The same old pie?:
The constancy hypothesis revisited
by Xabier Meil n-Pita1 and Haoming Denis Wu}
1PROEL (Promotora de Ediciones Electr"nicas), Madrid, Spain
}University of North Carolina at Chapel Hill
Address: P.O. Box 162, Chapel Hill, NC 27514
Phone: (919) 929-3696
Fax: (919) 962-0620
E-mail: [log in to unmask]
The same old pie?: The constancy hypothesis revisited
The constancy hypothesis, which purports that the percentage of income spent on
mass communications remains fixed over time, does not hold true with data
collected from 1959 to 1993. This hypothesis, first operationalized by McCombs,
was supported with data garnered in the period 1929-1968. The boom of
expenditures on audiovisual media and new technologies in the last decade, with
a faster growth rate than the economy, is the major factor that breaks the
assumed pattern of media consumption. In addition, at the personal level of
media use, this study finds no significant relationship between traditional
media use and new media use -- indicating that people's time and money allocated
to various media is not constant either.
The same old pie?:
The constancy hypothesis revisited
Will newspapers become museum artifacts in the near future? News professionals
as well as scholars have been pondering the very same question. Some point out
that it is unlikely since advertising expenditure on newspapers reached an
all-time high in 1994 and the circulation in general remains stable, according
to the Newspaper Association of America (NAA). However, on average thirteen
newspapers disappear each year in the United States since 1980. In addition, if
the circulation figures are compared to the fast growth of national population,
a steady decrease of the circulation per capita can be documented from as far as
1946. Have the newspaper readers been lured away to new media such as the
Internet and cable TV?
This paper aims to re-examine the constancy hypothesis of media use, which
postulates that "the amount of money spent on mass communication is a relatively
constant proportion of the available wealth" (McCombs, 1972, p.6). In more
detail, McCombs' seminal article states that consumers and advertisers devote
more resources to mass media when the economy is booming, and less during
recession periods. On the other hand, the article maintains that the advent of
new media do not necessarily signal that a new turf of audience shares will
ensue; on the contrary, the audience's expenditure on media will remain fixed --
all of the media, old-timers and newcomers, have to compete for the same old
There are two major reasons to undertake a new test of the relative constancy
hypothesis. First of all, a variety of new media appeared in the past few
decades since McCombs first operationalized the hypothesis using data collected
in the period of 1929-1968. It would be interesting to see whether McCombs'
conclusion still holds true in the 1990s. Secondly, no singular medium seems to
pose as powerful a danger for the future of the traditional media as the newly
born Internet. Among the newcomers, cable and satellite TV appear to affect
network TV to some extent. Video games and VCRs seem to meet some particular
needs of entertainment only. Yet the Internet is an entirely different
creation. It can simultaneously inform, entertain, and interact with the
audience. Its impact over the traditional media system should be
multi-dimensional and unprecedented.
The findings of this study will not only generate new knowledge about people's
media use in both aggregate and personal levels, but also help traditional media
make new strategies to survive. If the constancy hypothesis still holds true in
the 1990s, i.e., if the share of the total expenditures to each traditional
medium indeed shrinks because of the advent of more new media, the question then
will be whether money that might have been spent on traditional media is now
spent on new media instead. On the other hand, if the relative constancy
hypothesis turns out to be false when applied to the past thirty-five years, the
causes of the fluctuation of media expenditures above or below the trend of the
economy must be explored so as to uncover which media are most accountable for
the spending fluctuations.
When a new medium appears, fear always arises that another medium (or others)
may be totally replaced. That happened, for example, when VCRs hit the market
and became popular. At that time many predicted that movie audiences would stop
going to the theaters, and the cinema industry would receive a fatal coup de
gr ce. The public did not stop going to the theaters, though. As for the cinema
industry, it adapted to the new situation, and even took advantage of it. Films
that did not attract the people to the theaters could eventually benefit from
the release of video-cassettes.
Shaw (1991) investigated how and why mass media rise and fall in the U.S.
history and found some consistent patterns in their growth and diffusion.
According to him, all of the media examined went through an initial upward phase
characterized by an innovative technology and an open, progressive mentality.
Once the pinnacle of the medium's popularity is reached, there comes an
unavoidable decline -- and there is no way back. The media then go through a
period of adjustment, during which profits are not guaranteed and the audience
becomes smaller and smaller. In some cases, the old media may not survive, such
as the black-and-white film. In the downward slope, media struggle to adjust to
adverse situations, keeping a segment of loyal audience so as to sustain the
medium worth of advertisers' investments.
