Impact of Computing on the World Economy: A Position Paper

Frederick Harris Gordon Lee Stuart Rubin

T. C. Ting Billy Gaston Gongzhu Hu

Abstract can be seen with the inception and widespread use of the
As computing technologies continue to rapidly advance telephone, computers, databases and the Internet. The
since the last two decades, the knowledge economy has telephone provides a platform for local communication;
become an important part of the overall world economy. the computer provides computing power to the individual;
In addition to its significantly contribution to economy the database systems provide a local place to store and
growth, computing technologies also have profound impact organize data efficiently; and the Internet provides a wide-
on many aspects of society. A panel discussion was held scale platform for the aforementioned solutions where
at the 21st ISCA International Conference on Computer the best of each solution’s capability can be leveraged
Applications on Industry and Engineering (CAINE 2008) and used globally. The world has experienced enhanced
in Honolulu, November 12–14, 2008, where the panelists global communication, data organization and retrieval, and
and the audience exchanged views and ideas on this im- computing. These technologies have greatly affected the
portant issue. This position paper summarizes the panel economy of countries able to properly leverage them with-
discussion. out extreme restrictions. They have enabled individuals,
small groups, and small countries to have a voice by
providing just as much access, visibility and opportunities
as big businesses and advanced countries.
1 Technology and the New Economy
1.1 Computing and Economy Growth
Ubiquitous computing is here to stay and it has become
one of the main fibers of social, cultural, and economical Computing, as a general term for digital technologies,
life. It is an enabling technology that can increase the has greatly contributed to the economic growth in the last
productivity in a wide range of applications and econom- twenty years. According to an analysis of 110 economies
ical activities. The concept of “computing” has been in the world by Harvard economists Jorgenson and Vu [9],
continuously evolving. In early 80’s, it was the marriage the annualized growth of the world economy, measured in
of computing and communication technologies that cre- GDP, was 2.5% during the 1989-1995 period and 3.45%
ated the era of Internet-based information resources that for 1995-2003, a 38% increase. They included three
affect and penetrate into our daily activities in information main categories as the factors responsible for economic
accessing. Now, it is the polygamy of computing, com- growth: capital deployment, effective use of labor, and
munication, storage, sensory and displaying technologies total factor productivity (TFP). The contributions to the
that can impact almost all social, cultural, and economical overall economic growth are shown in Table 1.
development as well as our daily life. Action-oriented- It is seen in this table that the capital deployment
intelligent user/system interactive logic and system will be (including Information Technology and non-IT capital)
greatly improved for easy use as well as to augment our contributed almost 50%, while the othe two major factors,
intellectual capabilities. labor and productivity, contributed about 20-30% and 25%,
At the individual’s level, the impact of computing puts respectively. The contribution of IT capital deployment
opportunity into the homes and hands of many ordinary almost doubled from 0.27% to 0.53% during these periods,
people and assists in solving everyday problems, thus while the impact of non-IT capital and labor to economic
enhancing the lives of many on a global scale. This idea growth remained pretty much flat.

