|
Are Young People Reading the Newspaper?
A 25-Year Cohort Analysis
by
Nancy Cheever and Tony Rimmer
Department of Communications,
California State University, Fullerton
(714) 278-3217
[log in to unmask]
Paper submitted to the Association for Education in Journalism and Mass
Communications Annual Conference, New Orleans, August, 1999
Abstract
Are Young People Reading the Newspaper?
A 25-Year Cohort Analysis
by
Nancy Cheever and Tony Rimmer
Department of Communications
California State University, Fullerton
Fullerton CA 92834
(714) 278-3271
[log in to unmask]
The newspaper industry has long been concerned that it is losing the young
reader. The research informing this concern is derived largely from
cross-sectional studies. We argue that younger people become older, stable,
newspaper readers and that a cohort analysis approach is needed to understand
this phenomenon.
We look at newspaper readership and its predictors both cross-sectionally and
through cohort analysis in the 25-year cumulation (1972-1996) of the General
Social Survey. We conclude that cohort analysis is a useful tool for
understanding how young people age into the newspaper reading habit.
Are Young People Reading the Newspaper? A 25-Year Cohort Analysis -
Are Young People Reading the Newspaper?
A 25-Year Cohort Analysis
Introduction
A notion exists among today's newspaper editors and publishers that newspaper
reading among young adults has declined dramatically throughout the past two to
three decades. An outgrowth of this idea is the concern that if the newspaper
industry does not reach these young people now, it will lose them as readers
forever. Industry researchers have suggested that publishers target younger
audiences in order to attract them to the newspaper, asserting that if young
people do not pick up the paper now, they never will.
In order to attract the younger reader, news organizations have tried a variety
of tactics. These include re-design, in which more color and graphics, bigger
headlines and shorter stories are incorporated into the newspaper; writing types
of stories that represent teenagers, students and "gen x-ers"; and special
sections designed for young adults.
These tactics seem to have had limited success. The younger audience, in
general, is still not flocking to the newspaper for information.
Our premise is that they never have. The argument we develop here is two-fold -
conceptual and methodological. First, a conceptual idea: Young people have never
read the newspaper, at least at the levels attributed to older readers. We
suspect newspaper reading habits are acquired later in life, say, in a person's
30s. We argue that while the young reader has historically reported lower
newspaper reading levels than older people, as a person ages, his or her
newspaper reading levels increase and level off. Publisher attention, then,
might be more appropriately directed to nurturing the interest of older readers.
Second, a methodological idea: Publishers' concerns have been largely informed
by a cross-sectional, one-shot survey approach to analysis. Cohort analysis,
with its potential to inform through the aging process, might usefully add
insight to the debate. Our concern is that reliance on cross-sectional data
alone may lead to self-fulfilling prophecies of panic regarding the fate and
future of young readers while denying a role for maturation in the development
of newspaper readership. Cohort analysis can give us access to such maturation
processes.
Our data come from a 25-year cumulation of the National Opinion Research
Center's annual General Social Survey (GSS). The data extend from 1972 to 1996,
offering immediate opportunities for the inter-generational study of newspaper
readership and its prediction. Continued funding of the GSS might eventually
allow for the study of media use through lifetimes.
Using both cross-sectional and cohort approaches we consider two research
questions: What has been the pattern of newspaper readership in three different
age groups (18-29, 30-45, 46-89 years) over the past 25 years, and what role
have education and television viewing played in newspaper readership through the
period?
The paper first reviews previous literature on newspaper readership and
identifies reasons why young people might not read the newspaper as often as
older people do. We point out that this literature is based largely on
cross-sectional analyses and we propose that a cohort analysis approach might
qualify and supplement the extant literature. We then examine Americans as they
age and consider how and why newspaper reading levels might increase during this
aging process.
The study relies primarily on graphic displays to support its argument, with
some statistical analyses (correlation and regression) provided to clarify
differences and predictions. Our expectation is that we will find that young
readers differ more in degree than of kind from older readers. We do not doubt
that newer media pose important change problems for older media, but we expect
the addition of a cohort approach to analysis will encourage thinking in a
longer time frame rather than the immediate action called for by apocalyptic
pronouncements, which are informed largely by cross-sectional analysis.
Review of Literature
Studies on newspaper reading habits have shown a decline in newspaper readership
among all age groups in the past 30 years with young people (those under 30)
showing consistently lower levels of newspaper reading than older people (Cobb,
1986; Glenn, 1994; Kirsch, Jungeblut, & Rock, 1988; Robinson, 1978; Stone, 1987;
Times Mirror, 1990a & b). The data for younger readers have sometimes been
interpreted in apocalyptic terms: If the newspaper industry does not reach young
people now, it will lose them as readers forever (Carter, 1994; Fitzgerald,
1990).
The literature is derived almost entirely from cross-sectional studies in which
surveys are launched for one or two years at most and comparisons are made
across age categories within the same study. An assumption can arise from such
analyses - that as they age, the behavior of younger respondents will mimic that
of the older age categories reported in the same study and year. The reality is
that this younger age group will age into older age group categories at some
later date beyond the scope of the study under review. And the older age groups
reported on in the same study will have been in a younger age category at a
point in time earlier than the study reports. Evidence regarding aging, and
changes in media use consistent with that aging, can be better acquired through
cohort analysis. Cohort data may show that change through time is a more
gradual phenomenon than change across age categories at the same time, which is
what cross-sectional data provide.