All media inevitably will undergo this evolution. Historically speaking,
newspapers achieved its point of highest market penetration in the early 1920s.
Films hit its peak in the late 1940s, radio in early 1950s, magazines in the
late 1960s, and television in the 1980s. Many new media including cable TV are
growing in the 1990s. Interestingly, Shaw observed that the process of media's
rise and fall has become faster and faster as we approach the new millennium.
How do consumers allocate their money on mass media? The question was
addressed by McCombs (1972) twenty years ago. Using national data for the
period 1929-1968, McCombs found support for what he called the constancy
hypothesis, namely that "total consumer spending on mass communication shows no
real deviation from the general economic growth of the United States over four
decades." No matter how healthy the state of the economy is, consumers, as well
as advertisers, always allocate a certain amount of their income, and maybe
time, on mass media, in the same way as people spend an invariable share of
their income on food and housing.
Communication scholars have not been awfully enthusiastic on pursuing this
research topic. Among those who continued to explore, newspaper seems to be the
favorite medium to study. Tillinghast (1991) found evidence that newspaper
readership increases with the level of urbanization. Demographic
characteristics were found to be not very powerful predictors, for even the
combination of age, sex, education and income only accounts for 27% of the total
variance in news reading.
Stevenson (1994) has undertaken a cohort analysis of newspaper readership using
available data from various sources. First of all, he found that newspaper
circulation per capita in the period of 1946-1993 was declining. Based on the
result of General Social Survey, he discovered that people in every age group
are reading less now than ten years ago and, on the average, readers were older
in 1985 than ten years earlier. The decade of 1975-1985 is the period that has
the highest readership drop-off rate. Moreover, Stevenson found the younger the
generation, the less they read newspapers. Analysis of equivalent data from the
Carolina Poll for the period of 1979-1989 shows similar results.
The impact of demographic attributes over media use has been examined by
various researchers. For example, education is found to be a good predictor of
media use pattern -- highly educated people tend to read more newspapers
(Stevenson, 1994), and use more new media technologies (Dickerson & Gentry,
1983). Gender also plays a role in media use. McGrath (1993) found a steeper
decline of female newspaper readership in the past few decades. According to
the statistics, external factors were not at work on curtailing women's reading
interest. In fact, newspaper readership among working women is just as high or
even higher than among women who do not work outside home. On the other hand,
non-readers have been described by Sobal and Jackson-Beeck (1991) as older,
living mostly in rural areas, poorly educated and socially disadvantaged.
Finally, community ties have been proved to be one of the most powerful
predictors of newspaper circulation. Using the census data from the 1970s and
circulation figures from a sample of 200 daily newspapers, Stone (1977) found
that the number of households, population, and the number of singles living in a
community are positively correlated with daily newspaper's circulation. The
three variables were also considered by the author to be associated with any
city's level of resident stability.
Based on the literature above, it is apparent that audience's media use pattern
is influenced by many factors and media's popularity is changing over time.
However, most studies focus on newspapers and neglect the interactions between
newspapers and other new media. More importantly, there lacks an
across-the-board examination on to what extent people's monetary and time
resources allocated to different media are stable. Because of this gap in media
use knowledge, a new study is needed.
This study aims to re-examine the constancy hypothesis at the societal and the
personal levels. First of all, we wonder whether people spend a stable share of
their income on mass media along the past three decades, controlling for the
national economic index. Second, we would like to inspect what kind of impact
new media generate on traditional media such as newspaper, radio, or magazines
-- e.g., do people reduce their expenditure on newspapers so as to spend more
money on new media? Third, the relationship between media expenditure and the
U.S. economy will be examined -- do people spend less money on media when
recession strikes the economy? Lastly, users' characteristics and media use
patterns will be investigated to uncover their hidden relationships.
Data. The data used in this study were derived from two sources. The first
data set is composed of consumer expenditure on mass media for the period
1959-1993. The data were retrieved and compiled based on National Income and
Product Accounts (NIPA) of the United States, 1959-1988, and Survey of Current
Business, both of which are published by the U.S. Department of Commerce.