271
 Table 1. Contribution to the world economy growth. to the country’s economy future. The World Economic
Forum conducted an extensive Executive Opinion Survey
Source of Growth 2006–2007 that posted a questionnaire item “The govern-
Period GDP growth capital ment has a clear implementation plan for utilizing infor-
labor TFP
IT non-IT
mation and communication technologies for improving the
1989–95 2.50 0.27 0.91 0.89 0.53
country’s overall competitiveness (1 = strongly disagree, 7
1995–03 3.45 0.53 1.03 0.89 0.99
= strongly agree).” The survey statistics was included in
the Global Information Technology Report 2007-2008 [7],
from which we list the top ten countries with highest score
The growth in total factor productivity (TFP) is the
(among 127 countries/regions) in Table 2.
output growth of an economy not accounted for by the
growth in input, that was also almost doubled during Table 2. Importance of ICT to government vision for the
these two periods. It has been well documented [2, 4, future.
6] that such rapid growth in productivity is impacted
by the advancement of Information and Communication Rank Country/Region Score
Technologies (ICT) in a significant way. 1 Singapore 6.28
Undoubtedly, computing technologies will continue to 2 Portugal 5.73
make computers, communication links, and storage device 3 Malta 5.70
run faster and to be more available. However, the focus will 4 United Arab Emirates 5.65
be more on information and knowledge processing in wide
5 Malaysia 5.62
range of applications in almost all social and economical
6 Denmark 5.51
fields. The productivity and impact of computing should
7 Korea 5.50
be measured on the outcome of the increased economical
8 Qatar 5.46
productivities of all applications.
9 Estonia 5.40
10 Iceland 5.37
1.2 Network Readiness
The U.S. is ranked 28 and the BRIC nations (Brazil,
As the world economy becomes more knowledge- Russia, India and China) ranked 75, 103, 21, and 25,
based, connectivity has been recognized as a critical respectively.
factor not only to the economic growth but also a key Obstacles exist for the advances of ICT. Legal and
component of general public infrastructure. For example, regulatory changes will continue to lag behind that can
the Ten Principles for Digital Excellence suggested by the hinder the speed of development. Potential abuses and
Chicago Digital Access Alliance (CDAA) includes the security and privacy concerns may threaten and hinder
notion of “universal access to high-speed connectivity is the effective use of computing for social and economical
a public right and necessity” [1]. Countries around the development activities. So does the lack of focused ef-
world have made great efforts and investment to enhance forts, funding, and government cooperation. The research
the networking infrastructure. According to the Executive community, particularly universities and colleges, need to
Summary of the Global Information Technology Report make more efforts to apply their research results to real
2007-2008 [5] at the 2008 World Economy Forum, one world applications rather than letting the results sit on
of the benchmarks of a nation’s ability to foster economic shelves in libraries. Further research-oriented meetings
growth is unified communication, that is, the networking and conferences should make such events more visible to
infrastructure at the national level. The report shows a investors, industry and government officials.
close correlation between the network readiness index of a
country and its economic strength and competitiveness in 1.3 E-Skills
the world economy. Countries with high rankings in the
network readiness index are those developed countries in Networking is only a “hardware” measure of the
Europe, North America, and some Pacific regions. Several economy-enabling technology; an even more critical
countries with fast economic growth in recent years (such measure is a skilled labor force that is capable of using
as Brazil, China, India) also see their network readiness the technological infrastructure to strive for innovation
indexes rising as the governments of these countries and achieve higher productivity. ICT is now regarded as
increased capital investment in their communication a general purpose technology not only for ICT-producing
networks. industries but more importantly for the ICT-using sectors
A government’s vision of ICT is fundamentally critical [4].

272
 The term e-skills has been used to refer to the gen- 2.1 Current Educational Situation
eral knowledge and skill set related to computing and
information technology in this digital world. Today’s Many countries have put great emphases on education,
new knowledge-based economy requires a labor force particularly in fields with large gap between the availability
with e-skills to sustain the economic growth. In [11], of skilled workers and market demands, such as the STEM
Bruno Lanvin, the Director INSEAD e-Lab, emphasized fields – science, technology, engineering, and mathematics.
that “whether or not e-skills are available in sufficient There are quite a large disparities in the quality of individ-
numbers and quality will determine how countries succeed ual’s math and science education among countries. The
or fail as knowledge societies.” However, such availability education system in the U.S. is lacking more and more in
of e-skills is very much lacking today. The Global IT preparing students for the transition from high school to
Report 2007-2008 indicated that our economies fall short college and from college to industry.
of providing the necessary volume and levels of e-skills, According to the Global IT Report 2007-2008 [7], the
and the gap between the existing education systems and countries ranked in the top ten in math-science education
the requirements of the knowledge-intensive economy is in 2007-2008 are listed in Table 3. The scores in the
growing. The President of the European Commission, Mr. table were based on the World Economic Forum Executive
Jose-Manuel Barroso, recently warned [3], “There is some Opinion Survey 2006–2007 survey data (1 = lag far behind
urgency. Already today, millions of vacancies in Europe most other countries’ schools, 7 = are among the best in the
are unfilled because there are not enough people with the world).
right skills to fill them. For example, in one of Europe’s key
industry, advanced network technology – which includes Table 3. Quality of Math and Science Education.
mobile telecommunications, one in six vacancies cannot
be filled. Only three years ago, this was one in twelve Rank Country/Region Score
vacancies. This is a fast-growing sector for employment, 1 Singapore 6.34
and we cannot let this situation get worse.” 2 Belgium 6.29
3 Finland 6.17
Similar situation also exists in other countries such as
4 Hong Kong SAR 5.85
the United States. The Digital Skills Working Group
5 Switzerland 5.72
Institute for a Competitive Workforce at the U.S. Chamber
6 France 5.71
of Commerce categorizes digital skills in three levels:
7 Tunisia 5.62
literacy, fluency, and mastery. It claimed that “a growing
8 Taiwan, China 5.59
proportion of U.S. jobs require at least a basic level of
digital literacy, with many of the best jobs demanding 9 Czech Republic 5.53
increasing levels of digital fluency.” [8]. 10 Korea 5.46

