact of the su-
which can impact the employment rate, since it can indicate if the perfast broadband on employment is smaller than the effects of
area has good utility, transportation, education and other public fast broadband. This result suggests a diminishing return of Inter-
services. However, only 3 of the 8 states selected provide financial net speed on employment. Compared to normal broadband, fast
24 Y. Bai / Information Economics and Policy 40 (2017) 21–25

Table 2
Variables and descriptive Statistics.

Variables Labels Mean SD Year N Source

Dependent variable emp County employment rate 93.7% 0.03 2011–2014 1984 BLS
Independent variables superfast % county population with access to > 1 Gb/s download speed 9.0% 0.22 2011–2014 1984 NBM/SBI
fast % county population with access to 100 Mb/s to 1 Gb/s download speed 15.5% 0.28 2011–2014 1984 NBM/SBI
normal % county population with access to 3 Mb/s to 100 Mb/s download speed 71.2% 0.34 2011–2014 1984 NBM/SBI
no % county population with no broadband or Internet access 4.3% 0.10 2011–2014 1984 NBM/SBI
workingage % county population aged between 20–59 47.9% 0.06 2011–2014 1984 Census
highsch % county population with high school or higher education 81.3% 0.06 2011–2014 1984 Census
minority % county minority population 10.5% 0.12 2011–2014 1984 Census
minwage State minimum wage 7.59 0.54 2011–2014 1984 BLS
exppc State per capita government expenditure ($10 0 0) 6.54 1.53 2011–2014 1984 Census
stateemp State employment rate 94.2% 0.02 2011–2014 1984 BLS
∗
Strongly balanced panel, year = 2011, 2012, 2013, 2014.

Table 3
First-differenced regression results.

First difference estimator Heteroskedasticity–Robust SE

D_Superfast 0.020 0.0049∗ ∗ ∗

D_Fast 0.024 0.0047∗ ∗ ∗
D_Normal 0.015 0.0044∗ ∗ ∗
Workingage 0.036 0.0119∗ ∗
Highsch 0.043 0.0164∗ ∗
Minority −0.029 0.0072∗ ∗ ∗
Minwage 0.015 0.0021∗ ∗ ∗
Exppc 0.002 0.001 +
Stateemp 0.456 0.0332∗ ∗ ∗
Constant 0.0 0 0 0.0 0 03
R-Saqure = 0.395
Adjusted R-Squared = 0.392
∗∗∗ ∗∗ +
p < 0.001, P < 0.01, P < 0.1.

broadband was more effective in boosting county employment. On Fig. 1. Proportion overlap (POL) and the Rule of 50% Overlap. ∗ In Pair C, a POL of
the other hand, superfast broadband, compared fast broadband, 59% corresponds to a p value of 0.05. As a general rule, a POL smaller than 50% shows
produced smaller, though still positive, effect on employment. In two independent means are different at p < 0.05. See Cumming (2009) for more details.
the next section, two techniques were used to further test whether
the differences in the effects of the three broadband variables on
employment are statistically significant. increased by 13.5) and 0.12 when D_ Superfast was included (F in-
creased by 38.2).

The significance of the differences

Conclusion and discussions
To test if the three coefficients are statistically differ-
ent from each other, the rule of 50% overlap developed by Studies on the relationship between broadband availabil-
Cumming (2009) was applied. First, the confidence intervals for the ity/penetration and employment abound. Though many of them
three beta coefficients were obtained through bootstrapping. Sec- have shown a positive job-creating effect of the broadband, an
ond, the proportion overlap (POL) between two CIs was calculated. equal amount of evidence of no significant, or even negative im-
If the proportion overlap is smaller than 50%, then the two inde- pact is also found. The current study utilizes the annual county-
pendent means are significantly different at 0.05 level. Fig. 1 illus- level data provided by the National Broadband Map and State
trates three possible scenarios where the two independent means Broadband Initiatives and analyzes the employment impact of
are different at the 0.2, 0.1 and 0.05 level, respectively. broadband. As previous studies suggest, a variety of factors may
Using the confidence intervals given by the bootstrapping (20 0 0 impact local employment rate, and some of those factors may not
bootstrap samples), the POLs for the three broadband coefficients be observable. To control for the unobserved effects and other re-
were calculated. Specifically, the POL between D_ Fast and D_ Nor- lated factors, a first- differenced model was used, with other de-
mal is 100%, the POL between D_ Fast and D_ Superfast is 166%, mographic and employment-related variable included. The results
and the POL between D_ Normal and D_ Superfast is 150%. Thus, of this study show that broadband availability is positively associ-
based on this result, the three coefficients are not statistically dif- ated with the county-level employment.
ferent. In other words, the impacts of the three types of broadband A limited number of studies have differentiated the impact of
on county employment rate are not significantly different. different types of broadband. Therefore, the main contribution of
It is worth emphasizing that although the changes in the avail- this study is that it investigates the non- uniformity in the impact
ability of fast and superfast broadband produced no stronger posi- of different types of broadband technologies. By focusing on the
tive effects than did normal broadband, this fact does not mean the number of new jobs created, previous studies found evidence for
faster broadband are trivial in terms of boosting the county em- an increasing return of Internet speed: faster Internet can gener-
ployment. In order to test for the significance of the contribution ate more jobs than lower-speed broadband can (Singer and West,
to county employment produced by the fast and superfast broad- 2013; Shapiro and Hassett, 2012). However, focusing on the aggre-
band, a stepwise regression was conducted. The result shows a sig- gate employment data, this study found that the impacts of the
nificance increase of 0.08 in the R2 when D_ Fast was included (F super-fast and fast broadband on county employment rate are not
 Y. Bai / Information Economics and Policy 40 (2017) 21–25 25