The use of the first data set aims at replicating the tests of the constancy
hypothesis. However, several differences between the data McCombs originally
used and the current ones used for this paper need to be clarified. For one
thing, although McCombs' (1972) findings showed that "both consumer and
advertiser economic behavior basically has moved together over time", it should
be noted that this paper does not deal with advertising expenditures. In
addition, Personal Consumer Expenditures (PCE) which reflects the total
purchases of new goods and of services by individual, rather than Gross National
Product (GNP) employed by McCombs, is used to function as a basis to compare
people's expenditure on media.
It should be noted that most of available data on people's media spending are
grouped based on their natures, therefore it is not possible to separate the
components of an aggregated datum. For example, the variable PRINT2 includes
statistics for newspapers, magazines and sheet music. As a result, there is no
way to isolate the information of, say, newspapers that we are especially
The second data set used in this study comes from "Technology in the Home," a
national survey conducted by the Times Mirror Center (now known as The Pew
Research Center) in 1994, which contains substantial information on people's
uses of media and communication technologies. Particularly relevant to this
paper is that the survey questions give a special emphasis on the new media.
Therefore, this relatively up-to-dated data set allows us to examine the
relative constancy hypothesis at the individual level. The information of the
"Technology in the Home" survey was collected via telephone interviews, based on
a national sample randomly selected with an income-stratified method.
Therefore, households with different ranges of incomes would be reached equally.
Analyses. The first task of analyses with the first data set is to probe the
trend of how consumers' expenditure on mass communications (in constant dollars,
so as to eliminate the effect of inflation) has evolved over the time period.
Moreover, in order to gauge people's expenditure on mass communications more
correctly (holding the changing resources available to consumers constant), the
trend will be shown as percentage of total Personal Consumer Expenditures (PCE).
Another trend to be inspected is how each medium's share of the total media
expenditures has changed over time. If there were significant changes in any of
these three time-series data, then the constancy hypothesis should be
The second task for testing the constancy hypothesis is to run correlation
tests between years and expenditures on different media (using percentages of
the PCE along the years). If the slope of consumer's expenditure by years
indicates any trend (with statistical significance), then the constancy
hypothesis should be rejected. The question following this is which media
contribute to break the constancy.
This study is also intended to investigate the constancy hypothesis at the
individual level by using the Times Mirror survey data. After following the
evolution of consumer expenditures over more than three decades, this
cross-sectional data set can yield valuable information on how respondents
allocate their time and monetary resources among different media. To begin
with, the categories of new and traditional media were demarcated, and each
single medium interviewed in the survey was assigned to either of them. Two
media use indexes -- one for traditional media use, the other for new media use
-- were then developed to caliber to what extent people use either media.
The constancy hypothesis postulates that there is a weak correlation between
new and old media use -- since time and monetary resources are fixed,
respondents scoring high in the index of new media use would be rated low in the
traditional media use, and vice versa. Therefore, most of the respondents would
fall into either of the two categories: traditional media users who do not go
for new media, or new technology fans who would not have time and/or resources
to spend on traditional media. However, if this is not the case, and there
exists a significant percentage of respondents who are rated high or low in both
indexes, then the differences among the four possible categories of observations
ought to be investigated further.
I. The constancy hypothesis at the aggregate level
Table 1 shows personal consumption expenditures on mass media for the period
1959-1993. Figures are presented in constant dollars -- using U.S. dollar's
worth in 1987 as the base. It indicates that there has been a considerable
increase of consumer expenditures during the selected time interval. Americans
spent 132.1 billion dollars in 1993, as opposed to the 26.6 billion dollars
spent in 1959, a growth of almost 400%. The increases can be found at any group
of media, with one single exception -- motion picture (variable: MOVIE). This
is hardly surprising, as McCombs (1972) noted, box offices of movies reached
their peak in 1946, then a continuous decline followed.
Table 1 about here.
Although all other media have had a steady growth, the paces vary from one
medium to another. Among them, video and audio products, computers and their
lateral equipment, and musical instruments achieved the biggest expansion by far
(variable: AUDIOVI). The money spent on these items have skyrocketed along the
years (from 2.7 billion dollars in 1959 to 83.7 billion in 1993). This is
echoed by the figure of expenditure on all audio-visual media (variable:
TOTAUDIO), which includes both AUDIOVI and RTV, the expenditures on radio and
television repairs. RTV contributes far less than AUDIOVI to the spending
explosion on all audiovisual equipments, though.