Further, cross-sectional data can be acontextual. They are developed ad hoc in
response to crises perceived by sponsors. They arise, then, in unique contexts,
for example as an information-gathering tool during a perceived decline in
newspaper readership in a particular demographic. Decisions might be taken based
on such data. But the impact of these decisions beyond the crisis is problematic
because such issues as aging have not been appropriately addressed. Indeed, the
irony is that age as a category has been identified and considered, but aging as
a phenomenon cannot be addressed in such studies because of their
cross-sectional nature.
Exhibit 1 offers a graphic display locating the more important newspaper
readership studies across time. The studies are imposed on time series of
cross-sectional studies divided into three age groups. The data are from the GSS
cumulated data that is the basis of the present study. Apart from locating the
newspaper literature in time, our point in Exhibit 1 is to illustrate the idea
that cross-sectional analyses can be acontextual with regard to time. With the
exception of Glenn's (1994) cohort analysis of newspaper reading, television
viewing and verbal ability, all of the studies identified in Exhibit 1 are
cross-sectional in approach. Each of the studies is linked in our rendering
directly with the youngest age group, 18-29 years. In most cases this was the
age group the studies were concerned with. Most also report on newspaper use in
older age groups as well.
______________________
Exhibit 1 about here
Yankelovitch, Skelly and White's two studies (1976a & b) appear in Exhibit 1 to
have been executed in the middle of a precipitous decline in young people's
readership. Can the plateau in readership following from 1977 to 1986 be
attributed to strategies suggested by Yankelovitch et al? From this plateau,
Robinson's 1978 study asserts there has been no decline in young people's (18-24
years) newspaper reading in the 1965-1975 decade. Our GSS readership measure
would suggest otherwise. Robinson's 1980 study asserts the greatest decline in
postwar readership appeared in the 20-29 age group. Our data extend back to
1972. But we will attempt to make the case that there is no issue of greater or
lesser rates of decline between age groups. We will assert that all age groups
show similar rates of change. Cobb (1986), writing from the tail-end, in our
terms, of a plateau in young people's readership levels found a slight decline
in overall newspaper readership between 1961 and 1984, but that the decline had
been about the same in all age groups. A short-lived increase followed her
publication date.
Stone (1987) reported that industry research had noted that between the years
1967 and 1974 newspaper readership declined four percent among all adults, and
ten percent among "young people" (18-24 year-olds). Consistent with our data
in Exhibit 1, Stone reported that newspaper readership among young adults
"leveled off" in the 1980s. The exhibit also shows it went into a decline from
1987 through to 1991. Kirsch et al (1988), like Yankelovitch et at (1976a & b)
before them, appear caught on a downward trend in young people's readership
which raises questions about the longer term validity of the data they report.
Times Mirror (1990) report on relative readership levels by age over several
years, but their study was executed and published in a brief "oasis" in the
middle of a period when readership levels of young adults appeared to show a
substantially greater rate of decline than that for older age groups.
Bagby (1991) reported a prediction by newspaper design expert Mario Garcia that
in "about four or five years _ the transformation toward reader-driven
newspapers will be complete" (p. 23). Five years from 1991 is 1996, and the
tail-end of Exhibit 1 is not very encouraging towards Garcia's "reader-driven"
newspaper.
The cross-sectional data in Exhibit 1 do suggest a decline in overall readership
levels from 1972 to 1996. Our grand mean decreases by nearly 10 percent over
the period, but that is also about the same level as the standard deviation of
the grand mean's variation though the period. Again, we will assert that the
data for young people will show the same rates of change as do those for other
age groups. The data in Exhibit 1 do tend to confirm this consistency in
decline, but the graph for young people (18-29 years) does appear to show
considerably more variation than do those for the older age groups.
Young people appear to read the newspaper at significantly lower levels than
that for older people, but we assert that the decline in their readership has
matched that of older Americans. In other words, the differences between younger
and older readers may be more one of degree than of kind. From this perspective,
the task for publishers is to change the "degree." All this might require is
patience - waiting for younger people to become older.
Younger media consumers, while somewhat interested in news, are generally more
apt to look to media other than the newspaper to find specific topics that
interest them (Cobb, 1986). While they may occasionally pick up a newspaper for
sports or entertainment or the comics, the mainstream newspaper - which focuses
on political, economical, community and business issues - is not a resource the
younger reader turns to on a regular basis. It does not "speak" to them. The
older reader, on the other hand, is more likely to want a resource that provides
information about a broad range of topics, from economic and political issues,
to crime and human-interest stories - the kinds of material that a newspaper
might offer (Stone & Boudreau, 1995). Knowing this, it may be more important
for newspaper executives to target the age group that is beginning to read the
newspaper on a regular basis - those in their early 30s.
Most newspaper readership studies show young people reporting lower newspaper
reading levels than that for older people. For instance, a Times Mirror survey
(1990b) concluded that among 18-30 year olds, 40 percent read the newspaper. The
findings were offered for decades, showing roughly a 10 percent increase in
newspaper readership with each decade. For example, of 31-40-year-olds, roughly
50 percent read the newspaper; of 41-50-year-olds, 60 percent read the
newspaper.
In studies that attempt to look at trends in cross-sectional data across time,
most conclude there has been a decline in newspaper reading levels among all
people, with many researchers suggesting that the decline has been most drastic
for younger readers. Some, if not all, newspaper executives view this as a
continuing and relentless trend that will lead to a greater decline in overall
newspaper reading (Bagby, 1991; Carter, 1994; Gersh, 1990; Fitzgerald, 1990).
