Content-Type: text/html

This paper was presented at the Association for Education in Journalism and
Mass Communication in Toronto, Canada, August 2004.
        If you have questions about this paper, please contact the author
directly. If you have questions about the archives, email
<a href="/cgi-bin/wa?LOGON=A3%3Dind0410E%26L%3DAEJMC%26E%3D8bit%26P%3D320030%26B%3D--%26T%3Dtext%252Fplain%3B%2520charset%3Diso-8859-1" >[log in to unmask]</a> For an explanation of the subject line, send email to
<a href="/cgi-bin/wa?LOGON=A3%3Dind0410E%26L%3DAEJMC%26E%3D8bit%26P%3D320030%26B%3D--%26T%3Dtext%252Fplain%3B%2520charset%3Diso-8859-1" >[log in to unmask]</a> with just the four words, "get help info aejmc," in the
body (drop the "").
(Oct 2004)
Thank you.
Elliott Parker
************************************************************************

Quarter Position Effect during Super Bowl Broadcast : How adverting
effectiveness changes as a game goes on

Yong-Ick Jeong
Ph.D. Student
Address 105 BPW Club Rd. #A
Carrboro, NC 27510
Phone: 919-843-5860
Email: <a href="/cgi-bin/wa?LOGON=A3%3Dind0410E%26L%3DAEJMC%26E%3D8bit%26P%3D320030%26B%3D--%26T%3Dtext%252Fplain%3B%2520charset%3Diso-8859-1" >[log in to unmask]</a>

Koang-Hyub Kim
Ph.D. Student
And
Xinshu Zhao
Associate Professor

School of Journalism and Mass Communication
University of North Carolina at Chapel Hill





Paper submitted to advertising division
Association for Education in Journalism and Mass Communication 2004
 Abstract
The primary goals of this study are to investigate the ad effectiveness
originating from placing ads in different game segments and to suggest
marketing implications based on these game segment position effects,
especially in media planning strategy. The primacy effects were observed
from the results. The brands advertised in earlier quarters are more
remembered than those in later quarter. However, ad liking is not related
with the quarter-based position. Marketing implications for the results are
discussed.

 Order Effects in the Super Bowl: The Examination
of the Relationship between the Order of Ads and the Effectiveness of Ads

Since the first game in 1967, the Super Bowl has been positioned as the
most beloved game in the U.S. More than 100 million people watch the game
(Quindt, 2003), and television sets in four out of ten households are tuned
into the game (Kanner, 2004). In addition, on Super Bowl Sundays, more
people gather than on New Year's Eve (Kanner, 2004), and even crime rates
drop significantly (Lamb, 1992). The Super Bowl has become an American
institution.
The Super Bowl is also an event for celebrating advertising. New
advertisements are introduced, and those ads often become the topics of
discussion among people and the media the next morning. However,
advertising during the Super Bowl requires a much higher financial
investment than does advertising on other programs. In 2004, it cost $2.25
million for a 30-second commercial, a new high for Super Bowl advertising
(Linnett, et, al., 2004). Advertising is arguably one important factor that
makes the Super Bowl the biggest media event of the year.
The Super Bowl is divided into seven sectors based on quarter-based game
sequence. They are the pre-game show, first quarter, second quarter, half
time show/report, third quarter, fourth quarter, and post-game show.
Advertising placement is determined based on this quarter-based (game
sequence) segmentation. Thus, this quarter-based order division is
important not only to Super Bowl players, but also to Super Bowl
advertisers because ordering differences may influence the effectiveness of
the ad, which costs millions of dollars.
The quarter-based segmentation provides researchers with an ideal tool with
which to examine advertising effectiveness based on quarter differences in
the Super Bowl. Using the order effect theories of social psychology, this
study will explore whether quarter-based partitions influence ad
performance, and which sector (quarter) provides the best advertising
environment.
The primary goals of this study are first, to develop better conceptual
understandings of ad effectiveness on quarter-based order differences;
second, to understand insightful theoretical underpinnings of order effect;
and third, to produce useful marketing implications, especially in media
planning strategy.