Significant, although much smaller, increases are also found for the variables
PRINT1 and PRINT2 -- the former includes books and maps, and the latter is
comprised of magazines, newspapers, and sheet music. Their growth is also
reflected by the notable increase of their aggregate variable, TOTPRINT.
Consumers spent three times more on books and maps in 1993 than in 1959, and
twice more in newspapers and magazines. However, the growth of the print media
consumption is less significant than that of audiovisual media consumption.
All these changes can also be visually detected from Chart 1, which is
virtually created from those figures presented in Table 1. This chart makes the
overall trend more apparent: consumers' expenditures on audio-visual media
increased sharply since the 1980s, which probably draws the line of total media
spending upward too.
Chart 1 about here.
The relative constancy hypothesis, according to Charles E. Scripps, chairman of
the board of Scripps-Howard Newspapers, stated that "in spite of the increasing
complexity of mass communications with the advent of new media, the pattern of
economic support has been relatively constant, and more closely related to the
general economy than to the various changes and trends taking place within the
mass media field itself" (McCombs, 1972, p. 5). Therefore, increases of media
consumption by themselves do not necessarily refute the constancy hypothesis.
The growth of the economy in the U.S. has to be considered in the hypothesis
Personal Consumer Expenditures (PCE) should be able to resolve this problem,
because unlike raw figures of expenditure, it shows the percentage of
expenditure while holds the state of economy constant. Based on Table 2,
consumer expenditure on mass media (using PCE) has observed an uninterrupted
increase since 1980s. This increase is perfectly matched with a parallel growth
in the spending on all audiovisual media (see AUDIOVI and TOTAUDIO).
Table 2 and Chart 2 about here.
Chart 2, made with the same data from Table 2, does not look very different
from Chart 1. Expenditure on audiovisual media, one of the lowest among media
expenditures in 1959, started to take off in the beginning of the 1980s. By
1993 the audiovisual media expenditure has become the highest one, far above any
other spendings. As for print media, the percentage of people's PCE spent on
books and maps lingers around .45%, but the percentage spent on newspapers and
magazines drops steadily (from .86% in 1959 to .59% in 1993). Generally
speaking, the percentage of PCE allocated to print media has continuously
decreased along the years (from 1.28% in 1959 to 1.07% in 1993).
Another way to look at each medium's evolution in the marketplace is to examine
its percentage of the total media expenditures. Table 3 and Chart 3 present the
results, which again support what we reported above. Print media took the
biggest slice of the mass media pie in 1959 (56.77%), followed by movie and
theater (27.44%). Audiovisual media started small (15.79%) but gained the lead
position in 1985. Since then, audiovisual media have captured more and more
grounds, reaching an all time high 65.71% of the total media expenditure in
Table 3 and Chart 3 about here.
In summary, there is a strong and consistent evidence that the amount of money
spent on mass media has increased since 1959, not only in terms of constant
dollars, but also of the percentage of PCE. These findings therefore, do not
support the constancy hypothesis at the aggregate level. In addition, the
relative constancy of expenditures on newspapers and magazines that McCombs
discovered during the period of 1929-1968 has also disappeared from the results
presented in the above tables.
The line of constancy seemed to break down in the 1980s. As Chart 2 shows,
from 1982 on, the expenditures on audiovisual media and mass media in general
continue to expand. Print media expenditure had began to drop three years
earlier, and after leveling off a bit around 1989, it has kept decreasing since.
Therefore, 1982 seems to function as a breaking point of the mega trend of
media consumption in the U.S. Charts 4 and 5 present the results with the data
divided into two subsets using 1982 as the cutoff point. The focus of attention
here is restricted to three groups of media spending: newspapers and magazines,
audiovisual media, and total media expenditures, all presented as percentages of
PCE. The slope coefficients of the bivariate relationship give an approximate
measure of association between expenditure and year.
Charts 4 and 5 about here.
Regarding newspapers and magazines, the slope of expenditure by time indicates
a downward trend -- the coefficient is -.005 in both periods. On the other
hand, both audiovisual and total media spendings increase over the thirty-four
years. In the period of 1959-1982 (see Chart 4), both slopes of media spending
by years are positive. The slopes converted from the data in the period of
1983-1993, however, become much more steep; and interestingly, both coefficients
are the same for audiovisual and total media expenditures (.148).
All of the findings point to the fact that the money consumed on audiovisual
media did not come from the money that could have been spent on other
traditional media and/or from new money yielded from the growing economy.