Research by the American Society of Newspaper Editors (ASNE) has fueled this
sense of urgency to attract younger readers (McGrath, 1995). A 24-page booklet
titled "Beyond Survival: Keys to Forging Ties With Potential Readers," was
distributed to all ASNE participants outlining ways in which publishers can
restructure their papers to suit younger readers. A wave of readership studies
in the early 1970s spawned more alternative presses and "reader-friendly"
mainstream newspapers containing a lot of photographs and graphics to attract
the younger reader.
All this activity seemed to have limited success. Robinson (1978), using results
from national polls conducted in 1975-76, found no decline in newspaper use
among people aged 18-24 during those years, but he did find a decline among
older age groups of up to 13 percent. In his 1980 study using GSS data, Robinson
found that among those aged 20-29, there was a 22 percent drop in newspaper
readership. For those aged 30-39, there was a decline of 20 percentage points -
a two percentage point difference. Robinson (1980) suggests that newspapers'
strategies in the 1970s failed to "address the greatest source of attrition -
the older reader."
Industry leaders today worry that young people who do not read the newspaper
will never acquire the habit and that targeting that audience is the best way to
attract the young person to the newspaper. But the literature has suggested that
reaching out to the young reader might not work. Perhaps the steady newspaper
habit of older people should be the industry's main focus. Although we do not
develop the argument here, perhaps undue attention to the younger reader may
even alienate older readers.
So why is the industry so concerned with reaching the younger reader?
Cross-sectional studies appear to show bigger declines in young people's
newspaper reading levels than those of older groups. We have noted from Exhibit
1 that this might not be the case, though it would appear that the younger
newspaper reader does show more variation in their readership than do older age
groups.
The Young Media Consumer
Younger Americans historically report being turned off by the newspaper, and the
literature offers several reasons why.
Yankelovitch et al (1976a & b) identified some of the reasons why young people
do not read the newspaper in what they call "the balance sheet," a diagram from
Part 1 (1976a) of their massive study funded by the Harte-Hanks newspaper chain.
They assert that young people feel the newspaper is not relevant in their lives
and that it does not interest them. They also found that young people find the
newspaper boring, have little time to read it, think the newspaper has too much
information they don't want and too little they do want, never picked up the
newspaper reading habit, and only read the newspaper when they need it.
In Stone's 1987 review of research about newspapers, other reasons why young
people are turned off by the newspaper include lifestyle considerations,
competing media, and the fact that young people are less settled than older
people.
Content preferences, however, seem to prevail. "There is something about the
newspaper itself that has become a turn-off for young adults" Stone notes (1987,
p. 122). He cites studies in which younger people said newspaper reading is an
"old people's habit _ something you see old people do while waiting for a bus";
that they believe newspapers "speak for the status quo and against change in
society"; are cold and impersonal; and a "middle-aged product written and edited
by middle-aged people for a middle-aged audience" (p. 122). Aleff (1995) found
that young people do not trust the media, and think news is too negative and
irresponsible.
Our first research question proposes testing for the perceived decline in young
people's newspaper reading levels by examining data on reported newspaper
reading levels. The question also proposes examining how this apparent decline
compares to that for older age groups. This phase of the study will consider
the issue with cross-sectional data.
Research Question 1: Does newspaper reading among 18-29 year-olds change, in
cross-sectional terms, through the period from 1972 to 1996, and does it differ
from that of other age groups?
Aging Americans' Newspaper Reading Levels
As people age their newspaper reading levels typically increase. The literature
suggests that older age groups consistently show higher newspaper reading levels
than do younger age groups.
There are several reasons why this might be so. We address here ideas of
"settling down" as implied in the aging-stability and community ties hypotheses,
and the role of education. We also discuss a competing hypothesis that as the
population ages it watches more television and that this increase in viewing
might work against increases in newspaper readership.
As people age, attitudes and behaviors tend to stabilize. This phenomenon has
been referred to as the aging-stability hypothesis (Alwin, Cohen & Newcomb,
1991; Glenn, 1994). As people age, they also become less mobile and more
settled. As they settle people develop community ties and subscribe to the
newspaper more often (Demers, 1996; Stamm, Emig & Hesse, 1997). Because people
generally settle down after college, this suggests a higher rate of readership
by older, better educated, individuals (Reina, 1995). Further, those with
children in school tend to subscribe more often to the newspaper than do
younger, single people (Demers, 1993). The newspaper can be an important source
of information for these kinds of people, providing news about crime, community,
politics and economic issues.
Times Mirror (1990) found that those who are well educated are more likely to
read the paper than those who are less well educated. The study, in which nearly
4,900 people were interviewed, found that 55 percent of college graduates read a
daily newspaper; of those with some college education 41 percent reported
reading a daily newspaper; 41 percent of those with a high school diploma and 34
percent of those with less than a high school diploma read a daily newspaper.
Poindexter (1979) and Westley and Severin (1964) found non-readers are less
educated than those who read a newspaper regularly.
Because young people are more mobile, less educated and have more varied
lifestyles than the older audience, this should be an indication that the
newspaper is not a resource that they regularly turn to. But as people mature,
their interests change and the newspaper becomes a valued resource of
information.