The Super Bowl and Super Bowl Advertising: An Overview
The first Super Bowl was played at the Memorial Coliseum in Los Angeles,
California, in 1967, when National Football League (NFL) and American
Football League (AFL) were merged (Zimmerman, 1995). Since then, the Super
Bowl has become the most popular event not only in the sports, but also in
any televised program throughout the year (Kanner, 2004). In fact, eight of
the eleven most watched television programs were Super Bowl games,
indicating how popular this game is in American society (Kanner, 2004).
Americans love the Super Bowl regardless of their gender. Around 74% of men
and 61% of women watch the game (Hugick, 1992), and females are composed of
43% of Super Bowl viewers (McAdams, 1999). However, not all Super Bowl
viewers are considered die-hard football fans (Callahan, 1991). About three
out of four viewers are regarded non-football fans (Kanner, 2004).

Order Effects
The Super Bowl consists of seven segments based the quarter-based game
sequence: pre-game show, first quarter, second quarter, halftime
show/report, third quarter, fourth quarter, and finally, post-game show.
Thus, the advertisements seem to be simply positioned in the game based on
this classification. However, the partitions of Super Bowl advertising is
more complicated than the classification above because of the segue between
segments. For instance, what should one call the advertisements between the
first quarter and the second quarter? The advertisements placed in the
Super Bowl are classified again as the following: (1) ads before the
pre-game show, (2) ones in the pre-game, (3) ones between the pre-game and
the first quarter, (4) ones in the first quarter, (5) ones between the
first quarter and the second quarter, (6) ones in the second quarter, (7)
ones between the second quarter and the halftime show/report, (8) ones in
the halftime show/report, (9) ones between the half time show/report and
the third quarter, (10) ones in the third quarter, (11) ones between the
third quarter and the fourth quarter, (12) ones in the fourth quarter, (13)
ones between the fourth quarter and the post-game show, (14) ones in the
post-game show (15) ones after the post-game. This classification provides
a practical mechanism for investigating the effectiveness of Super Bowl
advertising based on its presenting order.
According to social psychology theories, generally, items (i.e., stimuli)
in the first and last positions are found to be more effective in
memory-based evaluations (e.g., recall and recognition) than those in the
middle positions (Burke & Drull, 1988; Dean, 1980; Haugtvedt & Wegener,
1994). The tendency to remember items in earlier positions in a sequence is
known as the primacy effect, and the tendency to remember items in last
positions is known as the recency effect. Both primacy and recency effects
have been actively used to explain the order effects in various areas:
sense of taste (Dean, 1980), survey questionnaire (Becker, 1954; Blunch,
1984; Coney, 1977; Friedman & Friedman, 1994; Krosnick & Alwin, 1987;
Landon, 1971), product test (Day, 1969), magazine readership (Sekely &
Blakney, 1994; Whipple & McManamon, 1992), and the relationship among
verbal, auditory, and visual stimuli (Sharps, et, al, 1996).
According to previous studies, order effects (i.e., primacy and recency)
are found to be moderated by various factors: emotional and cognitive
statuses, processing motivation, and message relevancy. According to Crano
(1977), a significant primacy effect occurs when people are under
conditions of low arousal or attention. Using an attention decrement
approach, Crano (1977) explained that, because attention to the stimuli
over a complete list progressively declines, the significance of later
stimuli is less heavily weighed than that of earlier stimuli during the
interpreting process. Krosnick and Alwin (1987) determined the relationship
between order effects and personal cognitive statuses. They found that
respondents whose cognitive sophistication was low were more influenced by
changes in response orders. For those respondents, items placed in the
first three on the list were more chosen as the three most important
qualities for a child.
Haugtvedt and Wegener (1994) examined the relationship between message
relevancy (i.e., high personal relevance vs. low personal relevance) and
order effects (i.e., primacy vs. recency) using the Elaboration Likelihood
Model (ELM). According to their study, the levels of personal elaboration
moderate the types of order effects. They found that when personal
relevance of a certain issue was high (i.e., high motivation to process),
the primacy effect dominated the entire cognitive process; therefore, an
initial message had a significant impact on final judgments and led to
better recall. On the contrary, when personal relevance of the issue was
low (i.e., low motivation to process), the recency effect occurred; thus, a
later message had a greater impact on final judgments and led to better
recall.
Order effects have been also investigated with the recall and recognition
of advertisements. Burke and Srull (1988) argue that associative
interference processes contribute to the primacy and recency effects on the
recall and recognition of television commercials because later stimuli
(e.g., competing commercials, other product information) inhibit viewers'
ability to remember advertised information. Pieters and Bijmolt (1997)
found the significant primacy effect in examining the serial order effects
of television commercials. In their study, when the recall of brand names
in the middle commercial was calculated as 100%, that of the first spot
commercial was 129% while that of the last one was 101%. Thus, although the
both brands placed in the first and last spot were more recalled than those
placed in the middle, the effect of the last-positioned brand was not as
significant as the one positioned in the first.
Krosnick and Alwin's explanation (1987) about the primacy effect provides a
useful theoretical framework to understand why the primacy effect occurs in
the interpretation of ads. First, stimuli presented earlier stages may
construct a cognitive framework that can be used to compare later ones.
Thus, earlier stimuli may be considered more significant determinants in
subsequent judgments. Second, stimuli presented earlier stages are likely
to be processed in a cognitively deeper level than later alternatives. This
is because stimuli presented in later stages are likely to be cluttered
with thoughts about prior alternatives that hinder cognitive consideration
in a deeper level. Thus, earlier stimuli are likely to take over during the
entire cognitive process. Additionally, people have a tendency to minimize
psychological costs by seeking for simply satisfactory or acceptable
alternatives as early as possible. Thus, instead of attempting to search
optimal alternatives, people seek for acceptable ones in the earlier
process (Simon, 1957).

