A Theoretical and Methodological Critique of the Principle of Relative
School of Communication
University of Miami
P.O. Box 248127
Coral Gables, FL 33124-2030
Phone: (305) 284-3500
E-mail: [log in to unmask]
Submitted for consideration to the
Media Management & Economics Division for the 1995 convention of
the Association for Education in Journalism and Mass Communication,
April 1, 1995
After 20 years of empirical research, this paper revisits the theoretical
assumptions of the principle of relative constancy (PRC). This
or hypothesis holds that consumers spend a constant fraction of their
income on mass media over time (constancy), although they may modify
patterns of spending within categories of mass media (functional
equivalence). A critique of the PRC literature reveals that whereas the
assumption of functional equivalence can be phrased in economic terms,
there is no such validation for the assumption of constancy.
Methodological factors that may explain why PRC studies have offered
conflicting evidence are also discussed.
A Theoretical and Methodological Critique of the Principle of Relative
Since the now-seminal McCombs (1972)'s "Mass Media in the Marketplace," a
growing number of studies have tested the principle of relative
(PRC) in the United States and abroad. This principle or hypothesis
that the proportion of income spent on mass media does not change
significantly over time (constancy), although compensatory variations
between mass media categories may occur during the same period to leave
room for the introduction of new mass media goods/services (functional
equivalence). But despite mass communication researchers' increasing
interest in this topic, several important theoretical and methodological
questions remain before establishing the PRC as a
empirically-determined hypothesis that characterizes the relationship
between consumer mass media expenditures and income. Taken as a whole,
this body of literature has revealed mixed support for the principle of
relative constancy, and this begs the question why it is so. But more
studies have yet to validate the PRC in economic terms and draw connections
between this principle and standard economic theory. Unlike the
ecology-based theory of the niche, which has been successfully applied to
the analysis of competition between mass media industries (Dimmick,
Dimmick & Rothenbuhler, 1984), the principle of relative constancy did
benefit from a solid theoretical foundation when it was initially
To remedy this gap in the literature, this paper examines whether the PRC
can pass the test of economic muster and why constancy studies have
produced conflicting results. Such validation, if supported, would
strengthen the theoretical arguments of the PRC and would acknowledge the
PRC as a sound economic proposition. This study first reviews the
PRC literature, then offers a critique of that literature primarily based
microeconomic theory, and concludes with a discussion of new directions for
mass media expenditures research. It focuses exclusively on the evolution
of consumer mass media expenditures, that is direct consumer spending on
mass media-related goods (e.g., newspaper, television set) and
(e.g., cable television).
The Principle of Relative Constancy
Although early studies have investigated the relationship between mass
media expenditures and the general economy (for reviews, see McCombs,
Son, 1990), it was not until the publication of a descriptive economic
study by Scripps-Howard Research in 1959 that the concept of relative
constancy became formally identified. Based on an analysis of mass media
expenditures from 1929 to 1957, Charles Scripps, chairperson of the
Scripps-Howard Newspapers board, concluded:
If we may suggest one broad generalization, it is that in spite of the
increasing complexity of mass communications with the advent of new
the pattern of economic support has been relatively constant, and
closely related to the general economy than to the various changes
trends taking place within the mass media field itself.
The consistency evident in the pattern of economic support for the mass
media seems significant. It suggests that mass communications
a staple of consumption in our society much like food, clothing,
shelter. Its stability in times of economic stress indicates that
consumers feel that mass communications is a necessary [sic] of life,
although their selection of media may vary [emphases added].
Support, 1959, p. 6)
Whereas Scripps offered several stimulating ideas for interpreting the
evolution of consumer mass media expenditures in the United States
constancy, relationship to the general economy, staple of
inter-media variations), he did not suggest any specific hypothesis or
statistical technique to test the viability of such assumptions.
McCombs (1972), and later Wood (1986), elaborated on Scripps' "explicit
point of view" (McCombs, 1972, p. 6), developed hypotheses, and
correlational and regression analysis to the various facets of the
As stated by Wood (1986), a key issue of the relative constancy principle
is the interpretation of Scripps' introductory remarks.
Interpretations and Analyses
Constancy of Mass Media Expenditures: In his pioneering work, McCombs
(1972) restated Scripps' concept of constancy as follows:
His hypothesis asserts that the amount of economic support provided for
mass communication media consistently follows the ebb and flow of
general economy. This means that a relatively constant proportion of
available wealth--Gross National Product, for example--will be
mass media. When they have more money [i.e., income], consumers
advertisers will spend more on mass media. When they have less, they
spend less on mass media. The important point is that this ebb
follows the general economy, not the competition and
within mass communication industries. (McCombs, 1972, p. 6)
In sum, a key assumption of the relative constancy is that total mass media
expenditures will grow or decline proportionate to the pace of the general
economy. Regardless of income changes, consumers will always devote about
the same share of income to the mass media; behaviorally, it is implicitly
assumed that they will react fairly quickly to rising or declining income
by adjusting their mass media spending accordingly. To test this
hypothesis, McCombs and his co-authors correlated categories of media s
pending with time, controlling for inflation, population growth, and
personal income (McCombs, 1972; McCombs & Eyal, 1980; McCombs & Son,
The sign and magnitude of the correlation coefficients (from -1 to +1)
indicates the direction and strength of the relationship between mass
expenditures and time. According to McCombs and Eyal (1980), a perfectly
monotonic (i.e., steady) increase and decrease across time would yield
trend of +1 and -1, respectively. McCombs (1972) establishes
significance for the relative constancy as follows:
If the probability (p value) of a given r's being a chance discrepancy from
zero is equal to or less than five times in a hundred (p _ .05), the null
hypothesis is rejected and the empirically reported value is
But, if p > .05 then the empirical value is ignored and the r
be zero. (p. 11)
In other words, if the coefficient is significantly different from zero at
the .05 level, the trend is considered non-zero. The reverse would
the principle of relative constancy, evidencing "absence of movement over
time" (McCombs, 1972, p. 18). Wood (1986) labeled this interpretation
the relative constancy principle the "time-trend constancy hypothesis"
because time (i.e., physical time) is the primary predictor.