significantly different from the benefits generated by the normal- Colombo, M., Croce, A., Grilli, L., 2013. ICT services and small businesses’ produc-
speed broadband. Noticeably, this finding is not necessarily at odds tivity gains: an analysis of the adoption of broadband internet technology. Inf.
Econ. Policy 25 (3), 171–189. doi:10.1016/j.infoecopol.2012.11.001.
with the previous studies. Faster broadband may create many new Czernich, N., 2014. Does broadband internet reduce the unemployment rate? Evi-
jobs in some sectors, but at the same time, it also can eliminate dence for Germany. Inf. Econ. Policy 29, 32–45. doi:10.1016/j.infoecopol.2014.10.
many jobs in other sectors, as shown in Crandall et al. (2007). Fo- 001.
Ericsson, Arthur D. Little, Chalmers University of Technology, 2013. Socioeconomic
cusing on the aggregate employment rate, this study cannot de- Effects of Broadband Speed Retrieved from http://www.ericsson.com/res/
tect if there exist differentiated impacts of broadband with vari- thecompany/docs/corporate- responsibility/2013/ericsson- broadband- final-
ous speeds on different industries. Also, faster Internet may have 071013.pdf.
Firth, L., Mellor, D., 2005. Broadband: benefits and problems. Telecommun. Policy
larger impact on some sectors such as entertainment, banking and
29 (2), 223–236. doi:10.1016/j.telpol.20 04.11.0 04.
education, which typically have fewer employees than traditional, Forman, C., Goldfarb, A., Greenstein, S., 2012. The internet and local wages: a puzzle.
labor-intensive sectors. It is possible that faster Internet did gen- Am. Econ. Rev. 102 (1), 556–575. doi:10.1257/aer.102.1.556.
Federal Communication Commission, 2015. A report on consumer fixed
erate more employment than slower Internet did in some sectors.
broadband performance in the United States Retrieved from https:
However, due to the small size of the workforce in those sectors, //www.fcc.gov/reports-research/reports/measuring-broadband-america/
the significant incremental effect of faster speed could be diluted measuring- broadband- america- 2015.
in an economy where the traditional, labor-intensive industries still Grimes, A., Ren, C., Stevens, P., 2009. The need for speed: impacts of internet con-
nectivity on firm productivity. J. Productivity Anal. 37 (2), 187–201. doi:10.1007/
dominate. Unfortunately, the data on industry-specific employment s11123-011-0237-z.
rate and sectorial distribution of workforce is not available at the Gillett, S.E., Lehr, W.H., Osorio, C.A., Sirbu, M.A., 2006. Measuring The Economic
county level, which prevents the researcher from further examina- Impact Of Broadband Deployment Final Report, National Technical Assistance,
Training, Research, and Evaluation Project No. 99-07-13829.
tion. Howell, B., Grimes, A., 2010. Productivity questions for public sector fast fibre net-
From a policy-making perspective, this study provides a differ- work financiers. Commun. Strategies 1 (78), 127–146.
ent view on the national or local level broadband promotion strat- Haller, S.A., Lyons, S., 2015. Broadband adoption and firm productivity: evidence
from Irish manufacturing firms. Telecommun. Policy 39 (1), 1–13. doi:10.1016/j.
egy. A broadband network can be built using different technolo- telpol.2014.10.003.
gies ranging from the cable modem to fiber optics, each with dif- Jayakar, K., Park, E., 2013. Broadband availability and employment: an analysis of
ferent costs to install, maintain, upgrade, etc. For instance, data county- level data from the national broadband map. J. Inf. Policy 3 181–200.
doi:10.5325/jinfopoli.3.2013.0181.
from the Costs Database of the U.S. Department of Transportation
Kwak, N., Skoric, M.M., Williams, A.E., Poor, N.D., 2004. To broadband or not
shows that the average per-mile cost to install a fiber-optic net- to broadband: the relationship between high-speed internet and knowledge