The role of competing media on newspaper reading levels is less clear, if only
because the diversity of these media has left the question unsettled. We
consider only television here. The Roper research organization claimed that
television first surpassed newspapers in 1963 as the public's preferred medium
for news (Roper, 1964; but see Rimmer & Weaver, 1987; Stempel, 1991). Times
Mirror (1990) concluded that in each age category (by decades), television news
watching was at least ten percentage points higher than newspaper reading among
all people interviewed. Poindexter (1979) found that the main reason given for
not reading the newspaper was because the respondents got their news from the
television. Sixty percent of her non-readers said they got their news from the
TV.
There is a notion abroad of late that the Internet is, or will soon be, the top
resource for news. The literature suggests that younger people do turn elsewhere
to get information that interests them. It has not been shown that that medium
is the Internet.
General television viewing (i.e., non-news viewing) is negatively associated
with education and with newspaper reading (Comstock, Chaffee, Katzman, McCombs &
Roberts, 1978). We might conclude, then, that television viewing levels will be
negatively associated with newspaper reading levels. If the typical American
daily newspaper reader is one who is older, better educated, and more settled
into his or her community (Reina, 1995; Stamm et al., 1997; Stone, 1987), it
would seem likely that even with cultural, environmental, and technological
changes, young people will grow into the newspaper reading habit. This is the
idea informing our second research question. The question explores newspaper
reading levels as people age through time and how education and television might
be associated with this newspaper reading. This question can be explored using
cohort analysis.
Research Question 2: What is the pattern of reported newspaper readership in
different age cohorts from 1972 through 1996?
a. Does the 18-29 year-old cohort show an increase in levels of newspaper
reading as it ages and how do the other age groups change as they age?
b. What have been the roles of education and television viewing as predictors of
newspaper reading?
Method
This study involves the secondary analysis of extant data. The data come from a
cumulation of the General Social Survey (GSS), an annual survey conducted by the
National Opinion Research Center since 1972 (Davis & Smith, 1994). The surveys
are designed as part of a program of social indicator research, replicating
questionnaire wording in order to facilitate time trend studies. The cumulation
here extends from 1972 to 1996. There was no GSS survey in 1979, 1981, 1992,
1995 and 1997.
The data reported here involve face-to-face personal interviews with people
representing the total non-institutionalized English-speaking population of the
United States, 18 years and older. The sample size each year is about 1,500,
with a total population in the data set reported here of more than 35,000
respondents. The samples are probability samples: block quota sampling was used
in 1972-74 and for half of the 1975 and 1976 surveys. Full probability sampling
was employed in 1977, 1978, 1980, 1982-1991, 1993-1994, 1996 and half of the
1975 and 1976 surveys.
A Black oversample was taken in 1982 and 1987. The oversample in 1982 was 354
persons beyond the average of 140 Blacks reported in the 1972 through 1982
samples. In 1987 it was 353 beyond the average 170 Blacks in samples from 1983
through 1987. We did not adjust for this oversampling in the present analysis.
Given lower education levels in such subsamples we might expect to see lower
newspaper reading levels and higher television viewing levels in the data for
1982 and 1987.
The GSS went into the field every two years in 1994 and 1996 following a change
in sampling and surveying strategy. Sample sizes were dramatically increased to
facilitate split sample survey practices. The sample sizes in 1994 (n=2992) and
1996 (n=2904) were twice the size of previous surveys. We did not adjust for
this change in sample size. We assumed standard errors would be reduced with the
larger sample sizes but felt this would have little impact on our findings.
This study uses four variables from the GSS: Our dependent variable is a
five-point report of newspaper reading. Our independent variables are: Age (in
years), education (years of schooling completed), and TV viewing (hours per day
respondent watches television). The age and education questions are available
for every year. The newspaper reading question was asked in the years 1972,
1975, 1977-1978, 1982-1983, 1985-1991, 1993-1994 and 1996. The TV viewing
question was asked in 1975, 1977-1978, 1980, 1982-1986, 1988, 1989-1991,
1993-1994. The TV viewing question was not asked of the Black oversample in
1987.
Question wording for most of our variables was the same each year. For example,
the newspaper reading variable was consistently asked for as: "How often do you
read the newspaper?" The answers are "every day," "a few times a week," "once a
week," "less than once a week," "never." In the GSS, the answers are coded so
that 1= "every day" and 5= "never." For this study, the codes were reversed so
that 1= "never" and 5= "every day." The variable was treated as interval.
For the age category, the question was changed after 1975 from "In what year
were you born?" to "What is your date of birth?" For television viewing hours,
the question wording in all years is, "On the average day, about how many hours
do you personally watch television?" The answers are from 0 to 24. For the
education variable, several questions were asked about level of schooling
completed. The answers are 0 to 20 years. We assumed a consistent pattern in the
response set.
This study uses cohort and cross-sectional analyses. Cohort analysis is a
longitudinal study approach which lends itself to prospective research.
Cross-sectional research is more in the nature of the one-shot, retrospective,
study. A cross-sectional study is one in which data are collected from a
representative sample at only one point in time (Wimmer & Dominick, 1997). Where
data are broken out by age, for example, comparisons between age groups are made
at one point in time, across the one data set. A cohort analysis is one in which
a characteristic of one or more age cohorts is compared at different points in
time (Glenn, 1977; Ryder, 1968). It might be seen as a repeated cross-sectional
study in which the analyst increments the groups of interest year by year
through the data. It is not a panel study where the analyst returns to the same
respondents. In cohort analyses, the analyst returns to the same classes of
respondents.