Functional Equivalence: McCombs also elaborated on the second key element
of Scripps' reasoning: functional equivalence (also called inter-media
variations). He relied on past research, especially Melvin DeFleur, to
hypothesize that consumers will vary in their selection of mass media.
Theories of Mass Communication, DeFleur and Ball-Rokeach (1989) use
term "functional alternatives" to describe a complementary relationship
between types of mass media that carry the same function; on a broader
level, this notion explains how consumers change their patterns of mass
media usage over time. "Each of these functional alternatives to the
newspaper has eaten into the circulation of the daily press. Each, in
sense, provides news, information, or entertainment in a way that once was
the exclusive province of the newspaper" (DeFleur & Ball-Rokeach, 1989, p.
60). Formally stated, the concept of functional equivalence holds that
"an entertainment or activity will be displaced by the newer one
it serves the same needs as the established activity but does so more
cheaply or conveniently" [emphasis added] (Himmelweit, 1977, p. 12).
McCombs (1972) posits that the sociological notion of functional
equivalence is consistent with the premises of the relative constancy
hypothesis. Limited by time and budget, consumers must decide which of
competing media--which serve the same communication function (e.g., news,
entertainment)--they should select. For McCombs (1972), functional
equivalence is an intrinsic part of the consumer decisionmaking process,
whereby consumers make choices among different but functionally
types of mass media after considering their value individually.
Consequently, the principle of relative constancy assumes "a tremendous ebb
and flow of money among the various media" (McCombs, 1972, p. 6). For
instance, an increase in electronic media expenditures would yield a
decrease in print media expenditures, but the overall proportion of income
devoted to mass media would remain unchanged. To succeed in the
marketplace, new communication technologies can only expand to the
detriment of old technologies (see McCombs, 1972).
Whereas the constancy assumption is generally tested using inferential
statistics, the functional equivalence assumption is substantiated
plots of mass media expenditure categories.
Wood's Critique: In a watershed article, Wood (1986) identified and
discussed three problems inherent in McCombs' analyses. First, because
mass media expenditures are treated as time-series data, those data are
likely to be serially correlated (autocorrelated); therefore,
analyses are inappropriate statistical techniques for testing the
constancy assumption of the PRC. If residuals at different points in time
are correlated, successive observations will then be related to each
(Ostrom, 1990). In the case of positively or negatively correlated
residuals, tests of significance and R} values can lead to erroneous
conclusions (Gujarati, 1988). Wood (1986) handled the problem of
autocorrelation by first detecting it using the Durbin-Watson test and then
modifying the original ordinary least squares regression equation to
Second, Wood questioned the validity of zero correlation for establishing
constancy of mass media expenditures over time. He argued that zero
correlation coefficients could conceal substantial variations in
expenditures within the analyzed period.
Finally, he recommends using disposable personal income (DPI) rather than
personal income. Because DPI consists only of private consumption and
savings, this variable is a more refined measure of income, thereby
decreasing measurement error and improving the testing of mass media
But perhaps more importantly, Wood (1986) offers a second model of
relative constancy. The "income-share constancy hypothesis" holds that
"when income changes, the share of income spent on media does not
significantly change" (Wood, 1986, p. 41). Unlike McCombs' partial
correlation approach and the time-trend constancy interpretation, the
income-share interpretation does not include a time variable. To test
hypotheses, Wood ran regression analyses using total mass media spending
as the dependent variable. According to Wood and O'Hare (1991),
constant income-share proposition would be supported if the intercept of
the fitted function is not significantly different from zero (C = _0 +
+ e). If it is statistically significant, there is evidence that media
spending as a fraction of income is not constant and that consumers
an increasing (or decreasing) portion of their DPI on mass media.
time-trend constancy, if the coefficient of the time variable is not
significantly different from zero, holding DPI constant (C = _0 + _1 time
_3 Y + e), this would signify the absence of significant increases or
decreases in mass media expenditures and would support the principle of
Research Findings in the United States
After this theoretical and methodological description, what can we
conclude about the principle of relative constancy from the existing
empirical research? In support of a movement of long-term constancy, some
studies have reported that consumers' mass media expenditures have
relatively stable over several decades (McCombs, 1972; McCombs & Eyal,
1980; Son, 1990; Wood & O'Hare, 1991). However, researchers also found
that consumer spending on mass media could deviate noticeably from
constancy when they broke down the analysis by decade (Wood & O'Hare, 1991)
or more contextually when they considered the impact of successful
consumer electronics innovations on total media expenditures (Fullerton,
1988; Glascock, 1993; McCombs & Son, 1986; Noh, 1994; Son, 1990; Wood
O'Hare, 1991). A brief review of U.S. PRC studies follows.
McCombs (1972) analyzed consumer mass media expenditures from 1929 to 1968
and found that the trend for total media spending (in constant dollars per
household) was (r =) .228, p = .08, suggesting near constancy of media
spending. Likewise, McCombs and Eyal (1980) found a modest decline in
media spending (in constant dollars per household) from 1968 to 1977
= -.324). Several differences within mass media categories were apparent,
though. The trend for periodicals (+.832) was opposite to that for books
(-.876). The authors considered these disparities as "short-term
perturbations" (p. 158) and concluded that "even with a proliferation of
audio-visual communication services and devices during the past
historical patterns of spending on print and other mass media remain much
the same" (McCombs & Eyal, 1980, p. 158).
But more recent studies have revealed findings that contradict the
relative constancy principle. For example, McCombs and Son (1986) found
that consumers increased their media spending between 1975 and 1984
= .652, p = .057). Whereas print media expenditures decreased
significantly from 1975 to 1984 (trend = -.752), electronic media
expenditures increased substantially during that 10-year period (trend =
+.802). The authors suggest that the diffusion of video cassette
is probably responsible for these deviations from the relative constancy.
McCombs and Son (1986), however, found support for the functional
equivalence assumption of the relative constancy principle. From 1975 to
1984, increasing expenditures on cable TV and electronic media were
by a sharp decrease in expenditures on print media.
