Dis­­cus­­si­­on Paper No.

16-056

The Economic Impacts of

Telecommunications Networks and
Broadband Internet: A Survey
Irene Bertschek, Wolfgang Briglauer,
Kai Hüschelrath, Benedikt Kauf, and Thomas Niebel
 Dis­­cus­­si­­on Paper No. 16-056

The Economic Impacts of

Telecommunications Networks and
Broadband Internet: A Survey
Irene Bertschek, Wolfgang Briglauer,
Kai Hüschelrath, Benedikt Kauf, and Thomas Niebel

Download this ZEW Discussion Paper from our ftp server:

http://ftp.zew.de/pub/zew-docs/dp/dp16056.pdf

Die Dis­­cus­­si­­on Pape­rs die­­nen einer mög­­lichst schnel­­len Ver­­brei­­tung von

neue­­ren For­­schungs­­arbei­­ten des ZEW. Die Bei­­trä­­ge lie­­gen in allei­­ni­­ger Ver­­ant­­wor­­tung
der Auto­­ren und stel­­len nicht not­­wen­­di­­ger­­wei­­se die Mei­­nung des ZEW dar.

Dis­­cus­­si­­on Papers are inten­­ded to make results of ZEW ­research prompt­­ly avai­­la­­ble to other
eco­­no­­mists in order to encou­­ra­­ge dis­­cus­­si­­on and sug­­gesti­­ons for revi­­si­­ons. The aut­­hors are sole­­ly
respon­­si­­ble for the con­­tents which do not neces­­sa­­ri­­ly repre­­sent the opi­­ni­­on of the ZEW.
 THE ECONOMIC IMPACTS OF TELECOMMUNICATIONS
NETWORKS AND BROADBAND INTERNET: A SURVEY

Irene Bertschek ″, Wolfgang Briglauer ♣, Kai Hüschelrath ∗,

Benedikt Kauf †, and Thomas Niebel °

August 2016

Abstract
We provide a structured overview of the quantitative literature on the economic impacts of
telecommunications networks and broadband internet. Differentiating between wireline and
wireless technologies as well as broadband availability and broadband adoption, respectively,
we review studies investigating the impacts on economic growth, employment and regional
development as well as productivity and firm performance. Eventually, the survey does not
only allow the identification of main research gaps but also provides useful information for
policy makers on the significance and importance of communication networks for social
welfare.

JEL Class: D24; J23; J24; L96; O33; O47

Keywords: Telecommunications; Broadband; Economic Growth; Employment; Regional
Development; Productivity

″
ZEW Centre for European Economic Research, MaCCI Mannheim Centre for Competition and Innovation
and University of Mannheim, P.O. Box 10 34 43, D-68034 Mannheim, Germany, E-mail: [email protected],
Corresponding author.
♣
ZEW Centre for European Economic Research and MaCCI Mannheim Centre for Competition and
Innovation, P.O. Box 10 34 43, D-68034 Mannheim, Germany, E-mail: [email protected].
∗ ZEW Centre for European Economic Research, MaCCI Mannheim Centre for Competition and Innovation
and University of Mannheim, P.O. Box 10 34 43, D-68034 Mannheim, Germany, E-mail:
[email protected].
†
ZEW Centre for European Economic Research, P.O. Box 10 34 43, D-68034 Mannheim, Germany, E-mail:
[email protected].
°
ZEW Centre for European Economic Research, P.O. Box 10 34 43, D-68034 Mannheim, Germany, E-mail:
[email protected].
We thank Benjamin Engelstätter for helpful comments and Julian Hidalgo for research assistance. Financial
support by the ZEW Sponsors' Association for Science and Practice is gratefully acknowledged.
1 Introduction

Among the technological advances of the last 50 years, the expansion and technological
improvements of telecommunications infrastructure have been some of the most crucial.
While the deployment of wireline telecommunications networks allowed large parts of the
world’s population, especially in developed countries, to communicate via fixed-line
telephony, the introduction of first and second-generation wireless telecommunications
networks since the 1980’s set another landmark for personal communications. However, the
progress in information and communications technologies (ICT) was not limited to voice
telephony only. Broadband internet technologies such as Digital Subscriber Line (DSL) or
Cable Internet have created unprecedented opportunities for worldwide data transmission. In
the last years, the deployment of so-called next-generation broadband networks has facilitated
much faster up- and download speeds as fiber-based wireline broadband access technologies
like FTTH (Fiber-to-the-Home), Fiber-to-the-Building (FTTB) or hybrid Fiber-to-the-Cabinet
(FTTC) and Fiber-to-the-Node (FTTN) technologies have started to replace the slower
entirely copper- or coax-based first-generation wireline technologies. The introduction of the
fourth generation (4G) mobile broadband technology Long Term Evolution (LTE) in 2010
brought substantial speed improvement for the wireless telecommunications networks.

Besides the benefits that these technologies have on the social lives of their consumers, their
possible economic benefits have been increasingly emphasized by economic research. In
particular, early estimations by Crandall and Jackson (2001) suggest that broadband
technologies can create substantial amounts of consumer surplus. Furthermore, broadband
internet is expected to generate new employment opportunities particularly in remote areas as
it enables a large amount of workers to work from home (so-called telecommuting) and thus
reduces the importance of distances.

Due to these advantages, broadband technologies have also gained the attention of policy
makers in recent times. In 2010, the Federal Communications Commission (FCC) released the
National Broadband Plan whose aims include that ‘every American should have affordable
access to robust broadband service’ and ‘at least 100 million U.S. homes should have
affordable access to actual download speeds of at least 100 megabits per second and actual
upload speeds of at least 50 megabits per second’ until 2020 (FCC 2010a, pp. 9-10).
Similarly, the European Commission launched the Digital Agenda for Europe (DAE) that
‘seeks to ensure that, by 2020, (i) all Europeans will have access to much higher internet
speeds of above 30 Mbit/s and (ii) 50% or more of European households will subscribe to
1
internet connections above 100 Mbit/s’ (European Commission 2010, pp. 19). While
achieving these goals promises considerable economic returns, they also go along with
substantial costs, in particular for the construction of the necessary new communications
infrastructure which is partly or entirely fiber-based. It is hence a necessity to carefully
evaluate whether these returns will exceed the accompanying costs or whether the expansion
of new (high-speed) broadband networks will go along with economic losses.

The central purpose of this survey is to provide a structured overview of the relevant
empirical literature focusing on the impacts of telecommunications and broadband
infrastructure and services on important economic indicators thereby focusing on
productivity, growth, employment and regional development. In order to obtain appropriate
guidelines for future public policies, we will only review quantitative research that attempts to
identify causal links between telecommunications/broadband infrastructure (availability or
coverage on the supply-side) or services (usage or adoption/penetration on the demand-side)
and key economic variables 1. Based on a thorough reading of the literature we aim to find
answers to the following research questions (RQs):

• Do telecommunications and/or broadband infrastructure and services have a

significant positive effect on relevant economic outcome indicators (RQ 1)?
• Do the economic impacts of wireline and wireless infrastructure and services differ
from each other (RQ 2)?
• Does the availability of infrastructure produce different economic impacts than the
actual usage (RQ 3)?
• Do the microeconomic studies confirm or challenge the results of the macroeconomic
studies (RQ 4)?

The answers to these questions will provide an important assistance for public policy makers
especially with regard to the further expansion of next-generations broadband infrastructures
and services. Furthermore, it allows us to identify main research gaps in the existing literature
that prevent more effective policy conclusions.

Previous surveys by Draca et al. (2007) and Cardona et al. (2013) focus on productivity as an
outcome measure and consider ICT as an aggregate input measure. In this survey, we focus on
telecommunication and broadband infrastructure as a specific type of ICT. Both fixed
(‘wireline’) and mobile (‘wireless’) networks will be included in the evaluation of the

1
Accordingly, we exclude qualitative as well as simulation-based studies.

2
economic impacts of telecommunications and broadband infrastructure and services.
Moreover, we will specify whether broadband or telecommunications infrastructure is merely
available to the public or whether the technology is actually being used. The actual
consumption of services provided by a certain technology will be referred to as usage or
adoption/penetration of the technology whereas the existence of infrastructure will be noted as
the availability or coverage of the technology. In general, the availability of infrastructure
quantifies real investment activities of telecommunications operators while the adoption of a
technology quantifies the extent of consumers’ willingness to pay since only those consumers
or businesses whose willingness to pay is sufficient will subscribe to the offered services. Due
to the often missing differentiation between broadband technologies, the broadband variables
will mostly capture a mix of broadband technologies. In case a study concentrates on a
specific high-speed broadband technology, we will indicate this technology accordingly. A
detailed definition of the most relevant wireline and wireless technologies can be found in the
Appendix.

The remainder of the article is structured as follows. Section 2 reviews the literature on the
economic impacts of narrowband telecommunications networks whereas Section 3 provides a
review of the empirical literature related to the economic impacts of broadband internet – with
both sections featuring interim conclusions and tabular summaries. Finally, Section 4 derives
overall conclusions as well as policy implications and identifies major research gaps.

2 The economic impacts of telecommunications networks

In this section, we focus on the economic impacts of telecommunications networks. Section

2.1 reviews studies focusing on the impact of telecommunications networks on economic
growth and Section 2.2 reviews studies that examine the impacts on productivity and
firm/market performance – both in ascending chronological order. Subsequently, Section 2.3
will feature interim conclusions.

2.1 Economic growth

Cronin et al. (1991) test the causal direction of the relationship between telecommunications
investment and economic growth by using data for the U.S. from 1958 to 1988. For this
purpose, Granger causality, Sims and Modified Sims tests with second-order lag structure are
performed on the first differences of the variables for telecommunications investment, gross
national product (GNP) and total output. The authors find significant evidence for

3
bidirectional causal relationships between telecommunications investment and GNP as well as
total output.

Madden and Savage (1998) analyze the relationship between telecommunications investment
and economic growth in Central and Eastern Europe using data for 27 countries from 1990 to
1995. The authors estimate static cross-country growth equations at the aggregate and the
sectoral level by ordinary least squares regressions (OLS). The results show that a higher
share of telecommunications investment in GDP significantly increases real GDP growth per
capita. For a subset of 8 countries, real telecommunications investment, as measured by the
growth rate of mainlines per 100 inhabitants, is associated with a significant positive increase
in GDP growth per capita in the industrial sector. Moreover, Granger-causality tests provide
evidence of mutual precedence between real economic growth and telecommunications
investment at the aggregate level.

Madden and Savage (2000) investigate the effects of telecommunications investment on GDP
growth using data for 43 countries from 1975 to 1990. Telecommunications investment is
measured by the share of telecommunications investment in GDP and the number of
mainlines per working age population. The authors employ ordinary least squares and
instrumental variable regressions to estimate a dynamic augmented supply-side growth model
based on Mankiw et al. (1992). The results show a significant positive impact of
telecommunications investment on the growth of GDP per capita for both telecommunications
measures.

Dutta (2001) analyzes the causal links between telecommunications infrastructure and
economic growth with data for 15 developing and 15 industrialized countries from 1960 to
1993. For this purpose, Granger causality tests are applied in which telecommunications
usage is quantified through the number of telephones (total and per 100 inhabitants). Using
both the original variables and their logarithmic transformations, the authors observe that
telecommunications infrastructure Granger-causes economic growth in over half of the
examined developing countries. Similar results are obtained for industrialized countries even
though Granger-causality is found in slightly less countries. In contrast, significantly fewer
countries show signs of a causal impact of economic growth on telecommunications. Based
on these findings, the authors conclude that a unidirectional causal effect of
telecommunications on economic growth is more probable than a bidirectional relationship
between the two variables.