Our "base cohort" is a group of young adults who are between the ages of 18 and
29 in the first year of our study, 1972. We call our cohorts "age cohorts"
because we are grouping them together by age and using their age as a
descriptive. Other authorities refer to cohorts as "birth cohorts" because they
label their cohorts according to their year of birth (Glenn, 1977). The concept
is the same in both instances.
Exhibit 2 graphically illustrates some of the differences between
cross-sectional and cohort data as viewed in our study.
______________________
Exhibit 2 about here
The upper line in the graph is a report of mean newspaper reading scores for the
18-29 year-old age cohort. The lower line is the mean newspaper reading score
for the cross-sectional 18-29 year-old age group. The cohort line indicates a
1972 start for the cohort, which ranges in age from 18 to 29. It might help to
consider the line itself as a mean age of 24 representing the cohort. Each year
the cohort ages one year through the 25 years through to 1996. The
cross-sectional line is a report of mean newspaper reading for the 18-24 age
group in each year through the 25 years of the study. So, the cross-sectional
group's age stays at 18-29 throughout the 25 years of the study, whereas the
cohort ages year by year. In this exhibit, the differences between the two lines
are substantial. The cross-sectional data show a decline over the 25 years from
highest to lowest of about 17 percent. The cohort data show a decline from
highest to lowest value of about half that: eight percent. An inspection of the
graph suggests the cross-sectional data do show a decline, whereas the cohort
data show no real change from 1972 through 1996.
In this study, we take cohort groups of interest and age them through the data.
Comparisons within cohorts are intra-cohort, comparisons between cohorts are
inter-cohort. The intra-cohort study is similar to a panel study, in which the
same people are compared at different points in time. The difference is that the
cohort study compares a sample of individuals, not the same individuals. The
panel study examines individual changes, while the cohort study looks at trends
(Glenn, 1977). The advantages of using cohort analyses are that one group can
be tracked through time and subsequently trends of a population can be asserted
by generalizing the reported behaviors of the cohort group.
Cohort analyses are typically examined through charts and graphs, which show
three types of effects: age effects, cohort effects and period effects. Our
interest here is in age effects - the effect of age on levels of newspaper
reading. However, in this type of analysis, age, period and cohort effects are
confounded with one another. This problem in cohort analysis has been referred
to by Glenn (1994) as the "age-period-cohort conundrum" and by Mason and her
associates (Mason, Mason, Winsborough & Poole, 1973) as the "confounding" of the
joint effects of aging, period and cohort effects. Glenn (1994) explains the
conundrum as follows: "Since each of the three variables is a perfect linear
function of the other two, linear effects of the variables are confounded with
one another in any relevant data, whether they are cross-sectional, longitudinal
or repeated cross-sectional" (p. 220). Exhibit 3 illustrates how the three
variables can entangle each other in this age, period, cohort conundrum.
______________________
Exhibit 3 about here
Exhibit 3 is a combination plot of three measures for the 18-29 year-old cohort:
newspaper reading, education, and television viewing. The age, period, cohort
conundrum might be illustrated with the following question: "Why does mean
education show an increase through the 25 years of the data? Shouldn't it be
level?" Education shows a steady increase from grade 12.5 in 1972 to grade 14
in 1996. Some of the younger members of the cohort were likely still pursuing
their formal education early in the study and that might help explain the
increase through the first half of the study. Call this an age effect. But
education continues to increase after this. Perhaps there is a problem in the
way GSS built its samples, such that the mean education of the sample increased
through the study. Perhaps education levels have increased in the population
through the period 1972-1996 such that we are a better educated society? Call
this a period effect. Perhaps the cohort has changed in makeup through the
course of the study? More women and minorities are coming to the GSS each year
better educated than before. This is reported in the study as an increase in
level of education of the cohort. Call this a cohort effect. So, again, why does
education appear to show an increase in Exhibit 3? A first glance might suggest
the cohort is better educated as it ages through 1996. But this apparent
increase is deceptive. The standard deviation for the cohort is nearly two
grades, more than the apparent increase shown in the exhibit. This variation
probably obscures any period effect. The apparent increase, then, may be due to
the entangled age, period and cohort effects. Or worse, the increase we think we
see may not be real at all. Glenn (1994) suggests that disentangling the
conundrum requires the introduction of external data streams to the analysis. We
do not have such data available to the present study.
Preliminary analyses were executed in the statistical program SPSS. As a means
of data reduction all cohort and cross-sectional data for newspaper reading,
education and television viewing were reduced to mean scores for groups and for
years. These means, their n's and standard deviations were transferred to
Microsoft Excel spreadsheets. Graphing and the correlation and regression
computations offered here were executed in Excel using the means as raw scores.
Lag options were attempted in the regression computations. It would seem
intuitive that education might not manifest itself as a predictor of newspaper
readership immediately. Newspaper readership therefore was lagged in the
regression computations from zero through two years. Serial correlation was not
addressed in these computations.
Where the GSS did not go into the field in particular years missing data were
substituted using a smoothing technique. Scores from years surrounding missing
years were averaged according to the number of missing cells in the interval.
These constructed data provided cleaner looking graphs. These constructed data
for missing years were excluded in statistical computations.
No smoothing was effected in the graphing of substantive data runs. Smoothed
grand means are offered in some graphs (e.g., in Exhibit 1). This smoothing was
effected with a three-point moving average, typically iterated two to three
times to deliver a useful smooth.