Another study reported that innovators and early adopters diverted
resources from other media in 1949-1950 to accommodate the price of
television receivers (Fullerton, 1988). In regard to the relative
constancy principle, the author states:
This sharp and brief interruption of the usual curve expected under the
principle of relative constancy need not be taken as an indication
principle is not generally supported. The fact that the trend line
quickly returned to a normal pattern indicates that there is a band
equilibrium within which all media spending tends to remain. The
departure from the trend indicates, however, that innovators and
adopters may represent exceptions to the principle of relative
(Fullerton, 1988, p. 82)
So despite these short-term departures, Fullerton (1988) concludes that the
results tend to support the principle of relative constancy. Furthermore,
the finding that some consumers diverted spending from other media to
television supports the functional equivalence assumption.
Using regression analysis with 1968-1981 data, Wood (1986) found that the
time-trend constancy and income-share constancy hypotheses differed in
their results and that mass media expenditures as a share of income
fluctuate at different points in time contrary to the constancy principle.
The author argued that a long-term regression analysis could mask changes
that took place in specific decades. This statement implies that U.S.
media expenditures may not be a fully linear function of DPI, but may ev
olve by bursts provoked by the successful adoption of new
Testing both the time-trend constancy and income-share constancy
hypotheses from 1929 to 1988, Wood and O'Hare (1991) confirmed the
long-term constancy of U.S. mass media expenditures for both models.
However, they also found that consumers devoted a larger share of their
income to mass media, especially new media, from 1979 to 1988 without
reducing their spending on print and electronic media.
Using a similar approach, Son (1990) found mixed support for the two
hypotheses (see also Son & McCombs, 1993). Before running his regression
analysis, he logarithmically transformed and first-differenced the
eliminate autocorrelation. He found that total mass media
from 1929 to 1987 did not increase as a share of income, vindicating
income-share constancy interpretation. On the other hand, the
hypothesis was not supported because consumer spending on mass media
ased over time when DPI was held constant. This study suggests that
introduction of cable TV and the VCR has caused mass media
rise between 1975 and 1987 (see also Glascock, 1993). Indeed,
the same Department of Commerce data from 1929 to 1974, Son (1990)
support for both relative constancy models.
Research Findings in Foreign Countries
Outside the United States, only two studies dealing with the principle of
relative constancy have been conducted. Werner (1986) found that
Norwegians' mass media expenditures were relatively constant from 1958 to
1982 and grew more slowly than expenditures on other leisure and
educational activities. She concludes that these findings support the
principle of relative constancy. However, the author's assessment stems
from a description of six household expenditure surveys in current
and not from a genuine analysis of aggregate media expenditures in
prices. No formal testing of the relative constancy principle using
either correlational or regression analysis was performed.
Dupagne (1994) tested both the income-share constancy (C = _0 + _1 Y + e)
and time-trend constancy (C = _0 + _1 time + _3 Y + e) hypotheses in the
United Kingdom, applying regression techniques (generalized least
estimation) to consumer mass media expenditures from 1963 to 1989.
analysis confirms a long-term constancy in mass media spending,
of variations in disposable personal income and the passage of time.
the results also showed that, in the 1980s, U.K. consumers devoted a
greater share of their income to mass media, largely due to the
introduction of video hardware and software. Thus, both U.S. and U.K.
consumers spent an increasing share of their income on new media in the
Critique of the Relative Constancy Literature
It is clear that the PRC studies reviewed here, especially McCombs (1972)
and Wood (1986), have generated fresh ideas and proposed new
all contributing to refine the analysis of mass media expenditures. But
despite this pioneering empirical work, several important theoretical
methodological concerns remain. For instance, recent research has
little to investigate new theoretical viewpoints about the relative
constancy or link it to a particular economic theory or consumption theory;
instead it has mostly relied on statements made by Scripps in 1959.
However, Scripps did not really develop a conceptual framework to support
his position. We are then left with a proposition with little
foundation. Another concern emerging from the literature review
the great variability of the findings. Whereas the assumption of
functional equivalence has gained ample support in PRC studies, the
evidence in support of the assumption of constancy has been mixed. Why is
it so? This section will address these and other theoretical and
Recently, Demers (1994) has questioned the relative constancy's
assumptions of fixed mass media expenditures and time over the years. He
contends that "there is no theoretical reason to justify the argument
spending on mass media should remain relatively constant through time,
if empirically it appears to have been that way" (p. 32). The author
maintains that the proportion of income spent on mass media would have
increased over time, had the mass media, like other goods, been widely
available to the public prior to the 20th century. He states:
Although it may be true that the proportion spent has remained relatively
constant during the 20th century, a longer historical view
force one to conclude that it has increased. The reason is
simple: For all
practical purposes, only the 19th and 20th centuries have consumers and
advertisers spent money on mass communication because mass media
available before then. Mass media are the product of an urban,
industrialized society. Thus, a long-term view suggests that relative
spending on media and advertising has increased, albeit the trend
take a monotonic form. (pp. 35-36)
In his second criticism, Demers (1994) recognizes that "time is a limited
resource" (p. 36), but argues that "there is little evidence to
that time spent with media has reached a threshold or has remained
stagnant" (p 36). Citing several studies, he suggests quite appropriately
that consumers have expanded the amount of time spent on the media in
last decades. But then he contends that if we accept the assumption
increasing time spent on the media translates into increasing mass
expenditures, consumers, contrary to the principle of relative
should devote a greater proportion of their income to the mass media.
Demers' first criticism is, of course, untestable because mass media is a
fairly recent consumption category, compared to food or housing. We
examine the notion of constancy as a valid economic proposition later
the paper. The second criticism warrants some discussion, because it
raises the important question of time ceiling.
Time Ceiling: To justify the relative constancy of mass media
expenditures, McCombs (1972) contends that time and income are scarce
resources that constrain the growth of consumer mass media expenditures.
His first point involves the scarcity of time--time spent on the mass
cannot be extended ad infinitum:
For a time the consumer can increase the amount or the number of goods
enjoyed per time unit. He sips his martini, scans his newspaper
listens to the stereo simultaneously. But there must be some limit.