work is $42,900, while the average per mile cost of building a DSL and participation. J. Broadcast. Electron. Media 48 (3), 421–445. doi:10.1207/
s15506878jobem4803_5.
network is only $4500. Based on the finding of this study, faster
Kolko, J., 2012. Broadband and local growth. J. Urban Econ. 71 (1), 100–113. doi:10.
broadband, which is, in most cases, more expensive to build, is not 1016/j.jue.2011.07.004.
better than its normal-speed counterpart in boosting local employ- Kline, Rex.B., 2016. Principles and Practices of Structural Equation Modeling, fourth
ment. Therefore, for the public sector, investing in faster broadband ed. The Guilford Press, New York, NY.
Lapointe, P., 2015. Does speed matter? The employment impact of increasing access
may not always be the best option when cost-efficiency is consid- to fiber Internet. Washington Academy of Sciences. J. Washington Acad. Sci. 101
ered. However, it is worth emphasizing again that the findings of (1), 9–33.
the current study do not suggest that fast broadband be ignored. It Najarzadeh, R., Rahimzadeh, F., Reed, M., 2014. Does the internet increase labor pro-
ductivity? Evidence from a cross-country dynamic panel. J. Policy Model. 36 (6),
is still a significant contributor to employment growth. Moreover, 986–993. doi:10.1016/j.jpolmod.2014.10.003.
employment is only one dimension of economic development. It Shapiro, R.J., Hassett, K.A., 2012. The Employment effects of advances in Internet and
is possible that fast-speed Internet is essential to other aspects of wireless technology: evaluating the transitions from 2G to 3G and from 3G to
4G Retrieved from http://www.sonecon.com/docs/studies/Wireless_Technology_
economic development such as the success of small businesses and and_Jobs-Shapiro_Hassett-January_2012.pdf.
the thriving of technological innovation. The most important take- Singer, H., West, J., 2013. Economic effects of broadband infrastructure deploy-
away from this study is that we cannot simply assume that fast ment and tax incentives for broadband deployment White paper, Fiber-To-
The-Home Council, Mar. 2, 2010. Accessed Apr. 23, 2013. Retrieved from http:
broadband always generates more economic benefits than normal-
//www.neofiber.net/Articles/Economic_Effects_of_FTTH.pdf.
speed broadband does. Faster is not always better. United Nations Conference on Trade and Development, 2012. Information Econ-
omy Report, 2011: ICTs As An Enabler For Private Sector Development 2012 IIS
References 4050-S33; UNCTAD/IER/2011; ISSN 2075-4396;ISBN 978-92-1-112833-8 (paper);
ISBN 978-92-1-055120-5 (internet).
United States. Executive Office of the President, 2015. Community-based Broadband
Cho, S., Byun, J., Sung, M., 2003. Impact of the high-speed internet on user behav-
solutions: The benefits of Competition and Choice For Community Development
iors: case study in Korea. Internet Res. Electron. Netw. Appl. Policy 13 (1), 49–
and High Speed Internet access. Executive Office of the President, Washington,
60. doi:10.1108/S10662240310458378.
D.C.
Crandall, Robert, Lehr, W., Litan, R., 2007. The effects of broadband deployment on
Wooldridge, J.M., 2009. Introductory Econometrics: A Modern Approach, fourth ed.
output and employment: A cross-sectional analysis of U.S. data White paper, Is-
Cengage Learning, Mason, OH: South Western.
sues in Economic Policy No. 6 (July 2007). Accessed Apr. 23, 2013. Retrieved from
Whitacre, B., Gallardo, R., Strover, S., 2014. Does rural broadband impact jobs and
http://www.brookings.edu/papers/2007/06labor_crandall.aspx.
income? Evidence from spatial and first-differenced regressions. Ann. Reg. Sci.
Cumming, G., 2009. Inference by eye: Reading the overlap of independent confi-
53 (3), 649–670. doi:10.10 07/s0 0168-0.
dence intervals. Stat. Med. 28 (2), 205–220. doi:10.1002/sim.3471.