Instead, the new expenditure pattern must have come from a changed consuming
behavior of Americans, whose allocation of "media money" ranges from 2.25% of
their PCE in 1983 to 3.82% in 1993.
II. The constancy hypothesis at the individual level
The results of "Technology in the Home" survey provide fairly substantial
information about media use or ownership at the individual level. Also included
is updated information about people's use of audiovisual and computer
equipments. The survey was completed in 1994, when consumer's expenditures on
new media products reached a new peak after ten years of uninterrupted increase.
The shortcoming of this data set is that people's expenditure on the new media
was not tackled in the survey, which makes it impossible to draw inferences from
a financial perspective.
Presented in Table 4 is a list of new media surveyed, and percentages of U.S.
households who reported owning any certain medium or having a given service
(such as cable TV or the Internet). For each question, there are three options
for respondents to choose -- "yes," "no," and "don't know/refused." The
percentages shown here are derived from the "yes" answers.
Table 4 about here.
In order to construct an index to succinctly display new media ownership, every
item on the list of new media was assigned a point whenever respondents checked.
The new media items include TV remote control, VCR, videogame, computer, video
camera, modem, fax machine, satellite dish, and subscription to cable or the
Internet service. Therefore, the possible points for each respondent range from
0 to 10. In the bottom of Table 6 the frequency and some descriptives of the
new media use index are displayed. Most respondents seem to have 3-4 items of
these new media. The distribution is slightly skewed to the right, although it
does not deviate too much from normality. A new collapsed variable was then
created -- those respondents who scored below the median are labeled as "low new
media users," and those above it as "high new media users."
Tables 5 & 6 about here.
A traditional media use index was also constructed in a similar way. However,
it should be noted the nature of this index is different, due to the limitations
of the survey's questionnaire. The questions asked about respondents' different
levels of frequency of traditional media use, instead of ownership as those
questions posed for new media. Table 5 displays the frequencies of responses
falling into each of the three categories (low, medium, and high) for each
traditional medium. Newspapers, magazines, television (news and entertainment),
radio, book, and movies are considered traditional here. The ownership is
supposed to be non-informative since all them have been in the market for long.
The range of the possible answers for traditional media use frequency is much
larger than that for the new media questions. The frequency estimates can be
any number from 1 to 8, except the cases of movies (from 1 to 6) and radio (from
1 to 9). Zero is assigned to the answer "don't know." In order to assign an
equal weight to each of the items in the index, the raw values of responses were
recoded into three categories: low frequency as 0, medium frequency as 1, and
high frequency as 2. The baseline for recoding the raw values is to make them
fall as evenly as possible into the three groups.
Table 6 also shows the distribution of the transformed index of traditional
media use. Its distribution is less skewed, and closer to normality than that
of the new media index. Again, a new variable was created to separate the
respondents between "low traditional media use" and "high traditional media
use." The dividing point, as the one in new media uses, is the median.
Table 7 about here.
A cross-tabulation test can be executed to examine the relationship between the
two nominal-scaled, recoded indexes. Table 7 presents a contingency table,
together with the chi-square test results, and the Pearson's correlation
coefficient. Although the differences among the cells are statistically
significant at the .05 level, the percentages shown in the table do not indicate
any systematic trend. If the constancy hypothesis could apply to time, in
addition to money, as it was originally postulated, it would be expected that
high new media users will tend to be low traditional media users, and low new
media users are more likely to be high traditional media users (i.e., only high
in one group of media but not in the other). Yet, this is not the case here.
The sum of the two diagonal cells (the matches of high in one and low in the
other) represents only 48.1% of the total respondents, which is even smaller
than the total percentage of the respondents who happen to be either high or low
users in both media groups (51.8%). As for the degree of association between
the new and traditional media uses, a weak positive correlation (r = .033) was
found, indicating again that constancy hypothesis is not supported.
The four different categories demarcated by the above contingency table are
presented in Table 8 with a demographic profile within each cell. The
demographic characteristics used to define each category are gender, age, income
and degree of urbanization. For example, the table shows that heavy new media
users (comparing the upper with the lower rows) tend to be younger, more
educated, more wealthy, and live in suburban areas. However, the most important
distinguishable characteristics between low and high old media users seems to be
age and education. For example, in the cell of high old media/low new media
users, the percentage of people who have a college degree (20.5%) is almost
twice as high as that for the low old media/low new media users (10.9%).
Tables 8 and 9 about here.
