Technical ReportPDF Available

How Much Information? 2009 Report on American Consumers



Content may be subject to copyright.
How Much Information?
How Much Information? 2009 Report on American Consumers
How Much Information? 2009
Report on American Consumers
Roger E. Bohn
James E. Short
Global Information Industry Center
University of California, San Diego
Date of Publication: December 2009
Last Update: January 2010
How Much Information? 2009 Report on American Consumers
How Much Information? 2009
Report on American Consumers
1 INTRODUCTION.........................................................................................................................................8
1.1 Data anD InformatIon ............................................................................................................................10
1.2 What Is InformatIon? .............................................................................................................................10
1.3 hoW many hours? .................................................................................................................................11
1.4 hoW many WorDs? ................................................................................................................................12
1.5 hoW many Bytes? ..................................................................................................................................13
1.6 storage vs. ConsumptIon ......................................................................................................................14
1.7 valuIng InformatIon ..............................................................................................................................14
2 TRADITIONAL INFORMATION IN U.S. HOUSEHOLDS .................................................................15
2.1 televIsIon ...............................................................................................................................................15
2.2 raDIo .......................................................................................................................................................17
2.3 telephone ...............................................................................................................................................17
2.4 prInt........................................................................................................................................................18
2.5 movIes ....................................................................................................................................................18
2.6 reCorDeD musIC .....................................................................................................................................18
3 COMPUTER INFORMATION IN U.S. HOUSEHOLDS ......................................................................19
3.1 CommunICatIng anD BroWsIng the Internet ..........................................................................................19
3.2 Internet vIDeo ........................................................................................................................................20
3.3 Computer gamIng ..................................................................................................................................21
3.4 off-Internet home Computer use ........................................................................................................22
3.5 smart phones .........................................................................................................................................22
4.1 analyzIng the groWth of InformatIon ..................................................................................................24
4.2 Where are the mIssIng Bytes? ..............................................................................................................25
4.2.1 Dark Data ...........................................................................................................................................25
4.2.2 tWo kInDs of QualIty: varIety anD resolutIon ...............................................................................26
4.3 analyzIng InformatIon ConsumptIon .....................................................................................................26
4.3.1 hoW muCh InformatIon Is DelIvereD vIa the Internet? ....................................................................27
4.3.2 the rIse of InteraCtIon ......................................................................................................................28
4.4 the future of Consumer InformatIon ...................................................................................................28
APPENDIX A: UC BERKELEY HMI? STUDIES ....................................................................................30
APPENDIX B: DETAIL TABLE ..................................................................................................................32
How Much Information? 2009 Report on American Consumers
Tables and Figures
Figure 1 Information Flows in a Home ......................................................................................... 10
Figure 2 INFO
Hourly Information Consumption ..................................................................... 11
Figure 3 Formula for Size Calculations ........................................................................................ 12
Figure 4 INFO
Consumption in Words ...................................................................................... 12
Figure 5 INFO
Consumption in Compressed Bytes ................................................................... 13
Figure 6 Evolution of Reading ...................................................................................................... 18
Figure 7 Average Daily Consumption of Bytes, INFO
.............................................................. 20
Figure 8 Example of a Graphics Processing Card ........................................................................ 21
Figure 9 Screen Shot from NBA 2K10 Game ............................................................................... 23
Figure 10 Shares of Information in Different Formats ................................................................. 26
Figure 11 Contrasting Measurements of INFO
and INFO
......................................... 26
Figure 12 Internet as a Source of Information .............................................................................. 27
Table 1 Three Measures of Information .......................................................................................... 9
Table 2 Partial Breakdown of Delivery Methods Analyzed ........................................................ 11
Table 3 Television and Radio Consumption ................................................................................. 16
Table 4 Telephone Consumption ................................................................................................... 17
Table 5 Conventional Media ......................................................................................................... 19
Table 6 Computer Use Non-Gaming ............................................................................................ 19
Table 7 Computer Game Playing .................................................................................................. 21
Table 8 Explaining the Gap Between Consumption and Capacity Growth .................................. 25
Table 9 Summary of Information for Major Groups .................................................................... 27
How Much Information? 2009 Report on American Consumers
This report is the product of industry and university collaboration. We are grateful for the support of our
industry partners, sponsor liaisons, university research partners, and administrative staff at the University
of California, San Diego. Special thanks for research and writing assistance provided by L. Lin Ong and
Doug Ramsey. Early support was provided by the Alfred P. Sloan Foundation of New York.
Financial support for HMI? research and the Global Information Industry Center is gratefully
acknowledged. Our sponsors are:
Cisco Systems
Intel Corporation
Seagate Technology
The authors bear sole responsiblity for the contents and conclusions of the report.
Questions about this research may be addressed to the Global Information Industry Center at the School of
International Relations and Pacic Studies, UC San Diego:
Roger Bohn, Director,
Jim Short, Research Director,
Pepper Lane, Program Coordinator,
Special thanks to Pepper Lane for major editing and all-around support.
Press Inquiries:
Doug Ramsey, Communications Director,, (858) 822-5825
How Much Information? 2009 Report on American Consumers
How Much Information? 2009
Report on American Consumers
Roger E. Bohn
James E. Short
Executive Summary
In 2008, Americans consumed information for about 1.3 trillion hours, an average of almost 12 hours per
day. Consumption totaled 3.6 zettabytes and 10,845 trillion words, corresponding to 100,500 words and
34 gigabytes for an average person on an average day. A zettabyte is 10 to the 21st power bytes, a million
million gigabytes. These estimates are from an analysis of more than 20 different sources of information,
from very old (newspapers and books) to very new (portable computer games, satellite radio, and Internet
video). Information at work is not included.
We dened “information” as ows of data delivered to people and we measured the bytes, words, and
hours of consumer information. Video sources (moving pictures) dominate bytes of information, with 1.3
zettabytes from television and approximately 2 zettabytes of computer games. If hours or words are used
as the measurement, information sources are more widely distributed, with substantial amounts from radio,
Internet browsing, and others. All of our results are estimates.
Previous studies of information have reported much lower quantities. Two previous How Much
Information? studies, by Peter Lyman and Hal Varian in 2000 and 2003, analyzed the quantity of original
content created, rather than what was consumed. A more recent study measured consumption, but estimated
that only .3 zettabytes were consumed worldwide in 2007.
Hours of information consumption grew at 2.6 percent per year from 1980 to 2008, due to a combination
of population growth and increasing hours per capita, from 7.4 to 11.8. More surprising is that information
consumption in bytes increased at only 5.4 percent per year. Yet the capacity to process data has been
driven by Moore’s Law, rising at least 30 percent per year. One reason for the slow growth in bytes is
that color TV changed little over that period. High-denition TV is increasing the number of bytes in TV
programs, but slowly.
The traditional media of radio and TV still dominate our consumption per day, with a total of 60 percent
of the hours. In total, more than three-quarters of U.S. households’ information time is spent with non-
computer sources.
Despite this, computers have had major effects on some aspects of information consumption. In the past,
information consumption was overwhelmingly passive, with telephone being the only interactive medium.
Thanks to computers, a full third of words and more than half of bytes are now received interactively.
Reading, which was in decline due to the growth of television, tripled from 1980 to 2008, because it is the
overwhelmingly preferred way to receive words on the Internet.
How Much Information? 2009 Report on American Consumers
The world is awash in information and data, the
‘raw material of information. The goal of the How
Much Information? Project is to create a census
of the world’s data and information in 2008. How
much did people consume, of what types, and
where did it go?
This rst report conveys our ndings about
information at the U.S. consumer level. In other
words, how much information was consumed
by individuals in the United States in 2008?
Our statistics include information consumed in the
home as well as outside the home for non-work-
related reasons, including going to the movies,
listening to the radio in the car, or talking on a cell
phone. It does not include information consumed
by individuals in the workplace. Future reports
will focus on information in companies and on a
global scale.
We have reached a variety of conclusions about the
uses of information in the digital age, especially in
the nearly 30 years since IBM launched its rst PC
in 1981 (which went on to become Time magazine’s
“Man of the Year”). A few highlights:
• Americans spend a huge amount of time at
home receiving information, an average of
11.8 hours per day.
• Bytes of information consumed by U.S.
individuals have grown at 5.4 percent
annually since 1980, far less than the growth
rate of computer and information technology
• Roughly 3.6 zettabytes (or 3,600 exabytes)
of information were consumed in American
homes in 2008. Americans spend 41 percent
of our information time watching television,
but TV accounts for less than 35 percent of
information bytes consumed.
• Computer and video games account for 55
percent of all information bytes consumed
in the home, because modern game consoles
and PCs create huge streams of graphics.
Our estimate of 3.6 zettabytes for U.S. household
information consumption is many times greater than
the ndings of previous studies. One zettabyte is
bytes, or 1,000 exabytes, while one exabyte is
bytes – a billion gigabytes (see inset). A 2007
IDC report estimated that total worldwide digital
data would not reach one zettabtye until 2010. (One
major factor accounting for this discrepancy is that
IDC probably did not include video gaming, or
most TV, in its calculations.)
Counting Very Large Numbers
Byte (B) = 1 byte = 1 = One character of text
Kilobyte (KB) = 10
bytes = 1,000 = One page of text
Megabyte (MB) = 10
bytes = 1,000,000 = One small photo
Gigabyte (GB) = 10
bytes = 1,000,000,000 =
One hour of High-Denition video, recorded on a
digital video camera at its highest quality setting, is
approximately 7 Gigabytes
Terabyte (TB) = 10
bytes = 1,000,000,000,000 = The largest current hard drive
Petabyte (PB) = 10
bytes = 1,000,000,000,000,000 =
AT&T currently carries about 18.7 Petabytes of data
trafc on an average business day
Exabyte (EB) = 10
bytes = 1,000,000,000,000,000,000 =
Approximately all of the hard drives in home computers
in Minnesota, which has a population of 5.1M
Zettabyte (ZB) = 10
bytes = 1,000,000,000,000,000,000,000
How Much Information? 2009 Report on American Consumers
The rate of growth
of information bytes
consumed, 5.4 percent
yearly on average, is
low in light of the more
familiar, exponential
growth linked to
Moore’s Law: the rate
of improvements in
computer processing
power, memory, storage,
and other digital
technologies. But over 28
years this growth adds up,
constituting a four-fold
increase in bytes and a 140 percent increase in words
“consumed” by Americans from 1980 to 2008.
Our results are based on our own denitions of
information and how to measure it. Appendix
A provides some comparisons with a previous
generation of studies, conducted by Professors Peter
Lyman and Hal Varian at UC Berkeley, published
in 2000 and 2003. We also compare some of our
numbers with two industry reports produced by
EMC and the International Data Corporation (IDC)
in 2007 and 2008. These studies asked different
questions, used different denitions, and got different
results. Comparisons with information ows in 1980
largely draw on the groundbreaking work of Ithiel
de Sola Pool, who used words as the metric for his
observations. For example, he quite literally counted
how many words were uttered in radio programs.
Pool did not analyze bytes, so we have reanalyzed
some of Pools data to make them more directly
comparable to those we use in 2008.
We have looked at words, bytes, and the number
of hours spent consuming information in the
household. These three measures show very
different pictures about the volume of information
in any given medium. (Table 1 Three Measures of
Information) Take radio: Americans spent nearly
19 percent of their information hours listening to
the radio, which accounts for 10.6 percent of words
received – but barely one-quarter of one percent
(0.3%) of the total bytes received each day. This
points to radio’s continuing role as a highly byte-
efcient delivery mechanism for information.
While this report looks at all three measures of
information consumption – hours, words and bytes
– probably the most attention will be paid to bytes,
given the prevalence of digital media. Measuring
bytes is bound to be controversial, because it
appears to emphasize types of information that
stream at very high rates (such as computer games)
yet account for only a fraction of the words or hours
we spend consuming information each day. As we
will show, moving pictures account for the vast
majority of bytes. Therefore, we report the other
measures as well.
The current report focuses on the U.S. household
sector, while subsequent HMI? reports will expand
the focus to a) the workplace, b) other regions, and
c) new types of data and information that have no
historical antecedent – notably the ‘dark data’ that is
increasingly transmitted from machine to machine.
This report is divided into ve sections:
Section 1 introduces our concepts and
measurement methods.
Section 2 looks at information consumed by
U.S. consumers from traditional sources of
Section 3 considers digital information and the
computer revolution.
Section 4 examines results, discusses some
interesting special topics, and outlines future
Appendices cover additional topics.
Table 1: Three Measures of Information
What is measured Variable name US 2008 Consumption
Hours of consumption INFO
1,273 billion hours
Words consumed INFO
10,845 trillion words
Compressed Bytes consumed INFO
3,645 exabytes
What does this report cover?
This study reports on consumers in the United States in 2008.
Where 2008 data was not available, we extrapolated from earlier
years. In some cases, behavior is changing so fast that January
2008 and December 2008 might be quite different; our goal was
to report usage for the entire year but we were not always able to
make an accurate adjustment for this situation. All of our data are
estimates; see endnotes for sources.
How Much Information? 2009 Report on American Consumers
1.1 Data and Information
We distinguish between data and information.
Information is a subset of data – but what is data?
For our purposes, we dene data as articial signals
intended to convey meaning. ‘Articial,’ because
data is created by machines, such as microphones,
cameras, environmental sensors, barcode readers, or
computer keyboards. Streams of data from sensors
are extensively transformed by a series of machines,
such as cable routers (location change), storage
devices (time shift), and computers (symbol and
meaning change). These transformations, in turn,
create new data.
Past high-level studies have generally measured
data of only two kinds: data that gets stored, and
data that is transmitted over long distances, such as
over the Internet backbone. We greatly expand on
these two categories. For example, we include data
that is transmitted over a local area network (LAN),
such as a home Wi-Fi (802.11) wireless network,
and data that is never stored in a permanent way.
Indeed, data in the 21st century is largely ephemeral,
because it is so easily produced: a machine creates
it, uses it for a few seconds and overwrites it as
new data arrives. Some data is never examined at
all, such as scientic experiments that collect so
much raw data that scientists never look at most
of it. Only a fraction ever gets stored on a medium
such as a hard drive, tape or sheet of paper. Yet
even ephemeral data often has ‘descendants’—
new data based on the old. Think of data as oil and
information as gasoline: a tanker of crude oil is not
useful until it arrives, its cargo unloaded and rened
into gasoline that is distributed to service stations.
Data is not information until it becomes available to
potential consumers of that information. On the other
hand, data, like crude oil, contains potential value.
1.2 What Is Information?
There are probably hundreds of denitions of
information, and even the way we use the term in
daily conversation changes depending on the topic.