Indeed, signs of the limit already are appearing. We are already in
age of half-read and unread newspapers. (McCombs, 1972, p. 62)
In other words, consumers will reach a time ceiling or threshold at which
point they will no longer be able to accommodate additional mass media
activities, given their time allocation to leisure. McCombs' argument
implies that although the amount of time spent on the mass media can be
extended in the short term, it cannot be stretched beyond a certain
in the long term because it is limited by the time devoted to other
activities, such as working, sleeping, and eating. Even if average
consumers work less in the decades to come, or even sleep less, time
devoted to non-media-related activities cannot be indefinitely compressed
within the notion of 20th-century Western lifestyle. Unless radical
changes occur, Western consumers will continue to spend a major part of
their day working, eating, sleeping, doing household chores, and the
Empirical research shows that leisure time or time spent on the media in
the United States has risen since the 1960s (see Hamilton, 1991;
Schlinger, 1981; Ogan & Kelly, 1986; Robinson, 1981, 1989; Vogel,
although this trend may have attenuated in the late 1980s (Robinson,
Son, 1990). But should we automatically posit that this increase
will cause mass media spending to increase as Demers (1994) seems to
suggest? The evidence on the relationship between increasing mass media
use and increasing mass media spending has been mixed. Whereas Wood
O'Hare (1991) found that U.S. consumers increased their share of
spent on the mass media, especially on new media (e.g., cable and VCR)
the 1980s, Ogan and Kelly (1986) found that respondents did not report
spending much money on these new communication technologies. More
importantly, Ogan and Kelly's findings showed no evidence of a relationship
between the amount of time respondents reported to allocate for mass media
use and the amount of expenditures they reported to spend on these media.
Whether these different conclusions are due to methodological variations
or other factors remains to be determined. Wood and O'Hare (1991)
secondary aggregate data from the Department of Commerce; Ogan and
(1986) relied on primary survey data collected in Indianapolis.
Although McCombs (1972) clearly perceives time as the ultimate constraint
on mass media expenditures, a second related reason for advocating the
principle of relative constancy is limited income. He states:
If indeed the goals or needs of each consumer are fixed, a voluntary
consumption maximum eventually has to be reached. As income grows,
and more wants will be gratified. Ultimately, the utility of
income and, especially consumption, will be zero. It can be
course, that the tremendous growth of mass communication--growth
constrained only by the consumer's ability to finance it--refutes
notion of fixed wants. (McCombs, 1972, p. 62)
While McCombs alludes to some economic principles (e.g., time scarcity,
income scarcity) to rationalize the principle of relative constancy,
than just taking it for granted based on the pattern of empirical
findings, he and other researchers fail to expose explicitly the connection
between economics and the principle of relative constancy. Wood (1986)
mentions in passing the relevance of the Engel law for the
constancy hypothesis, but he provides little explication why it should
so. This is the most fundamental weakness of the existing literature:
lack of any in-depth microeconomic or macroeconomic justification for
principle of relative constancy as a theoretically-driven economic
proposition. So far arguments for or against the constancy have originated
from the mass communication literature. Given that the principle of
relative constancy describes an economic relationship between consumer
media expenditures and income, it would seem logical to put the principle
of relative constancy and its assumptions to the test of economic
soundness. In other words, does economic theory confirm the validity of
the relative constancy principle?
Economic Theory - Scarcity: This economic principle holds that "goods are
scarce because there are not enough resources to produce all the goods
people want to consume" (Samuelson & Nordhaus, 1989, p. 26). Stated
differently, while consumers' wants are virtually unlimited, goods remain
scarce or finite in supply even for highly productive societies.
only produce so much. Unlimited wants are also constrained by limited
income and time (Eastwood, 1985).
McCombs (1972) is quite correct in using the scarcity rationale for
justifying that consumer spending on mass media will not be boundless.
scarcity--constraints on time, income, and supply of goods--does not imply
that consumers will be forced to spend a constant fraction of their income
on a particular good (e.g., beef, light bulb, newspaper) or consumption
category (e.g., food, housing, mass media). Such assumption does not
necessarily hold up over time, at least in the United States. Magrabi,
Chung, Cha, and Yang (1991) report that consumer spending on food has
decreased dramatically since the beginning of the 20th century; that
share spent on housing has barely changed from 1901 to 1987; and that
households have spent less of their budgets on apparel but more on
transportation over time. Rather the consequence of scarcity is that
consumers are compelled to make choices in their daily life based
on income, time, and preferences.
How consumers behave and purchase goods given these constraints falls
under the purview of demand (or consumer behavior) theory within
microeconomics. Demand theory assumes that consumers are rational--that
is, they seek to maximize utility (i.e., total satisfaction derived
the consumption of goods/services) subject to various constraints
income, price of goods, time) (see Eastwood, 1985). The next logical
in this economic exposition is to determine whether demand theory can
to the constancy and functional equivalence assumptions.
Economic Theory - Engel Law: The relative constancy hypothesis assumes
that the proportion of income devoted to mass media will remain
over time. Regardless of variations in income levels, this percentage
always be the same value, say 3% (this is the thrust of the income-share
constancy hypothesis). If this percentage is fluctuating--say 5% the
year, 2% the second year, and 8% the third year--the pattern observed
would be evidence against the principle of relative constancy. Another
to state the same idea is to say that mass media spending and income
covary--that is mass media expenditures do not change significantly
relative to income. When income increases by 1%, we would expect mass
media expenditures to increase by 1% as well. As McCombs (1972) pointed
out, for the constancy hypothesis to be supported, mass media
and income would be expected to move together--proportionately to each
other. As income rises, consumer spending on mass media should also
increase (and vice versa). But is McCombs' reasoning congruent with the
Engel law (or curve), the traditional microeconomic model of consumer
choice that addresses consumer responses to changes in income?