Hypotheses
This study examines the effectiveness of Super Bowl advertising,
considering the order effects (primacy and recency). According to previous
studies, the primacy and recency effects may occur in the process of a
series of stimuli, such as advertisements in a television program. Thus,
Super Bowl advertising, which is classified into the 15 quarter-based
order, offers a unique opportunity to determine the relationship between
the order of ads and ads performance. Using this approach, this study tests
which part of game provides more favorable advertising environment.
Based on the literature review, because attention to commercials should
gradually wane over the exposure of a series of stimuli, this study expects
the primacy effect of Super Bowl advertising. The factors found as
moderators of order effects (e.g., individual's emotional and cognitive
status, motivation, and message relevancy) also predict the primacy effect
of commercials placed in Super Bowl game. For instance, the Super Bowl is
perceived as a considerably important event by audiences. Thus, viewer's
emotional status should be higher as the game progresses, and their
cognitive ability and motivation to process the commercial information
during the game should be low. Moreover, a relatively long period of game
time (about four hours) also suggests primacy effect because numerous
memory inhibitors, such as competitive advertising and game fluctuations,
should decrease the ability to recognize all the advertisements throughout
the game. Thus, this study argues that the ads placed in the earlier
positions should earn more attention from the viewers than those in the
later positions. Therefore, this study proposes the following hypotheses:
H1: Respondents are more likely to remember brands advertised in the
earlier quarters than those in the later quarters.
H2: Respondents are more likely to view the advertised brands placed in the
earlier quarters favorably than those in the later quarters.
Methods