4
Röller and Waverman (2001) use data for 21 OECD countries from 1970 to 1990 to analyze
the influence of wireline telecommunications on GDP growth. In order to address potential
reverse causality, the authors apply a structural model, which endogenizes telecommunication
investment. Subsequently, all equations are estimated by nonlinear Generalized Method of
Moments. The results show that the increase in the number of mainlines per capita positively
impacts economic growth. Overall, telecommunications infrastructure is estimated to account
for approximately one third of annual GDP growth between 1970 and 1990. Furthermore, the
impact of telecommunications infrastructure is found to be nonlinear so that countries with a
penetration rate of more than 40 percent (close to universal service given 2 to 2.5 persons per
household) experience significantly higher growth than countries with low or medium
penetration rates. These findings provide evidence for a critical mass phenomenon as well as
network externalities and suggest that developing countries with low adoption rates would
need considerable advancement to reach growth effects similar to high penetration countries.

Yilmaz et al. (2002) assess the role of spillover effects of telecommunications infrastructure
investment on regional economic growth by utilizing state-level data for the U.S. from 1970
to 1997. The authors estimate a production-function model with network and geographical
spillovers variables by first-differenced generalized least squares and weighted two-stage least
squares regressions. The results indicate that while an increase in a state’s
telecommunications capital stock has a significant positive impact on its output growth, this
also produces significant negative spillover effects for other states. In the years from 1984 to
1997, geographical proximity to a state increasing its telecommunications investment further
adds to these negative spillover effects.

Datta and Agarwal (2004) employ a dynamic fixed-effects panel data model to examine the
impact of telecommunications adoption on economic growth for 22 OECD countries from
1980 to 1992. A higher number of access lines per 100 inhabitants is found to significantly
increase real GDP per capita for the examined country sample. The significant positive
coefficients for lagged values of the telecommunications infrastructure variable also suggest
that this impact is causal. Moreover, the authors find decreasing returns to
telecommunications infrastructure which implies a stronger impact of telecommunications
adoption in developing countries with few access lines.

Waverman et al. (2005) utilize data for 38 developing countries from 1996 to 2003 to
investigate the influence of mobile telecommunications adoption on GDP growth. As a first
approach, the authors adopt the simultaneous equations model approach of Röller and

5
Waverman (2001) which is estimated by Generalized Method of Moments. While the
estimations suggest a significant positive impact of mobile telecommunications, this impact
appears to be overstated and is not robust to the inclusion of different countries as well as to
adjustments of model specifications. Consequently, the authors estimate a cross-sectional
endogenous growth model similar to Barro (1991) estimated by ordinary least squares
regressions and employ data for 92 countries from 1980 and 1996 to 2003. The estimations
indicate that while a higher average mobile penetration level for the years from 1996 to 2003
led to a significant increase in the average growth rate of GDP per capita from 1980 to 2003,
the level of fixed-line penetration in 1980 did not have a significant effect. In particular, the
effect of mobile adoption was twice as large for low income countries relative to the high
income countries. Moreover, Hausman tests indicate that reverse causality did not present a
problem in the examined case.

Sridhar and Sridhar (2007) study the effects of telecommunications adoption on economic
growth in developing countries by using panel data for 63 developing countries from 1990 to
2001. Based on Röller and Waverman (2001), they estimate a system of equations that
endogenizes economic growth and telecommunications adoption using a three-stage least
square regressions methodology. When mainline and mobile phone adoption are evaluated
together, an increment in the number of telephones per 100 inhabitants is associated with a
significant increase in GDP which ranges between 0.1 (with country fixed effects) and 0.15
(without country fixed effects) percent for a 1 percent increase in telephone penetration.
Considering only landline penetration, the same increase in telephone penetration is estimated
to increase GDP by 0.14 percent in both specifications. Overall, growth in landline
penetration accounted for an average GDP increase of 1.62 percent in the observed countries.
Compared to mainlines, the positive impact of mobile telephones is considerably smaller with
a 1 percent increase in penetration leading to an increase in GDP of less than 0.01 percent in
the country fixed effects specification. The high growth rates of mobile phone penetration,
contributed on average 2.48 percent to GDP in the examined countries.

Shiu and Lam (2008) analyze the links between telecommunications adoption and economic
growth using a panel data for 22 Chinese provinces for the period from 1978 to 2004. The
authors employ a dynamic panel data model whose first-differenced form is estimated by
Arellano-Bond estimators and subsequently apply Granger causality tests. For the full sample
of provinces, a unidirectional causal effect of real GDP on telecommunications usage is
found. In a more detailed geographical analysis, a causal impact of telecommunications on

6
real GDP is found for high-income provinces in the Eastern region, but not in the low-income
central and western provinces. Similarly, telecommunications has a causal effect on economic
growth in provinces with high adoption rates while provinces with low penetration rates
experience no effect. These findings are in line with other studies that find returns on
telecommunications to be increasing in higher adoption rates and argue in favor of a critical
mass.

Lam and Shiu (2010) use panel data for 105 countries from 1980 to 2006 to evaluate the
relationship between telecommunications adoption and economic growth. For this purpose,
the authors estimate a dynamic panel data model in first-differenced form by Generalized
Method of Moments estimators and subsequently test for Granger causality. While no causal
impact of telecommunications on economic growth is found for the full country sample, this
impact is observed for Europe (bidirectional) as well as Asia and Oceania when regions are
examined individually. Moreover, the results also suggest a bidirectional causal relationship
in countries with high incomes. For the period from 1997 to 2006, the authors further obtain a
bidirectional causal relationship between mobile telecommunications and economic growth
for the full country sample and every categorization of countries that is particularly strong for
African countries.

Chakraborty and Nandi (2011) analyze the links between mainline access per 100 inhabitants
and GDP per capita by employing unit root tests, panel cointegration and Granger causality
tests on a panel data for 93 developing countries from 1985 to 2007. For the full country
sample, the authors find evidence for a unidirectional causal effect of GDP per capita on
telecommunications adoption in the short run and a bidirectional causal relationship between
the two variables in the long run. A more detailed analysis reveals that a bidirectional causal
link in both the short and the long run exists for less developed countries and countries with
high growth rates. High-growth countries display a particularly strong relationship between
mainline telecommunications and economic growth in the long run, whereas the more
developed countries in the sample exhibit the same causality pattern as the complete sample
of countries. Based on these results, the authors conclude that telecommunications
infrastructure investment could be a pivotal instrument in the catch-up process of countries
with lower levels of infrastructure and economic development.

Gruber and Koutroumpis (2011) employ data for 192 countries from 1990 to 2007 in order to
study the relationship between mobile telecommunications adoption and economic growth.
Following Röller and Waverman (2001), the authors use a simultaneous equations model with

7
a macro production function and demand and supply equations for mobile infrastructure as
their main approach. The obtained results are then compared to estimations from static fixed-
effect ordinary least squares and instrumental variable regressions. Regardless of
specification, the number of mobile lines has a significant positive impact on GDP that is
largest in the instrumental variable regressions while the number of fixed lines has a contrary
effect. However, the impact of mobile lines is nonlinear and increasing in a country’s mobile
penetration rate and income. Mobile telecommunications are estimated to cause average
growth returns of annually 0.2 percent in high income countries and 0.11 percent in low
income countries. In line with these findings, the authors identify a critical mass at a
penetration rate of 30 percent.

Lee et al. (2012) investigate the influence of landline and mobile telecommunications
adoption on the growth of GDP per capita using panel data for 44 sub-Saharan African
countries from 1975 to 2006. To account for reverse causality, the authors employ two-step
difference Generalized Method of Moments estimators to estimate a dynamic panel data
growth model similar to Datta and Agarwal (2004). For the period from 1975 to 2006, the
number of main telephone lines per 100 inhabitants is found to have a significant positive
impact on growth of GDP per capita. Contrary to that, the effect of the number of mobile
phone subscribers per 100 inhabitants is insignificant, which might be due to the relatively
recent introduction of mobile phones. Indeed, only mobile telecommunications are associated
with a significant increase in GDP growth when the analysis is confined to the years from
2000 to 2006. The negative coefficient for the interaction term between landline and mobile
telecommunications in this specification further suggests that the impact of mobile phones on
GDP growth is higher in countries with low levels of landlines. When landline and mobile
telecommunications are combined, the authors find no significant effect of
telecommunications on GDP growth.

Ward and Zheng (2016) evaluate the impact of fixed and mobile telecommunications
adoption on economic growth in China by employing industry- and province-level panel data
for 31 Chinese provinces from 1991 to 2010. The authors apply static two-way fixed-effect
ordinary least squares regressions as well as dynamic system Generalized Method of
Moments estimations in which telecommunications subscriptions are at first assumed to be
exogenous and later instrumented for in order to account for possible endogeneity. The
estimations indicate a significant positive effect of the number of mobile phone subscriptions
on the growth rate of GDP per capita that was particularly strong for the years from 1991 to

8
2000. Consistent with the theory that mobile telecommunications mainly affect less developed
regions, the authors further find that the underdeveloped western region of China benefitted
relatively more from mobile telecommunications than the wealthier eastern region for the
period from 1991 to 2000. For the years from 2001 to 2010 however, no such difference in
impact is observed. In a similar vein, complementarity between fixed and mobile
telecommunications is also found only for the initial period. Contrary to the positive impact of
mobile telecommunications, the number of fixed-line subscriptions negatively influenced
economic growth in both regions in the latter period.

2.2 Productivity and firm performance

Cronin et al. (1993) employ data for the U.S. from 1958 to 1991 to investigate the influence
of telecommunications investment on total factor productivity growth. In order to obtain
causal effects, the authors apply Granger causality tests and modified Sims tests in which lag
intervals are specified for two, four and six years. The results of these tests mostly suggest a
significant and unidirectional causal impact of telecommunications investment on private
business and manufacturing productivity. Further, the authors measure these effects for the
period from 1963 to 1991 by using industry-level data and comparing actual productivity
outcomes to hypothetical productivity outcomes which result from holding
telecommunications technology constant at the level of 1963. It is found that improvements in
telecommunications accounted for up to 37 percent of annual growth in total economy-wide
productivity and on average for 25 percent annual productivity growth between 1975 and
1991. Moreover, almost all industries experienced a significant increase in productivity
growth due to telecommunications with service sectors such as retail trade benefitting the
most.

Greenstein and Spiller (1995) study the effect of telecommunications infrastructure on the
finance, insurance and real estate (FIRE) sector and the manufacturing sector by employing
company-, county-, state-level data for the U.S. from 1986 to 1992. Results are obtained via
instrumental variable estimations of static pooled cross-section time series equations and
dynamic investment adjustment models. The authors find that infrastructure deployment has a
significant positive impact on real revenue in the FIRE sector. More specifically, the
estimations suggest that doubling the number of miles of fiber cable would result in an
increase in revenue in the FIRE sector of 10 to 45 percent depending on the specification. In
contrast, no significant impact of infrastructure deployment is found for the manufacturing
sector. The authors therefore argue that improved telecommunications infrastructure primarily

9
benefits technology intensive industries such as the FIRE sector whereas low-tech industries
like the manufacturing sector are less affected. Furthermore, the authors also caution that the
effects of telecommunications on the FIRE sector might be even larger since spillover effects
for other sectors and regions are not incorporated in their results.

Cieślik and Kaniewsk (2004) utilize panel data for 49 regions in Poland from 1989 to 1998 to
assess the impact of telecommunications adoption on regional economic development. For
this purpose, the authors use a static regional production function approach to set up a
theoretical model which is empirically estimated via ordinary least squares regressions as well
as fixed and random effects panel data techniques. The number of telephone subscribers per
100.000 inhabitants exerts a highly significant positive impact on retail sales per worker
across all specifications even though the size of the coefficient is considerably higher in the
ordinary least squares regressions. To ensure the causality of this effect, the authors also apply
Hsiao’s Granger tests. The results of all specifications suggest that the observed impact of
telecommunications infrastructure is causal. However, this study does not examine whether
the impact is bidirectional.