Simply stated, the Engel law raises the following question: Assuming that
the price of a good X is fixed, how much X would be consumed for a
of incomes? The Engel law is tied to two important concepts, the
line and the indifference curve, each deserving some development. In
standard economic terms, consumer behavior is driven by the interaction
preferences (indifference curve) and opportunities (budget line).
these conditions, consumers can make choices. The budget line shows which
baskets of goods consumers can afford to buy, given prices and income,
whereas the indifference curve depicts consumers' preferences or tastes
regardless of whether they can actually purchase those preferred goods
Landsburg, 1992). As a general rule, a change in income (increase or
decrease) will produce a parallel shift in the budget line. To
consumer choices, we must draw the indifference curve and the budget
on the same graph. Assuming that consumers are rational (i.e., they
in such a way to improve their welfare), they will then select the basket
at the point where their budget line is tangent to an indifference
This point, the consumer's optimum, is the most desirable basket of
baskets on the budget line (see Figure 1 for an illustration).
Generally, Engel curves exhibit an upward slope (Landsburg, 1992). As
consumers' income increases, they tend to buy more of all goods. For
normal goods, it is indeed the case. A "normal" good (also called a
superior good) is "a good that the consumer chooses to consume more of
his income goes up" (Landsburg, 1992, p. 95). For instance, with rising
income, it is likely that consumers would purchase more consumer
electronics products. On the other hand, a good is said to be "inferior"
when the consumer chooses less of the good even though his or her
goes up. In this case, the Engel curve will slope downward. For
a consumer is likely to consume less margarine or potatoes as his or her
The Engel curve provides a useful framework for analyzing the evolution of
mass media expenditures in economic terms and describing how consumers
would adjust their level of mass media expenditures in reaction to
changes. Furthermore, it can be directly linked to the income
of demand which estimates the percentage change in quantity in
a percentage change in income, thereby enabling researchers to
mass media goods into inferior goods (income elasticity ax < 0),
goods (0 < ax _ 1), and luxury goods ((ax > 1). The Engel law does
stipulate, however, that the relationship between expenditures on a
and income ought to be proportional or constant.
Economic Theory - Constancy: Earlier in this section, Demers (1994) was
quoted to state that "there is no theoretical reason to justify the
argument that spending on mass media should remain relatively constant
through time, even if empirically it appears to have been that way" (p.
32). In view of the foregoing economic discussion, we must conclude
he is right. The constancy assumption of mass media expenditures has
economic foundation. The Engel law would suggest that increasing or
decreasing changes in income translate into corresponding changes in mass
media spending, but it does not imply that the share of income on mass
media should be a constant. In hindsight, few economists would
a constant relationship between income and expenditures on a good over
time, whether this good is inferior, a necessity, or a luxury. For the
constancy assumption of the relative constancy, we are then left with
previous empirical results for guidance. But that does not mean that the
entire PRC is atheoretical--in fact, as discussed below, the
functional equivalence is supported by economic theory.
Economic Theory - Functional Equivalence: The concept of functional
equivalence can be explained in economic terms by relying on the previous
discussion. Because wants are limited by such scarce resources as
income, consumers must choose among alternatives. They must reconcile
their ability to buy (budget constraint) and their willingness to trade
(indifference). Let us suppose that consumers enjoy a fixed budget for
mass media and, for the sake of argument, that there are only two media
goods, newspapers and magazines. The ability to purchase media goods
be determined by the budget share devoted to mass media and the prices
newspapers and magazines. Assuming that consumers spend all their
budget on the mass media, the only way to buy more of magazines would
buy less of newspapers (see Eastwood, 1985). Within the budget
constraint, market prices (relative prices--e.g., how many newspapers
consumers can purchase for a magazine) determine at what point consumers
will substitute units of newspapers for units of magazines or vice
(marginal rate of substitution). A key assumption of consumers'
willingness to trade (preferences) is that the locus of all bundles (all
combinations of two goods) on a particular indifference curve implies
consumers derive the same overall level of utility. So it is quite
possible for consumers to substitute more magazines for newspapers and
still obtain the same degree of utility.
The concept of functional equivalence demonstrates why consumers alter
their mix of mass media goods assuming a fixed mass media budget.
Functional equivalence can serve as a useful theoretical point of departure
to predict and analyze the pecuniary impact of new communication
technologies on traditional technologies (see Noh  for a discussion
of functionality of VCR technology).
We could wonder why some studies reviewed in the first section offer
conflicting results. Second only to lack of theoretical justification,
issue of multiple constancy conceptualizations muddles the results even
further and casts more doubts on the validity of that assumption. As
indicated in the literature review, the principle of relative constancy
be conceptualized and estimated in various ways and at different levels of
sophistication. Although it is beneficial to use a multi-model approach
to test the relative constancy, a lack of a primary model can be
especially if income-share/time-trend interpretations produce different
results (Wood, 1986). If those two models are on equal footing and
in their findings, then how do we conclude there is support or
for the principle of relative constancy? Which interpretation is
So far we lack a unified operationalization of the constancy assumption.
Future empirical research should establish at the onset which model is
primary test of the constancy hypothesis and provide a rationale for
Four other factors may explain these discrepancies (see Son, 1990): DPI
versus PI; correlation versus regression; short-term constancy versus
long-term constancy; and the handling of serial correlation problems.
rationale for using disposable personal income and regression
instead of personal income and correlational analysis, has already been
exposed above. But the fourth and fifth reasons deserve some
Son (1990) has argued that running analyses with such a small sample size
as 10 observations may lead to different findings than running the
analyses with 50 observations. Indeed, studies using more than 25
observations uphold the principle of long-term constancy (Dupagne, 1994;
McCombs, 1972; Son, 1990; Wood & O'Hare, 1991). However, for analyses
10 observations, decreases or increases in media spending--which result in
deviations from the relative constancy--often happen (Dupagne, 1994;
McCombs & Son, 1986; Son, 1990; Wood & O'Hare, 1991). Statistically,
analyses with only 10 cases invite criticism because the sampling error
very large. McCombs and Nolan (1992) have critiqued Wood's approach
break down the constancy analysis per decade: "One can dismiss these
departures from constancy as random variations in small data sets (n = 10)
given the arbitrary division of the data into sets without any
historical unity in terms of mass communication" (p. 47). Ironically,
McCombs has analyzed 10-year mass media expenditures data on two
(McCombs & Eyal, 1980; McCombs & Son, 1986) without necessarily
an adequate justification for selecting starting and ending data
way he did (McCombs & Eyal, 1980). Whereas in theory, collection of
expenditure data must depend upon historical and economic considerations,
in practice it is often the availability of data that guides the
data-gathering process for these secondary analyses.