A more thorough examination of the relationships among the media use and
demographic variables can be found in Table 9. The correlation matrix includes
the predictors of mass media use identified by literature, and the two media use
indexes constructed in this research. Among the highest correlation
coefficients, two are associated with income -- .46 and .16 respectively for new
and traditional media use. In other words, the more affluent the people are,
the more likely they are to use media in general -- and that is especially the
case for new media. Other coefficients that are particularly worth paying
attention to are education, age, and residential locality. Education plays a
conducive role to both media uses, whereas age's function is strikingly
different in predicting uses of these two media groups -- older people are more
likely to use traditional media but far less likely to use new media.
Interestingly, the factor of residential locality yields a consistent result
with traditional and new media -- living in suburban area is positively related
to media use while living in towns or rural areas are negatively related
(although residential locality correlates with income level, well-to-do
Americans tend to live in suburban areas).
Finally, Table 10 presents a compilation of partial correlation coefficients
between the new and old media indices, controlling for each variable listed in
the table. When controlling for the effect of income and education
respectively, both the associations between old and new media use decreases from
.11 to .04. However, with a control on income, the correlation between two
media uses is significant at roughly .03 level, indicating that income generates
a stronger effect than education does. Partialling out the effect of all the
predictor variables together, the association between the two indices become
insignificant (r = .01, with a probability value of .527).
Table 10 about here.
In essence, education, income, and to a less extent, age correlate with media
use. When they were controlled for, the correlation between the two media
indexes drops significantly. This phenomenon demonstrates that the equilibrium
of different media uses should be examined more deeply than was originally
thought. But traditional media use and new media use are positively related
with each other, which indirectly refutes the constancy hypothesis.
With the solid evidence of people's increasing expenditure on media over the
last thirty-five years garnered from the Department of Commerce and the findings
of people's media use yielded from the Times Mirror survey, the constancy
hypothesis that McCombs raised three decades ago should be rejected. McCombs
(1972) in his pioneer article stated that "what Americans spend on mass
communication has not increased with the advent and spread of new media such as
radio and television" (p. 9), which is obviously at odds with the latest
observation in the 1990s.
Consumer expenditure on mass media in 1993 is three-time higher than that in
1959. As indicated earlier, people's spending on audiovisual media took off in
early 1980s and remains at a high growth rate since then, which directly raises
the total media expenditure along the years. The recent boom of audiovisual
media expenditure probably results from an increasing media use or needs from
consumers rather than the outcome of the growing economy or money spared on
other media. In other words, it is the changing pattern of media use and
spending that accounts for the growth of media expenditure in the past three
decades. Personal computer, and other new audiovisual products such as CD
player and video game equipment are probably the factors that contribute to the
transformation of the old media use pattern. Unfortunately, the available data
cannot provide the answer to the question.
Time allocation is another component of the constancy hypothesis.
Unfortunately, the testing of the time part in this study is rather indirect.
Yet, an initial observation based on the Times Mirror survey result is that new
media use and traditional media use seem to be independent with each other. For
example, the percentage of people who happen to be heavy users on both new and
traditional media is much larger than any other percentage figures.
Media use at the personal level is more complicated than the oversimplified
constancy theorem. A great number of factors play roles in determining media
use -- some of the determinants discovered in this study nullify that constancy
mechanism is actually at work on personal use of media. According to the result
of the Times Mirror survey, income is the most important predictor of new media
use. Since information and entertainment provided by new media technologies are
more expensive than ever, income level becomes a threshold to the access. It is
foreseeable that the gap between the media-rich and the media-poor will become
much wider and deeper. Other than income, education and age were also found to
be important predictors of new media use.
Believing in the constancy hypothesis, some people have been awfully concerned
that traditional media such as newspapers and magazines will disappear soon.
With the solid evidence presented in this paper that rejects the constancy
hypothesis, it seems that the concern, out of the constancy theorem, is not to
the point. Traditional media will, as Shaw (1991) points out, find their niche
in the marketplace to meet special needs. However, the competition among
traditional and new media deserves more and longer observations.
Future research should investigate how media consumers allocate their money and
how important media and their products mean to them in their everyday lives,
compared with other goods and services. Is "media money" well planned on
personal and/or household budget in advance or just spent intuitively? Do
consumers purchase less of other goods and services, such as clothing, food, or
transportation in the same period? Answers to these questions can help us
understand better how people use media.