While most of advertising studies conduct controlled experiments or
conventional surveys, this study uses a naturalistic quasi-experiment. The
data were from two years of Super Bowl broadcasts in 2002 and 2003.
        Because the respondents knew nothing about the study until at least
twenty-four hours after the game, their viewing (or non-viewing) behavior
may be assumed to be perfectly natural. Such behavior may include a reduced
exposure and attention to television during a commercial break, as a result
of zapping, room-leaving, socializing, and participation in other non-TV
focused activities. This natural setting carries particular importance in
advertising effectiveness studies regarding clutter or order.
If the baseline attention/involvement is artificially intensified, as it
might in a laboratory setting, the true advertising effects could be
distorted. Hence, in comparison with most of the forced viewing experiments
of advertising research, this study appears naturalistic in several
important aspects, including naturally varying length of commercial pods,
broader range of commercial qualities, more representative sample of
viewers, home viewing environment, and longer time between exposure and
memory test. Generalizability, however, is only relative. In comparison
with a hypothetical "dream study,"  this study's generalizability may be
also limited in a number of regards.  The quality of the Super Bowl
advertisements, especially the network advertisements, may be far higher
than the quality of average advertisements. The annual media hoopla
surrounding the two "super championships" – the football championship in
stadiums and the advertising championship on the air – may have also
contributed to an unusually high exposure or attention to both the
broadcast and the advertisements.  Nevertheless, advertisers and
commentators suspect that the uniqueness of the Super Bowl as an
advertising event has been overstated, and they cite the lower than
expected recall scores of Super Bowl advertisements to support their
suspicion (Deveny, 1993; Moore, 1993; Goldman, 1994).
        A trade-off for a higher generalizability is a lower degree of internal
control. Like most field studies, this study is correlational rather than
causal. Although various measures were taken to guard against some of the
methodological threats, causal inference should be made with caution. The
design in itself is no more valid than the more often used controlled
experiments. It is strong where the controlled experiments are often weak,
and it is weak where the others are typically strong. This difference,
however, may be one of this study's main contributions.  By comparing field
results with laboratory findings, our collective understanding of
advertising effects may become more valid externally and internally.
        The unit of analysis in this study is each brand advertised during the two
games. The dependent variables were measured through telephone interviews
and then aggregated across respondents by brands. The independent variables
were measured by analyzing the content of the television commercials taped
during the games.
        Telephone Interviews: Measuring Memory and Liking. A telephone survey was
conducted from Monday evening through Thursday evening following each of
the two Super Bowl games played in 2002 and 2003. Graduate and
undergraduate students enrolled in research classes at a major university
used random digit dialing to reach the local residents of Orange County of
North Carolina. Guided by a computerized questionnaire, the interviewers
asked for the person who had the next birthday. If a call yielded a
machine-recorded answer or no answer, that number was re-dialed at least
three times before being discarded. A total of 528 interviews were
completed, with an average response rate of nearly 60%.  Each year, more
than two thirds of the respondents reported having watched the game.
        Dependent Variable 1: Unaided Brand Recall. The interviewers asked each
respondent whether he or she had watched the Super Bowl game, and which
part.  Those who watched any part were then asked to list all
advertisements they remembered seeing during the game.  Two coders coded
independently the responses, which had been recorded verbatim during the
interviews.  The recall rates were then calculated according to:
                          Rb
        Recall Rate =   __ _ 100
                          Ws

Where Rb is the number of respondents who recalled the brand, and Ws  is
the number of respondents who watched the segment(s) in which the brand was
advertised.

        To see if any of the recalls may have been a false alarm _ a respondent
could mistakenly recall a brand he or she had seen elsewhere but not during
the Super Bowl advertising _ we searched the responses for brands that were
not advertised during the game segments that a given respondent reportedly
watched.  No such false alarms were found in two years.
        Dependent Variable 2: Brand Recognition. After the unaided recall measure,
each respondent was given a list of brand names.  Students, teaching
assistants, and the instructor had compiled the list by observing the
advertisements aired during the game. The observations were cross-verified
via video tapes. Respondents were asked if they remembered seeing an
advertisement for that brand during the game. The recognition rates were
calculated according to:
                                          Gb
            Recognition Rate =  __ _ 100
                                          Ws

Where Gb is the number of respondents who recognized the brand, and Ws  is
the number of respondents who watched the segment(s) in which the brand was
advertised.

        The threat of false alarms is typically larger for recognition measures
than for unaided recall. To address this concern, interviewers emphasized
to respondents that the brands listed may or may not have been advertised
during the game. Further, false-alarm tests were conducted for each
year.  The 2002 and 2003 questionnaire included several brands that had not
been advertised during the game but were major competitors of the
advertised brands.
The recognition rate and liking score were then recalculated after being
weighted by each respondent's correction rate in the false-alarm tests. The
correlation between the weighted and unweighted scores is .99. This result
suggest that false alarms, while a low frequency phenomenon, also occurred
rather randomly and distributed quite evenly among brands. Therefore, false
alarm should not significantly affect correlation between variables, which
is the basis of our analysis. However, we use the weighted scores as the
major basis of analysis and reporting.  A parallel analysis based on the
unweighted scores gave essentially the same results.
        Dependent Variable 3: Advertisement Liking. Advertisement liking was
measured by asking those respondents who remembered seeing an advertisement
how good or poor they thought the advertisement was. Likert scales (1- 7
for each year) were used.  To facilitate interpretation, all liking scores
were linearly transformed to a 0-100 scale (100: the best;  0: the poorest):


                      OL _ 1
        Liking  =   ___ _ 100
                        6

Where OL is the original liking score.  Those scores were then averaged
across respondents for each brand of each year.