Actions must be taken to to address this problem of lack The United States is ranked 43th. There has been a co-
of e-skilled workers. Education is the key. ordinated effort from government, academia, and industry
in the U.S. to improve and reform the education systems
to put greater emphasis on the STEM education at the
precollege levels to attract more young people into these
2 Computing and Education fields. Several U.S. government agencies, including the
Department of Education and the National Science Foun-
dation, have been actively promoting STEM education.
One of the great impacts of computing will be in
the areas of human learning and cognitive and logical 2.2 The Need for Educational Reform
inference activities that it can and will inevitably change
the way we learn, work and live. Life long learning from Educational reform is served by computer-based train-
birth at all levels will be the focus for long term and ing, which is vital to our national economic infrastructure
continuous economical development of our society. In in the medium-term on out. Success here is contingent
the Recommendation of the Europe Parliament and the upon success in computational creativity. It is understood
Council on key competences for lifelong learning [13], that many details and justifications are necessarily missing
“digital competence” and ”learning to learn” are among the from this overview for the sake of brevity.
eight key competences recommended for citizens to adapt The then-President-candidate Barack Obama said on
flexibly to a rapid changing and highly interconnected October 15, 2008 that no nation in history has ever main-
world. tained a strong military in the absence of fiscal solvency

273
[12]. One aspect of fiscal solvency is a relatively full em- that computing is changing the human society and
ployment, in keeping with the Phillips curve (i.e., allowing what once was noted as fundamental in one age
for 4 percent unemployment) [14]. But this is not the may be obsolete in another, which suggests that the
complete picture. We need to maintain and thus invest in domain of fundamental skills is in a gradual process
an ever-more educated workforce to remain economically of redefinition.
competitive in the global economy. How do we do this?
It used to be held that our schools served the role of 3.1 The Next Revolution
education. However in the current form of our education
systems, the schools cannot effectively reach students on Leverage computing power, communication, storage ca-
the current or foreseeable budget. What is needed is the in- pability and collaboration techniques provide some insight
troduction of affordable computer technologies to capture into the next revolution in computing. All of these areas are
good instructional practice to deliver quality education to driven by the immense need to access, analyze, process,
each and every individual where it is needed and when it and distribute information more intelligently. Informa-
is needed. We know that this can be done. We found from tion is the common denominator. It is sometimes said
our experience that intelligence is a phenomenon of scale. that information/knowledge is power; however, it should
That is, we can expect intelligent tutoring and training for be said that correctly applied information/knowledge is
the military and country overall will cost less per delivered power, assuming that the information being applied is of
instructional unit with scale – until a point is reached where the necessary quality. Over the past years, steps have
it becomes the most cost-effective method for delivery. been initiated by a number of government and commercial
entities to capture a plethora of data to address the issue of
providing a more informative, preventative and predictive
3 Challenges in Computing analysis to today’s fleet management effort. A large
number of such efforts consist of capturing more data.
Computing will continue to impact the economy in the Unfortunately, whether it is historical data stored in legacy
years to come. To sustain such impact, we will need systems, current data captured by mechanics, or future
to address some issues concerning the development of data acquired by the use of sensor technology, some of
computing technologies: the underlined issues still remain. These issues include
but are not limited to making sense of the captured data,
(1) It should be a technology to foster more collaboration acquiring the appropriate combinations of data, making the
with others versus one that removes human interaction proper relationships between data, as well as determining
and social development, unless human life is at stake. the relevance of said data. One of the biggest problems of
(2) Researchers must be careful to determine if efforts data collection is determining what exactly one can keep
should be focused on leveraging technology to think and/or disregard. This creates a problem in an unnecessary
for us, to interpret and understand for us, to instruct cost of storage, which can be high when processing a large
us, to predict/forecast the future for us. Some may amount of data. Discovering and defining relationships
deem that as creating a sort of god for mankind. between similar documents or sets of information has
(3) One must ensure that technology must not prevent become a rather common and needed task within all areas
users from knowing and experiencing fundamental of government, business and industry. Entities needing this
concepts and learning experiences. For example in capability range from potentially small scale systems, such
speaking with a tutor for grade school students, she as a small elementary school library book management sys-
noted that some students experienced difficulty in tems, to large scale systems, such as military government
trying to find the definition of a word by searching information systems servicing several international logistic
a dictionary. Apparently, the students had practiced depots. The right information is often needed to obtain
more on typing a word in a website form and after real time situational awareness, mission capability, damage
clicking ’Submit’, the definition of the work was assessment, and risk analysis. These issues affect all
provided. The skills necessary to traverse a dictionary types of business and government organizations and may
had not been grasped. prove beneficial in business decisions, medical discoveries,
(4) The technology should enable the enhancement of advances in information retrieval and organization, envi-
communication, computing, production, and thus eco- ronmental management, aircraft maintenance management
nomic opportunities. It is important to leverage and the like.
this enhancement byproduct without eliminating the Throughout the business world, whether commercial or
fundamental skills that should be learned by an indi- non-commercial, there exists the need for a solution that
vidual. It is noted that this is said cautiously, knowing can capture input from disparate data sources, analyze the