For looking at consumers, we choose to dene
information as data that is delivered for use by a
person. Our measures of information include all
data delivered directly to people at home, whether
for personal consumption (such as entertainment),
for communication (e.g., email) or for any other
reason. Some data delivered to machines could also
be considered information, but only if it is factored
into a decision or action. We will not analyze it in
this report. Figure 1 Information Flows in a Home
shows some of the data ows around a home. The
data displayed directly to consumers, shown by the
wide arrow, is the information we are interested in.
As we will show, there are a wide variety of types
of information consumed daily, such as:
• Text in readable form such as on a printed
page or cell phone display;
• Moving pictures on a TV, in a movie theater
or on a computer screen;
• An MP3 audio track received through
earphones or speakers;
• An electronic spreadsheet
For the purposes of this study, information is
‘useful’ in itself, while data is only a means to
ultimately produce information. In many situations,
a lot of data is created and then ltered and
manipulated to produce a relatively small amount
of information. For example, the “signal strength”
bar on your cellphone is the result of continuously
monitoring radio signals from a cellphone tower. A
30 second TV commercial is the result of shooting,
converting, and editing hours of raw footage. Data
Flows versus stocks
of information
Our denition emphasizes ows of data – data in motion. We
count every ow that is delivered to a person as information.
Another approach goes to the opposite extreme: it counts data
that is stored somewhere, such as a book, whether or not it is
subsequently used.
Figure 1: Information Flows In A Home
Web pages, blogs etc.
Internet Communication
Computer Games
Digital Video Recorder
DVD player
Physical media library
Home PC
External hard drive
Cable TV
Broadcast TV
Broadcast Radio
Telephone Line
Internet eg email
Printed Media
Digital Storage eg DVD
Photos, videos
eg email
Phone Calls
(video + sound
Made Here
How Much Information? 2009 Report on American Consumers
can also be expanded, as when that TV commercial
is sent to millions of TVs simultaneously.
1.3 How Many Hours?
In focusing on U.S. household consumption of
information, a natural question is how much
time Americans spend with different sources
of information. Our time statistics for U.S.
households in 2008 – including use of mobile
phones and movie-going – are tabulated in Figure
Hourly Information Consumption. We
estimate that an average American on an average
day receives 11.8 hours of information a day.
Considering that on average we work for almost
three hours a day and sleep for seven, this means
that three-quarters of our waking time in the home
is receiving information, much of it electronic.
is, indeed, the “Information Age.” We dene “hours
spent receiving information” as INFO
, one of our
three measures of information.
Our calculations for measuring information used
by consumers start by breaking information down
into about 20 categories of delivery media. (Table
2 Partial Breakdown of Delivery Methods
Analyzed) For each medium, we estimate the
number of people who use it, and the average
number of hours per user each year. The data on
numbers and hours comes from various sources,
including the US Census and other government
sources, Nielsen and other industry sources, and a
variety of studies on special topics. These sources,
in turn, used a variety of surveys and observation/
Our hourly statistics conrm that a
large chunk of the average American’s day is
spent watching television. We estimate that on
average 41 percent of information time is watching
TV (including DVDs, recorded TV and real-
time watching). An additional 19 percent of our
information time involves listening to the radio –
even though this activity is increasingly relegated
Simultaneous information
We do not adjust for double counting in our analysis. If someone
is watching TV and using the computer at the same time, our
data sources will record this as two hours of total information.
This is consistent with most other researchers. Note, though,
that this means there are theoretically more than 24 hours in an
information day!
The use of multiple simultaneous sources of information is
analyzed extensively in Middletown Media Studies: Media
Multitasking ... and how much people really use the media by
Robert A. Papper, Michael E. Holmes, and Mark N. Popovich.
Figure 2: INFO
Hourly Information Consumption
Recorded Music
Computer Games
All TV
Hours Per Day
Recorded Music
Computer Games
All TV
Table 2: Partial Breakdown of
Delivery Methods Analyzed
Cable TV – SD (Standard Denition)
Over air TV - SD
Cable TV – HD (High Denition)
Over air TV - HD
Satellite - HD
Satellite - SD
Mobile TV
Other TV (Delayed View)
Internet video
Print Media
Satellite Radio
AM Radio
FM Radio
Fixed Line Voice
Cellular Voice
High-end Computer gaming
Computer gaming
Console gaming
Handheld gaming
Internet including email
Ofine programs
Movies in theaters
Recorded Music
How Much Information? 2009 Report on American Consumers
to our daily commute. In other words, traditional
media still dominated U.S. households in 2008
based on how much time we spent consuming
information: more than seven hours watching
TV and listening to the radio, for more than 60
percent of total information hours. By comparison,
computers accounted for 24 percent of INFO
time (including browsing the Internet, playing
computer games, texting, watching videos on the
PC, and so on). So more than three-quarters of
U.S. households’ information time is spent with
non-computer sources – despite the widespread
belief that the seemingly ubiquitous computer now
dominates modern life. (Figure 2 INFO
Information Consumption)
Of course, our hypothetical “average American on
an average day” is a composite of many different
people. For example, although adults frequently
complain about how much time children spend
watching TV, the facts show otherwise: American
teenagers watch less than four hours per day while
the largest amount is watched by older Americans,
those 60 to 65, who watched more than seven hours
per day.
How do we compare with Americans in the past?
Not surprisingly, INFO
has gone up. The per
capita time spent consuming information has risen
nearly 60 percent from 1980 levels – from 7.4
hours per day in 1960, to 11.8 in 2008. The forms
of information media have also changed. When de
Sola Pool did his analysis in 1980, he included a
variety of media that either don’t exist any more or
are very small for consumers today. They included
Direct mail, First-class mail, Telex, Telegrams,
Mailgrams, and Fax.
1.4 How Many Words?
In 1960, digital sources of information were
non-existent. Broadcast television was analog,
electronic technology used vacuum tubes rather
than microchips, computers barely existed and were
mainly used by the government and a few very large
companies, music recording used vinyl disks called
“records,” and newspapers and magazines had
black and white pictures, if they had any at all.
concept that we now know as bytes barely existed.
Early efforts to size up the information economy
therefore used words as the best barometer for
understanding consumption of information.
Using words as his only metric, Pool estimated that
4,500 trillion words were consumed in 1980.
We calculate that words consumed grew to 10,845
trillion words in 2008, which works out to about
100,000 words per American per day. This measure
of information, words consumed, is our second
metric, which we label INFO
. We calculate it
by multiplying the amount of consumption time
, by the rate of information consumed
per unit of time. (Figure 3 Formula for Size
Calculations) To get total consumption, we
sum over the various media. All our numbers
are estimates - see the on-line appendix and the
endnotes for more information about data sources
and methods. <
Figure 3: Formula For Size Calculations
Total information for a year from technology Z, population segment M
= Average daily hours of Z use per person in segment M
x Total number of people in M who use Z
x 365 days per year
x 3600 seconds per hour
x Information per second for Z (bytes or words)
Total for technology Z = Sum over all population segments M
Comparing the 2008 statistics by type of media, TV
remains the single largest source of information –
over 45 percent of all words consumed. (Figure 4
Consumption in Words) In many
categories, the percentage distribution of INFO
and INFO
is similar, such as with computers (24%
, 27% of INFO
). A bigger difference
between words and hours is for radio: in 2008 radio
accounted for about 10.6 percent of our daily
information intake in words, even though we spent
nearly 19 percent of our information time listening
Figure 4: INFO
Consumption in Words
Computer Games
Recorded Music
All TV
Percentage of Words
Recorded Music
Computer Games
All TV
How Much Information? 2009 Report on American Consumers
to the radio. The reason is simple: a lot of radio
programs are mostly music, with comparatively few
words per minute.
1.5 How Many Bytes?
While the statistics based on hours and words are
useful, especially when trying to draw conclusions
about long-term trends, we now live in a digital age
when most of the information we consume comes
in the form of 0s and 1s, of bits and bytes. Music
is consumed via MP3 devices, ‘newspapers’ can be
read online, and virtually all electronic devices are
now based on digital integrated circuits.
So it stands to reason that in the digital age, an
appropriate way to measure information is by the
number of bytes consumed. We call this measure
, where the C stands for “Compressed bytes.”
Much of our research has gone into estimating
. Our formula for measuring bytes of
information starts from INFO
, the measure of
hours. For each media type, such as high denition
TV, we estimated the rate at which information is
delivered, called the “bandwidth,” traditionally
measured in bits per second. Multiplying the
bandwidth by the number of hours, and adjusting
for the conversion between seconds and hours and
between bits and bytes, gives the number of bytes
for that category.
Determining the correct bandwidth to use, however,
is quite literally “tricky.” The reason is that
computer and communications engineers use a
variety of tricks to transmit information as rapidly
and economically as possible. The denition we
use for INFO
is the rate at which compressed
information is transmitted over the link between
the originator and the consumer. This rate is
sometimes only one percent of the uncompressed
rate, as we will discuss in the section on television.
But not all information is actually “transmitted”
in the usual meaning of the term. For example
newspapers and movies are, for the most part, still
delivered physically on analog media (paper and
lm, respectively). In these cases, we developed
measures of bandwidth “as if” the information were
transmitted over a digital link.
Whatever the precise denitions used for measuring
, one fact stands out: when measured by
bytes, moving pictures dominate all other types of
consumer information. Even photographs are tiny
by comparison with most video. A high-resolution
digital picture might be 10 megabytes, but this is
equivalent to only 20 seconds of a standard TV
This led to a big surprise: only three activities
contribute a signicant amount of information based
: television, computer games, and movies
in theaters. Everything else adds up to less than
one percent! (Figure 5 INFO
Consumption in
Compressed Bytes)
In total, we estimate that an average American
consumed about 34 gigabytes (3.4 x10
) bytes
per day in 2008. 34 gigabytes would t on about 7
DVD disks, or 1.5 Blu-ray disks, or about one fth
of an average notebook computers hard drive –
depending on when you last purchased a computer.
About 35 percent was from television, 10 percent
from movies, and 55 percent from computer games.
Computer games are a big story in themselves, and
we will discuss them extensively in Section 3.
Compared to the 140 percent increase in total words
consumed from 1980 to 2008, there was a 350
percent increase in the number of bytes consumed,
to 3.6 zettabytes. The higher growth of bytes than
words reects faster growth in visual media (TV and
computers) than in verbal and textual media (radio
and print). We will discuss these growth rates fully
in Section 4.1.
How much is 3.6 zettabytes?
If we printed 3.6 zettabytes of text in books, and stacked them as
tightly as possible across the United States including Alaska, the
pile would be 7 feet high.
Figure 5: INFO
Consumption in Compressed Bytes
Computer Games
Recorded Music
All TV
Percentage Consumed
Recorded Music
Computer Games
All TV
How Much Information? 2009 Report on American Consumers
1.6 Storage vs. Consumption
One implication of our denition is that stored data
is not necessarily information. Storage is vital to
shift data consumption forward in time, because
someday it may be useful to create information. Some
previous studies dene stored data as “information.”
But we classify it as data, until such time as the data is
transmitted to the consumer for use.
For the purposes of this study, we measure data
as information each time consumers use it.
This measurement is feasible in the household
sector, where the primary storage media include
books, DVDs, CDs, MP3 players, computers
and, increasingly, digital video recorders
(DVRs). Indeed, our statistics for consumption
of information are many times larger than total
storage of data in those devices. According to
some estimates, the total amount of hard disk
storage worldwide at the end of 2008 was roughly
200 exabytes. In other words, the 3.6 zettabytes
of information used by Americans in their homes
during 2008 was roughly 20 times more than what
could be stored at one time on all the hard drives in
the world.
The data ‘footprint’ of a storage device is not just
how many bytes it holds, but how many bytes are
created (both reads and writes) over time. Hence,
for most storage devices, their nominal capacity is
much smaller than the data that can be housed on
the device over a period of time (as les are erased
and replaced).
1.7 Valuing Information
Hours, words and bytes measure the volume of
information, not its value. There are many potential
criteria for measuring the value of a stream of
information, including subjective judgment,
selling price, willingness to pay by consumers,
development cost, and audience size. But there is
no clear way of comparing value, especially when
comparing information of different kinds, and
Recording devices can process more than their capacity
Digital video recorders in 2008 typically store between
80 and 160 gigabytes (GB) of recorded video. But
consumers erase programs after viewing them, and
overwrite the data with more recent programs. So the
nominal storage capacity in a DVR is almost irrelevant
when measuring how much information is accessed by
members of a household.
Similarly, take the example of a home video surveillance
system with four cameras. A DVR stores the video
streams, and new video over-writes older images. How
long it stores the images before overwriting depends on
the ratio between the size of the DVR’s hard disk drive
and the bit rate of each surveillance camera stream.
In turn, the bit rate is determined by the quality of the
original video: is it in color or black-and-white? How
much video compression is used in storing the feed? A
typical medium-quality video stream occupies about one
gigabyte per hour of video. So if the DVR has a 160 GB
hard drive, the system will hold approximately the most
recent 36 hours of video frames.
So how do we classify these data streams? The data
stored on the DVR is not yet information, because those
bytes are not necessarily used by anyone. In the case
of the surveillance system, only a fraction of the data
becomes information, i.e., data delivered for use. This
would include any time the homeowner is watching the
live feed (rare if ever), or that a recent period is played
back for evidence in the event of a burglary. The same
size of DVR used in the home to store TV programs (to
zip through commercials or simply time-shift viewing)
is likely to produce much more information, and as a
program is viewed, the same bytes can be written over
with future programming. In both DVR examples, we
measure only what the consumer sees.
To complicate matters further, in the home video
surveillance example when a security camera creates
a frame, it actually creates at least four frames of
data – the original plus three descendants. Because of
compression, the three descendant frames have fewer
bytes than the original.
If the frame is later recalled and
displayed on a monitor, two additional descendants are
created. So according to our measurement, the original
and most descendants are data; only the nal descendant
is information.
This explanation oversimplies issues such as the byte-equivalent
of analog images.
How Much Information? 2009 Report on American Consumers
particularly from different time periods. Take for
example a landmark speech, Abraham Lincoln’s
Gettysburg Address versus a current TV series, a
2008 episode of “Heroes” on NBC.
The Gettysburg Address took roughly 2.5 minutes
to deliver and was 244 words in length, i.e., 1,293
characters, or bytes of text. Nobody is sure exactly
what Lincoln said; his handwritten texts do not
match contemporary accounts. (On the other
hand, a presidential speech today will be recorded
electronically for posterity, as they have been
since the early Fifties). The direct cost of writing
and delivering the Address was probably less than
$5,000 (valuing Lincoln’s time at $200 an hour in
today’s dollars).