Content Analysis: Measuring Position.  Each independent variable involved
at least two coders who did independent coding using video tapes recorded
during the games. If disagreement happens, consensus on the case was
reached quickly after re-examining the tape.
        When a brand had just one advertisement in the Super Bowl broadcasting,
the general concept of clutter was measured by the number of other ads in
the pod, that is, the total number of commercials in the pod minus one.
Here, a pod means a position where a series of ads are placed between
program contents. When a brand was advertised twice or more, the number of
other ads in all pods in which the brand was advertised were summed to
measure the total amount of clutter competing with the advertisements of
the given brand.
        Identifying and Measuring Control Variables.  An obvious confounding
factor is ad frequency, defined as the number of advertisements promoting
the same brand during a given game. Higher frequency may be associated with
better memory.  Our one of independent variable, clutter, is also
associated with frequency; a brand airing three advertisements tends to
have more ads than a brand advertised only once. Prior studies addressed
this problem by restricting frequency to one advertisement per brand.  This
study controlled frequency statistically, rather than physically.
Conceptually, it is equivalent to examining the effects within each level
of frequency, and then averaging the magnitude of the effects across
frequency levels.
        Another possible source of contamination was the year variable. The ads in
2002 Super Bowl broadcast should be different from those in 2003 as well as
games themselves. When the data were pooled together, there was a chance
that differences between the years could confound advertising effects.  One
dummy variable for 2002 was created.  The brands from 2002 served as a
comparison group.
         Product categories posed yet another problem requiring statistical
control.  Brands in certain product categories might be more easily
remembered than others, and some of those brands might happen to be in
certain positions.  Therefore, seven dummy variables were created to
represent seven product categories.  An eighth category, entertainment
advertisements, served as a comparison group.
        Finally, while there could different classification methods for a game,
for the clarity of this study, five quarter-based segments were selected
and used as the classification in this study. They are (1) the first
quarter, (2) the second quarter, (3) the third quarter, (4) the fourth
quarter, and (5) other game segments. The most important reason for this
classification is the small number of ads in other game segments. With the
very small number, it was difficult to include those other game segments in
the model. Also, when only considering the main game, these four quarters
contain most ads from Super Bowl broadcast. The other game segments was
averaged into one segment and used as the basis for other four dummy
variables of four quarters.
Results
The results supported the first hypothesis regarding brand memory. The
primacy effect prediction was supported among four quarters in a whole
game. Recognition and recall rates were the highest for the ads in the
first quarter and decreased gradually afterward. However, liking rates for
ads in each quarter were not different by each quarter and similar to ads
in the other game segments.
Quarter-based order effects were found to be more important than ad
frequency effects –placing an advertisement in the first quarter could be
almost five times beneficial than running an additional advertisement in
the other game segments, the basis for the comparison of the dummy
variables. As shown in TABLE 1, recall and recognition rates for ads in
earlier quarters were significantly higher than those in later quarters.
Means for recognition, recall, and liking scores for ads in 15 game
segments were described in TABLE 2, FIGURE 1 and FIGURE 2. These rates for
ads are peaked during each quarter and decreased gradually. Also, TABLE 1
shows means for five classification categories. The scores for ads in a
earlier quarter is higher than those in a later quarter, and the scores for
ads in all four quarters are higher than the scores for ads in other game
segments.
-----------------------------------------------------
Table 1 and FIGURE 1 and 2 about here
-----------------------------------------------------