Yet, McCombs and Nolan (1992) have a point here: It is important to ground
mass media expenditures analyses in a relevant historical context. If
communication researchers present a compelling reasoning for selecting
their sample of years the way they do, they make it more defensible to
a small number of observations because the period under scrutiny may
exhibit certain unusual media-related or general spending characteristics
that are worth statistical exploration. For example, several
analyses have isolated the period from the late seventies to the
when the diffusion of video hardware and software in the marketplace
to investigate the evolution of mass media spending during that period
(Dupagne, 1994; McCombs & Son, 1986; Son, 1990; Wood & O'Hare, 1991).
A fifth reason for these mixed results might be the way the presence of
autocorrelation has been handled. Although Son (1990) and Wood and
(1991) use the same data sets and regression models, they dealt with
autocorrelation differently and arrived not unexpectedly at different
results. The specification of a correct model that meets the assumptions
of bivariate or multiple regression is essential in econometric
The existing constancy literature has focused primarily on eliminating
serial correlation, but has done little to refine models or identify the
source of serial correlation. Serial correlation more than anything
means that the model is misspecified and/or that relevant variables
have been left out. If omitted, important variables will be relegated
the error term. To the extent that these excluded variables (e.g.,
of magazines), now part of the error term, might influence variables
included in the model (e.g., price of newspapers), a systematic pattern of
the residuals might appear, which would indicate a (false) problem of
serial correlation (Gujarati, 1988). Unfortunately, mass communication
researchers have yet to expand constancy or mass media expenditure
beyond the basic mass media expenditures-income relationship.
It is vital to realize from the onset that the relative constancy
hypothesis is not an economic theory per se; it is a proposition (or a set
of generalizations) that originates from the mass media literature,
been conceptualized and tested within the communication discipline. As
demonstrated in the previous section, the principle of relative
is a fairly atheoretical approach to the characterization of mass
expenditures over time; it is primarily supported by empirical
Are we then reaching a cul-de-sac or is it possible to ameliorate our
understanding of this important topic--mass media expenditures-income
relationships--in media economics by investigating mass media spending
an angle other than through the PRC? This section discusses some
theoretical and methodological suggestions for future research on consumer
mass media expenditures.
Constancy: Because the constancy assumption of the PRC is not grounded in
any economic theory, one may argue that it should not even be tested.
lack of theoretical foundation should at least signal to communication
researchers to consider abandoning (or downplaying) the study of this
assumption in favor of more promising theoretically-based mass media
expenditures-income inquiries. This is not to say that the testing of the
constancy proposition is useless or irrelevant; as exploratory
has its role to illuminate potential relationships between mass media
spending and income. But the constancy research has limited
implications due to its lack of proper theoretical grounding and
opportunities to advance the subject of media economics theoretically and
open new avenues for mass media expenditure research. Since the
analyses of consumer mass media expenditures have been conducted within
realm of the PRC. This paper has reappraised the significance of the PRC
in media economics, and suggests if not to sever ties between the PRC
consumer mass media expenditures at least focus greater attention on
alternative models of consumer mass media expenditures.
Future studies should explore other linkages between mass media
expenditures and consumption theories. This is where major theoretical
contributions to the mass media expenditures research could be
To a certain degree, the principle of relative constancy falls under
broad umbrella of mechanistic Keynesian income-consumption tradition
because it assumes that consumers respond immediately to changes in income
levels. If we apply the Keynesian macroeconomic theory to the
microeconomic context of the constancy hypothesis, the absolute income
hypothesis in the long term would suggest a near proportional
between DPI and mass media expenditures (see Froyen, 1990). Likewise,
Friedman's permanent income hypothesis also assumes that consumers
in a backward-looking fashion as they revise their estimate of
income "based on how last period's actual income differed from last
period's estimate of permanent income" (Froyen, 1990, p. 398). The
permanent income hypothesis posits that
the reaction of current consumption to an increase or decrease in current
income depends on the individual's expectations about whether the
or decrease in income is likely to continue in future periods or
a windfall gain or loss. Permanent changes in income levels are
directly affect expected consumption, whereas transitory changes are
assumed to have no effect on expected consumption. (Doran, 1989, p.
In contrast, when we consider the rational expectations perspective in
conjunction with the permanent income hypothesis, we assume that
estimate permanent income by using "all information available prior to
current period" (Froyen, 1990, p. 398), thereby forming
(rational) expectations. The rational expectations perspective
that household consumption follows a stochastic (i.e., random) path
1978). Because of the random walk nature of consumption, no current
variable but the current consumption level should have predictive power
future consumption. Future work testing the permanent income hypothesis
and/or rational expectations perspective with aggregate measures of
media expenditures and wealth would offer valuable theoretical and
ical insights to predict consumer behavior relative to mass media
Functional Equivalence: The principle of relative constancy cannot be
completely rejected on theoretical grounds because the functional
equivalence assumption can be validated in economic terms. Functional
equivalence is an important economic concept (product substitution) that
has practical implications for today's electronic media environment.
Consumer electronics firms and media companies in the United States and
other Western countries are poised to market such innovations as
high-definition television and interactive television (as soon as technical
issues are resolved). Assuming little expansion of household budgets and
time for mass media in the near future, functional equivalence
suggest that "cannibalizing" among electronic media activities and
may reach its full potential in the 1990s, because of the introduction of
new and competing services. For instance, Levy and Pitsch (1985)
that cable TV and the VCR were substitutes. More recently, a survey
conducted by Broadcasting & Cable indicated that of all cable subscriber
respondents interested in interactive TV (53.8%), nearly half (47.2%)
reported that they would drop some existing cable services to offset the
cost of interactive services (Jessell, 1994). In the same vein,
ary studies suggest that direct broadcast satellite (DBS) service and
service are competing products, with cable subscribers switching to DBS
for improved signal quality and expanded programming options (e.g.,
Additional Regressors: If we accept the reasoning for shifting emphasis
away from the PRC to other models of mass media expenditures, it is
imperative to extend mass media expenditures models beyond a simple mass
media expenditures-income relationship, so that we can assess the
economic and non-economic predictors on the evolution of mass media
expenditures, as well as forecast short-term mass media spending based on
the values of those predictors. Whereas the existing PRC research has
ignored the potential influences of variables other than income on mass
media expenditures, future studies should repair this omission by
incorporating such literature-guided independent variables as price,
interest rate, and unemployment rate in models of mass media expenditures.