In contrast, a 2008 episode of “Heroes” on NBC
ran 44 minutes in length (without commercials), the
master version occupies 10 GB of digital storage,
and each special effects-laden “Heroes” episode
cost an estimated $4 million to produce. So by any
quantitative measure, the popular TV program
would be considered much more information than
the Gettysburg Address offered. Yet Lincoln’s words
were far more important, and most of the world would
agree that they were, in most senses of the word, far
morevaluable and worthy of saving for posterity.
In this report on information, we measure neither
original delivery time nor bytes of storage, but total
bytes of all copies across all recipients. “Heroes”
episodes in the 2008-09 season had just over 10
million viewers, including those views on DVRs
within a week of the broadcast. Reruns could push
that gure to 18 million per episode.
Optimistically, let’s guess that the Gettysburg
Address has been read twice by every American
who reached 6th grade since Lincoln uttered the
words in 1863. This is approximately 500 million
people, multiplied by two readings, which equals
one billion readings. So measured by the pure
number of information consumers, Lincoln’s one-
billion readership trumps “Heroes”’ 10 million
average weekly viewership by a factor of 100. On
the other hand, looking at bytes, a compressed
episode of “Heroes” on an average TV comes in at
about 500 MB, times 10 million views, which adds
up to 5 petabytes. In contrast, the 1 billion readings
of the Gettysburg Address are only 2.4 terabytes.
Looked at this way, NBC’s “Heroes” wins by a
factor of 2,000.
Another approach to measuring information then
and now is to calculate the amount of time people
spend receiving different kinds of information.
The Gettysburg Address takes barely 2.5 minutes
to read, but more time is spent understanding the
background: the American Civil War, the carnage
of the battle, the political importance of Lincoln
rallying the North to continue the war, and so forth.
So let’s call it 20 minutes. Now Lincoln’s Address is
measured at 40 billion minutes, or 0.7 billion hours.
In contrast, a “Heroes” episode is watched only
about 14 million hours. So the Address is bigger by
a factor of 50.
So which of the two ‘information’ events is larger in
an absolute sense? Perhaps none of our quantitative
measures captures this. The pure volume of
information does not necessarily determine its value
or impact. The right information, delivered at a key
time and place, can move mountains. At the other
extreme, raw bytes are now so inexpensive that we
often pay only minor (or zero) attention to them. So
this study eschews efforts to determine the value
of one type of information over another, in favor of
estimating the volume of information consumption.
Information can be roughly classied intoinformation
for consumption,primarily in households and mobile
uses, andinformation for production” in workplaces
and between machines (both of which will be the
subject of future HMI? reports).
This section discusses traditional information
in U.S. households – information delivered and
consumed from media that preceded the home
computer era. While most of these media are
increasingly digital, thanks to the power of modern
computing and networking technologies, they
remain “traditional” in the sense that the content
and the consumption experience are conventional –
think of people watching television, speaking on the
telephone, reading a book or magazine, or going out
on a Friday night to the movies.
2.1 Television
Americans are heavy users of TV, and on two of our
three measures of information (hours and words),
TV is by far the largest source, although it is only
second as measured by bytes. However, television
usage measured in hours per person is rising only
slowly. After all, whether you have 150 channels on
digital cable or just a handful of channels of over-
the-air broadcast TV, you still have only a limited
number of hours to watch TV. The total time has
not changed dramatically despite today’s broader
channel choices and higher-denition TV reception.
How Much Information? 2009 Report on American Consumers
The estimated 292 million U.S. viewers average
nearly ve hours of TV viewing per day.
Total TV viewing accounts for 41 percent of total
hours of information consumption, and nearly 35
percent of total bytes.
(Table 3 Television and Radio Consumption) We
receive TV programs over a wide variety of media,
including cable, satellite, and plain old broadcast, in
descending order. Digital television is compressed
for transmission and then uncompressed for
viewing, and we measure the compressed bit rate.
And if two people are watching
the same show on the same TV
set, it will show up twice in our
measurements, because we use
A.C. Nielsen for TV data, and
thats how they do their counting.
While HDTV began to take
off with consumers in 2008,
more homes had HDTV sets
(53 percent in January 2009,
according to estimates from
the Consumer Electronics
Association) than actually get
HDTV signals (approximately
40 percent, although estimates
vary). It is quite common for
TV owners to not realize that
their “high denition” TV set is
actually showing only standard
denition TV signals. For those
households that do receive HDTV, we don’t have
good data, but in 2008 roughly 40 percent of their
viewing hours were high denition.
Even so-called “high denition” television
programs vary considerably in quality. One reason
is that original content varies, but another is that
cable companies often choose to offer a higher
number of channels, with lower bandwidth and
lower quality per channel, rather than the reverse.
Over the air, cable, and satellite (digital) TV are
transmitted at an average of 4 megabits per second,
although this depends on what compression
methods are used. We estimate that high denition
TV averages about 12 megabits per second. Putting
all of this together, we use an estimate of 4 megabits
per second for standard TV, and 7.2 megabits per
second for the weighted average bandwidth of TV
into homes that receive HDTV.
Not surprising given TV’s still-dominant role in
consumer information, it has spawned a variety of
special delivery methods beyond cable and satellite
(which, in their time, were novel). These include
video cassette recorders (VCRs), digital video discs
(DVDs), and home video recorders that use hard
drives, called DVRs or PVRs (digital or personal
video recorders). While roughly 80 million video
cassette recorders (VCRs) remain in U.S. home,
their usage is so low that VCRs are no longer
broken out as a separate category in home-video
playback statistics. Meanwhile, the high quality of
DVDs led to their use in essentially all American
households. And according to a recent estimate,
DVR penetration was 28 percent in late 2008, and
33 percent in 2009.
Nielsen lumps most DVR use
into its “live” TV hours, but it reports DVD use
separately with an average per user viewership time
of 9.2 hours every month. DVDs have a variable
bandwidth averaging 5 megabits per second,
slightly better than a standard denition TV, and
contribute 2.2 percent of INFO
hours, and 1.9
percent of INFO
in bytes. In comparison, “live”
television was 41 percent of hours and 35 percent of
in bytes.
DVD usage is expected to increase as the price of
Blu-ray high-denition players declines and more
people buy HDTV sets on which to watch Blu-ray
discs. Ironically, until now the biggest purchasers
of Blu-ray discs have been gamers, because Sony
built Blu-ray technology into its Playstation 3 game
console, making it the most widely used Blu-ray
player in the world (This decision probably raised
the price of the Playstation 3 signicantly, hurting
its competition with other game consoles). But in
2008, sales and rentals of Blu-ray disks had little
impact on home TV viewing, and they are not
included in our DVD viewing estimates for this report.
The leading TV ratings service, A.C. Nielsen,
began collecting and reporting on the use of mobile
telephones to view video content in 2008, and we
Table 3: Television and Radio Consumption
Total #
of Users
per User/
Total Info.
% of Total
% of Total
Television (incl. Delayed View)
292 148.5 1,197 39.22% 32.83%
DVD Players
254 9.3 70 2.21% 1.91%
Mobile TV
10 3.6 0 0.03% 0.00%
1,266 41.47% 34.74%
AM/FM Radio
233 80.6 10 17.62% 0.27%
Satellite Radio
19 65.8 1 1.17% 0.04%
11 18.79% 0.30%
TOTAL 1,277 60.25% 35.04%
How Much Information? 2009 Report on American Consumers
have used their new measurements to calculate the
information received over this mobile data service.
Just over 10 million U.S. subscribers watched
video content on their mobile phones, and they
averaged 3.6 hours of video viewing per month.
Since both the hours per user and number of users
are low relative to the juggernaut of mainstream
TV, this works out to only about 0.04 percent of
words of information INFO
. Because of the small
screen size and the scarcity of mobile telephone
bandwidth, the bit rates and resolution of these
signals can be quite low, less than a quarter of even
a conventional TV signal. Therefore, their total
impact on bytes is even smaller – we estimate .002
percent of total bytes INFO
. On the other hand,
iPods and similar devices have the ability to watch
TV programs and movies via downloads from the
Internet, using a service like iTunes. These can
certainly be considered “mobile TV,” but currently
Nielsen provides no viewer data for these devices.
2.2 Radio
Video never did “kill the radio star” as the British
pop group, The Buggles, warned in their 1979 chart-
topping single that famously became the rst video
on MTV when it began broadcasting in 1981. Radio
today is thriving on new technology, including HD
audio, satellite transmission, online radio and other
new services. But in a census of total information
consumed in U.S. households, audio requires very
low data rates. Even without factoring HDTV into
the equation, video requires roughly 30 times more
data throughput than audio. Or to compare satellite
services, the throughput of satellite TV (1,800
megabytes per hour) compares to only 8 megabytes
per hour for satellite radio.
The country’s 233 million AM-FM radio listeners
received 10 exabytes of information in 2008. This
includes radio in and out of the home (mostly at
home on weekends) and in the car during weekday
morning and afternoon commutes. Satellite radio,
with nearly 19 million listeners, is still in its
infancy, but users of services such as Sirius listen
to more than 2 hours a day on average (almost
as much as listeners to standard AM-FM radio),
pushing satellite radio information to more than 1.3
exabytes. Data for Internet radio – a new and small
but potentially important segment of the market –
are not yet reliable, and it was not included in the
radio totals.
2.3 Telephone
While most U.S. households had a telephone 25
years ago, today it is common to have at least one
landline in the home and more than one mobile
phone. There were 263 million wireless users in
2008, versus only 154 million wired lines.
the other hand, on average wired lines are used
for almost twice as many minutes per day, so
information words INFO
are slightly higher for
xed lines. For the sake of accuracy, therefore, this
report divides the ‘telephone’ sector into two parts:
the traditional or conventional phone usage covered
in this section, and wireless phone service that is
fast evolving into mobile computing, and therefore
is covered in Section 3 below. (Table 4 Telephone
First though, some comparisons of wired versus
wireless voice telephony. It is quite likely that by
2010, the total number of hours that Americans
spend on their cell phones will overtake their use
of landline phones in the household. As a factor
of total hours of information consumed by U.S.
households in 2008, it was
already a close race: xed-
line phones accounted for
3.2 percent of total time
consuming information, while
mobile phones accounted for
2.9 percent. Translated into
bytes though, landline calling
per person per day remained
12 times greater. This is due
to two factors: much more
sophisticated compression
and lower voice quality of
wireless phones.
Our calculations of information consumed by
telephone xed landline users in 2008 (also known
as ‘POTS,’ for ‘plain old telephone service’) are for
voice trafc – not DSL nor dial-up Internet service
supplied through a wired telephone connection to
the home. We calculate that occupants in every U.S.
household used their home phone for an average
of 22.5 hours each month. Using these numbers,
we calculate 1.2 exabytes as the total voice-trafc
information consumed by people using landline
Table 4: Telephone Consumption
# Subscribers
Hours per
Total Info.
% of Total
% of Total
Fixed Line Voice 154 22.5 1.19 3.26% 0.03%
Cellular Voice 263 11.9 0.17 2.94% 0.00%
TOTAL 1.36 6.20% 0.04%
Note 1: Fixed Line Voice includes residential, business and most VoIP subscribers.
Note 2: Cellular Voice includes residential and business subscribers.
How Much Information? 2009 Report on American Consumers
telephone service in 2008. Adding mobile voice
trafc to the mix, total voice communications
created and consumed 1.4 exabytes of information.
2.4 Print
The most traditional media consumed in the home
are the different avors of print publications.
Taken together, U.S. households in 2008 spent
about 5 percent of their information time reading
newspapers, magazines and books, which have
declined in readership over the last fty years. From
the perspective of the information measured in
words INFO
, printed media account for almost 9
percent of all words consumed. However, translated
into bytes, they barely register: two-hundredths
of a percent (0.02%) of INFO
. The alphabet is a
very compact way to transmit words, and although
magazines have color photographs, they are only
still images. (Table 5 Conventional Media)
It is expected that readership of print publications
will continue to decline, even if newspapers and
magazines are able to nd a sustainable model
for publishing their content on the Web. Our print
data do not take into account any Internet editions,
which are instead included as computer information
in the home (see Section 3). Printed books – on
which Americans spent barely 2 percent of their
information time INFO
, and 4 percent of words
– may someday be displaced by digital devices
such as the Amazon Kindle, but the electronic book
platforms had more potential than actual readers in
2008. Yet, in many ways electronic documents have
already taken over for paper – see the sidebar “The
Evolution of Reading.”
2.5 Movies
Although Americans spend much more time
watching movies on television, broadcast and
through DVDs, watching movies in a theater
remains a popular attraction. No other medium
offers anything like the data throughput of a large-
screen theatrical projection – roughly 250 million
bits per second, which is 20 times the bandwidth of
high denition TV. How is this possible, especially
since movies are shown at only 24 frames per
second, versus 30 for television? Movies have
the advantage on the three other determinants of
bandwidth: they have larger screen resolutions
with more pixels, they have ner color rendition
corresponding to more bits per pixel, and they use
arguably less compression since they use lm and
not electronic transmission.
So even though the average American spends less
than one hour per month at the movies, it adds up to
3,300 megabytes of information INFO
per person
per day — a surprising ten percent of the total daily
Digital projection is gradually coming to movie
theaters, but the penetration in 2008 was limited.
At present, IMAX technology which is based on 70
mm (analog) lm provides the highest quality.
2.6 Recorded Music
While the technology for listening to recorded
music has changed dramatically, and retail sales
of recorded music have declined, the popularity
of this medium appears to be intact. We estimate
that Americans spend an average of 14 hours per
The Evolution of Reading
The use of different media has changed dramatically over time.
It is a cliché that reading is in decline. But on the other hand we
get considerable information from the Internet, which is a heavy
print medium. Do we really read less?
We show this evolution in Figure 6 Evolution of Reading.
Conventional print media has fallen from 26 percent of INFO
in 1960 to 9 percent in 2008. However, this has been more
than counterbalanced by the rise of the Internet and local
computer programs, which now provide 27 percent of INFO
Conventional print provides an additional 9 percent. In other
words, reading as a percentage of our information consumption
has increased in the last 50 years, if we use words themselves as
the unit of measurement.
Figure 6: Evolution of Reading
Fraction•of words INFO
from different sources
Recorded Music
Computer Games
All TV
How Much Information? 2009 Report on American Consumers
month listening to recorded music – on CDs and
MP3 players. That is nearly 4 percent of all the
hours spent consuming information, contributing to
a volume of 8.8 exabytes of information. Although
the amount of time INFO
used for recorded music
is much lower than radio, the compressed bytes
are almost as large, due to higher audio
quality of recordings.