Jensen (2007) studies the effect of mobile phone availability on price dispersion and social
welfare in fish markets in the Indian state of Kerala based on weekly survey data from 1996 to
2001. The individual markets are assorted into 3 regions according to their district affiliation.
Difference-in-differences estimations are performed in which treatment is defined as the
introduction of mobile phone availability in a region. Mobile phone availability is associated
with a significant decrease in price dispersion across all regions and a 38 percentage point
decline of the coefficient of variation for pooled treatments. This reduction in price dispersion
also causes most markets to converge closely to the Law of One Price. Moreover, the
availability of mobile phones significantly improves static efficiency such that the amount of
wasted unsold fish is effectively reduced to zero. Further estimations suggest that while
average prices for fish decreases, larger amounts of fish are sold and average daily profits are
around 9 percent higher. This positive effect on profits is also partially experienced by sellers
without mobile phones. The introduction of mobile phones thus provides positive external
effects for non-users. Finally, the reduction in consumer retail prices has a significant
positive, albeit small, effect on consumer surplus.

Thompson and Garbacz (2007) assess the influence of telecommunications adoption on

productive efficiency by using panel data for 93 countries from 1995 to 2003. For this
purpose, the authors employ a one-stage stochastic-frontier production function approach to

10
derive inefficiency functions for different groups of countries. The estimated inefficiency
functions show clear differences between regions regarding the impact of
telecommunications. While productive efficiency in low income and sub-Saharan African
countries can significantly benefit from increasing mobile phone and fixed-line
telecommunications adoption, only increased fixed-line adoption in Latin American countries
and increased mobile phone adoption in Asian countries are found to have a significantly
positive impact. Furthermore, high-income OECD countries which already have near-
universal telecommunications adoption do not experience significant positive effects from
further increases in telecommunications penetration. These findings support the notion that
the deployment of telecommunications infrastructure can be a helpful instrument for
particularly stimulating economic development in poor and developing regions.

Muto and Yamano (2009) analyze the influence of mobile phone coverage and adoption on
market participation and market outcomes of farmers in developing countries by using
household- and community-level survey panel data for Uganda from 2003 and 2005. By
applying static fixed-effect regressions and static fixed-effect instrumental variable
regressions, the authors find that farmers that cultivate bananas and live in remote areas, are
more likely to sell their crop if their community has access to a mobile phone network. In
particular banana farmers with small amounts of land benefit more from mobile phone
coverage than large banana farmers. The smaller farmers improve access to information more
than larger farmers who already had better access to information and more social contacts
before. However, these findings are only limited to the sales of bananas (perishable good)
whereas no significant impact on the sales of maize (non-perishable good) is observed.

Aker (2010) utilizes trader- and market-level data for grain markets in Niger from 1999 to
2007 to evaluate the impact of mobile phone introduction on price dispersion. The author uses
a difference-in-differences approach with pooled treatments in which a pair of markets has
received treatment if both markets have access to the services of a mobile phone tower. To
further test for robustness, propensity score matching techniques and a falsification test are
employed. Across almost all specifications, the introduction of mobile phone availability
significantly decreased price dispersion between markets by at least 10 percent with the
majority of the reduction being realized after the first three months. With the addition of
interaction terms, the impact of mobile phones is found to be the strongest between markets
that face high transport costs such as distant markets and markets connected by unpaved
roads. Further analysis provides evidence that the decrease in price dispersion is not caused by

11
spillover effects or collusive behavior of traders but rather due to lower search costs and
improved marketing strategies.

Forero (2013) uses panel data for 23 low-income and 18 high-income countries from 1980 to
2009 in order to analyze the impact of mobile telecommunications adoption on technical
efficiency. The author applies a stochastic-frontier production function approach similar to
Thompson and Garbacz (2007) but employs a flexible transcendental logarithmic production
function instead of a Cobb-Douglas production function and includes fixed effects. The
estimation results show that higher numbers of subscribers to mobile phone services per 100
inhabitants significantly contribute to decreases in technical inefficiency with the effect being
particularly strong in Asian and Latin American countries. The inclusion of lagged values of
mobile phone usage further provides evidence that the impact of mobile telecommunications
on technical efficiency is causal in low-income countries whereas no causality is found for
high-income countries.

2.3 Conclusions on the economic impacts of telecommunications networks

The main results of the reviewed papers studying the economic impacts of
telecommunications networks are presented in ascending chronological order in Table 1
(‘impact on economic growth’) and Table 2 (‘impacts on productivity and firm/market
performance). In both tables, we apply a further separation into studies examining wireline
and wireless telecommunications networks.

Overall, a majority of 13 studies reviewed provide evidence that economic growth measured
by GDP/GNP or GDP per capita is significantly and positively affected by landline
telecommunications with all of the 10 studies with data at the country level finding such an
effect for at least some of the examined countries. However, some studies also indicate that
the positive effect of telecommunications is mostly present in developed countries with a high
number of mainlines whereas evidence for developing countries with a lower number of
mainlines is slightly less conclusive. In contrast to wireline telecommunications, mobile
telecommunications networks are found to significantly enhance economic growth in
developed but even more so in less developed countries. Lee et al. (2012) further demonstrate
that the impact of mobile telecommunications is stronger in developing countries with less
developed wireline telecommunications infrastructure. Based on these findings, developing
countries should put their focus on the further deployment of mobile telecommunications if
they want to stimulate economic growth effectively.

12
All of the studies focusing on both wireline and wireless telecommunications find at least
partially significant positive impacts on productivity, firm performance or market
performance. On the sectoral level, Cronin et al. (1993) and Greenstein and Spiller (1995)
provide evidence that the benefits of wireline telecommunications are particularly
concentrated in service sectors which generally require higher skilled workers. In contrast,
studies by Jensen (2007) and Aker (2010) suggest that mobile telecommunications can help
agricultural markets in developing countries to be more efficient and increase social welfare.

In sum, the reviewed studies provide strong evidence that telecommunications networks
(fixed-line and mobile) exert positive effects on economic growth as well as national and
sectoral productivity (RQ 1). Regarding research question 2 (RQ 2), we find that mobile
telecommunications networks have a stronger positive impact on developing countries while
wireline telecommunications networks rather benefit developed countries. Furthermore, we
observe no distinguished difference between the availability (infrastructure investment) and
the adoption of both wireline and wireless telecommunications networks (RQ 3). Almost all
microeconomic studies regarding the effects of wireline and wireless telecommunication
networks on growth and productivity find positive effects in line with the macroeconomic
country-level studies (RQ 4).

13
Table 1: Telecommunications’ impact on economic growth

Authors Data Methodology Main results

Wireline telecommunications

Cronin USA; 1958- Granger, Sims and modified • Bidirectional causal relationships
et al. 1988; Sims tests performed on first- between telecommunications
(1991) country-level differenced variables investment* and GNP

Madden & 27 Central and Static growth equations • Monetary/real wireline

Savage Eastern estimated by ordinary least telecommunications investment
(1998) European squares regressions; Hsiao’s has a positive impact on growth of
countries; Granger causality tests to GDP per capita for the whole
1990-1995; determine the timing of effects economy/the industrial sector
country-level • Real investment precedes
(Granger causes) GDP growth
while monetary investment does
not

Madden & 43 countries; Dynamic augmented growth • Both monetary and real wireline
Savage 1975-1990; model estimated by ordinary telecommunications investment
(2000) country-level least squares and instrumental positively affect growth of GDP
variable regressions per capita

Dutta 15 industrial. Granger causality tests for all • A causal effect of

(2001) and 15 countries individually telecommunications usage* on
develop. economic growth is found for
countries; more countries than a causal effect
1960-1993; in the other direction
country-level

Röller & 21 OECD Simultaneous equations model • Wireline telecommunications

Waverman countries; with a macro production adoption positively affects GDP
(2001) 1970-1990; function and demand and supply growth
country-level equations for • Evidence for a critical mass of
telecommunications investment telecommunications infrastructure
estimated by nonlinear at an adoption rate of 40 percent
Generalized Method of
Moments

14
(Table 1: Continued)

Authors Data Methodology Main results

Wireline telecommunications

Yilmaz USA; Static classical production • Telecommunications

et al. 1970-1997; function approach estimated by infrastructure investment* has a
(2002) state-level first-differenced generalized and positive effect on output growth
weighted two-stage least squares • But: a state’s telecommunications
regressions investment negatively impacts
output growth in the other states

Datta & 22 OECD Dynamic fixed-effect panel data • Wireline telecommunications

Agarwal countries; growth model adoption positively affects GDP
(2004) 1980-1992; per capita
country-level • Decreasing returns imply larger
effects in developing countries

Sridhar & 63 countries; Simultaneous equations model • Wireline telephone adoption has a
Sridhar 1990-2001; estimated by three-stage least significant positive impact on
(2007) country-level squares regressions GDP growth

Shiu & 22 Chinese Dynamic panel data model • Wireline telecommunications

Lam provinces; estimated in first-differenced adoption causally impacts GDP
(2008) 1978-2004; form by Arellano-Bond only in provinces with high
province-level estimators; Granger causality income/adoption rates
tests

Lam & 105 countries; Dynamic panel data model in • Bidirectional causal links between
Shiu 1980-2006; first-differenced form estimated telecommunications adoption*
(2010) country-level by Generalized Method of and economic growth in European
Moments, Granger causality test and high income countries
• Causal impact of GDP on
telecommunications adoption in
lower income countries

15
(Table 1: Continued)
Authors Data Methodology Main results

Wireline telecommunications

Chakraborty 93 developing Unit root tests, panel • Bidirectional causal relationship

& Nandi countries; cointegration and Granger between wireline
(2011) 1985-2007; causality tests telecommunications adoption and
country-level economic growth in less
developed or fast growing
countries
• The effect is weaker in more
developed countries

Lee 44 sub-Saharan Dynamic panel data growth • Significant positive impact of

et al. countries; model estimated by two step wireline telecommunications
(2012) 1975-2006; difference Generalized Method adoption on GDP per capita
country-level of Moments

Ward & 31 Chinese Static two-way fixed-effect • Between 1991 and 2000, fixed-line
Zheng provinces; ordinary least squares and subscriptions had a positive impact
(2016) 1991-2010; dynamic system Generalized on growth of GDP per capita in the
industry- and Method of Moments regressions well-developed eastern provinces
province-level (with and without instruments for • In contrast, fixed-line subscriptions
possible endogeneity of the had a negative effect from 2001 to
telecommunications variable) 2010

16
(Table 1: Continued)

Authors Data Methodology Main results

Wireless telecommunications

Waverman 38 developing Simultaneous equations model • First approach: positive results are
et al. countries/ 92 estimated by Generalized not robust
(2005) countries; Method of Moments; cross- • Second approach: mobile
1996- sectional endogenous growth telecommunications adoption has
2003/1980- model estimated by ordinary a significant positive effect on
2003; least squares regressions growth of GDP per capita that is
country-level twice as strong in low income
countries

Sridhar & 63 developing Simultaneous equations model • Mobile phone adoption has a
Sridhar countries; estimated by three-stage least significant positive impact on
(2007) 1990-2001; squares regressions GDP growth
country-level

Lam & 105 countries; Dynamic panel data model in • Bidirectional causal links between
Shiu 1997-2006; first-differenced form estimated mobile telecommunications
(2010) country-level by Generalized Method of adoption and economic growth for
Moments, Granger causality test the complete country sample and
high/low income countries

Gruber & 192 countries; Simultaneous equations model • Significant positive impact of
Koutroumpis 1990-2007; estimated by three-stage least mobile telecommunications
(2011) country-level squares; static fixed-effect infrastructure on economic growth
ordinary least squares and • Critical mass at a penetration rate
instrumental variable regressions of around 30 percent

Lee 44 sub-Saharan Dynamic panel data growth • Between 2000 and 2006,
et al. countries; model estimated by two step significant positive effect of
(2012) 1975-2006; difference Generalized Method mobile phone adoption on GDP
country-level of Moments that is negatively correlated with
wireline infrastructure

Ward & 31 Chinese Static two-way fixed-effect • Significant positive impact of

Zheng provinces; ordinary least squares and mobile phone subscribers on
(2016) 1991-2010; dynamic system Generalized growth of GDP per capita
industry and Method of Moments regressions • Between 1991 and 2000, the
province-level (with and without instruments impact was significantly stronger
for possible endogeneity of the in less developed provinces
telecommunications variable)

* These studies either include a telecommunications variable that combines wireless and wireline technologies or
do not specify which form of telecommunications is utilized. However, we assign these studies to wireline
telecommunications as either their data sets often end before 2000 when wireline telecommunications networks
were still the dominant technology or their data on mobile telecommunications is incomplete.