Results
In this section we first report some of the general characteristics of our data,
then we describe the data associated with each of our three cohorts. Finally a
series of exhibits and statistical tests are offered in association with the
reporting of our analyses of the study's research questions.
Exhibit 4 offers descriptive statistics for our three age groups and three
variables of interest. Differences in the n's for the three variables are
artifacts of the number of years the questions were asked.
______________________
Exhibit 4 about here
Exhibit 4 shows that the mean newspaper reading score was more than 3.5 for all
of the age groups studied. More impressively, the mean newspaper reading score
for all three age groups was greater than 3.4 in each year in the data. If a
score of three means the respondent read the newspaper once a week, then the
average person aged 18 to 29 reported reading the newspaper at least "once a
week" to "a few times a week." So young people report they are reading the
newspaper. They are just not reading newspapers as often as older Americans are.
Further, the oldest age group (age 46-89 years) reported mean newspaper reading
levels of at least 4.0 throughout the data, meaning this oldest group reported
reading the newspaper at least "a few times per week" throughout the tested
years.
Our first research question dealt with the perceived drastic decline among the
18-29-year-old group. The question was as follows: Does newspaper reading among
18-29 year-olds change, in cross-sectional terms, through the period from 1972
to 1996, and does it differ from that of other age groups? The short answer is
yes and yes.
Exhibit 1 compares the mean newspaper reading scores of the three age groups
from 1972 to 1996. Newspaper reading among the 18-29 year-old age does change -
it shows a decline. The 18-29-year-old group started with a mean newspaper
reading score of 4.07 in 1972 and ended with a score of 3.42 in 1996 - a
difference of .53 (16 percent).
In order to compare the rates of change between the three age groups a Pearson
correlation was computed. No adjustment was made for any serial correlation. The
correlation coefficients are offered in Exhibit 5. We speculated that a
significant correlation between the age groups would indicate similar rates of
decline, while no relationship would indicate the age groups were independent of
each other.
______________________
Exhibit 5 about here
Exhibit 5 shows a strong positive correlation (r=.85, df=23, p<.05) between the
18-29 and the 30-45 year-old age groups, and a moderate positive correlation
(r=.43, df=23, p<.05) between the 18-29 and the 46-89 year-old age groups. This
indicates that the three age groups have tended to decline at about the same
rate. We had earlier spoken of issues of degree and of kind. The correlation
analysis suggests there is no difference in kind among the three age groups with
regard to rates of decline.
Exhibit 1, however, does indicate differences of degree. The 18-29-year-olds -
the "young" group - show a consistently lower pattern than the older groups.
Using a mean of the grand means in Exhibit 1 as our base (mean score 4.13) we
find that the two older age groups cluster nearer to the grand mean (30-45
mean=4.14; 46-89 mean=4.23) than does the 18-29 year-olds with its substantially
lower mean of 3.73.
Research Question 2 asked the following:
What is the pattern of reported newspaper readership in different age cohorts
from 1972 through 1996? Two additional questions were posed:
a. Does the 18-29 year-old cohort show an increase in levels of newspaper
reading as it ages and how do the other age groups change as they age?
b. What have been the roles of education and television viewing as predictors of
newspaper reading?
Exhibit 6 is an expanded plot of mean newspaper reading levels for our three age
groups.
______________________
Exhibit 6 about here
Research Question 2 first asked about patterns of readership amongst the three
cohorts. The general pattern is one of substantial differences in 1972 with
convergence through 1996. A further characteristic of this general pattern is
that the convergence is not one in which the two younger cohorts' newspaper
reading levels moving up through time to the newspaper reading levels of the
oldest cohort. Rather we see something more akin to a regression, in which all
three cohorts converge to a more central level. This is not what we expected. We
had assumed that reading levels would increase with age, and consistent with
that idea was the assumption that the oldest age group would serve as the model
toward which the younger cohorts would move. The rates of change are somewhat
deceptive as offered in the expanded plots in Exhibit 6. A closer inspection of
the vertical scale in the exhibit reminds us that the changes are not as
dramatic as they might appear to be.
Part a of Research Question 2 asked about patterns of readership in the 18-29
year-old cohort. For the purposes of interpretation, assume a mean age for the
cohort of 24. By 1978 the cohort is aged 30 and at this point a gradual increase
in newspaper reading is reported through to age 45 in 1993. This is consistent
with our expectations that as people age their newspaper reading levels would
increase. (The consistent drop in reading levels across all three cohorts for
1994 and 1996 is problematic so we choose not to incorporate it into the
discussion here).
Part a of the question also asked about patterns of newspaper readership in the
other two cohorts.
As we expected, the older age group (46-89 year-olds) showed the most
consistency in levels of readership. We attribute the substantial dip in 1987
for this cohort to the Black oversample in this year. We had anticipated that
the oversample might have some impact on newspaper reading.
The most dramatic pattern is that for the 30-45 year-old cohort. It shows a
substantial decline through the period. Again, for the purposes of
interpretation, assume a mean age for this cohort of 38 in 1972. This group
reports the highest reading levels near 4.6 in 1972. They age to 63 by 1996 and
at that point report reading levels of about 4.1. Yet our data analyses show
that across the data set as a whole, respondents aged 65 showed the highest
levels of newspaper readership. The lower levels of readership reported by our
30-45 year-old cohort when they reach 63 years of age may be a reminder that
this cohort is somewhat unique and should not be regarded as typical of all
cohorts. We address this point further in our discussion section.