New digital technologies continue to remake the
American home. Ten years ago 40 percent of U.S.
households had a personal computer, and only
one-quarter of those had Internet access. Current
estimates are that over 70 percent of Americans now
own a personal computer with Internet access, and
increasingly that access is high-speed via broadband
Adding iPhones and other ‘smart’ wireless phones,
which are computers in all but name, personal
computer ownership increases to more than 80
percent. Many households now boast dozens of
digital devices for entertainment, information and
other purposes: 3G phones, PDAs, MP3 players,
television sets, DVRs, home computers, game
devices, and so on.
In this section we report on
ve major categories of home
computer use:
• Accessing the Internet such as web
browsing, communications (including email)
and social networking;
• Uploading, downloading and watching
videos on the Internet;
• Playing computer games;
• Mobile devices and applications; and
• Ofine computer activities that don’t require
Internet access; such as writing a letter in
Word, putting together an Excel spreadsheet,
or editing home photos.
The average American spends nearly three hours
per day on the computer, not including time at
work. That is 24 percent of total information hours,
and over 55 percent of all information bytes INFO
We estimate that 2,000 exabytes of information, or
2 zettabytes, were consumed by Americans using
home computers, gaming consoles and mobile
computing devices in 2008. The vast majority of this
information is attributed to computer games, whereas
the majority of the time Americans spend on the
computer involves the less graphics-intensive but
more commonplace Web browsing, email and such.
3.1 Communicating
and Browsing the
The Internet has
revolutionized the way
Americans communicate. In
1980, email was non-existent
in U.S. households, and
sending a fax was the hot
new way to send messages
faster and more cheaply than
Table 5: Conventional Media
# of Users
per User/
Total Info.
% of Total
% of Total
Movies 295 0.9 356.31 0.25% 9.78%
Books, Newspaper, Magazine 295 32.8 0.67 5.09% 0.02%
Recorded Music 295 13.8 8.85 3.83% 0.24%
365.83 9.2% 10.04%
Table 6: Computer Use Non-Gaming
# of Users
per User/
Total Info.
% of Total
% of Total
and web browsing 226 65.7 8.01 13.99% 0.22%
Internet video 95 1.8 0.89 0.16% 0.02%
Ofine programs 226 11.1 0.68 2.37% 0.02%
TOTAL 9.58 16.51% 0.26%
How Much Information? 2009 Report on American Consumers
via Telex or rst-class mail. Today, 220 million
Americans spend 14 percent of their information
hours INFO
on the Internet, almost all of it on
applications such as web browsing and email (Table
6 Computer Use, Non-Gaming).
In 2008 email remained the most widely used
application, accounting for nearly 35 percent of all
hours on the Internet. Studies show that the average
user can process 30 to 60 emails an hour, involving
a sequence of read, respond, assign, delay or delete
actions for each message.
However, because email is
largely text-based, it accounted for relatively few bytes.
By comparison, Americans spent fewer hours on web
browsing (30% of our time on the Internet). Studies
show that people cycle quickly through Web sites and
doing searches to nd content, and they estimate that
most users spend only 8-9 seconds looking at most
Web pages. Users tend to continue this behavior until
they nd the page of interest, change their minds, get
bored or shift to another task.
Web pages generally include both photos and text,
and rapid browsing behavior creates delays as each
page is loaded. Internet use continued to evolve
rapidly during 2008. Web use was changing due to
the rapid uptake of social networks such as Facebook
and MySpace. Facebook reported over 175 million
users worldwide with an average Facebook user
spending 27.5 minutes a day on the site.
For our byte measure INFO
we track the amount
that actually moves across the “pipe” into the
home. This is limited by the average download
speed, which varies considerably by technology,
by region, by what plan the consumer is signed up
for, and even by time of day.
However, bandwidth
levels are increasing over time as people sign up
for higher levels of service, and as Internet service
providers strengthen their networks. We assume an
average speed of 100k bits per second, which gives
an estimated total of 8 exabytes in 2008. All of the
text Internet applications combined represent a drop
in the bucket when estimating the total number of
bytes of information consumed in 2008 by U.S.
households – just two-tenths of a percent (0.2%),
even though Americans spend 76 percent of their
Internet time on email and other text. The reason:
Internet video and especially computer gaming
involve computer graphics that deliver much higher
data throughput to the users computer screen.
3.2 Internet Video
We measure Internet video, such as YouTube, in
its own category. Although there were 95 million
viewers in 2008, their average viewing time was
less than 2 hours per month. Hulu and other sites
for viewing “regular” television shows may have a
big effect in the future, but were used only sparingly
in 2008.
Furthermore, the resolution of Internet
video was very low. Again, the speed of the pipe
into the house limits how much can be received
while the consumer is actively trying to watch.
Although in principle delayed download methods
such as peer-to-peer and Apple TV (from iTunes
or similar web sites) can increase video download
sizes, surveys of consumers don’t yet indicate much
use. Furthermore, whatever the pipeline into the
home, providing high quality video costs more for
the provider, be it YouTube, Hulu, or otherwise,
because they must pay for all of the bandwidth
used at their end. YouTube only made so-called
HD video available late in 2008, and even that
has a much lower resolution than high denition
As a result, Internet video is still small by most
measures. Time consumption INFO
was only .2
percent of the total, and bytes INFO
were under
1 exabyte, less than text-based Internet use.
higher bandwidth of video compared with web
browsing is more than counterbalanced by the
smaller number of users (95 million versus 226
million) and much smaller number of average hours
per user (1.8 versus 65.7 hours per user per month)
In the future, the small role of Internet video may
change. YouTube and other video sites are growing
exponentially in both the number of unique visitors
to the sites each month, and in the number of videos
uploaded and viewed daily.
We return to this topic
in the conclusion.
Figure 7: Average Daily Consumption of Bytes, INFO
0 4 8 12 16 20
Gigabytes Per Person Per Day
Recorded Music
Computer Games
Computer Games
All TV
All TV
How Much Information? 2009 Report on American Consumers
3.3 Computer Gaming
While non-game activities account for more of the
time Americans spend on computers, computer
gaming has come to dominate the total number
of information bytes – for a total of nearly 2,000
exabytes (2 zettabytes) in 2008. That is the lion’s
share of total bytes from all home computing and
all sources in general (Figure 7 Average Daily
Consumption of Bytes, INFO
), even though
gaming accounted for less than 8 percent of total
information hours INFO
In 2008, an estimated 70 percent of adults in the
U.S. played computer games, averaging slightly
less than one hour a day. Players were split roughly
evenly between men and women (although gender
played a role in the differing types of games
played). Another estimate in 2009 was that 87
percent of males of all ages, and 80 percent of
females, play some form of computer game.
Approximately 15 million dedicated gaming
machines (consoles and
portables) are sold annually in
the US.
It is difcult to talk about
computer gaming in aggregate,
because there are many
different categories of gaming
and each type is associated
with different percentages
of players as well as hours
and bytes consumed. Our
headcounts and estimated
hours of play are from a 2008
industry report on computer
gaming, which described seven types of gamers,
ranging from “extreme gamers” (3% of the gaming
population) to casual gamers (20%).
gamers play on more than one type of machine,
which is not surprising since almost everyone owns
a cellphone and most cellphones in 2008 had the
capability to play at least simple games.
Hardware is the critical factor in determining the
volume of information generated by videogames
and computer games. We report hardware in four
• High performance gaming computers, which
were used by 21 million players in 2008;
• Standard computers – 124 million users;
• Console game machines, such as Microsoft’s
Xbox, Sony’s Playstation and Nintendo’s
Wii – 89 million users; and
• Portable game machines, including the Sony
PSP, Nintendo DS, and others – 129 million
For each hardware type, we estimated the video
throughput for an “average machine” in the class,
playing an “average game.” High-performance
gaming PCs use the most powerful processors
in the world, called Graphics Processing Units
(GPUs), to generate graphics. Some GPUs have
over one billion transistors, and more than 200
parallel processors running at once. (Figure 8
Example of a Graphics Processing Card) We
estimate the effective compressed bandwidth of
these machines at approximately 100 megabits per
second – eight times that of high denition TV.
An estimated 21 million users spend an average
of 87 hours every month playing games on these
computers. (Table 7 Computer Game Playing)
They account for a huge share of all information
bytes consumed by U.S. households: 1,400 exabytes
(1.4 zettabytes) annually – or approximately 39
percent of all INFO
. This large role of high-end
computer gaming is particularly surprising, because
Table 7: Computer Game Playing
# of Users
per User/
Total Info.
% of Total
% of Total
PC (high performance) 21 86.9 1,405 1.70% 38.56%
PC (standard) 124 18.1 194 2.10% 5.33%
Consoles 89 30.3 368 2.53% 10.09%
Handheld Devices 129 12.6 24 1.53% 0.64%
TOTAL 1,991 7.86% 54.62%
Figure 8: Example of a Graphics Processing Card
How Much Information? 2009 Report on American Consumers
it accounts for less than 2 percent of the hours
Americans spend consuming information. The
quality of visual effects on high-end machines and
the rapidity with which the player is confronted
with changing scenes on the screen are why these
devices and games represent such a huge portion
of total information to U.S. households, as well as
why the games are so immersive to play. Figure
9 Screen Shot from NBA Live 10 shows a screen
shot from the Playstation 3 version of a recent
computer game. (This resolution is considerably
below the best possible from computers in 2008.)
By contrast, six times more Americans play games
on standard PCs than on high-end PCs, with an
average of 18 hours a month. The quality of their
on-screen graphics on these PCs is on average
far inferior, so this translates into 194 exabytes,
barely 10 percent of the total amount of the INFO
from games.
Nearly 90 million Americans play
games on dedicated game machines, and the
average player uses a console 30 hours a month.
By our calculations, game consoles account for
368 exabytes – 10 percent of total household
information. Most of the consoles are used ofine,
but increasingly, users are playing over the network
as well, so the line that divides online and off-the-
Internet gaming is rapidly fading. Even handheld
game devices, used by 129 million players, created
24 exabytes of information in 2008, or about triple
the total bytes of information received in the form
of recorded music (primarily CDs and MP3s).
Looking at all the game platforms, we calculate that
total information from this relatively young form
of entertainment (2 zettabytes) is 50 percent larger
than the volume of information from established
media that are more than 50 years old, TV and radio
(1.3 zettabytes). (Of these 2 zettabytes, 70 percent
is from 21 million high-end gaming computers.)
In the short run, TV’s share of bytes may increase
as the percentage of U.S. households with HD
television reception grows at a rapid pace. But
manufacturers of high-end gaming computers
and consoles are already working on even more
powerful new machines and photorealistic games
– so in the long run, gaming is likely to continue
accounting for the bulk of information consumed
by U.S. households, as measured by INFO
. On the
other hand, measured by words and hours, computer
games are a modest 2.5 percent and 8 percent of
total consumption, respectively.
3.4 Off-Internet Home Computer Use
In many households, considerable computer time is
spent locally, without going online except perhaps
to send an email. After all, fewer than 60 percent
of adult Americans had broadband connections in
Off-line use includes activities like updating
a resume, editing photos, or running a household
nance program. Time-use statistics for such off-
Internet, non-gaming computer use are no longer
reported directly by U.S. government or industry
sources. We relied on partial data provided by the
American Time Use Study (ATUS) conducted by
the Bureau of Labor Statistics (BLS), and time-of-
use studies published by the Center for Research in
Information Technology and Organization (CRITO)
at the University of California, Irvine.
We estimate that non-Internet, non-gaming
home computer use was very widespread, but
averaged only 17 minutes per day per average
American. Because these applications are primarily
text based, they add up to only 0.7 exabytes per year.
3.5 Smart Phones
The growth of new media, viewing video, sending
text messages, or playing games on a feature phone
or smartphone are growing quickly,
driven by
consumer sales of new phones and the provision
of new content services, both free and subscriber
The contributions of these new devices and
their use, however, do not gure prominently in our
information totals – their numbers are still too low
to be a signicant fraction of the total information
consumed by Americans when compared to the
information volume delivered by traditional media.
Approximately 263 million Americans carry cell
phones, and in 2008 approximately 50 million
of these phones were smartphones such as the
Apple iPhone.
With rst-generation analog cell
phones and 2G digital handsets, consumers were
largely limited to using their phones for voice
calling (cellular phone voice trafc was discussed
in Section 2, 2.3 on telephones). So while voice
communication accounted for over two-thirds of
cell phone hours in 2008, the spoken word is such
an efcient medium for conveying information that
voice trafc, measured in bytes, accounted for only
0.2 exabytes, a negligible fraction of total INFO
While Americans spent approximately 7 billion
hours text messaging in 2008, because SMS
text messages are so small, their byte total is
For now, new media volumes are
small compared with traditional media volumes, but
this is changing.
How Much Information? 2009 Report on American Consumers
Figure 9: Screen Shot from NBA 2K10 Game
How Much Information? 2009 Report on American Consumers
4 Trends, Perspectives
and the Future of
U.S. Information
Our analysis, while incomplete, has uncovered
a variety of trends and patterns, and also some
4.1 Analyzing the Growth of
While at one level, the estimated ve-fold increase
from 1980 to 2008 in INFO
bytes consumed is
impressive, this is an annual growth rate of only
5.4 percent. This is far less than the rate of increase
in most measures of digital technology, which
tend to be driven by Moore’s Law: the number
of transistors on an integrated circuit doubles
approximately every two years. For example,
the cost of hard disk storage in an 1982 personal
computer was about $50 per megabyte for a 10
megabyte drive; today it was less than $1 per
gigabyte for a drive of 100 gigabytes, a 50,000-fold
William Nordhaus studied long-term
trends in the cost of computation, and found that
it fell faster than 60 percent per year from the mid
1980s to 2006. The total reduction was ve orders
of magnitude, a cost/performance improvement by a
factor of 200,000.
If the total revenue of an industry is constant, then
the quantity of its output, measured in terms of total
performance, must grow in inverse proportion to
its price/performance ratio. And in fact, revenue for
both the semiconductor industry and the electronics
industry grew between 1980 and 2008. So the
capacity to process bytes must have grown at a rate
somewhere between 30 percent (the lower end of
Moore’s Law estimates) and 60 percent per year.
How is it possible that consumption INFO
grew at
only 5.4 percent per year, less than twice as fast as
growth in GDP over the period?