If variables that could intervene in the process of consumer spending on
mass media are omitted, the mass media expenditure model is likely to
misspecified and produce erroneous estimates.
Static Versus Dynamic Regression: Future studies should develop dynamic
regression models to assess whether changes in mass media expenditures
react immediately to changes in income (or in other variables) or
the effect of income is spread over several time periods. A
model is said to be dynamic if it contains a lagged dependent variable;
dynamic behavior patterns are also represented when the model comprises
lagged independent variables (Doran, 1989). On the other hand, static
models, which best characterize the existing constancy literature, are
those assuming "an instantaneous adjustment to the new equilibrium
when prices and income change" (Phlips, 1983, p. 155). Empirically,
means that static models do not contain lagged variables (i.e.,
at time-1, time-2...). But in reality, consumption, especially
expenditures on such durables as consumer electronics products, often
responds to income changes incrementally over several time
or years. Phlips (1983) has argued that a static approach to
theory does not really offer a realistic picture of how consumers
increases in income. In sum, dynamic models should be favored over
models because they open the door to new conceptual avenues of mass
expenditures research (e.g., testing the permanent income hypothesis
rational expectations perspective requires the inclusion of lagged
variables), as well as new statistical approaches to mass media
expenditures modeling (e.g., use of autoregressive and distributed-lag
Brophy, L. (1995, March 13). Sky high on DBS. Cable World, p. 20.
DeFleur, M. L., & Ball-Rokeach, S. J. (1989). Theories of mass
communication (5th ed.). New York: Longman.
Demers, D. P. (1994). Relative constancy hypothesis, structural pluralism,
and national advertising expenditures. Journal of Media Economics,
Dimmick, J. (1993). Ecology, economics, and gratification studies. In A.
Alexander, J. Owers, & R. Carveth (Eds.), Media economics: Theory and
practice (pp. 135-156). Hillsdale, NJ: Lawrence Erlbaum.
Dimmick, J., & Rothenbuhler, E. (1984). The theory of the niche:
among media industries. Journal of Communication, 34(1), 103-119.
Doran, H. E. (1989). Applied regression analysis in econometrics. New York:
Dupagne, M. (1994). Testing the relative constancy of mass media
expenditures in the United Kingdom. Journal of Media Economics, 7(3),
Eastwood, D. B. (1985). The economics of consumer behavior. Boston, MA:
Allyn and Bacon.
Economic support of mass communications media: 1929-1957. (1959). New York:
Froyen, R. T. (1990). Macroeconomics: Theories and policies (3rd ed.). New
Fullerton, H. S. (1988). Technology collides with relative constancy: The
pattern of adoption for a new medium. Journal of Media Economics,
Glascock, J. (1993). Effect of cable television on advertiser and consumer
spending on mass media, 1978-1990. Journalism Quarterly, 70, 509-517.
Gujarati, D. N. (1988). Basic econometrics (2nd ed.). New York:
Hall, R. E. (1978). Stochastic implications of the life cycle-permanent
income hypothesis: Theory and evidence. Journal of Political Economy,
Hamilton, R. F. (1991). Work and leisure: On the reporting of poll results.
Public Opinion Quarterly, 55, 347-356.
Hanke, J. E., & Reitsch, A. G. (1992). Business forecasting (4th ed.).
Boston: Allyn and Bacon.
Himmelweit, H. T. (1977). Yesterday's and tomorrow's television research on
children. In D. Lerner & L. M. Nelson (Eds.), Communication research--a
half-century appraisal (pp. 9-36). Honolulu: The University Press of
Hornik, J., & Schlinger, M. J. (1981). Allocation of time to the mass
media. Journal of Consumer Research, 7, 343-355.
Houthakker, H. S., & Taylor, L. D. (1970). Consumer demand in the United
States: Analyses and projections (2nd ed.). Cambridge, MA: Harvard
Jessell, H. A. (1994, May 23). New yardstick for interactive TV.
Broadcasting & Cable, pp. 6, 76-78.
Landsburg, S. E. (1992). Price theory and applications (2nd ed.). Fort
Worth, TX: Dryden.
Levy, J. D., & Pitsch, P. K. (1985). Statistical evidence of
substitutability among video delivery systems. In E. M. Noam (Ed.),
media competition: Regulation, economics, and technology (pp.
York: Columbia University Press.
Magrabi, F. M., Chung, Y. S., Cha, S. S., & Yang, S. J. (1991). The
economics of household consumption. New York: Praeger.
McCombs, M. E. (1972). Mass media in the marketplace. Journalism Monographs
, No. 24.
McCombs, M. E., & Eyal, C. H. (1980). Spending on mass media. Journal of
Communication, 31(1), 153-158.
McCombs, M., & Nolan, J. (1992). The relative constancy approach to
consumer spending for media. Journal of Media Economics, 5(2), 43-52.
McCombs, M. E., & Son, J. (1986, August). Patterns of economic support for
during a decade of electronic innovation. Paper presented at the annual
convention of the Association for Education in Journalism and Mass
Communication, Norman, OK.
Noh, G. Y. (1994, August). New media departure in the principle of relative
constancy: VCRs. Paper presented at the annual convention of the
Association for Education in Journalism and Mass Communication, Atlanta,
Ogan, C., & Kelly, J. (1986, August). Time and money spent on the mass
media in an age of new communication technologies: A market study.
presented at the annual convention of the Association for
Journalism and Mass Communication, Norman, OK.