17
Table 2: Telecommunications’ impacts on productivity and firm/market performance

Authors Data Methodology Main results

Wireline telecommunications

Cronin USA; 1958- Granger, Sims and modified • Significant causal effect of
et al. 1991; Sims tests; Peterson Indices telecommunications investment*
estimated by interindustry on private business and
(1993) industry- and economics and translog manufacturing productivity
country-level production functions • Wireline telecommunications
investment increases productivity
in most industries

Greenstein & USA; Estimation of static pooled • Deployment of

Spiller 1986-1992; cross-section time series telecommunications
(1995) company-, equations and dynamic infrastructure* has a significant
county- and investment adjustment models positive effect on total revenue in
state-level the technology-intensive finance,
insurance and real estate sectors
(FIRE) but no effect on the
manufacturing sector

Cieślik & Poland; Regional production function • Telecommunications adoption*

Kaniewsk 1989-1998; approach estimated by ordinary has a significant positive and
(2004) region-level least squares as well as fixed- likely causal impact on retail sales
effect and random effect panel per worker
data regressions; Hsiao’s
Granger tests

Thompson & 93 countries; One-stage stochastic-frontier • Wireline telecommunications

Garbacz 1995-2003; production function approach adoption significantly reduces
(2007) country-level estimated by random effects productive inefficiencies only in
regressions low-income countries but has no
significant effect in high income
countries

18
(Table 2: Continued)

Authors Data Methodology Main results

Wireless telecommunications

Jensen Fish markets in Difference-in-differences • Mobile phone availability

(2007) the Indian state estimations with the introduction significantly reduces price
of Kerala; of mobile phone availability as dispersion and waste of unsold
1996-2001 treatment fish
individual- • Mobile phone availability
level (survey) increases sellers’ profits and
consumer surplus

Thompson & 93 countries; One-stage stochastic-frontier • Wireless telecommunications

Garbacz 1995-2003; production function approach adoption significantly reduces
(2007) country-level productive inefficiencies only in
low-income countries

Muto & Uganda, Static fixed-effect regressions • Rural communities’ mobile phone
Yamano 2003-2005, and static fixed-effect coverage increases the probability
(2009) community- instrumental variable regressions to sell and the quantity sold of its
and household- farmers with regard to bananas but
level not with regard to maize

Aker 33 grain Difference-in-differences • The introduction of mobile phone

(2010) markets in estimations with the introduction towers is associated with a
Niger; of mobile phone towers as significant decrease in price
1999-2007; treatment, inclusion of fixed dispersion between markets
trader- and effects and lagged dependent • The effect is particularly in distant
market-level variables and difficult to access markets

Forero 23 low income Estimation of a fixed-effect • Mobile telecommunications

(2013) and 18 high stochastic-frontier production adoption significantly reduces
income function approach with a technical inefficiencies, especially
countries; translog production function in Asian and Latin American
1980-2009; countries
country-level • The effect is likely causal in low-
income countries but not in high-
income countries

* These studies either include a telecommunications variable that combines wireless and wireline technologies or
do not specify which form of telecommunications is utilized. However, we assign these studies to wireline
telecommunications as either their data sets often end before 2000 when wireline telecommunications networks
were still the dominant technology or their data on mobile telecommunications is incomplete.

19
3 The economic impacts of broadband internet

In this section, we focus on the economic impacts of broadband internet. Section 3.1
investigates the effects of broadband on economic growth. Since broadband has been
considered as a main driver of rural development, Section 3.2 focuses on the regional
development, household outcomes and employment effects. Section 3.3 then analyzes the
influence of broadband on labor and firm productivity as well as on firm performance. In all
three sections, we again review the respective literature in ascending chronological order.
Eventually, Section 3.4 provides interim conclusions.

3.1 Economic growth

Koutroumpis (2009) examines the relationship between broadband adoption and GDP growth
utilizing data for 22 OECD countries from 2002 to 2007. Using a simultaneous equations
model based on Röller and Waverman (2001), estimates are obtained by employing single
instrumental variable and three-stage least squares Generalized Method of Moments
estimators, both including country fixed effects. The author finds a significant positive impact
of broadband penetration on GDP with a one percent increase in broadband adoption
generating a 0.023 percent increase in GDP growth according to the fixed-effects instrumental
variable specification. The estimations further imply that improvements in broadband
infrastructure were responsible for nearly 10 percent of the annual growth rates of the OECD
countries in the observed period. In addition, the author identifies a critical mass at a
broadband adoption rate of 30 percent after which the growth returns to broadband
infrastructure investments are considerably larger than before.

Czernich et al. (2011) measure broadband adoption as the number of broadband subscribers
per 100 inhabitants and investigate its influence on economic growth with panel data for 25
OECD countries from 1996 to 2007. The authors apply a two-stage instrumental variable
approach in which the rate of broadband adoption is estimated in the first stage through non-
linear least squares based on a logistic diffusion model. Subsequently the estimated broadband
penetration rates are employed as broadband variables in two second-stage models: in the first
model, broadband introduction is used as a dummy variable in difference-in-differences
estimations with fixed effects; in the second model, the broadband adoption rate is used as a
continuous variable in a static cross-country growth model with fixed effects. The authors
find that broadband introduction increases per capita growth significantly such that the GDP
per capita after the introduction of broadband was 2.7 to 3.9 percent higher than before. A 10

20
percentage point increase in the broadband adoption rate increases the annual per capita
growth by 0.9 to 1.5 percentage points. The authors also find evidence for a critical mass
phenomenon at a broadband adoption rate of 10 percent.

Thompson and Garbacz (2011) evaluate the impact of mobile and fixed broadband adoption
on economic growth utilizing panel data for 43 countries from 2005 to 2009. Fixed-effect
instrumental variable regressions are applied in which static equations are estimated for high
and low income countries individually and combined. In all specifications, the authors find
that mobile broadband has a significant positive impact on GDP per household. However, the
magnitude of this impact is larger in low income countries. In contrast, the coefficient for
fixed broadband is negative but marginally significant in low income countries only.

Arvin and Pradhan (2014) establish a causal relationship between broadband adoption and
economic growth employing panel data from 1998 to 2011 for 19 of the former G20 countries
which are subsequently divided into developed and developing (‘emerging’) countries. Based
on static and dynamic regressions that include variables for GDP, the number of broadband
users, the degree of urbanization and the real inflow of foreign direct investment, the authors
apply panel cointegration and Granger causality tests to verify causal relationships. Initially,
cointegration is found between all four variables suggesting long-run causal relationships
between these variables. In contrast to that, the Granger causality tests do not provide
evidence for any kind of long-run causal relationship between GDP and broadband
penetration. In the short run, the authors find that a bidirectional causal relationship between
economic growth and broadband penetration exists only for developed countries; in
developing countries, only economic growth causally impacts broadband penetration.

Gruber et al. (2014) use data for 27 European Union (EU) countries from 2005 to 2011 to
evaluate the benefits and costs of the Digital Agenda for Europe (DAE). To estimate the
return parameters of broadband infrastructure, the authors set up a simultaneous equations
model, similar to Röller and Waverman (2001), and apply fixed-effect three-stage least
squares regressions. The estimates suggest that broadband adoption rates had a significant
positive effect on GDP in the observed period. Moreover, the results indicate that this effect is
significantly larger for a broadband adoption rate greater than 15 percent.

3.2 Employment and regional development

Kandilov and Renkow (2010) use ZIP code- and county-level data for the U.S. from 1994 to
2007 for ZIP codes with a population of 20.000 inhabitants or less to examine the economic

21
impacts of two programs of the U.S. Department of Agriculture on rural areas, including the
Pilot Broadband Loan Program as well as the subsequent Broadband Loan Program.
Employment, annual payroll and the number of business establishments are used as dependent
variables in the baseline regression and estimations are obtained on the ZIP code and county
level. In order to estimate the baseline regression, the authors apply a difference-in-
differences fixed effects approach, including year dummies, and difference-in-differences
propensity score matching estimators for nearest neighbor matching as well as other
robustness checks. The authors find that only the initial Pilot Broadband Loan Program had a
significant positive effect on employment, annual payroll and the number of business
establishments on the zip code level whereas the subsequent Broadband Loan Program had
mostly no significant impact. Overall, it cannot be concluded that the Broadband Loan
Program helped to promote economic development in rural areas.

Forman et al. (2012) use U.S. county-level data to examine the relationship between Internet
investment and wage growth from 1995 to 2000. The authors employ various instrumental
variables to identify causal effects and find that business investment in advanced Internet
technologies (such as ERP or e-commerce) comes along with significant wage and
employment growth. However, this holds only for six percent of counties that were
characterized by intensive IT usage, high income, high skills and high population density
already before 1995. In turn, broadband internet and wage growth appear unrelated in the
other counties. Thus, although the Internet lowers the cost of engaging in economic activity in
geographically isolated regions, better performing regions benefit significantly more from
investment in advanced Internet technologies and applications.

Kolko (2012) uses ZIP code- and county-level data for the U.S. from 1992 to 2006 to analyze
the impact of the number of broadband providers on local economic outcomes. The author
applies static ordinary and weighted two-stage least squares regressions in which the number
of broadband providers is instrumented by the average slope of terrain within an area.
Employment growth is found to be significantly and positively affected by the number of
broadband providers in both specifications and also when population growth is being
controlled for. Technology-intensive sectors such as scientific and technical services appear to
benefit the most whereas other sectors such as mining and public administration seem not to
be significantly impacted. Moreover, the significantly negative coefficient of population
density implies that the impact is larger in less densely populated areas. Aside from an
increase in population growth, the effects of a higher number of broadband providers on other

22
household outcomes on the county level are more ambiguous: for the period from 1992 to
1999, broadband expansion positively influences employment rate, average pay per employee
and median income. In contrast, no significant effect on the first two indicators and even a
negative impact on median income is obtained for the period from 1999 to 2006. These results
in combination with the increase in population growth suggest that the growth in employment
due to higher broadband availability may initiate an influx of workers which causes workers’
per capita outcomes to be un- or negatively affected. Employing household survey data, it is
further shown that a higher number of broadband providers did not lead to an increase in
telecommuting or home-based businesses.

Atasoy (2013) studies the labor market effects of broadband availability in U.S. counties with
county-level panel data from 1999 to 2007. Using the share of population living in ZIP codes
with at least one broadband provider as a measure of broadband availability, the author
estimates static ordinary least squares and fixed-effect models. In addition, regressions similar
to Granger causality tests as well as a falsification test are employed to check for causality.
Broadband is found to have a significant and causal positive effect on the employment rate
that is robust to different time trends and mainly stems from an expansion of workforce in
existing firms rather than from the foundation of new businesses. The inclusion of an
interaction term between broadband and education levels further shows that broadband
increases employment and wages solely for workers with a college degree. In contrast, the
impact of broadband on the employment rates and wages of workers without a college degree
is significantly negative in this specification. In line with these results, the author also finds
that the positive impact of broadband availability is more pronounced in counties with a
higher number of skilled workers and in industries that require highly skilled and educated
workers. On a regional level, rural areas benefit more than urban areas from increased
broadband availability although the magnitude of this advantage is rather small.