We analyze this by decomposing growth in INFO
into three components. Total information consumed
is the product of three factors: the American
population, average hours per person spent
consuming information, and average information
per hour. We decomposed total growth into these
components for the period 1980 to 2008:
• Population grew at 0.95 percent per year,
from 226 million to 295 million (ages 2 and
• Average hours of information consumption
per person grew at 1.7 percent per year, from
7.4 hours to 11.8 hours of INFO
• Average bandwidth (across all media)
grew at 2.8 percent per year from 2.9 Mbps
(megabits per second) to 6.4 Mbps. This is
a measure of “information intensity” of our
• Gigabytes per person per day grew at an
annual rate of 4.4 percent, from 9.8 to 33.8
Gigabytes of INFO
. Not coincidentally,
4.4 percent is the sum of the growth rates in
hours per person and in average bandwidth
If there is one major surprise in this study, it is
that INFO
consumption and information intensity
per hour grew at these low rates from the dawn
of personal computing in 1980 to today, despite
Moore’s Law and the revolutionary shift from
analog to digital technology in most information
media. Slow growth in the US population is well
known, and the 1.6 percent per year growth in
hours of consumption per person is understandable
given the constant 24 hour length of a day. But the
2.8 percent compound annual growth rate in bytes
consumed per hour remains a drop in the bucket
compared to the doubling every two years in the
number of transistors on an integrated circuit.
Given how cheap information processing is today
compared with 1980, why aren’t we consuming
hundreds of times more bytes per hour than we did
in 1980?
There is one basic mathematical reason for this
result: very slow growth of INFO
from television.
The dominant source of bytes in 1980 – color
television – remained largely unchanged until the
very recent switch to high denition TV in the U.S.
market. And because high denition TV in 2008
was in less than half of households, and accounted
for less than half of the TV viewing hours in those
households, it had little impact on average bytes
per hour from TV. Finally, TV viewing time as a
share of our information day was approximately
unchanged. Putting together the slow growth in
hours of TV and the minimal change in the quality
of TV signals, bytes from TV grew slowly.
But this arithmetic does not get at the essence
of the issue. First, why did TV picture quality
stay stagnant for so long? Second, the capacity
of information technology has been increasing
at Moore’s Law speeds. Intel and the rest of the
semiconductor industry sell more devices, and more
transistors per device, every year, and America’s
How Much Information? 2009 Report on American Consumers
share of worldwide consumption has been roughly
constant. So if these transistors were not being used
to consume more bytes, where did they go? Third,
personal computers now occupy a major share of
our information consumption, and depending on
what measure is used so does the Internet. Will their
growth raise the historical trajectory for the future?
4.2 Where are the Missing Bytes?
We have tentatively identied four places where the
missing bytes have gone, although further research
will be needed to conrm and measure them. (Table
8 Explaining the Gap between Consumption
and Capacity Growth) First, we have measured
information consumed by consumers, but the
amount of information available to them has grown
much faster.
Second, this report looks only at consumer
information. We are working on a study of
information in enterprises, which follows different
growth patterns. Third is the reduction of load
factors. Our houses today are full of electronic
devices that we use for only hours or minutes a
month. Even devices that we use every day, such as
cell phones, contain transistors that have capabilities
that we may never use, such as built-in GPS and
4.2.1 Dark Data
A nal factor is the rise of “dark data.” When
electronics were expensive, devices were naturally
reserved for high-value activities. People and
information worked closely together. But now one
million transistors costs less than one cent, yet
people’s time is still valuable. We can no longer
afford, nor do we need, to have people closely
scrutinizing data as it is created and used. Instead,
we hypothesize that most data is created, used, and
thrown away without any person ever being aware
of its existence. Just as cosmic dark matter is detected
indirectly only through its effect on things that we
can see, dark data is not directly visible to people.
Examples of dark data occur in the home, although
most of it is elsewhere. Data can be created in
an automated fashion without the consumer
intervening. For example, a consumer can set a
DVR just by specifying the name of the program,
not when it is broadcast. Information is exchanged
over the Internet between the cable company’s
computer and the DVR, and the DVR decides when
to record, and what channel. We recognize the
results of the dark data when
we turn on the DVR and it is
converted to information on
our TV screen.
The family auto (or
automobiles) is a more
typical example of dark data.
Luxury and high-performance
cars today carry more than
100 microcontrollers and
several hundred sensors,
with update rates ranging
from one to more than
1,000 readings per second.
One estimate is that from 35 to 40 percent of a
cars sticker price goes to pay for software and
As microprocessors and sensors ‘talk’
to each other, their ability to process information
becomes critical for auto safety. For example,
airbags use accelerometers, which measure the
physical motion of a tiny silicon beam. From that
motion, the cars acceleration is calculated,
approximately 100 times each second, this data
is sent to a microprocessor, which uses the last
few seconds of measurements to decide whether
and at what intensity to inate the airbag in the
event of a collision. Over the life of an auto, each
accelerometer will produce more than one billion
measurements. Yet in a crash, only the last few data
points are critical.
Each sensor creates several
gigabytes of data without a single byte that counted
as “information” in our analysis of consumer
The phenomenon of dark data permeates modern
digital technology, and goes far beyond the range
of this report. We hope to analyze it carefully in
the future.
Table 8: Explaining the Gap Between Consumption and Capacity Growth
Cause Explanation Example
Growth of information available
over information consumed
We have far more choices of
what to consume
Average household now receives
120 TV channels, but still watches
only about 10 hours per day
Dark data
Much of the world’s data now ows
between machines, without human
intervention or awareness
Automobiles now contain more
than 50 processors each
Enterprise information
This report only considers consumer
Low load factor
We can afford multiple redundant
TVs in the kids’ bedrooms
How Much Information? 2009 Report on American Consumers
4.2.2 Two Kinds of Quality: Variety
and Resolution
Some of the benet of cheaper information
technology has been in the form of more choices of
what to consume. The number of TV channels per
average household has now reached about 130, of
which the average household actually watches 18.
Both numbers are considerably higher than they
were in 1980. This is an example of a more general
phenomenon: the ratio of information available to
information consumed grows over time.
The additional channels of TV, however, have come
at a cost: higher compression and therefore lower
video resolution for the channels we receive. The
issue is straightforward: bandwidth costs money (all
those transistors). For a xed budget, a cable TV
company, and especially a satellite TV company,
have only a xed total capacity in megabits per
second. Suppose it allocates 600 Mbps to broadcast
TV. If it divides this capacity into 130 channels,
their bandwidth must average 4.6 megabits per
second. Total bandwidth can be split between high
denition channels (at roughly 12 Mbps each) and
standard denition channels (4 Mbps each), but
most of the 130 will have to be standard denition.
Or, they could provide half as many channels,
and double the average bandwidth, or any other
combination as long as the total is 600 Mbps. For
example, CSPAN or weather could be given 2
Mbps, while a sports channel could receive 16.
It appears that most TV carriers have chosen to
go for lower bandwidth per channel and more
channels. Almost no broadcasts are close to the
full resolution 1080i that many TV sets are now
capable of receiving.
In fact, channels advertised
as “HDTV” are sometimes so compressed that the
pictures are far below the theoretical capability of
the TV set.
The same issue comes up for broadcast
stations, which are each given the use of 16 Mbps
of bandwidth, and typically divide it into two or
three different channels.
Assuming that this accurately reects what
TV viewers want, this tells us that American
consumers generally prefer variety (more choices)
over sheer visual quality. But one result has been
the very slow growth in average bytes per hour
bandwidth. Presumably over the next
ten years the mass migration to HDTV-capable
sets will gradually lead to an increase in average
bandwidth and information consumption. It’s not
clear how quickly carriers, networks and display
manufacturers will give consumers the full HD
experience that many consumers assume they are
already getting.
4.3 Analyzing Information
We have discussed each medium of information
in turn, using three different measures (hours,
compressed bytes, and words), and a range of
reference points including percentages, yearly
totals, and daily consumption. Appendix B provides
much of the underlying detail, from which the
summary numbers were drawn. (However, it does
not include details of calculations for the more
complex topics, such as computer games.)
As Figure 10 Shares of Information in Different
Formats illustrates, INFO
bytes are completely
dominated by video sources: movies, TV, and
computer games. Consumption time, INFO
on the
Figure 10: Shares of Information in Dierent Formats
Per Average American, Per Day
0% 20% 40% 60% 80% 100%
Video Audio Text
Figure 11: Contrasting Measurements
and INFO
All TV
How Much Information? 2009 Report on American Consumers
other hand, is primarily used for video and audio
(radio, telephone, and recorded music). Words,
nally, come heavily from text sources (newspapers,
magazine, books, and Internet use).
Figure 11 Contrasting Measurements shows in
more detail how different media dominate each
measure of information. Only television is a large
contributor to all three measures.
Table 9 Summary of Information for Major
Groups aggregates the information in Appendix B
by major categories, such as television and print.
4.3.1 How Much Information is
Delivered via the Internet?
Another question we investigated is the quantitative
importance of the Internet: how much does it
contribute to our information consumption?
Our basic nding is that the Internet provides a
substantial portion of some kinds of information,
but very little of others. Measuring with hours or
words, the Internet provided a signicant fraction
of our information, although less than television.
(Figure 12 Internet as a Source of Information)
We spent 16 percent of our information hours using
the Internet (versus 41 percent for TV), and receive
25 percent of our words INFO
from it (versus
45 percent from TV). The Internet was the source
of only 2 percent of our INFO
bytes (versus 35
percent for TV).
Yet surveys show that many of us view the Internet
as very important, to the extent that we will cut
spending on cable TV before we cut Internet access.
How can this importance be reconciled with its
smaller quantitative measurements? Our analysis
explains why the unique properties of the Internet
make it considerably more useful per byte or word
of information for certain purposes.
We classify our information consumption into
three mutually exclusive purposes: two-way
communication, entertainment, and research/
current events. Two-way communication is self
explanatory. Before the Internet, the only ways to
have a two-way exchange without being in the same
room were telephone and rst-class letters. The
Figure 12: Internet as a Source of Information
41.6% 34.7% 44.8%
15.6% 1.83% 24.7%
Table 9: Summary of Information for Major Groups
Total per year
(entire population) % of Total
Per average American
per day
in hours
in bytes
in words % Hrs % Bytes % words Hours
bytes Words
All TV 5.30E+11 1.27E+21 4.86E+15 41.62% 34.77% 44.85% 4.91 11.75 45,100
Radio 2.39E+11 1.10E+19 1.15E+15 18.79% 0.30% 10.59% 2.22 0.10 10,645
Phone 7.89E+10 1.36E+18 5.68E+14 6.20% 0.04% 5.24% 0.73 0.01 5,269
Print 6.49E+10 6.72E+17 9.34E+14 5.09% 0.02% 8.61% 0.60 0.01 8,659
Computer 2.08E+11 8.69E+18 2.93E+15 16.35% 0.24% 26.97% 1.93 0.08 27,122
Computer Games 1.00E+11 1.99E+21 2.65E+14 7.86% 54.62% 2.44% 0.93 18.46 2,459
Movies 3.24E+09 3.56E+20 2.14E+13 0.25% 9.78% 0.20% 0.03 3.30 198
Recorded music 4.88E+10 8.85E+18 1.20E+14 3.83% 0.24% 1.11% 0.45 0.08 1,112
TOTALS 1.27E+12 3.64E+21 1.08E+16 100.00% 100.00% 100.00% 11.80 33.80 100,564
5.3E+11 means 5.3 x 10
= 530,000,000,000
How Much Information? 2009 Report on American Consumers
Internet adds multiple additional methods, including
email, social networking, and instant messaging.
We estimate that Americans averaged 1.6 hours
per day conducting two-way communication, of
which 57 percent was via the Internet, with the
rest of the time on cellular or landline telephones.
Correspondingly, the Internet provides 79 percent
of the bytes and 73 percent of the words in two-
way communication. The Internet is so important
for two-way communications because of its
unique technical characteristics, including a nearly
universal network, very low variable costs, and the
ability to handle both real-time and delayed activity.
The other uses we classied information into were
entertainment and research/current events, by
which we mean gathering factual information of
any kind – basically any non-ction information,
to distinguish it from entertainment. We calculate
that Americans average 6.5 hours per day on
entertainment and 3.7 hours on research/current
The Internet’s contribution to pure entertainment
information is very small: less than 2 percent,
whether measured by hours, bytes, or words. The
reasons stem from entertainment’s dominance by
video activities: TV shows, movies, and computer
games. Video requires very high bandwidth,
and Internet speed to most Americans is still
far below what is needed to watch conventional
live television. A standard TV program requires
approximately 4 megabits per second of bandwidth,
while most Internet connections can deliver only
a fraction of that or less at peak times. Broadband
providers in many areas do offer premium-priced
service levels, but the speed is not sufcient for
live TV, for several reasons. Even when the “last
mile” to a house is capable of adequate speeds,
this is based on statistical multiplexing, meaning
that it assumes that only a fraction of users will be
operating at this speed at the same time. If everyone
turned on their “Internet TV” at 7pm, many parts
of the network would be unable to handle the load.
On the other hand, video on the Internet is growing
rapidly. The popularity of video download sites
indicates that demand exists, even with lower visual
quality than standard television.
In our third and nal use category, research and
current events, the Internet provides 23 percent of
our hours and 31 percent of our INFO
. It connects
to vast amounts of factual information, making it
very good for current events that can be delivered
in the form of text. We classify about one third
of television programming as research or current
events (including not only news but also reality
shows, talk shows, and the like), so television
dominates the total bytes in this category. Given the
much higher bandwidth of TV, the Internet provides
only 1.3 percent of our research/current event bytes.
4.3.2 The Rise of Interaction
Most sources of information in the past were
consumed passively. Listening to music on the
radio, for example, does not require any interaction
beyond selecting a channel, nor any attention
thereafter. Telephone calls were the only interactive
form of information, and they are only 5 percent of
words and a negligible fraction of bytes. However,
the arrival of home computers has dramatically
changed this as computer games are highly
interactive. Most home computer programs (such as
writing or working with user generated content) are
as well. Arguably, web use is also highly interactive,
with multiple decisions each minute about what to
click on next.
As a result, we estimate that a full third of our
in words is now received interactively,
and 55 percent of our INFO
bytes. This is an
overwhelming transformation, and it is not
surprising if it causes some cognitive changes.
These changes may not all be good, but they will be
On the other hand, we are only measuring
articial forms of information. For most of human
evolution, we spent most of our days interacting
with our environment and with each other, without
articial assistance. In fact, if we include “personal
conversation” as a source of information, it is
possible that we receive fewer bytes INFO
our ancestors did 100 years ago. The reason is
that conversation is very “high bandwidth.” A
full delity video link between two locations,
including stereo vision and sound is not possible
with present technology – the observer will realize
they are not physically in the location. If we could
do it, however, it would require conservatively 100
million bits per second. Three hours of personal
conversation a day at this bandwidth would be 135
gigabytes of INFO
, about four times the average
daily consumption today.