274
information, discover new information of relevance, link possible to better enable domain-specific computation.
related information, and from said information allow the Currently, the best unclassified computers operate at about
necessary personnel to make intelligible decisions. The 55,000 times the speed of a high-end PC. That is no small
term “necessary” is used because such a solution should achievement. In the short-term, we should be able to
possess the capability to adapt the distributed information up that factor to 5.55 × 106 using high-temp Josephson
to each end-user based on ability, role, clearance, learning superconducting circuits. In the medium term, 5.55 × 109
capacity, and level of need. This involves some knowledge seems plausible using optronic hybrids, and in the long-
and interaction with the user. This technology should term, 5.55 × 1015 seems plausible – additionally based on
enable intelligent information access and interaction and a distributed computational paradigm. This is fast enough
not information overload. In specific it should provide the to outperform all of the combined mental facilities of the
necessary understanding to make optimal decisions and not planet, given appropriate software. Such systems hold the
hinder decision making. Furthermore with such an increase promise of medical advances, engineering advances, and
in processing and computation, this computing technology the like; but, one capability stands out among the rest – that
will benefit from the current concentrated effort to develop is, the capability to revolutionize education and training.
clean, green energy alternatives.
3.3 Object-Oriented Intelligence
3.2 Josephson/Optronic Hybrid Computing
The vast majority of intelligent tutors and trainers these
One thing is clear – we need to innovate, or be rendered days have good graphics and multimedia support. They are
obsolete. Computing currently supports a significant per- lacking in intelligence, however. It is our opinion that the
centage of the GDP growth as stated in Section 1.1; but best architecture for delivering intelligence in tutoring and
it can contribute more if we commit to the development training applications is to arrange all multimedia materials
of faster, greener, more powerful devices. These devices as small objects that are maintained in an object-oriented
can be small like PCs, but the greatest growth in power relational database and brought up by a simple rule-based
will come from the development of large computer utilities production system. The input to this expert system can be
running hybrid optical and Josephson-Junction based cir- quiz questions (even connected to simulation programs),
cuits [10]. Josephson Junction devices already run at 50 to meant to elicit student feedback. The pattern of responses
100 times the speed of silicon circuits due to their much serves as the context to the intelligent system. While such
faster switching times, lower power, less heat production, a system can be prototyped and will certainly work, there is
less cross-talk, and far less parasitic capacitance. High- a problem. If n is the number of objects in the system, then
temperature supercomputing promises the development of there can be n! paths through the training system and thus
Josephson chips that need run no cooler than the tem- it becomes impractical for an expert to precognate most
perature of liquid nitrogen and can be switched/shunted response patterns.
with far less current. Another implication is less thermal
stress at the relatively warmer temperatures. Also, liquid 3.4 The Creative KASER
nitrogen is far easier to produce than liquid helium – the
current gold standard for such devices. Moreover, recent At this point, we need to introduce the KASER fam-
advances in alloys promise the development of thermopiles ily of intelligent systems [15]. They have theoretically
that are closer to the theoretical limit for electronic heat- and empirically applied computational analogy to map
pumping. They can be used for the last stage of cooling necessarily limited supplied knowledge to an appropriate
because unlike silicon circuits, Josephson devices consume response – along with a squelchable possibility of error.
virtually no power and produce virtually no heat to be We know that the cost of knowledge acquisition is the
pumped out as a consequence. limiting factor in constructing such systems. KASERs
Optronics allows for a speedup by a factor of one can, for example, allow for the development of a million
thousand to one million times where inexact computation dollar expert system for about a thousand dollars (i.e., the
is permissible (e.g., ensembles of neural networks). Much square root concept). They are also amenable to massive
of AI works best using heuristic, fuzzy processing. For distributed heterogeneous and parallel processing and they
example, the Traveling Sales Problem can be solved on can learn from a trusted user. All this comes together to
a PC in about a second if we relax the requirement for allow for the definition of computational architectures that
optimality to within one percent for about 200 cities. allow for the delivery of educational content in a reusable,
Conversely, optimality would take about one year on a augmentable, cost-effective manner that offers to make
supercomputer – and that is just for 200 cities. Clearly, text-based instruction all but obsolete. We can show that
one must take advantage of heuristic processing where there will invariably be a net return on the educational