Part b of Research Question 2 asked the following: What have been the roles of
education and television viewing as predictors of newspaper reading? This
question was resolved in two ways: graphically, and with a series of multiple
regression computations executed on data from the 18-29 year-old cohort.
Exhibit 3 shows a combination plot for the 18-29 year-old cohort of mean
newspaper reading, television viewing and education. A caution: the ordinate
(Y-axis) is a combination scale. A forced break has been introduced on the
ordinate to accommodate the lower numerical values associated with television
viewing and newspaper reading and the larger values associated with levels of
education.
The plot of newspaper reading in Exhibit 3 appears to be relatively static
through the 25-year period of the study. Education shows an apparent increase
through the period so we might expect it to be a positive predictor of newspaper
reading. Television viewing levels show a decline in the plot, so we expect that
if it is a predictor of newspaper reading it will be a negative predictor.
Newspaper reading was regressed on education and television viewing in a series
of lagged computations. We review here the zero, one- and two-year lags of
newspaper reading. At zero lag the regression equation accounted for nearly 27
percent of the variance (R2= 26.75) in newspaper reading. Education was a
significant predictor (Regression coefficient=.25, p<.05) but television viewing
was not. The regression equation failed to reach significance when newspaper
reading was lagged by one year, but did show significance when newspaper reading
was lagged two years (R2= 50.86, Regression coefficients: Education=.37, p<.05,
TV viewing=.39, p<.05). The finding that education in particular might have an
impact on newspaper reading two years later is intuitively appealing. Given we
did not address the issue of serial correlation in our analyses, however, this
issue needs more work before any support can be claimed for the finding.
Conclusion
This study's goal was to better understand the newspaper reading behavior of
young adults. We asked two general questions: What are the patterns of
readership in cross-sectional and cohort-derived data for young readers and how
might they differ from those of older age groups? We have found differences in
the patterns between cross-sectional and cohort-derived data and between the
readership behaviors of younger and older cohorts. These differences are such
that we now feel ready to offer some preliminary conclusions to the industry.
Differences are more of degree than of kind. We mean by this that younger
readers do report reading the newspaper but at lower levels than do older age
groups. Their newspaper reading does not appear to be so different in its
patterns that we might venture it as a different kind of reading. This should be
reassuring to the industry. Our expectation that younger people will come into
the newspaper reading fold appears to be supported.
The differences there are between cross-sectional and cohort data show up most
dramatically with younger readers. Exhibit 2 shows this clearly with the
diverging plots of cross-sectional and cohort data. Other plots we have done on
the same data for the two older age groups show markedly less divergence. This
suggests to us that cohort analyses can be most useful in understanding how
younger people age into their 30s and the life stage that we have argued signals
the beginning of stable newspaper reading. We have found the newspaper reading
habit to be most stable in our oldest age group. So, cohort analysis might not
be so useful a research tool with older age groups, at least with regard to
newspaper reading.
We have two recommendations for future research. Our first recommendation is to
develop better measures and controls in the data we have worked with here. Our
newspaper readership measure is a five-point variable we have treated as an
interval measure. There is a need for more diversity in type and scale in the
media use measures and their predictors in the GSS. Age might usefully be
extended back beyond the GSS start of 18 years. The media behaviors we are
interested in get their start long before respondents turn 18. GSS samples need
to deliver more younger and older people. For example, across the 25 years of
the cumulated data set there were just 115 18-year olds and 47 88-year olds.
With regard to controls, several methodological issues arise that could be
addressed in future research. We did not address the issue of serial
correlation, or weighting for the oversampling and split sampling that
characterizes many of the GSS surveys. And we did not consider a point Glenn
reported he adjusted for in his 1994 cohort study - that the GSS samples by
households rather than by individuals, yet it is individuals we are attempting
to study.
Our second recommendation is that we need to consider cohorts more as serial
phenomena. In the present study we have evaluated just three parallel cohorts,
all of which we started at one point in time, 1972. This narrow approach limits
the potential that these rich GSS data have to offer. There is a need for future
analysts to develop the facility to simultaneously analyze many cohorts over
many years. Exhibit 7 might be seen as the beginnings of an attempt to address
this idea. The exhibit shows newspaper readership data for single-year cohorts
with starts at 18 years of age in 1962, 1972, and 1982, and ages these cohorts
through the 25 years of the GSS data set.
______________________
Exhibit 7 about here
References
Aleff, A. (1995). Research finds kids do read newspapers. ASNE Bulletin. No.
766, 5-7.
Alwin, D.F., Cohen, R.J., & Newcomb, T.M. (1991). Political attitudes over the
life span: The Bennington women after 50 years. Madison: University of Wisconsin
Press.
Bagby, M.A. (1991, April). Transforming newspapers for readers. Presstime,
18-25
Carter, M. G. (1994, October). How to think young. Presstime, 31-39.
Cobb, C. J. (1986). Patterns of newspaper readership among teenagers.
Communication Research, 13 (2), 299-324.
Comstock, G., Chaffee, S., Katzman, N., McCombs, M., & Roberts, D. (1978).
Television and human behavior. New York: Columbia University Press.
Davis, J. A., & Smith, T. W. (1994). General social surveys, 1972-1994:
Cumulative codebook. Chicago: National Opinion Research Center.