Mack and Faggian (2013) use county-level data for the U.S. to assess the role of broadband
availability in 1999 on changes in productivity between 2000 and 2007. To account for spatial
variations between counties, the authors employ fixed-effect spatial lag models. Further,
productivity is measured by earnings and interaction terms between broadband availability
and workers’ skill/education levels are constructed. Specifying broadband availability as a
dummy variable, it is initially found that broadband availability increases productivity
significantly. When interaction terms for education and skill are added, the coefficients for
these terms are positive and significant while the effect of broadband availability alone

23
becomes negative and significant. Similar results are obtained when the number of broadband
providers is used as a continuous variable. In this case however, the coefficient for broadband
availability is also significantly negative when interaction terms are excluded. These findings
suggest that broadband can only develop its potential benefits if workers are sufficiently
trained regarding its use and public policies should be designed accordingly.

Czernich (2014) examines the impact of DSL-based broadband availability on unemployment

rates by using municipality- and county-level data for Germany from 2002 and 2006. The
author initially applies a dynamic ordinary least squares regression model in which the
unemployment rate in 2006 is regressed on the percentage of households with DSL
availability in 2006, the unemployment rate in 2002 and various control variables. In these
specifications, broadband availability has a significant negative effect on the unemployment
rate even though the coefficient of the DSL rate becomes smaller and less significant when
more control variables are taken into account. Similar findings are obtained when the
unemployment rate of under 25 year-olds is considered instead of the overall unemployment
rate although the coefficient of the DSL rate variable remains highly significant for all
specifications. When the distance of a municipality to the next main distribution frame is
employed as an instrument for the DSL availability rate, these results no longer hold. Instead,
the dynamic instrumental variable regressions suggest that DSL availability increases
unemployment albeit this effect is not statistically significant. Likewise, no significant
influence is observed for the unemployment rate of under 25 year-olds. Based on these
results, the author concludes that higher broadband availability does not causally imply a
lower unemployment rate. However, it is also cautioned that the instrumental variable
regressions only include municipalities without an own main distribution frame. Since these
municipalities are mostly located in rural areas, the insignificance of the impact may be
limited to rural municipalities.

Mack and Rey (2014) investigate the relationship between broadband availability and the
presence of knowledge intensive businesses in 2004 by utilizing ZIP code-level data for 54
U.S. metropolitan areas. Estimating an individual model for each metropolitan area, the
authors employ static ordinary or two-stage least squares regressions for areas without spatial
dependence as well as spatial lag or spatial error models for areas with spatial dependence.
The regression results show that broadband availability had a significantly positive impact on
the number of knowledge intensive firms in 49 of the 54 metropolitan areas which suggests a
generally beneficial effect of broadband. Given the industrial and geographical variety of the

24
affected areas, the authors argue that broadband may not only further foster areas with already
existing knowledge firm concentration but may also help areas which have no such
agglomerations or suffer from large distances and high commuting costs.

Whitacre et al. (2014) employ county-level data for the U.S. from 2001 to 2010 in order to
obtain a causal link between fixed residential broadband availability/adoption and rural
economic development. For this purpose, the authors group non-metropolitan counties
according to their performance regarding several broadband speed thresholds and
subsequently perform nearest neighbor and kernel propensity score matching with logit
propensity scores that are based on county characteristics in 2001. Several significant effects
are observed: Firstly, counties with higher levels of broadband adoption experience faster
growth in median household income and reduced growth in unemployment while counties
with lower levels of broadband adoption suffer from lower growth of employment and
number of firms. Moreover, high download speed is associated with lower levels of poverty
and more creative class workers. On the downside, counties with higher levels of broadband
availability experience lower growth in non-farm proprietor income whereas a higher growth
in median household income is, to a marginal extent, prevalent in counties with low levels of
broadband availability and download speed. These results suggest that both the increase of
broadband adoption and the increase in available broadband speed should be emphasized by
broadband policies.

Canzian et al. (2015) investigate the effects of a government initiative for ADSL2+ broadband
deployment on rural communities in the Italian Province of Trento with panel data for firms
(outside the agricultural and public service sector) and municipalities for the years from 2008
to 2012. Treating the deployment as a quasi-experimental set-up, the authors apply a fixed-
effect difference-in-differences approach in which the provision of ADSL2+ broadband is
used as a dummy variable. As an alternative measure, the number of days that ADSL2+
broadband had been available in a municipality is employed as a continuous variable. In the
binary specification, the authors find that the availability of broadband had a significant
positive influence on annual sales turnover and annual value added with broadband
availability increasing the annual turnover by around 40 percent and the annual value added
by around 25 percent over the years from 2010 to 2012. Similar results are obtained for the
continuous variable specification even though the treatment effect for the annual sales
turnover is nonlinear. Placebo analysis for the years from 2008 to 2010 suggests that the
effects are causal. On the sectoral level, the authors find that the majority of observed sectors

25
significantly benefit from broadband availability and that these benefits are particularly large
for hotels and restaurants as well as professional, scientific, technical and administrative
services. However, no specific increase in impact is observed for the high-tech sector.

Ahlfeldt et al. (forthcoming) employ transaction- and local exchange-level data for England
from 1995 to 2010 and individual speed test data from 2009 and 2010 to investigate the
relationship between fixed residential broadband speed and housing prices. The authors apply
a hedonic pricing approach and a boundary discontinuity design to construct static fixed-
effect regressions that utilize broadband availability variations over time within local
exchange areas as well as variations over time across local exchange boundaries. The
estimations suggest a significant positive impact of broadband speed on housing prices with
diminishing returns to speed. The speed elasticity of property prices amounts to 3 percent at
the real mean speed of 2.2 Mbit/s and enhancing a broadband connection from low to fast
speed would increase the value of a property by at most 1 percent. Based on an expansion of
their baseline model, the authors subsequently discuss the benefits and costs of the DAE. For
urban areas, the benefits of the speed upgrades are found to easily outweigh the costs
regardless of whether FTTB technologies or the more expensive FTTH technologies are used
to achieve the upgrades. In contrast, the benefits of the speed upgrades for suburban and rural
areas are not sufficient to cover the costs so that a considerable share of the benefits of the
coverage upgrades would be needed as well.

Ivus and Boland (forthcoming) analyze the influence of fixed broadband coverage on local
employment and wage growth employing data for 4344 Canadian communities for the years
from 1997 to 2011. The authors classify the communities into 76 economic regions and utilize
the impact of the variation in elevation within a region on the rate of broadband deployment
to use the logarithm of the former measure as the instrument in static instrumental variable
regressions. While the baseline estimations do not suggest a significant effect of increases in
broadband coverage, this changes when an interaction term between broadband coverage and
rural areas is included. In this case, higher broadband coverage has a significant positive
impact on employment growth in rural areas and a marginally significant negative impact on
employment growth in urban areas. Furthermore, these effects are particularly prominent in
IT-intense industries and can also be observed when only wireless broadband technologies are
examined. In contrast, the baseline estimations with wage growth as the dependent variable
suggest a positive influence of broadband coverage on wage growth. However, no specific
impact on rural areas is found when an interaction term for rural areas is specified. Moreover,

26
the authors find that the benefits of increased broadband coverage regarding employment and
wage growth mainly seem to stem from the service sector whereas no effect was found for
manufacturing firms.

3.3 Productivity and firm performance

Grimes et al. (2012) examine the impact of firms’ broadband adoption on labor productivity
using firm-level data for 6060 firms in New Zealand collected in 2006. As a first approach,
the authors apply strata and kernel propensity score matching. It is found that broadband
adoption has a significant positive effect on labor productivity within a firm implying a
productivity increase of around 7 to 10 percent. However, due to high standard deviations of
productivity within sectors, this increase does not translate to a strongly improved
performance relative to the other firms in the sector. Furthermore, the authors observe no
significant variations in productivity increases between firms in rural and urban areas as well
as between firms in high and low knowledge industries. Employing dynamic instrumental
variable regressions as a second approach, a significant positive effect is found that is
considerably larger than the one found by propensity score matching.

Bertschek et al. (2013) employ firm-level data for Germany to analyze the relationship
between firms’ broadband adoption (leased line or DSL) and changes in their labor
productivity and innovation activity between the years 2001 and 2003. For labor productivity
(measured as sales per employee), the authors utilize a production function approach and
apply the duration of DSL availability as an instrument for broadband adoption in static
instrumental variable regressions. While broadband adoption initially appears to have a
marginally significant positive impact on labor productivity, no such impact is found when
regional differences are taken into account. Based on these findings, the authors conclude that
the effects of broadband adoption vary between firms and more extensive data might be
needed to observe the long-run benefits of broadband usage in its entirety. The probability of
realizing a process or product innovation is estimated by a recursive binary probit model in
which the duration of DSL availability is used as the exclusion restriction in the equation for
broadband adoption. For both innovation types, broadband adoption is found to have a
significant positive effect with average treatment effects estimated to be in the order of 41 to
45 percent for product innovations and in the order of 28 to 35 percent for process
innovations. However, the authors caution that these effects might be overstated due to
negative correlation between broadband adoption and unobserved factors that increase
innovation activity.

27
Colombo et al. (2013) examine the influence of broadband adoption on the productivity of
small and medium enterprises utilizing firm- and province-level panel data for 799 Italian
firms from 1998 to 2004. The authors employ principal component analyses to construct
categories of broadband applications and accentuate enterprises that implemented strategic or
organizational changes. Subsequently, augmented Cobb-Douglas production functions
including year-fixed effects are estimated by a two-step system Generalized Method of
Moments estimator. Wald test statistics suggest that the adoption of basic broadband
applications such as e-banking do not significantly influence firm productivity even in
combination with strategic or organizational changes. In contrast, more advanced applications
are shown to be capable of significantly increasing productivity but only if they are of
importance for the specific sector and are conjoined with appropriate changes in business
conduct. In particular, sizeable productivity gains are experienced by firms in the
manufacturing sector when ‘supply chain and customer management’ applications like e-mail
or e-learning are adopted together with strategic changes. Similarly, firms in the service sector
can raise their productivity by adopting ‘advanced communications’ applications like file-
sharing systems in combination with organizational changes. These results provide evidence
that firms can only realize the potential benefits of broadband if the right set of broadband
applications is combined with sufficient changes in doing business. Hence, the authors argue
that public policies should actively help firms achieve these targets rather than merely provide
them with monetary aid.

De Stefano et al. (2014) use plant- and firm-level panel data for the region around the city of
Kingston-upon-Hull (UK) from 2000 to 2004 in order to investigate the influence of firms’
ADSL broadband adoption on their performance. For this purpose, the authors initially
employ static ordinary least squares regressions. The results suggest a significant positive
relationship between broadband adoption and firm size, sales and sales per worker. However,
no relationship between broadband adoption and sales growth or exit of firms is found. In
order to obtain a causal link, a static fuzzy regression discontinuity approach is applied. This
approach makes use of a spatial discontinuity in broadband availability within the city of
Kingston-upon-Hull caused by two broadband providers. In this specification, the positive
results of the previous estimations cease to hold and no significant impact of broadband
adoption on the economic outcomes is found. Since these findings are sustained under various
robustness checks, the authors conclude that ordinary least squares regressions are likely to
overestimate the effects of broadband and no causal link between broadband adoption and
firm performance exists for the sample under consideration.

28
Akerman et al. (2015) analyze the effect of broadband availability and adoption on labor
productivity with special focus on workers’ skill levels using Norwegian firm-level data from
2000 to 2008. For this purpose, the authors initially conduct intention-to-treat analysis and
find that increased broadband availability had a significant positive impact on the hourly
wages and employment of skilled workers. Contrary to that, the hourly wages of low skilled
workers were negatively affected. Similarly, broadband availability significantly increased the
output elasticities of skilled workers while it significantly reduced the output elasticities of
low-skilled workers. These results are in the following confirmed for broadband adoption
through estimations of production functions by fixed–effect ordinary least squares and
instrumental variable regressions. Moreover, broadband is adopted faster by firms that rely
heavily on skilled workers. In addition to the effects mentioned above, broadband availability
is also found to have a positive effect on firms’ outputs. In particular, the estimations indicate
that increasing broadband availability by 10 percent raises output by 0.4 percent for a fixed
amount of inputs.