275
dollars invested from at least a national macroeconomic Table 4. Key actions to enhance e-skills readiness.
standpoint. Then too there will be desirable spin-off
benefits, which include: the development of KASERs, Key action Expected leader
the development of faster/greener computational systems, Share a compelling e-vision government
creative design and engineering, medical discoveries, sci-
government (visibility)
entific discoveries, educational superstars (i.e., excellence Strengthen e-skills
industry (inclusion, action)
in teaching is an under-compensated skill set), and the like
industry (needs)
– all of which may pale in comparison to the benefits atten- Formulate e-curricula academia (formulation)
dant to the transformation of the educational infrastructure government (regulatory)
in the United States and beyond. academia (curricula)
We are lucky – lucky because the project outlined above Promote math and science government (vision)
need not begin with a giant step; rather, it can be phased in industry (sponsoring)
and guided by project milestones. We have demonstrated industry (content, rewards)
and can further validate each of the claims made herein. Enhance life-long learning government (fiscal)
Capabilities attendant with scale can be limited to smaller academia (e-learning)
studies for empirical validation. The goal is to make the
student an active, rather than passive learner. This idea has
not loomed on the horizon as a cost-effective possibility what the research community in computing should do in
prior to our R&D on the KASER family of engines. the next decade.
The NSF has supported small studies on intelligent
(1) Develop more focused efforts.
tutoring systems, but they essentially used COTS (Com-
mercial off-the-shelf) – not KASERs. Most importantly, (2) Specifically market government and commercial busi-
the early progenitors did not think in the large – presum- nesses to come to conferences (e.g. discount registra-
ably because they could see no benefits of scale here and tion, create incentives for teaming with universities,
conventional expert systems invariably experience perfor- program committee should consist of industry part-
mance degradation with scale; whereas, KASERs perform ners).
better and better with scale. An intelligent system that (3) Team with small businesses. Small businesses are
cannot utilize all of the processor power made available positioned to create cost effective solutions, more
to it is fundamentally flawed. Again, Josephson/Optronic innovation, more enhanced/focused solutions. There
hybrid computing offers the promise of extreme processor is funding specifically set aside for small business and
power. The program, if stood up, needs a political bridge university researcher teaming.
from the academic to the superintendent and that needs to (4) Lobby to prevent government intervention without
be understood from the inception. PERT charts may be understanding.
employed in this regard. You get low cost, low risk, and (5) Much more research in security and confidentiality
virtually unlimited military (training) and societal benefits areas need to be emphasized.
with success. (6) Standards and compatibility for the massively con-
nected computational and information processing sys-
tems need more research and policy attentions for the
4 New Directions in Computing information oriented and dependent society.
(7) Research in intelligent information processing need to
In the competitive world, a strong cooperation between
focus on holistically and symbiotically for both static
government, academia, and industry play a vital role crit-
and dynamic data and knowledge.
ical to the health and prosperity of our economies and
to improve the quality of life of the citizens. Bruno (8) Software efficiency and reliability including systems
Lanvin suggested five key actions to take in work force as well as application software systems need to be
enhancement [11], as shown in Table 4. emphasized.
Although the five key actions above were targeted to (9) Intelligent and logical user/system interaction, includ-
the European Union, we believe they are also universally ing multi-media and sensory interfaces needs to be
applicable to other countries including the U.S. As the greatly improved and developed.
researchers and practitioners in the field of computing, we (10) Representatives from computing research and de-
are an integral part of the national and international coor- velopment communities need to be engaged more
dinated effort for the advancement of computing to sustain actively participating in policy, legal, and regulatory
the economy growth. In particular, the following highlight processes.