Findings from Univariate Analysis. 75 brands in 2002 and 65 brands in 2003
were advertised in each game, totaling 140 for analysis.  Those brands were
from eight product categories (Table 3).  The most successful brands had
recall and recognition rates around 80%, as is shown in Table 4.  The least
successful brands recorded no recall or very low recognition rates. Those
who recognized the brands also tended to like the advertisements
moderately, giving an average score of nearly 60 on a scale of 0-100.  The
number of ads in each pod ranged from one to seven. The total number of
other ads indicating general clutter ranged from 1 to 22.
-----------------------------------------------
Table 2, 3, and 4 about here
-----------------------------------------------

Two of our dependent variables, recognition and liking, had reasonably
bell-shaped distributions. Deviation from normality is nevertheless
inevitable for unaided recall considering many respondents with no ad
recall. So, recall distribution tends to be skewed toward the positive
end.  Because of the robustness of regression, however, such deviation is
deemed within the tolerable range. The other variables we used are
independent or control variables, therefore their normal distribution is
not concerned.
While advertising researchers often use ANOVA to analyze experimental data,
this study chose the multiple regression method, a technique most often
used in field studies for its power and flexibility (Cohen & Cohen, 1983).
This method enables effective controlling of confounding factors.
The analysis procedure was the same for each dependent variable.  Year,
product categories, and advertising frequency were entered first as a
control block (nine entries because of dummy coding. See Section 1 of Table
5). On top of those controls, two blocks of the independent variables were
entered.  The first block contained general clutter, the number of ads in a
same pod. Quarter-based orders (four entries because of dummy coding of the
first, second, third and fourth quarter) were included in the final
regression model as the last block. The results of the multiple regression
analysis are summarized in Table 5.
--------------------------------
Table 5 about here
--------------------------------