Demers, D.P. (1996, Summer). Does personal experience in a community increase or
decrease newspaper reading? Journalism and Mass Communication Quarterly, 73 (2),
304-315.
Demers, D.P. (1993). Community attachment, social priming and newspaper reading.
Paper presented at the annual meeting of the Association for Education in
Journalism and Mass Communication, Kansas City, Mo.
Fitzgerald, M. (1990). Most young people are not reading newspapers: latest
survey warns newspapers that the situation is worse than it was originally
believed to be, Editor & Publisher, 123 (17), 13.
Gersh. D. (1990). Older people read newspapers more. Editor & Publisher, 123
(30), 17.
Glenn, N. D. (1977). Cohort Analysis. Series: Quantitative Applications in the
Social Sciences No. 07-005. Beverly Hills, Ca.: Sage Publications.
Glenn, N. D. (1994). Television watching, newspaper reading and cohort
differences in verbal ability. Sociology of Education, 67 (3), 216-229.
Kirsch, I. S., Jungeblut, A., & Rock, D.A. (1988, April). Reading newspapers:
The practices of America's young adults: a summary. Condensed version of a
report presented at the meeting of the Literacy Committee of the American
Society of Newspaper Editors, Washington, D.C.
Mason, K. O., Mason, W. M., Winsborough, H. H., & Poole, W. K. (1973). Some
methodological issues in cohort analysis of archival data. American Sociological
Review, (38), 242-258.
McGrath, K. (1995). Beyond survival: keys to forging ties with potential
readers. American Society of Newspaper Editors Bulletin, 3-32.
Poindexter, P. M. (1979). Daily newspaper non-readers: Why they don't read.
Journalism Quarterly, 56 (4), 764-770.
Reina, L. (1995, November). Who's reading newspapers? USA weekend survey offers
some answers. Editor & Publisher, 24-25.
Rimmer, T., & Weaver, D. (1987). Different questions, different answers? Media
use and media credibility. Journalism Quarterly, (64), 1, 28-36.
Robinson, J. P. (1978, September). Daily news habits of the American public.
ANPA News Research Report No. 15.
Robinson, J. P. (1980, Winter). The changing reading habits of the American
public. Journal of Communication, (30), 141-152.
Roper, E., & Associates (1964). New trends in the public's measure of television
and other media. New York: Television Information Office.
Ryder, N. B. (1968). Cohort analysis. In Sills, D. L. (Ed.), International
encyclopedia of the social sciences, Vol. 2. pp. 546-550.
Stamm, K. R., Emig, A.G. & Hesse, M. B. (1997). The contribution of local media
to community involvement. Journalism Quarterly, 74 (1), 97-107.
Stempel, G.H. (1991). Where people really get most of their news. Newspaper
Research Journal, 12 (4), 2-9.
Stone, G. (1987). Examining newspapers: what research reveals about America's
newspapers, 20. Audience attention to newspapers (pp. 108-126). SAGE
Publications.
Stone, G., & Boudreau, T. (1995). Comparison of reader content preferences.
Newspaper Research Journal, 16 (4), 13-28.
Times Mirror Center for The People & The Press. (1990a, June 28) The age of
indifference.
Times Mirror Center for The People & The Press. (1990b, July 15). The American
media: Who reads, who watches, who listens, who cares.
Westley, B., & Severin, W.J. (1964). A profile of the daily newspaper nonreader.
Journalism Quarterly, 41, 45-50.
Wimmer, R.D., & Dominick, J.R. (1997). Mass media research: An introduction. New
York, Wadsworth.
Yankelovitch, Skelly & White, Inc. (1976a, May) Young people and newspapers: An
exploratory study, part I. Harte-Hanks Newspapers, Inc. San Antonio, Texas.
Yankelovitch, Skelly & White, Inc. (1976b, May) Young people and newspapers: An
exploratory study, part II. Harte-Hanks Newspapers, Inc. San Antonio, Texas.
Exhibit 4
Descriptive Statistics for Three Cross-Sectional Age Groups
(18-29 years, 30-45 years, 46-89 years), 1972-1996.
Newspaper Reading
Television Viewing
Education
Age
n
mean
st dev
n
Mean
st dev
n
mean
st dev
18-29
5,387
3.71
1.15
4,610
3.21
2.25
8,153
12.66
1.91
30-45
7,694
4.11
1.10
6,529
2.64
1.90
11,559
13.07
2.79
46-89
10,177
4.21
1.11
8,664
3.21
2.05
15,335
10.83
3.24
Totals
23,258
4.11*
19,803
2.98*
35,047
12.36*
* Total means offered as grand means, not as average of age category means
Exhibit 5
Correlation Matrix Showing Zero Order Inter-Cohort
and Cross-Sectional Associations
Grouped Age Cohorts
(Age in 1972)
Cross Sectional
Age Groups
18-29 Years
30-45 Years
46-89 Years
18-29 Years
30-45 Years
46-89 Years
18-29 Age Cohort
1.0
30-45 Age Cohort
-.23
1.0
46-89 Age Cohort
-.19
.29
1.0
18-29 Age
Cross Section
.03
.71*
.28
1.0
30-45 Age
Cross Section
-.02
.80*
.34
.85*
1.0
46-89 Age
Cross Section
-.22
.51*
.96*
.43*
.54*
1.0
* Pearson correlation coefficient significantly different from zero, p<.05
|