4.4 The Future of Consumer
There are some patterns of information consumption
in the rst half-decade of the twenty-rst century
that may be considerably changed by 2015. The
signicance of these changes, however, is not clear
and may not become clear for some time.
How Much Information? 2009 Report on American Consumers
Perhaps the most visible is shifts in television. We
have already discussed rapid changes in the delivery
of television from 2005 to today, including the shift
to digital broadcasting, the mass acceptance of high
denition TV sets (although not high denition
programming), and digital video recorders becoming a
mass-market product. On the other hand, the number of
TV channels has grown steadily for 50 years, and actual
video quality has not grown nearly as fast as a simplistic
theory of technological progress (Moore’s Law)
seemingly predicted.
Two nascent developments might also cause signicant
dislocations: mobile television and video over the
Internet. So far, mobile TV has low utilization and is
very much a niche product. On the other hand, video by
Internet is quite widespread, but as a complement rather
than a substitute for conventional TV program delivery
mechanisms. YouTube and its cousins have made a huge
variety of novel and specialized video material available
to anyone with a mediocre broadband connection. But
at least in the US, the quality of video over the Internet
is far below what is available by more “conventional”
means such as cable TV. The reason again is basically
bandwidth constraints. A minimal standard denition TV
signal requires 4 megabits per second, and a “medium”
version of so-called high-denition TV requires double or
triple that. The result is that Internet videos are generally
small, or grainy, or downloaded gradually rather than
streamed. If and when a substantial number of Americans
are able to receive streaming video at sustained speeds of
roughly 10 megabits per second and low latency, it may
dramatically alter the way they receive video. Internet-
based television, rather than being reserved for material
where low quality is compensated for by a very wide
selection (the “long tail effect”) might become common
for mainstream programming as well.
Beyond television, computer games will be an area for
growth of consumption INFO
. The performance of
GPUs follows Moore’s Law, and will continue to do so.
In consequence, game-playing enthusiasts will consume
rapidly increasing numbers of exabytes. Casual gamers
have shown little interest in high-resolution graphics so
far. But at least for a few years, rapid growth in consoles
and high-end computers will drive faster growth in
Consumption of words and hours, INFO
and INFO
are destined to continue their slow growth. They are
contrained by human physical limits, including the length
of a day and reading speed. Their growth will never
exceed a few percent per year.
How Much Information? 2009 Report on American Consumers
Appendix A: UC Berkeley HMI? Studies
How Much Information? 2009 follows two University of California, Berkeley research reports, HMI? 2000 and HMI? 2003,
conducted by Professors Peter Lyman and Hal Varian. HMI? 2009 builds on the two Berkeley studies, but there are important
differences. First, Lyman and Varian report on World and U.S. information totals for calendar years 1999 and 2002. HMI? 2009
reports only on the U.S. for calendar year 2008. Second, the two studies measure information differently and use different methods
to count it. Lyman and Varian measured “original” information - that is, the rst instance of new information being created, such as
a voice telephone call, or someone composing an email. They analyzed the quantity of original content created – how many hours of
radio broadcasting were produced worldwide, how many books were published, and so on. HMI? 2009 dened information as ows
of data delivered to people. We measured the amount of information delivered to people for consumption. Our contrasting denitions
led to differences in calculating information totals, and later we work through two examples to illustrate the importance of these
differences. Third, Lyman and Varian divided total information into two measures and reported them separately - the rst, the annual
size of the “stock” of new information contained in storage media; the second, the volume of information seen or heard each year in
information ows. We measured information consumption as the number of hours information was received by people (INFO
), the
number of bytes delivered (INFOc), and the number of words consumed (INFOw). We reported annual totals for each measure.
We also consulted industry sources, including two reports on digital information growth completed by the International Data
Corporation (IDC) published in 2007 and 2008.
HMI? 2000
The rst UC Berkeley report estimated that in 1999 the world produced between 2 and 3 exabytes of new information, or roughly 500
megabytes for every man, woman, and child (we dene an exabyte of information elsewhere in this report).
Lyman and Varian identied three key conclusions in summarizing their 2000 report:
• First, they referred to the “paucity of print.” Printed materials of all kinds made up less than .003 percent of the total amount of
annual information produced in the world. They cautioned that this number did not mean print was insignicant. On the contrary, they
noted it simply meant the written word was a very efcient way to convey information.
• Second, they referred to a growing “democratization of data” – the fact that a vast amount of new information is created and stored
by individuals. For example, original documents created by ofce workers were more than 80% of all original paper documents (the
other 20% included original copies of newspapers, books, magazines, and other print material). And photographs taken by consumers
and X-rays together were 99% of all original lm documents.
• Third, they noted the increasing “dominance of digital” content. Not only was digital information production the largest in total, it
was also the most rapidly growing. They concluded that while unique content produced on print and lm was hardly growing at all,
magnetic storage was by far the largest medium for storing information and was the most rapidly growing medium, with shipped hard
drive capacity doubling every year.
HMI? 2003
In 2003, Lyman and Varian extended their earlier study. They added a new section on the Internet, sampling the World Wide Web to
estimate the size of the surface web and to determine the source and content of Web pages. And they added an analysis of desktop disk
drives, to determine how people consumed information received on the Internet. They concluded:
• Print, lm, magnetic, and optical storage media produced about 5 exabytes of new information worldwide in 2002. Ninety-two
percent of the new information was stored on magnetic media, mostly on hard disks.
• Information owing through electronic channels – telephone, radio, TV and the Internet – contained almost 18 exabytes of new
information worldwide in 2002, three and a half times more than was recorded on storage media. Ninety eight percent of this total was
the information sent and received in telephone calls – including both voice and data on both xed line and wireless phones.
• They estimated that the total amount of new information stored annually on paper, lm, magnetic, and optical media worldwide had
doubled in the last three years.
Lyman and Varian drew a number of implications from their 2003 study. Perhaps most important, they noted that our ability to store
and communicate information was far outpacing our ability to search, retrieve and present it.
How Much Information? 2009 Report on American Consumers
Comparing HMI? 2009 with HMI? 2003 and 2000
As noted, our contrasting denitions and measures produce different annual information totals, and these totals are not directly
comparable. For example, in Television and Radio we calculated total annual information in the U.S. (INFOc) was 1,277 exabytes per
year. Lyman and Varian’s total for the U.S. was seven one-hundreds of an exabyte. Why? They counted a television (or radio) program
once, the rst time it was aired. We counted every time a television viewer watches a program, which could be 20 million people.
Here is how each respective total was calculated: Berkeley estimated that in 2002 there were approximately 3.6 million hours of
original information broadcast by U.S. television stations, and 19.8 million hours of original information broadcast by U.S. radio
stations (reported in their Table 1.11). Using a conversion factor of 1.3 GB to 2.25 GB per hour for television, and 0.05 GB per hour
for radio, they calculated total U.S. Television and Radio information was between a lower bound estimate of 5,718 terabytes and an
upper bound estimate of 9,175 terabytes, of new information in 2002. In HMI? 2009, we counted the number of television viewers
(292 million people), the amount of time they view television (on average 148.5 hours a month), and calculated 1,197 exabytes of
data was delivered to their television screens that year. Adding in DVD players and Radio brought the total to 1,277 exabytes (Table 3
Television and Radio Consumption). We also contrast Telephone information, where a more thorough explanation is necessary, in an
extended endnote.
International Data Corporation (IDC) 2007 and 2008
International Data Corporation (IDC) published two reports on the growth of digital data in 2007 and 2008. IDC’s denition of
digital information and their methods for counting it were not explained in sufcient detail to reliably compare their totals with the
HMI? 2000 and 2003 reports, or HMI? 2009. IDC’s numbers for the entire world were 12 times less than our 2009 numbers for
U.S. households alone. But it is not clear whether the large discrepancy was due to our including more types of information sources
(such as non-Internet computer use and game consoles), our inclusion of analog as well as digital sources, our different approach to
measuring bytes, or for other reasons.
The main conclusions of IDC’s 2008 report include:
• The amount of digital data created in 2007 was 281 billion gigabytes (281 exabytes), equivalent to 45 gigabytes per capita, roughly
the size of a Blu-Ray disc. (The maximum capacity of the new Blu-ray HD format is 50 GB on a dual-layer disc.)
• Digital data was projected to grow at a compound annual growth rate of almost 60%, reaching 1.8 zettabytes (1,800 exabytes) by
• More than 80 percent of bytes are images: pictures, surveillance videos, TV streams, and so forth.
• Individuals’ “Digital Shadows” – information generated as a result of activities such as web surng and shopping, but not by them
directly – surpasses the amount of digital information individuals create themselves.
Peter Lyman and Hal R. Varian, How Much Information, 2000.
Peter Lyman and Hal R. Varian, How Much Information, 2003.
John F. Gantz, et. al., The Diverse and Exploding Digital Universe: An Updated Forecast of Worldwide Information Growth Through
2011, IDC (March 2008).
32 33
Appendix B: Detail Table
Users Throughput
Total per year
(entire population)
Per User / Per Day
Per average
American / per day
% of Total
# of Users
bits per sec.
(bps) - comp.
per minute
Hours (billion)
Words (trillion)
Hours Megabytes Words Hours Gigabytes Words % Hrs % Bytes % words
Cable TV - SD 95.7 4,000,000 153 163.0 293.0 1,493 4.66 8,380.0 42,740 1.51 2.71 13,843 12.8% 8.0% 13.8%
Cable TV - HD* 69.3 7,200,000 153 118.0 382.0 1,081 4.66 15,085.0 42,740 1.09 3.54 10,024 9.3% 10.5% 10.0%
Over air TV - SD 27.8 4,000,000 153 47.0 85.0 434 4.66 8,380.0 42,740 0.44 0.79 4,027 3.7% 2.3% 4.0%
Over air TV - HD * 20.2 7,200,000 153 34.0 111.0 314 4.66 15,085.0 42,740 0.32 1.03 2,916 2.7% 3.0% 2.9%
Satellite - SD 45.5 4,000,000 153 77.0 139.0 710 4.66 8,380.0 42,740 0.72 1.29 6,586 6.1% 3.8% 6.5%
Satellite - HD* 33.0 7,200,000 153 56.0 182.0 514 4.66 15,085.0 42,740 0.52 1.68 4,769 4.4% 5.0% 4.7%
DVD 253.8 5,500,000 153 28.0 70.0 258 0.30 751.0 2,787 0.26 0.65 2,394 2.2% 1.9% 2.4%
Other TV (delayed view) 50.0 3,000,000 153 3.9 5.3 36 0.21 289.0 1,966 0.036 0.05 333 0.31% 0.14% 0.33%
Mobile video 10.3 300,000 153 0.4 0.1 4.1 0.12 16.0 1,089 0.004 0.00 38 0.03% 0.002% 0.04%
Internet video 94.7 1,000,000 153 2.0 0.9 18 0.06 26.0 527 0.018 0.01 169 0.16% 0.024% 0.17%
Newspapers 51.2 18,235 240 9.0 0.4 124 0.46 3.8 6,628 0.080 0.00 1,149 0.68% 0.011% 1.14%
Magazines 250.0 18,000 240 29.0 0.2 421 0.32 2.6 4,616 0.27 0.00 3,906 2.3% 0.007% 3.9%
Books 250.0 1,330 240 27.0 0.0 389 0.30 0.2 4,261 0.25 0.00 3,605 2.1% 0.000% 3.6%
Satellite Radio 18.9 192,000 80 15.0 1.3 71 2.16 186.0 10,354 0.14 0.01 662 1.2% 0.035% 0.66%
AM & FM Radio 232.5 96,000 80 224.0 10.0 1,077 2.64 114.0 12,686 2.08 0.09 9,982 17.6% 0.27% 9.9%
Conventional Telephone (POTS) 154.0 64,000 120 41.0 1.2 299 0.74 21.0 5,311 0.38 0.01 2,768 3.3% 0.033% 2.8%
Cellular Voice 263.0 10,000 120 37.0 0.2 270 0.39 1.8 2,809 0.35 0.00 2,501 2.9% 0.005% 2.5%
High-end Computer gaming** 20.8 Varies 50 22.0 1,405.0 65 2.85 185,100.0 8,548 0.20 13.03 602 1.7% 38.6% 0.60%
Computer gaming** 123.7 Varies 50 27.0 194.0 80 0.59 4,299.0 1,777 0.25 1.80 744 2.1% 5.3% 0.74%
Console gaming** 88.8 Varies 50 32.0 368.0 97 0.99 11,349.0 2,980 0.30 3.41 896 2.5% 10.1% 0.89%
Handheld gaming** 128.9 Varies 20 20.0 24.0 23 0.41 500.0 497 0.18 0.22 217 1.5% 0.64% 0.22%
Internet text (email, web, etc.) 226.3 100,000 240 178.0 8.0 2,564 2.16 97.0 31,032 1.65 0.07 23,771 14.0% 0.22% 23.60%
Ofine programs 226.3 50,000 200 30.0 0.7 361 0.36 8.0 4,375 0.28 0.01 3,352 2.4% 0.019% 3.3%
Movies 295.5 244,737,638 110 3.2 356.0 21 0.03 3,304.0 198 0.03 3.30 198 0.25% 9.8% 0.20%
Recorded Music inc. MP3
295.5 403,200 41 49.0 9.0 120 0.45 82.0 1,112 0.45 0.08 1,112 3.8% 0.24% 1.11%
MASTER SUM 1,273 3,645 10,845 11.80 33.80 100,564 100.0% 100.0% 100.0%
* HD numbers are a blend of High Denition and Standard Denition use in HD households.
**Computer gaming users and bandwidths are averages from more detailed calculations.
All our numbers are estimates - see the on-line appendix and the endnotes for more information about data sources and methods.
How Much Information? 2009 Report on American Consumers
A 40-hour per week job is 22 percent of a year. Slightly less than
half of the US population is employed. Therefore an “average person”
is at work 2.7 hours per day. Source: Bureau of Labor Statistics 2008. <>.