Haller and Lyons (2015) utilize firm-level panel data for 2290 Irish manufacturing firms from
2002 to 2009 to investigate the relationship between firms’ DSL broadband adoption and
productivity as well as productivity growth. To obtain a causal link, the authors apply static
fixed-effect instrumental variable regressions in which broadband adoption enters the
regression equation as a dummy variable and geographical DSL broadband availability is
used as an instrument for broadband adoption. Although pairwise correlations indicate that
broadband adoption positively impacts firm productivity and negatively impacts productivity
growth, the regressions mentioned above are unable to support these results. Instead, the
authors find that neither firm productivity nor productivity growth are significantly affected
by broadband adoption even though the coefficients have the expected signs. These findings
continue to hold when only broadband speeds ≥ 2 Mbit/s or broadband technologies other
than DSL are considered as well as when firms are classified in terms of their respective size.
Furthermore, the coefficient remains insignificant when labor productivity instead of total-
factor productivity is employed as the dependent variable. Due to these results, the authors
conclude that broadband adoption did not lead to higher firm productivity or productivity
growth in the examined case.

Bertschek and Niebel (2016) analyze whether employees’ access to mobile internet improves
firms’ labor productivity. The analysis is based on a firm-level data set containing 2143 firms
from the German manufacturing and services industries and referring to the year 2014. At this

29
time, high-speed mobile internet was still at a relatively early stage of diffusion within firms.
Only 26 per cent of employees in the sample were equipped with mobile devices being able to
connect to the internet. The econometric analysis shows that firms' labor productivity
(measured as sales per employee) significantly increases with the share of employees with
mobile internet access. ICT use at the workplace other than mobile internet is controlled for
by including measures of the use of computers and access to fixe line internet. An
instrumental variables approach suggests that mobile internet use does cause higher labor
productivity. Positive aspects of mobile internet use such as improving information flows and
supporting a more flexible organization of work with respect to time and location seem to
outweigh negative aspects like higher monitoring cost.

3.4 Conclusions on the economic impacts of broadband internet

The main results of the reviewed papers studying the economic impacts of broadband internet
are presented in ascending chronological order in Table 3 (‘impact on economic growth’),
Table 4 (‘impacts on employment and regional development’) and Table 5 (‘impact on firm
performance’). In each table, we apply a further separation into studies examining the impacts
of broadband availability (infrastructure investment) and broadband adoption (usage).

Overall, four out of five studies that examine the impacts of broadband adoption on GDP and
GDP growth with country-level data find a positive and significant effect (RQ 1). Arvin and
Pradhan (2014) represents the only study with inconclusive results on the relationship
between broadband and GDP. Contrary to that, the clear results in the other studies ascertain
that broadband adoption indeed causes GDP growth. Given that two of the cross-country
studies provide evidence for a critical mass or a stronger influence of broadband in developed
countries, the impact should, however, not be assumed to be linear but rather to be increasing
in the penetration rate of a country. Moreover, it seems to be necessary for countries with low
broadband penetration to reach a certain penetration level if they want to experience the
largest possible benefits of broadband.

The majority of studies (6 out of 7) that focus on overall employment find positive links
between broadband and employment growth (RQ 1). Again, this impact is not constant:
several studies that take into account differences in urbanization and workers’ skill levels
suggest that the effect of broadband is significant mostly in rural areas and for high-skilled
workers. The stronger impact on remote areas lends support to the notion that broadband can
help these regions to catch up with more economically developed urban areas.

30
On the firm level, the evidence for a positive impact of broadband on productivity is mixed.
While 2 studies find a significant positive effect, 2 more recent studies argue against the
existence of a significant effect. Colombo et al. (2013) verify a positive impact only for
advanced broadband applications in combination with strategical and organizational changes.
These results provide evidence that broadband adoption does not necessarily increase firm
productivity and that ICT-intensive businesses may be favored with regard to productivity
growth. Furthermore, they support the theory that firms need to incorporate structural changes
in business conduct in order to achieve the potential positive impact of ICT on productivity. 2

In sum, we find strong evidence for positive impacts of broadband networks on economic
outcomes (RQ 1). With regard to our research question referring to differences in wireline and
wireless infrastructure and services (RQ 2), however, there is no conclusive evidence as only
very few studies employ data related to mobile broadband networks. Concerning our third
research question (RQ 3), we also find no clear evidence that broadband availability and
adoption produce systematically different economic outcomes although research by Whitacre
et al. (2014) indicates that certain differential impacts may exist. Moreover, it is not
necessarily the fact of using broadband that matters but the type of application that is
employed. Lastly, in contrast to the throughout positive impact of broadband found at the
macro level, microeconomic studies do not find clear evidence for positive productivity
effects of broadband on the firm level (RQ 4).

2
For related literature on ICT, see e.g. Brynjolfsson and Hitt (2000) and Cardona et al. (2013).

31
Table 3: Broadband's impact on economic growth

Author(s) Data Methodology Main results

Broadband adoption
Koutroumpis 22 OECD Simultaneous equations model • Broadband adoption positively
(2009) countries; estimated by instrumental impacts GDP growth
2002-2007; variable regressions and three- • Critical mass at 30 percent
country-level stage least squares Generalized adoption
Method of Moments with fixed
effects

Czernich 25 OECD Two-stage instrumental variable • Broadband introduction and

et al. countries; approach: broadband adoption have a
(2011) 1996-2007; 1. Two-stage least squares positive effect on growth of GDP
country-level estimation of a logistic per capita
diffusion model • Critical mass at an adoption rate
2. Difference-in- of 10 percent
differences estimations
and static fixed-effect
cross-country growth
model estimation

Thompson & 43 countries; Static fixed-effect instrumental • Mobile broadband adoption

Garbacz 2005-2009; variable regressions positively influences GDP per
(2011) country-level household
• Fixed broadband adoption has an
insignificant negative impact on
GDP per household

Arvin & 19 of G20 Panel cointegration and Granger • No conclusive evidence for a
Pradhan countries; causality tests causal relationship between
(2014) 1998-2011; broadband penetration and GDP
country-level growth in the long run
• Bidirectional causal relationship
between broadband penetration
and GDP growth for developed
countries in the short run

Gruber 27 EU Simultaneous equations model • Broadband adoption has a

et al. countries; estimated by fixed-effect three- positive impact on GDP
(2014) 2005-2011; stage least squares to obtain • Realizing the goals of the DAE
country-level return parameters; varying would generate economic
assumptions about economic benefits for the whole EU and
development most member countries

32
Table 4: Broadband’s impact on employment and regional development

Author(s) Data Methodology Main results

Broadband availability
Kandilov & USA; Fixed-effect difference-in- • Some positive effects of the pilot
Renkow 1994-2007; ZIP differences approach and program on employment, annual
(2010) code- and nearest neighbor propensity payroll and number of business
county-level score matching establishments
• No effects of the subsequent loan
program

Kolko (2012) USA; Static ordinary and weighted • A higher number of broadband
1992-2006; ZIP two-stage least squares providers positively impacts
code- and regressions employment and population
county-level growth
• Varying effects on employment
rates, average pay per employee
and median income

Atasoy USA;; Estimation of static ordinary • Broadband availability increases

(2013) 1999-2007; least squares and fixed-effect employment for workers with
models, Granger causality test college degree but decreases
ZIP code- and and a falsification test employment for workers without
county-level college degree
• Employment growth mainly stems
from existing firms rather than
newly founded firms

Mack & USA; Estimations of spatial lag • Only workers with high
Faggian 1999, 2000, models; interaction terms education/skill level benefit
(2013) 2007; between broadband and significantly from increased
county-level workers’ skill broadband availability
(partially education/education level • Uneducated/unskilled workers are
aggregated) negatively affected

Czernich Germany; 2002- Dynamic regression equations • No evidence for a significantly

(2014) 2006; estimated by ordinary least causal impact of DSL broadband
municipality- squares and instrumental availability on unemployment
and county-level variable regressions rates

Mack & USA; Static ordinary and two-stage • Increased broadband availability
Rey 2004, 2006; ZIP least squares regressions as is associated with higher presence
(2014) code-level well as spatial error and spatial of knowledge intensive businesses
lag models in almost all observed
metropolitan areas

33
(Table 4: Continued)

Author(s) Data Methodology Main results

Broadband availability
Whitacre USA; Nearest neighbor propensity • Broadband availability negatively
et al. (2014) 2001-2010; score matching based on logit affects median household income
county-level propensity scores and non-farm proprietors income
in non-metropolitan areas

Canzian Province of Quasi-experimental set up, • Introduction and expansion of

et al. Trento (Italy); estimations via a fixed-effect ADSL 2+ broadband availability
(2015) 2008-2012; difference-in-differences in rural areas positively affected
firm- and approach firms with regard to annual sales
municipality- turnovers and annual value added
level • Positive impact on the number of
employees per firm only in rural
areas

Ahlfeldt England; Boundary discontinuity • Available broadband speed has a

et al. 1995-2010; approach estimated by static significantly positive but
(forthcom- individual-, fixed-effect regressions decreasing impact of property
ing) transaction- prices
and local • Achieving solely the speed goals
exchange-level of the DAE will cover the costs
only in urban areas

Ivus & Canada; Grouping of municipalities into • Significantly positive impact of

Boland 1997-2011; economic regions; static broadband availability on
(forthcom- municipality- instrumental variable regressions employment growth in rural areas
ing) level and overall wage growth
• Positive impact stems from the
service sector

Broadband adoption
Forman et al. USA; Estimations with instruments • Investment in advanced Internet
(2012) 1995-2000; and timing of regional wage technologies comes along with
county-level divergence to identify causal significant wage and employment
links growth in certain counties
• Already better performing
.
regions benefit significantly more

Whitacre USA; Nearest neighbor and kernel • More broadband adoption is

et al. (2014) 2001-2010; propensity score matching based associated with faster growth of
county-level on logit propensity scores employment, businesses and
median household income

34
Table 5: Broadband’s impact on productivity and firm performance

Author(s) Data Methodology Main results

Broadband availability
Akerman Norway; Intention-to-treat analysis, • Broadband availability (and
et al. 2000-2008; estimations of production adoption) positively affects the
(2015) firm-level functions by ordinary least employment and wages of skilled
squares and fixed-effect workers but negatively affects
instrumental variable regressions wages of low-skilled workers
• Further positive impact on firms’
output

Broadband adoption
Grimes New Zealand; Strata and kernel propensity • More broadband adoption is
et al. 2006; score matching with probit associated with increased labor
(2012) firm-level propensity scores, instrumental productivity
variable regressions as • No differences between
robustness checks urban/rural areas and high/low
technology industries
Bertschek Germany; Production function approach • Broadband usage has a significant
et al. 2001-2003; estimated by static instrumental positive impact on firms’
(2013) firm-level variable regressions to estimate innovation activities
changes in labor productivity; • No effect on labor productivity
recursive binary probit approach
to estimate the probability of
innovations

Colombo Italy; Augmented Cobb-Douglas • Adoption of basic broadband

et al. 1998-2004; production functions estimated application does not increase
(2013) firm-level by two-step system Generalized firms’ productivity
Method of Moments including • More advanced broadband
year-fixed effects applications can be beneficial if
they are combined with strategic
or organizational changes

De Stefano et Kingston-upon- Ordinary least squares • Increased ADSL adoption does

al. Hull (UK); regressions and a fuzzy not lead to significant changes in
(2014) 2000- discontinuity regression firm performance
2004;plant- and approach are used to estimate
firm-level the static regression equations

Haller & Ireland; Static regression equations • No effect of DSL broadband

Lyons (2015) 2002-2009; estimated by fixed-effect adoption on firm productivity and
firm-level instrumental variable regressions productivity growth

35
(Table 5: Continued)

Authors Data Methodology Main results

Broadband adoption
Bertschek & Germany; Production function approach • Firms’ labor productivity
Niebel (2016) 2014; estimated by static instrumental significantly increases with the
firm-level variable regressions to estimate share of employees with mobile
changes in labor productivity internet access

36
4 Conclusions and policy recommendations

In this article, we have provided a survey of quantitative research on the economic impacts of
telecommunications networks and broadband internet. Differentiating between wireline and
wireless technologies as well as broadband availability and broadband adoption, respectively,
we have reviewed studies investigating the causal effects on productivity and various
measures of firm performance, economic growth, employment and regional development.