276
(11) Most importantly, we need to actively engage in [5] Soumitra Dutta. Executive summary: the
the development learning strategies and activities in global information technology report 2007-2008.
schools as well as in society for preparing individ- http://www.weforum.org/pdf/gitr/2008/
uals for effective use of computing and information Summary.pdf, 2008.
systems for learning and working in our current and
future information oriented society. [6] John Fernald and Shanthi Ramnath. The acceleration
in U.S. total productivity after 1995: the role of
information technology. Economic Review, Q1:1–15,
2004.
5 Summary
[7] World Economy Forum. The global
At The Impact of Computing on the World Economy information technology report 2007-2008.
panel at the CAINE 2008 Conference, the participants http://www.weforum.org/en/initiatives/
expressed and exchanged their views on this important gcp/Global%20Information%20Technology%
topic and pointed out that such discussions need to continue 20Report/index.htm, 2008.
and actions must be taken by all sectors of our societies,
particularly on education and new computing technologies. [8] Digital Skills Working Group. http:
This position paper summarized their discussions. //www.uschamber.com/icw/strategies/
On the education front, we as researchers and educators digitalskills.htm.
need to be creative. Rote instructional presentation does [9] Dale W. Jorgenson and Khuong Vu. Information
not work any better than does fixed micro-managed lesson technology and the world economy. In Proceedings.
plans. Then, to cost-effectively scale human creative Federal Reserve Bank of San Francisco, 2005.
instruction, you need to develop computational creativity
to the same level. Thus, software systems like KASER can [10] Brian David Josephson. The discovery of tun-
be properly viewed as enabling the automation of the true nelling supercurrents. Reviews of Modern Physics,
instructional process – all of which can snowball into an 42(2):251–254, 1974.
improvement in American Society and concomitantly, the
ever-present American Spirit. This spirit is perhaps best [11] Bruno Lanvin. E-skills, competitiveness and
captured by the Walt Disney’s phrase on the wall of the GE employability: knowledge societies next frontier. In
Carousel of Progress exhibit at the 1964 New York World’s European E-skills 2008 Conference, 2008.
Fair, “If you can dream it, you can do it.” [12] Barack Obama. Remarks in Londonderry,
New Hampshire: October 15, 2008. http:
//www.presidency.ucsb.edu/2008_election_
References speeches.php?candidate=44.

[1] Chicago Digital Access Alliance. Ten [13] European Parliament and the Council. Recommen-
principles for digital excellence. http: dation of the Europe Parliament and the Council of
//www.digitalaccessalliance.org/ 18 December 2006 on key competences for lifelong
principles-for-digital-excellence, 2007. learning. Offical Journal of the European Union, L
394:10–18, December 2006.
[2] Robert D. Atkinson and Andrew S. McKay. Digital
prosperity: understanding the economic benefits [14] Alban William Phillips. The relationship between un-
of the information technology revolution. Special employment and the rate of change of money wages
Report of Information Technology and Innovation in the United Kingdom 1861-1957. Economica,
Foundation, March 2007. 25(100):283–299, 1958.
[15] Stuart H. Rubin, S. N. Jayaram Murthy, Michael H.
[3] José Manuel Barroso. Lisbon: a strategy for all
Smith, and Ljiljana Trajković. KASER: knowledge
seasons. speech at Lisbon Council Growth and Jobs
amplification by structured expert randomization.
Summit, Brussels, March 2008.
IEEE Transactions on System, Man, and Cybernetics
[4] Susanto Basu and John G. Fernald. Information — Part B: Cybernetics, 34(6):2317–2329, 2004.
and communications technology as a general purpose
technology: evidence from U.S. industry data.
Economic Perspectives, pages 52–67, 2008.

277