HMI? 2009 draws on an unusually large number of data sources
from university research, government and industry. Reconciling the
many differences in denitions, sample populations and measurement
approaches has been a major preoccupation of the research team,
especially where sample populations may vary in age or other
demographic characteristics, or where double-counting could take place
in cases where multiple measurements have been taken of the same
population. We have done the best we can in isolating such cases and
accounting for them. We have also consulted other large-scale media
studies facing the same methodological challenges, for example, the
Video Consumer Mapping (VCM) Study conducted by the Council
for Research Excellence and Ball State University’s Center for Media
Design (CMD). <
Teenage viewing is analyzed in Nielsen, How Teens Use Media:
A Nielsen report on the myths and realities of teen media trends, June
2009. Statistics for various age groups are from The Council for
Research Excellence, Video Consumer Mapping Study: Appendix -
Additional Findings & Presentation Materials, June 2009.
In 1960, transistors were used only in a few applications, including
some computers and a new kind of consumer electronics, “portable
radios.” Integrated circuits were not even invented until later in the
We have adjusted Pool’s numbers for some differences in
Analog integrated circuits are also very important, but even devices
with analog circuitry such as radios generally are controlled by digital
Standard Denition TV (SDTV).
Lincoln’s salary at the time was $25,000 per year, or about $8 an
hour. The salary today is $400,000, or about $200 an hour. < http://www.>.
Nielsen, A2/M2 Three Screen Report, January 2009. Based on data
collected in 4Q 2008, Nielsen reported U.S. viewers watched an average
of 151 hours per month. This number probably has some seasonality in
Over the air analog (NTSC) television is not compressed. NTSC
is an analog color TV standard developed in the U.S. in 1953 by
the National Television System Committee. Television signals that
are compressed and then uncompressed for viewing are MPEG-2 or
higher. MPEG-2 is a standard for the coding of moving pictures and
associated audio information. It describes a combination of lossy video
compression and lossy audio data compression methods.
Further adding to the confusion, in 2009 all US broadcasters shifted
from analog to digital broadcasting. Some cable companies and most
satellite broadcasters made the shift years before, but there are still some
cable signals that are analog. In any case, digital TV signals can have
a number of different resolutions, so whether a show is high denition
does not depend on whether it is broadcast in digital or analog.
Bill Carter, “DVR, Once TV’s Mortal Foe, Helps Ratings,” New
York Times 1 November 2009.
Our source for radio data is Arbitron Inc. We reviewed Arbitron’s
Radio Today: How America Listens to Radio, 2007, 2008 and 2009
Editions, and Arbitron Radio Listening Report, The Innite Dial 2008:
Radio’s Digital Platforms Online, Satellite, HD Radio and Podcasting.
Arbitron reports AQH (average quarter hour) listenership by location.
For the most popular radio formats, for example News/Talk/Information,
at work listener share was 12.8 percent in 2008, and for other popular
formats, averages 20 percent or less. We did not deate average listening
hours by format by location in our estimates.
Our cellular and xed line telephone numbers include both
residential and business lines. Additionally, our xed line number
also includes most Voice over IP (supplied by the cable TV
companies). Voice over IP (also referred to as VoIP, IP telephony,
and Internet telephony) refers to technology that enables routing
of voice conversations over the Internet or a computer network.
Sources: CTIA-The Wireless Association; Federal Communications
Commission, Local Telephone Competition: Status as of December 31,
2007, Industry Analysis & Technology Division, Wireline Competition
Bureau, September 2008; Federal Communications Commission, Local
Telephone Competition: Status as of June 30, 2008, Industry Analysis &
Technology Division, Wireline Competition Bureau, July 2009.
The 154 million gure includes residential and business lines and
most VoIP (which is supplied by the cable TV companies). What it
does not include is “over the top” VoIP like Vonage and Skype. Our
telephone numbers, therefore, may overstate consumer information by
approximately 30 percent. Estimates differ as to the number of VoIP
subscriber lines in 2008. By the end of 2008, the top 10 ISPs (Internet
Service Providers) had approximately 19.6 million residential customers
in the US. If we use xed line usage as an approximation for VoIP usage,
VoIP subscribers would have spent approximately 5.3 billion hours
making VoIP telephone calls in 2008. Our estimate was calculated by
adding up the total number of VoIP customers listed in annual reports
and in SEC disclosures by the top ten ISPs providing VoIP services in
2008. We have not included these calculations in our voice telephony
information totals. We also do not include international calls. Sources:
Customer data obtained from SEC 10-K and 10-Q disclosures and
Annual Reports for Comcast, Time Warner, Vonage, Cox, CableVision,
Charter, Insight Communications, Mediacom, SureWest and CBeyond.
Industry sources included, ISP-Planet, BusinessWire
and information services companies including Pike and Fischer, Nielsen,
TeleGeography, iLocus and In-Stat.
There is no easy way to rate the “bits per second” of lm. For
example, lm resolution is measured in a different way than video –
lines per inch, rather than pixels. And the quality of 35 mm lms actually
shown in theaters degrades over time as the negatives get scratched.
Even when rst shown, theater lms are usually third generation copies
of the original negative, and since the reproduction process is analog,
resolution is lost from the original. See Vittorio Baroncini, Henry Mahler
and Matthieu Sintas, The Image Resolution of 35mm Cinema Film in
Theatrical Presentation, for details of a human-observer study of lm
resolution. Even different brands of lm differ.
Since 1998, American households went from less than 10 percent of
homes owning a personal computer, to over 70 percent of homes having
How Much Information? 2009 Report on American Consumers
personal computers wired with Internet access. In High Denition
television, HD ownership has doubled in the last two years - a quarter
of all US households owned HD in 2007, to just under 50 percent of
American homes in 2008. In the ubiquitous cell phone market, sales of
smartphones such as Apple’s iPhone were over 20 percent of all new
handset sales in the US in 2008, up from 12 percent in 2007. Sources:
U.S. Census, Computer and Internet Use in the United States: 2003,
October 2005; Nielsen Wire, Household TV Trends Holding Steady:
Nielsen’s Economic Study 2008, 24 February 2009; ComScore, Key
Trends in Mobile Content Usage & Mobile Advertising, 12 February
Microsoft Email productivity consultants state that effective email
users can view and handle 30 percent of their incoming email box in 2
minutes, based on Microsoft Productivity Study (MPS) Statistics. MPS
statistics show that on average, people can process up to 60 e-mail
messages an hour, where “process” means to complete the full action
necessary (not just scan/read – the full sequence is read, respond,
assign, delay, or delete). Sources:<http://of
help/HA011464801033.aspx>; <
manageinfo/email.mspx> ; <
Studies of web behavior and navigation nd high variability of
document display and view time. For example, Weinreich et. al. report:
“Our data conrms the rapid interaction behavior with heavy tailed
distributions already reported in previous studies… participants stayed
only for a short period on most pages. 25 percent of all documents
were displayed for less than 4 seconds, and 52 percent of all visits were
shorter than 10 seconds (median: 9.4s). However, nearly 10 percent
of the page visits were longer than two minutes. Figure 4 shows the
distribution of stay times grouped in intervals of one second. The peak
value of the average stay times is located between 2 and 3 seconds;
these stay times contribute 8.6 percent of all visits.” See Weinreich et
al., “Not Quite the Average: An Empirical Study of Web Use,” ACM
Transactions on the Web 2, no. 1 (2008): p. 5:18 <ISSN:1559-1131>.
Worldwide user total reported by Facebook at <http://www.>.
Average daily time use reported by Silverbean, “Mobile users visit
Facebook ‘3 times per day’ - 18th February 2009,” Online Marketing
News <
There are many technology factors affecting the average download
speed to a home, including backbone network speed, access connection
speed, web server speed, the home network itself, and physical factors
such as inside wiring.
Depending on who is counting, Hulu had either 9 million or 42
million viewers in May 2009. See Brian Stelter, “Hulu Questions Count
of Its Audience,” New York Times 14 May 2009. <http://www.nytimes.
This is with an assumed average video download speed of 1 Mbps.
YouTube’s number of unique visitors grew nine-fold between March
of 2006 and March of 2007, and video page-views grew at a rate of
25 times over the same period. In July of 2008, YouTube reported 72
million unique visitors to its US site, 4.7 billion page-views per month,
and hundreds of millions of videos viewed daily. Source: YouTube. You
Must Know – July 2008.
United States National Gamers Survey 2009 available at <http://>.
Anita Frazier, The Games People Play, NPD Group, July 2008.
We analyzed computer games in more detail than is reported here.
We used a total of 12 categories, which we have summarized down to 4
categories in our tables. For example, our fastest computer runs a screen
resolution of 2080 by 1024 at 60 frames per second. This is based on
data from and other computer game sources. For a low-end
laptop, we estimated 800 by 600 at 15 frames per second.
There are no estimates of bandwidths for high-end computer games,
and our estimates are therefore plus or minus 25 percent.
See for example John Horrigan, Home Broadband Adoption 2009,
Pew Internet & American Life Project. Available at <http://pewinternet.
Chuan-Fong Shih and Alladi Venkatash, “A Comparative Study of
Home Computer Use in Three Countries: U.S., Sweden, and India,”
Center for Research on Information Technology and Organizations,
University of California, Irvine, Paper 378, 2003; Alladi Venkatesh,
“Smart Home Concepts: Current Trends,” Center for Research on
Information Technology and Organizations, University of California,
Irvine, Paper 377, 2003.
We reviewed multiple sources for data on mobile Internet, text
messaging, and mobile gaming use. For mobile Internet, the Council
for Research Excellence (CRE) reported mobile web use of 1 minute
per day for the average media consumer in 2008. ComScore M:Metrics
reported the average U.S. smartphone user spent 4.6 hours per month
browsing the mobile web. When we calculated total annual hours for
the U.S. population, we obtained on the order of 2 billion hours for
2008. We therefore did not include this category. Sources: Council for
Research Excellence (CRE), A Day in the Media Life: Some Findings
from the Video Consumer Mapping Study, April 3, 2009. ComScore
M:Metrics, “Americans Spend More Than 4.5 Hours Per Month
Browsing on Smartphones, Nearly Double the Rate of the British,”
ComScore Press Release, 21 May 2008.
For mobile gaming, our primary data sources did not break out
gaming on smartphones from gaming on dedicated handheld devices.
We also reviewed secondary sources on mobile gaming for 2008. All
indications are that it is quite small.
Luke Simpson, “Smartphones vs Feature Phones: What’s the
Difference?” WirelessWeek, February 28, 2009 <http://www.
We estimate that Americans spent 7 billion hours text messaging
in 2008. We calculated this amount as follows: Nielsen reported that in
mid 2008, the average US mobile customer sent or received 357 text
messages a month. Assuming that each text message is sent or received
in 30 seconds, and that approximately 200 million American cell phone
users subscribed to or paid for text messaging service, multiplying the
number of users (200M) by the number of messages (357 per month)
by the average time per message (30 seconds) works out to an estimated
7.14 billion hours for the year. Because SMS text messages are so small,
the byte totals are insignicant. Sources: Nielsen Wire, “In U.S., SMS
Text Messaging Tops Mobile Phone Calling,” Insights, 22 September
2008; Nielsen Telecom Practice Group, “Flying Fingers: Text-messaging
How Much Information? 2009 Report on American Consumers
overtakes monthly phone calls,” Insights, November 2008.
For an analysis of storage costs over time see E. Grochowski and
R. D. Halem, “Technological impact of magnetic hard disk drives on
storage systems,” IBM Systems Journal 42, No 2, 2003. < http://www. >.
William D. Nordhaus, “Two Centuries of Productivity Growth in
Computing,” The Journal of Economic History 67, No.1, March 2007.
(Tables 5 and 6).
This observation was rst made by Ithiel de Sola Pool, and
was studied in 2005 by Russell Neuman and colleagues. See W.
Russell Neuman, Yong Jin Park and Elliot Panek, “Tracking the
Flow of Information into the Home: An Empirical Assessment of
the Digital Revolution in the U.S. from 1960 – 2005,” International
Communications Association Annual Conference, Chicago, IL. 2009.
Robert N. Charette, “This Car Runs on Code,” IEEE Spectrum,
February 2009. Available at <>.
For a description of airbags and how they are activated, see
“Inside the Toyota Prius: Part 1 - The airbag control module,”
Automotive DesignLine, 16 April 2007. Available at <http://www.>.
100 Hz x 5000 hours of life x 3600 sec/hour = 1.8E+9 = 1.8
Nielsen’s 2008 Television Audience Report states that the average
U.S. television household received a total of 130.1 station channels as
tuning options that year (the total includes digital cable and satellite
channels, and 17.7 channels of over the air broadcast). Growing digital
cable and satellite penetration has increased the tuning options for the
average household. In 2006 the average total available was 104 channels.
In 2008 the average household actually watched 18 channels or
approximately 14 percent of the total station channels available. Source:
Nielsen, 2008 Television Audience Report. Available at
“Resolution” is more than the number of pixels. It includes frames
per second, and the degree of compression. A program can theoretically
be 1080i, but still be so heavily compressed that it is no more attractive
visually than a standard denition (480i) program.
Hard data on this topic is, probably not surprisingly, hard to come by.
Lyman and Varian’s original estimate was between 1 and 2 exabytes
of information, published in their 2000 report. In their 2003 report, they
updated their earlier estimate to 2 to 3 exabytes of information.
Our total for annual xed line voice information in the U.S. is
similar to the totals reported by Lyman and Varian in HMI? 2003.
Our approaches were different, however. Lyman and Varian asked
the question, how much storage would be needed to store all of the
xed line voice calls taking place in the U.S. in 2002? They reported
two answers: 9.25 exabytes of uncompressed storage; and 1.2 to 1.5
exabytes of compressed storage (assuming compression would reduce
storage requirements by a factor of 6 to 8). To calculate these totals,
they consulted Federal Communications Commission (FCC) sources
reporting the number of xed wirelines in the U.S. (approximately 190
million in 2002), and the total number of minutes of use of these lines
(4,819 billion DEMS, or Dial Equipment Minutes). Dial Equipment
Minutes (DEMS) are measured by telephone switching equipment as
“calls enter and leave telephone switches so two dial equipment minutes
are recorded for every conversation minute.” As Lyman and Varian
counted the production of original information, they were interested
only in measuring the time of the phone call itself, not how many callers
were on the call. Therefore in using DEMS to estimate time usage, they
divided the dial equipment minutes in half. They then multiplied usage
time by a conversion factor they had previously dened for storing audio
information on storage media – 64,000 bytes per second (uncompressed).