Aiming at relating the main insights of our survey to the four research questions identified in
the introduction, we first find that both telecommunications and broadband infrastructure and
services tend to exert a positive impact on economic growth as well as on productivity (RQ
1). On a macroeconomic level (RQ 4), increased broadband deployment not only translates
into higher economic growth and productivity but also helps to create new jobs and, at least
partially, facilitates economic development in rural areas. However, the impact on regional
development is mixed as several studies also find insignificant effects. Overall, focusing on
these findings only suggests that promoting broadband deployment is an effective measure to
realize welfare gains (RQ 1). The microeconomic studies under review (RQ 4), however,
paint a more differentiated picture. Most notably, broadband adoption is accompanied by a
pronounced skill bias regarding workers’ labor market outcomes that strongly favors highly
skilled workers. While skilled workers enjoy higher wages and employment rates as well as a
rise in productivity as a consequence of broadband adoption, workers with lower skill levels
experience higher unemployment and lower wages. Accordingly, business establishments that
are ICT intensive and require a relatively large amount of skilled workers such as the service
sector are observed to be the main beneficiaries of broadband. In contrast, industries such as
the manufacturing sector that rely heavily on manual labor are not or even negatively affected.
Furthermore, firms may need to implement changes in doing business in order to materialize
the possible economic advantages of broadband usage.

Regarding the differential impact of fixed and mobile broadband technologies (RQ 2), there is
hardly any empirical evidence available. Wireline and mobile telecommunications differ,
however, with regard to the country-specific distribution of their benefits. Wireline
telecommunications are found to have the strongest effect in developed countries exhibiting
almost nation-wide coverage. Furthermore, the reviewed studies provide evidence that mobile
telecommunications are particularly advantageous for economic growth and productivity in

37
developing countries with much less developed wireline infrastructure. Since developing
countries can often still achieve sizeable increases in telecommunications usage and
consequently economic growth, our main policy implication in this regard is therefore that
developing countries should mostly focus on designing and implementing policies that
facilitate the deployment of mobile telecommunications or directly deploy wireline next
generations broadband networks especially if those countries currently possess a low level of
old wireline telecommunications infrastructure.

The estimated coefficients in the investment or infrastructure-related empirical specifications

represent the impact of broadband availability on economic outcomes but not the effect
related to actual usage of broadband and telecommunications services (RQ 3). Whereas the
former measures the intention-to-treat effect, the latter directly impacts on economic
outcomes measuring the average treatment effect. Apparently, adoption of services is a
function of availability of the underlying infrastructure. Accordingly, the empirical studies
employing measures of availability only estimate a reduced-form model where the estimated
coefficients represent a proportional effect which is smaller than the effect via broadband
usage (Czernich, 2014). In that sense, impact assessments of the infrastructure availability
actually tend to underestimate economic impacts and realized welfare gains. At the same time
these studies are more informative from a policy perspective which is mainly targeted at the
supply-side, i.e. investment by broadband providers, via regulation or public funding. In fact,
according to our review of the literature, almost all adoption-related broadband studies find
positive effects on economic outcomes whereas this finding is clearly less evident in
availability-related broadband studies (RQ 3).

Our findings have further specific implications for policy makers. The existence of network
effects and the often small incentives for broadband providers to supply rural areas due to
high costs and low revenues suggest that public initiatives and financial aid should be
considered as instruments to achieve increased broadband development, particularly in
countries with a large share of the population living outside of urban areas. At the same time,
however, policy makers will have to take into account the distributional effects that are caused
by the skill bias of broadband adoption. Accordingly, it appears necessary to accompany
broadband policies with increased investments in education for both individuals (general
education and advanced training on the job) and firms (to facilitate changes in business
conduct). Such policies would increase not only the share of workers benefitting from

38
broadband by enabling them to find employment in ICT-intensive jobs but also the overall
impacts of broadband on economic growth and employment.

Turning from the main insights and policy implications of our survey to the identification of
main research gaps, despite the great variety of data sets and methodologies applied in the
reviewed studies, several limitations still remain. Firstly, a large number of studies focusing
on the impact of broadband on regional development employ measures of the mere existence
of broadband infrastructure instead of actual broadband adoption as explanatory variables.
Albeit previous literature (e.g. Cava-Ferreruela and Alabau-Munoz, 2006; Akerman et al.,
2015) generally indicates that the relationship between broadband availability and broadband
adoption is positive, broadband availability only serves as a pre-condition for broadband
adoption. In this regard Whitacre et al. (2014) suggest that the influence of broadband
availability and actual broadband adoption can differ considerably. Moreover, the FCC data
which is used in most U.S. studies often measures broadband availability as the number of
broadband providers in a ZIP code. This measure has been repeatedly criticized in the related
literature (e.g. Grubesic, 2008; Holt and Jamison, 2009) for not accurately measuring
broadband availability.

Another gap in the existing research on the economic impacts of broadband is that only very
few studies specifically investigate the impact of broadband speed on economic outcomes. In
fact, most studies define broadband very broadly as internet connections with download
speeds greater than 200 Kbit/s and do not account for faster broadband speeds or different
broadband technologies. However, it seems likely that different speeds will have different
effects and that some industries might benefit in particular from higher broadband speeds.
Future research should hence focus on the impact of broadband speed, in particular, with
respect to the current deployment of next-generation broadband networks which will
gradually replace old broadband networks and provide broadband speeds as targeted in the
Digital Agenda for Europe (DAE) in the EU or the National Broadband Plan in the U.S.
Meanwhile these targets have been implemented in most EU member states in national
broadband plans which are in many cases targeted at even more ambitious coverage and
adoption levels 3.

Accordingly, assessing the costs and benefits appears to be of high policy relevance. In order
to estimate costs and benefits of reaching the DAE goals, Gruber et al. (2014) examine several

3
See https://ec.europa.eu/digital-single-market/en/broadband-member-states (last accessed on 29 July 2016)
for further information.

39
scenarios to account for different economic developments. The authors further find that
returns exceed costs for the whole European Union in all cases with base case estimates
projecting net cumulative gains to be in the magnitude of €282 billion. Given that costs are
estimated to be €213.6 billion, this implies that the net gains will exceed costs by 32 percent
in the base case. On an individual level, most countries should benefit from achieving the
goals of the DAE although the extent of these benefits varies considerably among the member
states depending on their already existing broadband infrastructure stock and rate of
urbanization. The authors conclude that public subsidies might be necessary to realize the
complete benefits of broadband as private firms would not fully supply the infrastructure
otherwise. Ahlfeldt et al. (forthcoming) also found that the speed upgrades easily outweigh
the costs regardless of which fiber technologies are actually used. However, this holds only
for urban areas, whereas the benefits of the speed upgrades for suburban and rural areas are
not sufficient to cover the costs.

Although Gruber et al. (2014) explicitly include high-speed next generation broadband
technologies, reliable and broad evidence on economic impacts of high-speed wireline or
wireless broadband infrastructure and adoption is still largely missing so far but of great
policy relevance, in particular for the public funding of such networks.

While a multitude of studies examine the effects of wireline and wireless telecommunications
on macroeconomic growth and productivity with country-level data, the number of studies
that investigate the impacts of telecommunications and broadband infrastructure on firm or
household level is much smaller. For that reason, the existing literature so far only allows
valid conclusions on the aggregated effects of telecommunications and broadband networks
but not on the distribution of benefits within the observation units. A more granular analysis is
also necessary to gain insights on whether infrastructure availability and services help rural
areas to catch up with urban areas, i.e. whether closing the digital divide also closes the
economic divide. Moreover, further research is needed with respect to the labor market effects
of telecommunications and broadband networks, in particular on employment and wage
growth, in order to further examine whether the deployment of networks is accompanied by
complementarities related to education or the adoption of other ICT services.

40
Appendix: Relevant telecommunications networks and technologies

With regard to the distinction between wireline and wireless networks, we follow the
International Telecommunication Union (ITU) which defines fixed telephone lines as ‘active
line[s] connecting the subscriber’s terminal equipment to the public switched telephone
networks [PSTN] and which has a dedicated port in the telephone exchange equipment’ (ITU
2010, pp. 1) and accordingly includes ‘analog fixed telephone lines, Integrated Services
Digital Network (ISDN) channels, fixed wireless, public payphones and VoIP [Voice over
Internet Protocol] subscriptions’ (ITU 2010, pp. 1). Given that active lines had to be utilized
within three months prior to the ITU data collection, studies that employ the number of
mainlines will be defined as examining telecommunications adoption (actual usage) and not
merely telecommunications availability (the existence of infrastructure).

Mobile phone subscriptions refer to ‘the subscriptions to a public mobile telephone service
and provides access to Public Switched Telephone Network using cellular technology,
including number of pre-paid Subscriber Identity Module (SIM) cards actively during the past
three months. This includes both analogue and digital cellular systems (IMT-2000 Third
Generation, 3G and 4G subscriptions), but excludes mobile broadband subscriptions via data
cards or USB [Universal Serial Bus] modems’ (ITU 2010, pp. 3).

Our definition of fixed broadband will include both fixed wired technologies (such as cable
modem, xDSL and FTTH/B/C/N technologies) as well as fixed wireless hybrid technologies
such as Wi-Fi (WLAN). It should be noted at this point that most of the reviewed studies
employ very broad definitions of fixed broadband with regard to broadband speed. While the
lowest speed threshold in a study is set at 128 Kbit/s, a majority of the research that analyzes
the effects of broadband availability in the US employs data from the FCC which, until 2010,
defined broadband as an internet connection with either up- or downstream speeds of at least
200 Kbit/s (see FCC 2010b). Similarly, most cross-country studies as well as several other
publications make use of OECD (2016) and ITU (2010) standards which associate broadband
with download speeds of at least 256 Kbit/s. Studies utilizing German data, on the other hand,
apply slightly higher thresholds at 384 or 768 Kbit/s.

Finally, mobile broadband will, in line with ITU (2010), be comprised of both satellite and
terrestrial mobile wireless subscriptions that achieve download speeds of 256 Kbit/s or more.
Accordingly, mobile broadband includes, in particular, Universal Mobile

41
Telecommunications System (UMTS) 3G, 4G, Worldwide Interoperability for Microwave
Access (WiMAX) and Long Term Evolution (LTE) technologies.

As a consequence, fixed and mobile technologies below the above-mentioned bandwidth

thresholds are defined as narrowband telecommunications infrastructures and services which
we simply refer to as telecommunications networks in the following.

42
References

Ahlfeldt, G., Koutroumpis, P., & Valletti, T. (forthcoming). Speed 2.0-Evaluating Access to
Universal Digital Highways. Journal of the European Economic Association.

Aker, J. C. (2010). Information from Markets Near and Far: Mobile Phones and Agricultural
Markets in Niger. American Economic Journal: Applied Economics, 2(3), pp. 46-59.

Akerman, A., Gaarder, I., & Mogstad, M. (2015). The Skill Complementarity of Broadband
Internet. Quarterly Journal of Economics, 130(4), pp. 1781-1824.