Ostrom, C. W., Jr. (1990). Time series analysis: Regression techniques (2nd
ed.). Newbury Park, CA: Sage.
Phlips, L. (1983). Applied consumption analysis (rev. ed.). Amsterdam:
Pindyck, R. S., & Rubinfeld, D. L. (1991). Econometric models and economic
forecasts (3rd ed.). New York: McGraw-Hill.
Robichaux, M. (1993, November 29). Highway of hype: Despite many claims for
500-channel TV, long road lies ahead. Wall Street Journal, pp. A1, A5
Robinson, J. P. (1981). Television and leisure time: A new scenario.
Journal of Communication, 31(1), 120-130.
Robinson, J. P. (1989, July). Time's up. American Demographics, pp. 32-35.
Samuelson, P. A., & Nordhaus, W. D. (1989). Economics (13th ed.). New York:
Son. J. (1990). The impact of new electronic media on audience support for
mass media. Dissertation Abstracts International, 51, 3263A.
Microfilms No. 91-05,666)
Son, J., & McCombs, M. E. (1993). A look at the constancy principle under
changing market conditions. Journal of Media Economics, 6(2), 23-36.
Thomas, R. L. (1987). Applied demand analysis. New York: Longman.
Vogel, H. L. (1990). Entertainment industry economics (2nd ed.). Cambridge:
Cambridge University Press.
Werner, A. (1986). Mass media expenditures in Norway: The principle of
relative constancy revisited. In M. L. McLaughlin (Ed.), Communication
yearbook 9 (pp. 251-260). Beverly Hills, CA: Sage.
Wood, W. C. (1986). Consumer spending on the mass media: The principle of
constancy reconsidered. Journal of Communication, 36(2), 39-51.
Wood, W. C., & O'Hare, S. L. (1991). Paying for the video revolution:
Consumer spending on the mass media. Journal of Communication, 41(1),
 Trends are described as long-term increasing or decreasing move
ments in a time series
(Hanke & Reitsch, 1992).
ve serial correlation, the most frequently encountered type of serial
correlation, occurs when "a positive error in one period....[is]
associated with a
positive error in the next period" (Pind
yck & Rubinfeld, 1991, p. 49); negative serial
on the other hand, occurs when "negative errors in one time period are
associated with positive errors in the next, and vice versa" (
 Virtually all U.S. PRC studies use national aggregate data fro
m the Department of
Commerce. Mass media categories includ
e: (1) newspapers, magazines, and sheet music; (2)
d maps; (3) radio, TV receivers, records, and musical instruments; (4) radi
TV repairs; (5) motion picture admissions; and (6) ot
her admissions (theaters, opera, and
entertainment of non-p
rofit institutions). Whereas McCombs and his associates performed
correlational analyses using these individual categories as well as
a combination of all
six categories (total mass media), Wo
od ran regression analyses using total mass media
as the only dependent variable.
 Negative and positive constant terms i
ndicate increasing and decreasing shares of DPI
mass media, respectively.
 VCRs, radio sets, TV sets, and musical instr
uments were all included in the electronic
hardware category. So this ca
tegory is actually an aggregate of old and new
. Typically, national statistical offices provide aggregate data for
categories that include several recreational products. This sit
uation illustrates the
difficulty of isolating VCR expendit
ures from other types of electronic media expenditures
and assessing its
effect on the overall mass media spending.
 A 1989 survey of 7,154 resp
ondents in the Houston metropolitan area revealed that
e subscribers spent about the same amount of time with the mass media as
non-subscribers, supporting the contention that the adoption
of cable TV did not lead to a
dramatic increase in time spent on the mass
media (Son, 1990). Unexpectedly, he also
found that VCR o
wners spent less time with the mass media than non-owners, although time
devoted to VCR use was not included in the analysis.
us assume that the prices of X and Y are kept constant at $1 and $2 per un
respectively. Figure 1.A shows the budget lines for in
comes of $6, $10, and $20. With an
income of $10, a consumer can afford
(1) 10Xs, (2) 5Ys, or (3) a mix of Xs and Ys (e.g.,
3Xs), as long as the sum of his/her purchases does not exceed $10. When w
the indifference curves, we can draw the consumer's o
ptima for the three income levels
(for the sake of illustra
tion let us assume that these tangential points represent
uantities of 4, 5, and 8Xs, respectively). For instance, with an income of
optimal consumption level of Xs would be 5; with
an income of $20, an individual would
consume 8Xs. To esti
mate the Engel curve for X, we report on another graph (quantity of X
income) how many Xs individuals consume for incomes of $6 (4Xs), $10 (5Xs)
, and $20
(8Xs) (Figure 1.B).
 Houthakker and Taylor
(1970) estimated the short-term income elasticity (_) for
arious media-related expenditure items in the United States: books and maps
newspapers and magazines, 0.38; radio and televisio
n receivers, records, and musical
instruments, 4.20; radio
and television repair, 0.64; and motion pictures, 0.81. To
illustrate, an income elasticity of 1.67 for books and maps means that whe
income (personal consumption expenditures) inc
reases by 10%, we expect spending on books
and maps to incr
ease by 16.7%. Long-term elasticity figures (_') were: books and maps,
1.42; newspapers and magazines, (not reported); radio and tele
vision receivers, records,
and musical instruments, 2.99; r
adio and television repair, 5.20; and motion pictures,
. The difference between short-term and long-term elasticity lies in the d
temporal adjustment needed to respond to income ch
anges. Landsburg (1992) explains:
"Following an increase in
income, it usually takes time for people to fully adjust their
spending patterns. Thus, we can estimate both a short-run and a long-
elasticity, reflecting an initial partial respon
se to an income increase and the ultimate
full response. W
e expect the long-run elasticity to exceed the short-run elasticity" (p.
109) (see also Thomas, 1987).
 It was recently reported t
hat the U.S. average household income in the early 1990s
clined by $2,000 from its 1989 level and that consumers had less leisure ti
me in 1989
than in the mid 1980s (Robichaux, 1993).
hereas an autoregressive model includes one lagged dependent variable among
its predictor variables, a distributed-lag model contains on
e or several
lagged independent variables in the equation (Gu