Using these numbers and adjusting for the different units of measure,
they calculated total annual storage for voice trafc in the U.S. of 9.25
exabytes. They then ran a second calculation for compressed bytes,
noting that “compression could reduce storage requirements by a factor
of 6 to 8, resulting in a total of 1.2 to 1.5 exabytes.” Our methodology
has been to use wherever possible actual device usage time by people
in all of our calculations. Importantly, this led us to interpret DEM
data differently in estimating conversational minutes of use for phone
subscribers. In our approach, two people speaking on the telephone
for one hour is counted as two hours – one hour each for each phone
caller. Therefore, our information total was calculated as follows: we
relied on similar FCC documents as Lyman and Varian to estimate
the number of wireline subscribers in the U.S. in 2008 (154 million,
including residential, business and some VoIP). To estimate the time
usage of these lines, we reviewed studies conducted by the FCC on
household wireline penetration and conversational minutes of use. In
one study, “Recent developments in US wireline telecommunications,”
Paul Zimmerman, an FCC economist, reported that conversational use
of wireline phones averaged 900 minutes a month in 2003 (Zimmerman
reported this data in Figure 2 Average Monthly Wireline and Wireless
Usage by Year (1993-2003), and in Footnote 24, p. 430). We used
Zimmerman’s data to estimate an average use time of 22.5 hours per
month per subscriber (further details available upon request). We then
calculated our annual information total by multiplying the total number
of subscribers (154 million), by the average rate of use per subscriber
(22.5 hours per month), by the compressed throughput of wireline
telephone calls (64,000 bits per second, or 64 kbps). Adjusting for the
different units of measurement we calculated a total of 1.2 exabytes
a year of information (compressed) for xed line voice (see Table 4
Telephone Consumption). Interestingly, our total for compressed bytes
of annual information is similar to the totals calculated by Lyman and
Varian, but we arrived at them by different means. Sources: Federal
Communications Commission, Trends in Telephone Service, August
2003; Federal Communications Commission, Trends in Telephone
Service, August 2008; Federal Communications Commission, Trends in
Telephone Service, July 2009; Paul Zimmerman, “Recent developments
in US wireline telecommunications,” Telecommunications Policy 31
(2007), pp. 419-437; Paul Zimmerman, “Strategic incentives under
vertical integration: the case of wireline-afliated wireless carriers and
intermodal competition in the US,” Journal of Regulatory Economics 31
(2008), pp. 282–298.
Special Thank You
There are many people to thank for their involvement in and support of the
HMI? 2009 Report on American Consumers including:
Jonathan Aronson, USC
Kristen Brooke Backor, Stanford
Margaret E. Beck, University of Iowa
Vint Cerf, Google
Jaideep Chandrashekar, Intel
Richard Clarke, AT&T
Chris Cookson, Sony Pictures
Tom Coughlin, Coughlin Associates
Mary Czerwinski, Microsoft Research
Mahmoud F. Daneshmand, AT&T LABS
James Danziger, UC Irvine
Gary Delp
Shawn DuBravac, Consumer Electronics Association
Brian Dunne, UC San Diego
Lorraine Eakin, UNC
Blake Ellison, UC San Diego
Elaine Fleming, UC San Diego
Scot Hastings, Qualcomm
Ron Hawkins, San Diego Supercomputer Center
Xiaobin He, Stanford
William Huber, UC San Diego
Theresa Jackson, Orchard View Color
Richard Kowalski, Consumer Electronics Association
Xiaomei Liu, Cisco
John Longwell, Computer Economics
David Luebke, Nvidia
Gloria Mark, UC Irvine
Norman Nie, Stanford
Andrew M. Odlyzko, University of Minnesota
Ryan Pfeiffer, UC San Diego
Ellen Quackenbush
Amy Robinson, UC San Diego
Frank Scavo, Computer Economics
Jurgen Schulze, UC San Diego
Eve Schooler, Intel
Abigail Sellen, Microsoft Research
Jeff Smits, Intel
Alladi Venkatesh, UC Irvine
David Wasshausen, Department of Commerce
Morley Winograd, USC
Global Information Industry Center
UC San Diego
9500 Gilman Drive, Mail Code 0519
La Jolla, CA 92093-0519
... Based on extrapolations from a study done at University of California, Berkeley, it is estimated that the deep web contains approximately 91,850 terabytes and the surface web is only about 167 terabytes in 2003 [1]. More recent studies estimated that 1.9 zettabytes were reached and 0.3 zettabytes were consumed worldwide in 2007 [2], [3]. An IDC report estimates that the total of all digital data created, replicated, and consumed will reach 6 zettabytes in 2014 [4]. ...
Full-text available
As crawler performs deep web operation at a very fast pace, there has been increased interest in techniques that help efficiently locate deep-web interfaces. Because of large volume of web resources and the dynamic nature of deep web, achieving wide coverage and high efficiency is a challenging issue. Two overcome this a two-stage framework, namely advance Crawler is proposed, for efficient harvesting deep web interfaces. At initial stage, advance Crawler performs site-based searching for center pages with the help of search engines, avoiding visiting a large number of pages. To get more accurate results for a focused crawl, Advance Crawler ranks websites to prioritize highly relevant ones for a given topic. In the final stage, advance Crawler achieves fast in-site searching by excavating most relevant links with an adaptive link-ranking. To eliminate bias on visiting some highly relevant links in hidden web directories, a link tree data structure has been design to achieve wider coverage for a website. Our experimental results on a set of representative domains show the agility and accuracy of our proposed crawler framework, which efficiently retrieves deep-web interfaces from large-scale sites and achieves higher harvest rates than other ones.
... A modo de ejemplo, sólo en 2007 se transmitieron 1.9 zettabytes de información en el mundo [1], siguiendo una histórica tendencia donde cada vez se producen, transmiten y almacenan más datos. Otros estudios y publicaciones indican números crecientes para distintos años sucesivos [2,3], pero no es necesario ahondar en detalles para dejar en claro que las sociedades modernas producen enormes cantidades de información, ya sea en conjunto o de manera aislada. ...
... In the context of a single individual, it is not new that personal information (the input) exceeds the processing capacity of the system (the individual). It was reported that the average U.S. person spent 12 hours of leisure time per day consuming 100,500 words and 34 gigabytes (Bohn and Short 2009). This includes information consumed through mobile phones, the Internet, email, television, radio, newspapers, books, etc. ...
Typical Internet users today have their data scattered over several devices, applications, and services. Managing and controlling one's data is increasingly difficult. In this thesis, we adopt the viewpoint that the user should be given the means to gather and integrate her data, under her full control. In that direction, we designed a system that integrates and enriches the data of a user from multiple heterogeneous sources of personal information into an RDF knowledge base. The system is open-source and implements a novel, extensible framework that facilitates the integration of new data sources and the development of new modules for deriving knowledge. We first show how user activity can be inferred from smartphone sensor data. We introduce a time-based clustering algorithm to extract stay points from location history data. Using data from additional mobile phone sensors, geographic information from OpenStreetMap, and public transportation schedules, we introduce a transportation mode recognition algorithm to derive the different modes and routes taken by the user when traveling. The algorithm derives the itinerary followed by the user by finding the most likely sequence in a linear-chain conditional random field whose feature functions are based on the output of a neural network. We also show how the system can integrate information from the user's email messages, calendars, address books, social network services, and location history into a coherent whole. To do so, it uses entity resolution to find the set of avatars used by each real-world contact and performs spatiotemporal alignment to connect each stay point with the event it corresponds to in the user's calendar. Finally, we show that such a system can also be used for multi-device and multi-system synchronization and allow knowledge to be pushed to the sources. We present extensive experiments.
Full-text available
World Wide Web is developing rapidly; there are large number of Web databases available for users to access. This fast development of the World Wide Web has changed the way in which information is managed and accessed. So the Web can be divided into the Surface Web and the Deep Web. Surface Web refers to the Web pages that are static and linked to other pages, while Deep Web refers to the Web pages created dynamically as the result of specifi c search. In the same way the Tweets are being created as short text message. Tweets are shared for each users and knowledge analysts. Twitter that receives over four hundred million tweets per day has emerged as a useful supply of reports, blogs, and opinions and additional. In general, tweet summarization and third to observe and monitors the outline-based mostly and volume based variation to supply timeline mechanically from tweet stream. Implementing continuous tweet stream reducing a text document is but not an easy task, since an enormous range of tweets are paltry, unrelated and raucous in nature, because of the social nature of tweeting. However, due to the large volume of web resources and the dynamic nature of deep web, achieving wide coverage and high effi ciency is a challenging issue. In this paper, we have a tendency to introduce a unique summarization framework known as summarization (continuous summarization by stream clustering) and also propose a two-stage framework, namely Smart Crawler, for effi cient harvesting deep web interfaces. In the fi rst stage, Smart Crawler performs site based searching for center pages with the help of search engines, To achieve ranks websites to prioritize highly relevant ones for a given topic. In the second stage, Smart Crawler achieves fast in-site searching by excavating most relevant links with an adaptive link-ranking crawlers.
Recent studies have suggested that the stability of peer-to-peer networks may rely on persistent peers , who dwell on the network after they obtain the entire file. In the absence of such peers, one piece becomes extremely rare in the network, which leads to instability. Technological developments, however, are poised to reduce the incidence of persistent peers, giving rise to a need for a protocol that guarantees stability with non-persistent peers. We propose a novel peer-to-peer protocol, the group suppression protocol , to ensure the stability of peer-to-peer networks under the scenario that all the peers adopt non-persistent behavior. Using a suitable Lyapunov potential function, the group suppression protocol is proven to be stable when the file is broken into two pieces, and detailed experiments demonstrate the stability of the protocol for arbitrary number of pieces. We define and simulate a decentralized version of this protocol for practical applications. Straightforward incorporation of the group suppression protocol into BitTorrent while retaining most of BitTorrent’s core mechanisms is also presented. Subsequent simulations show that under certain assumptions, BitTorrent with the official protocol cannot escape from the missing piece syndrome, but BitTorrent with group suppression does.
How does language affect cognition? Is it important that most of our concepts come with linguistic labels, such as car or number? The statistical distributions of how such labels co-occur in language offers a rich medium of associative information that can support conceptual processing in a number of ways. In this article, I argue that the role of language in conceptual processing goes far beyond mere support, and that language is as fundamental and intrinsic a part of conceptual processing as sensorimotor-affective simulations. In particular, because linguistic association tends to be computationally cheaper than simulation (i.e. faster, less effortful, but still information-rich), it enables an heuristic mechanism that can provide adequate conceptual representation without the need to develop a detailed simulation. I review the evidence for this key mechanism – the linguistic shortcut – and propose that it allows labels to sometimes carry the burden of conceptual processing by acting in place of simulated referent meanings, according to context, available resources, and processing goals.
Full-text available
This paper presents a machine learning tool "HaarFilter" that can be used for analyzing data of JPEG/PNG image formats. It tries to solve the problem of big data phenomena of present age, produced as a side effect of image data generation in astronomical units.[8] We have tried to combine and harness the power of two rich open source softwares namely Hadoop and OpenCV[1] using Hadoop Image Processing Interface. "HaarFilter" is java based tool that will process and filter the required image data from the given storage location. Next, using Haar Cascading machine learning technique it will do the object detection for analysis work and stores the resultant images in Hadoop environment with specialized data format (HIB). Processing at such a large scale using a single machine can be very time consuming and costly. HaarFilter can add much a relief while processing and performing analytics with image data. It will facilitate efficient and high-throughput image processing with MapReduce style parallel programs typically executed on a cluster. It provides a solution for how to store a large collection of images on the Hadoop Distributed File System (HDFS) and make them available for efficient machine learning.[10]
Full-text available
This study analyzes the increasing dominance of electronic media in the American media diet and a growing discrepancy between supply and demand in the digital cornucopia. Drawing on the communication flow methodology pioneered by Ithiel Pool in the 1980s, the study tracks U.S. industry data on technology penetration and household behavior from 1960 to 2005 to reveal a transition from "push" to "pull" media dynamics.
Magnetic hard disk drives have undergone vast technological improvements since their introduction as storage devices over 45 years ago, and these improvements have had a marked influence on how disk drives are applied and what they can do. Areal density increases have exceeded the traditional semiconductor development trajectory and have yielded higher-capacity, higher-performance, and smaller-form-factor disk drives, enabling desktop and mobile computers to store multi-gigabytes of data easily. Server systems containing large numbers of drives have achieved unparalleled reliability, performance, and storage capacity. All of these characteristics have been achieved at rapidly declining disk costs. This paper relates advances in disk drives to corresponding trends in storage systems and projects where these trends may lead in the future.
The present study analyzes computer performance over the last century and a half. Three results stand out. First, there has been a phenomenal increase in computer power over the twentieth century. Depending upon the standard used, computer performance has improved since manual computing by a factor between 1.7 trillion and 76 trillion. Second, there was a major break in the trend around World War II. Third, this study develops estimates of the growth in computer power relying on performance rather than components; the price declines using performance-based measures are markedly larger than those reported in the official statistics.
A Comparative Study of Home Computer Use in Center for Research on Information Technology and Organizations Smart Home Concepts: Current Trends
  • Chuan-Fong Shih
  • Alladi Venkatash
Chuan-Fong Shih and Alladi Venkatash, " A Comparative Study of Home Computer Use in Three Countries: U.S., Sweden, and India, " Center for Research on Information Technology and Organizations, University of California, Irvine, Paper 378, 2003; Alladi Venkatesh, " Smart Home Concepts: Current Trends, " Center for Research on Information Technology and Organizations, University of California, Irvine, Paper 377, 2003.
Pew Internet & American Life Project
  • See For Example John Horrigan
See for example John Horrigan, Home Broadband Adoption 2009, Pew Internet & American Life Project. Available at <http://pewinternet. org/Reports/2009/10-Home-Broadband-Adoption-2009.aspx>.
Smartphones vs Feature Phones: What's the Difference? " WirelessWeek
  • Luke Simpson
Luke Simpson, " Smartphones vs Feature Phones: What's the Difference? " WirelessWeek, February 28, 2009 <http://www. What%E2%80%99s-the-Difference-/>.
and video page-views grew at a rate of 25 times over the same period YouTube reported 72 million unique visitors to its US site, 4.7 billion page-views per month, and hundreds of millions of videos viewed daily
YouTube's number of unique visitors grew nine-fold between March of 2006 and March of 2007, and video page-views grew at a rate of 25 times over the same period. In July of 2008, YouTube reported 72 million unique visitors to its US site, 4.7 billion page-views per month, and hundreds of millions of videos viewed daily. Source: YouTube. You Must Know – July 2008. 25 United States National Gamers Survey 2009 available at <http://>.
Hulu had either 9 million or 42 million viewers in See Brian Stelter Hulu Questions Count of Its Audience
Depending on who is counting, Hulu had either 9 million or 42 million viewers in May 2009. See Brian Stelter, " Hulu Questions Count of Its Audience, " New York Times 14 May 2009. <http://www.nytimes. com/2009/05/15/business/media/15nielsen.html>.