Arvin, B. M., & Pradhan, R. P. (2014). Broadband Penetration and Economic Growth Nexus:
Evidence from Cross-country Panel Data. Applied Economics, 46(35), pp. 4360-4369.

Atasoy, H. (2013). The Effects of Broadband Internet Expansion on Labor Market Outcomes.
Industrial & Labor Relations Review, 66(2), pp. 315-345.

Barro, R. J. (1991). Economic Growth in a Cross Section of Countries. Quarterly Journal of

Economics, 106(2), pp. 407-443.

Bertschek, I., & Niebel, T. (forthcoming). Mobile and More Productive? Firm-Level Evidence
on the Productivity Effects of Mobile Internet Use. Telecommunications Policy.

Bertschek, I., Cerquera, D., & Klein, G. J. (2013). More Bits–More Bucks? Measuring the
Impact of Broadband Internet on Firm Performance. Information Economics and Policy,
25(3), pp. 190-203.

Brynjolfsson, E., & Hitt, L. M. (2000). Beyond Computation: Information Technology,

Organizational Transformation and Business Performance. The Journal of Economic
Perspectives, 14(4), pp. 23-48.

Canzian, G., Poy, S., & Schüller, S. (2015). Broadband Diffusion and Firm Performance in
Rural Areas: Quasi-Experimental Evidence, IZA Discussion Papers no. 9429. Retrieved
from: http://www.econstor.eu/bitstream/10419/124938/1/dp9429.pdf (Last accessed on 29
July 2016).

Cardona, M., Kretschmer, T., & Strobel, T. (2013). ICT and Productivity: Conclusions from
the Empirical Literature. Information Economics and Policy, 25(3), pp. 109-125.

Cava-Ferreruela, I., & Alabau-Munoz, A. (2006). Broadband Policy Assessment: A Cross-

national Empirical Analysis. Telecommunications Policy, 30(8), pp. 445-463.

43
Chakraborty, C., & Nandi, B. (2011). ‘Mainline’ Telecommunications Infrastructure, Levels
of Development and Economic Growth: Evidence from a Panel of Developing Countries.
Telecommunications Policy, 35(5), pp. 441-449.

Cieślik, A., & Kaniewsk, M. (2004). Telecommunications Infrastructure and Regional

Economic Development: The Case of Poland. Regional Studies, 38(6), pp. 713-725.

Colombo, M. G., Croce, A., & Grilli, L. (2013). ICT Services and Small Businesses’
Productivity Gains: An Analysis of the Adoption of Broadband Internet Technology.
Information Economics and Policy, 25(3), pp. 171-189.

Crandall, R. W., & Jackson, C. L. (2001). The $500 Billion Opportunity: The Potential
Economic Benefit of Widespread Diffusion of Broadband Internet Access. Washington,
DC: Criterion Economics, LLC. Retrieved from:
http://www.ntia.doc.gov/legacy/ntiahome/broadband/comments/verizon/ExhibitA.pdf
(Last accessed on 29 July 2016).

Cronin, F. J., Colleran, E. K., Herbert, P. L., & Lewitzky, S. (1993). Telecommunications and
Growth: The Contribution of Telecommunications Infrastructure Investment to Aggregate
and Sectoral Productivity. Telecommunications Policy, 17(9), pp. 677-690.

Cronin, F. J., Parker, E. B., Colleran, E. K., & Gold, M. A. (1991). Telecommunications
Infrastructure and Economic Growth: An Analysis of Causality. Telecommunications
Policy, 15(6), pp. 529-535.

Czernich, N. (2014). Does Broadband Internet Reduce the Unemployment Rate? Evidence for
Germany. Information Economics and Policy, 29, pp. 32-45.

Czernich, N., Falck, O., Kretschmer, T., & Woessmann, L. (2011). Broadband Infrastructure
and Economic Growth. The Economic Journal, 121(552), pp. 505-532.

Datta, A., & Agarwal, S. (2004). Telecommunications and Economic Growth: A Panel Data
Approach. Applied Economics, 36(15), pp. 1649-1654.

De Stefano, T., Kneller, R., & Timmis, J. (2014). The (Fuzzy) Digital Divide: The Effect of
Broadband Internet Use on UK Firm Performance. University of Nottingham Discussion
Papers in Economics, (14/06). Retrieved from:
http://beta.nottingham.ac.uk/economics/documents/discussion-papers/14-06.pdf (Last
accessed on 29 July 2016).

44
Draca, M., Sadun, R., & Van Reenen, J. (2007). Productivity and ICT: A Review of the
Evidence. In R. Mansell (Ed.), The Oxford Handbook of Information and Communication
Technologies (pp. 100–147). Oxford University Press.

Dutta, A. (2001). Telecommunications and Economic Activity: An Analysis of Granger

Causality. Journal of Management Information Systems, 17(4), pp. 71-95.

European Commission (2010). A Digital Agenda for Europe. Retrieved from: http://eur-
lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:52010DC0245&from=EN (Last
accessed on 29 July 2016).

FCC (2010a). Connecting America: The National Broadband Plan.

Retrieved from: https://transition.fcc.gov/national-broadband-plan/national-broadband-
plan.pdf (Last accessed on 29 July 2016).

FCC (2010b). Sixth Broadband Deployment Report. Retrieved from:

https://apps.fcc.gov/edocs_public/attachmatch/FCC-10-129A1_Rcd.pdf (Last accessed on
29 July 2016).

Forero, M. D. P. B. (2013). Mobile Communication Networks and Internet Technologies as

Drivers of Technical Efficiency Improvement. Information Economics and Policy, 25(3),
pp. 126-141.

Forman, C., A. Goldfarb & Greenstein, S. (2012). The Internet and Local Wages: A Puzzle.
American Economic Review, 102, 556-575.

Greenstein, S. M., & Spiller, P. T. (1995). Modern Telecommunications Infrastructure and

Economic Activity: An Empirical Investigation. Industrial and Corporate Change, 4(4),
pp. 647-665.

Grimes, A., Ren, C., & Stevens, P. (2012). The Need for Speed: Impacts of Internet
Connectivity on Firm Productivity. Journal of Productivity Analysis, 37(2), pp. 187-201.

Gruber, H., Hätönen, J., & Koutroumpis, P. (2014). Broadband Access in the EU: An
Assessment of Future Economic Benefits. Telecommunications Policy, 38(11), pp. 1046-
1058.

Gruber, H., & Koutroumpis, P. (2011). Mobile Telecommunications and the Impact on
Economic Development. Economic Policy, 26(67), pp. 387-426.

Grubesic, T. H. (2008). Spatial Data Constraints: Implications for Measuring Broadband.

Telecommunications Policy, 32(7), pp. 490-502.

45
Haller, S. A., & Lyons, S. (2015). Broadband Adoption and Firm Productivity: Evidence from
Irish Manufacturing Firms. Telecommunications Policy, 39(1), pp. 1-13.

Holt, L., & Jamison, M. (2009). Broadband and Contributions to Economic Growth: Lessons
from the US Experience. Telecommunications Policy, 33(10), pp. 575-581.

ITU (2010). Definitions of World Telecommunication/ICT Indicators. March 2010, ITU,

Geneva. Retrieved from: https://www.itu.int/ITU-
D/ict/material/TelecomICT_Indicators_Definition_March2010_for_web.pdf (Last
accessed on 29 July 2016).).

Ivus, O., & Boland, M. (forthcoming). The Employment and Wage Impact of Broadband
Deployment in Canada. Canadian Journal of Economics. Retrieved from:
https://www.researchgate.net/publication/266739303_The_Employment_and_Wage_Impa
ct_of_Broadband_Deployment_in_Canada (Last accessed on 29 July 2016).

Jensen, R. (2007). The Digital Provide: Information (Technology), Market Performance, and
Welfare in the South Indian Fisheries Sector. Quarterly Journal of Economics, 122(3), pp.
879-924.

Kandilov, I. T., & Renkow, M. (2010). Infrastructure Investment and Rural Economic
Development: An Evaluation of USDA's Broadband Loan Program. Growth and Change,
41(2), pp. 165-191.

Kolko, J. (2012). Broadband and Local Growth. Journal of Urban Economics, 71(1), pp. 100-
113.

Koutroumpis, P. (2009). The Economic Impact of Broadband on Growth: A Simultaneous

Approach. Telecommunications Policy, 33(9), pp. 471-485.

Lam, P. L., & Shiu, A. (2010). Economic Growth, Telecommunications Development and
Productivity Growth of the Telecommunications Sector: Evidence Around the World.
Telecommunications Policy, 34(4), pp. 185-199.

Lee, S. H., Levendis, J., & Gutierrez, L. (2012). Telecommunications and Economic Growth:
An Empirical Analysis of Sub-Saharan Africa. Applied Economics, 44(4), pp. 461-469.

Mack, E., & Faggian, A. (2013). Productivity and Broadband The Human Factor.
International Regional Science Review, 36(3), pp. 392-423.

46
Mack, E. A., & Rey, S. J. (2014). An Econometric Approach for Evaluating the Linkages
between Broadband and Knowledge Intensive Firms. Telecommunications Policy, 38(1),
pp. 105-118.

Madden, G., & Savage, S. J. (1998). CEE Telecommunications Investment and Economic
Growth. Information Economics and Policy, 10(2), pp. 173-195.

Madden, G., & Savage, S. J. (2000). Telecommunications and Economic Growth.

International Journal of Social Economics, 27(7/8/9/10), pp. 893-906.

Mankiw, N. G., Romer, D., & Weil, D. N. (1992). A Contribution to the Empirics of
Economic Growth. Quarterly Journal of Economics, 107(2), pp. 407-437.

Muto, M., & Yamano, T. (2009). The Impact of Mobile Phone Coverage Expansion on
Market Participation: Panel Data Evidence from Uganda. World Development, 37(12), pp.
1887-1896.

OECD (2016), Households with Broadband Access (Indicator). Retrieved from:

https://data.oecd.org/broadband/households-with-broadband-access.htm (Last accessed on
29 July 2016).

Röller, L. H., & Waverman, L. (2001). Telecommunications Infrastructure and Economic

Development: A Simultaneous Approach. American Economic Review, 91(4), pp. 909-923.

Shiu, A., & Lam, P. L. (2008). Causal Relationship between Telecommunications and
Economic Growth in China and its Regions. Regional Studies, 42(5), pp. 705-718.

Sridhar, K. S., & Sridhar, V. (2007). Telecommunications Infrastructure and Economic

Growth: Evidence from Developing Countries. Applied Econometrics and International
Development, 7(2), pp. 37-56.

Thompson, H. G., & Garbacz, C. (2007). Mobile, fixed-line and Internet service effects on
global productive efficiency. Information Economics and Policy, 19(2), pp. 189-214.

Thompson, H. G., & Garbacz, C. (2011). Economic Impacts of Mobile Versus Fixed
Broadband. Telecommunications Policy, 35(11), pp. 999-1009.

Ward, M. R., & Zheng, S. (2016). Mobile Telecommunications Service and Economic
Growth: Evidence from China. Telecommunications Policy, 40(2-3), pp. 89-101.

Waverman, L., Meschi, M., & Fuss, M. (2005). The Impact of Telecoms on Economic
Growth in Developing Countries. Vodafone Policy Paper Series, 2, pp. 10-23. Retrieved
from: https://www.vodafone.com/content/dam/vodafone/about/public_policy/

47
 policy_papers/public_policy_series_2.pdf (Last accessed on 29 July 2016).

Whitacre, B., Gallardo, R., & Strover, S. (2014). Broadband’s Contribution to Economic
Growth in Rural Areas: Moving Towards a Causal Relationship. Telecommunications
Policy, 38(11), pp. 1011-1023.

Yilmaz, S., Haynes, K. E., & Dinc, M. (2002). Geographic and Network Neighbors: Spillover
Effects of Telecommunications Infrastructure. Journal of Regional Science, 42(2), pp. 339-
360.

48