Effects of Control Variables.  Average brand memory and advertisement
liking did not change significantly from 2002 to 2003. They varied somewhat
significantly, however, among product categories. Another important factor
in the control block was the number of advertisements that each brand aired
during each game. When an advertiser runs an additional advertisement, he
should expect it to bring additional memory for his brand. The Results show
that an additional advertisement tends to increase recall by more than six
percentage points. But advertising frequency was not statistically
significant in affecting advertisement liking and recognition. However, the
coefficients for these variables were in the positive direction.
Considering the statistical significance on recall rates, which is one of
measures of brand memory, the sample size is too small to have enough power
to detect the statistically significant relationship between ad frequency
and recognition. While liking was usually believed to be negatively related
to ad frequency, this relationship was not found in this analysis.
There is no significance difference in brand memory for ads between 2002
and 2003 after controlling other variables. However, ads for 2002 were
liked five percent more than those for 2003. The effects of product
category appear different in this data set.  Advertisements for health and
beauty, for example, have a lower liking score than entertainment
advertisements by over 11 points on a scale of 0-100.
The first control block with year, product categories and advertising
frequency control explain over 39% of the variances in recall, 44% in
recognition and nearly 31% in liking (Section 4 in TABLE 5).  They appear
to be reasonable to be included in the model because they have accounted
for a good amount of variances. However, these variables were confounding
factors in this study.
General Clutter Effects.  Advertisements placed in longer pods were
believed to generate lower brand memory for the advertised brand. The
direction of clutter effects showed up; coefficients for clutter effects
were all negative. However, in this study, it didn't show statistically
significant results for this variable in recognition, recall, and liking.
This might be due to the small sample size, or quarter-based order effects
might take most of clutter effects. This indicates that probably general
clutter effects in previous studies might be inflated without controlling
order effects in a whole program.
Quarter Position Effects.  There were five quarter positions in this study
including the first, second, third, fourth quarter, and other game
segments. Due to the low number of ads in other game segments, ads in other
game segments were combined and used as the comparison basis for the ads in
quarters. As shown in TABLE 1, there were quite large differences in
recognition and recall rates for ads in the four quarters and in other game
segments. If we only consider mean scores of brand memory, the first
quarter has the highest scores among all followed by the second, third, and
fourth quarter. The mean of liking seemed reasonably similar among all game
segments.
With the multiple regression model, we could see the unique effects
originating from quarter-based orders. The ads in the first quarter had
almost 20 points higher recognition rates and 8 points higher recall rates
than the ads in other game segments. The ads in the second quarter also had
significantly higher recognition (12%) and recall rates (6%) compared to
the ads in other game segments. The Ads in the third quarter only had
significantly higher recognition (10%) but not recall rates. Recognition
and recall rates for the ads in the fourth quarter were very similar to the
ads in other segments.
Based on these results, several important findings were observed. First of
all, the ads in quarters achieved higher brand memory scores than the ads
in other game segments. In other game segments, audiences are likely to be
ready for upcoming ads and could use various defense mechanisms to avoid
those persuasive messages. Second, the size of quarter-based order effects
was considerably huge compared to the effects of other variables. Third,
the primacy theory was working from the first to fourth quarter successively.
Comparing Quarter-based Order Effects with Effects of Frequency and Other
Variables. The number of advertisements to be aired is decided by the
advertiser; and there is almost always a cost for any additional
advertisement. Interestingly, our data suggest that quarter-based order
effects may be stronger than the effects of frequency.
By comparing the appropriate regression coefficients (in Sections 1 & 3 of
TABLE 5) we can see that placing ads in an earlier quarter significantly
improve the recognition and recall rates for ads. However, liking rates
were not different by quarters. If the objective of ads is an improved
brand memory, placing ads in the first quarter might exceed running one
more ad.
The beta coefficients are often used for comparing the effects of different
independent variables (Zhao, 1997).  The beta coefficients in Table 5 show
that advertising frequency variable is among the best predictor. However,
coefficients for quarter dummies also seem to be important predictors, too.
Predictive Power of Quarter Position Variables.  Quarter positions of ads
are good predictors of the brand memory and advertisement liking. That is,
if an advertisement is aired in a better quarter-based order, we should not
only predict that the advertised brand may be remembered by more people,
but we should also be fairly confident that our prediction may be quite
close to the actual effects.
Our confidence comes from the (incremental) R squared statistics. On top of
the more than 31%-44% variance in memory already explained by the control
variables, the block of quarter-based order variables can add another
5%-14% (Sections 4 & 6 of Table 5). Both results represent substantial
predictive powers of the quarter-based order variables, and they are both
statistically significant in the models. Incremental R squared in the model
for liking as a dependent variable was not statistically significant. So,
liking was not affected by quarter-based orders.
Discussion
The primary objective of this study is to examine the order effects on
advertising effectiveness. A basic premise of the present study was that
viewers' attention on the advertisements would be lowered as the game
progresses. Therefore, this study expected that the primacy effect would
occur during the game, and, consequently, advertisements placed in the
earlier position would be more effective than that placed in the later
position. The following discussion section contains a short summary of the
key findings of this study, followed by a discussion of the implications of
the results, limitations of this study, and finally directions for future
studies.
This study proposed two hypotheses, of which one was supported. The result
indicated that the brands in advertisements placed in the earlier position
(e.g., the first quarter) yielded the higher memory scores than those in
ads place in the later position (e.g., the second through fourth quarter).
This finding is consistent with previous studies that suggest
advertisements shown early are processed at a deeper level than later ads
because they are not in competition with other alternatives and are
established as significant determinants to compare other ads. Therefore,
the likelihood of memorizing the brand names in ads seen earlier should
have increased.
The quarter-based order effect on advertising effectiveness was
considerably huge compared to the effects of other variables. This study
found that the ads placed in any quarters were found to achieve higher
brand memory scores than the ads in other game segments. However, the
result indicated that ads in the first quarter were more effective than
those in other quarters by at least 20% in terms of memory. Moreover, the
scores of ads shown during the first quarter were as higher as five times
than those in the other game segments. Thus, this study found a strong,
significant primacy effect occurs during the Super Bowls.
The context effect study provides a useful framework to understand the
primacy effects in this condition. A number of studies found that the
effectiveness of ads are higher when ads are embedded in a low involvement
context than when they are embedded in a high involvement one (Bello, et
al., 1983; Bryant & Comisky, 1978; Gunter, at al., 1997; Lord & Burnkrant,
1988; Norris & Colman, 1992; Park & McClung, 1986; Pitts, 1986; Soldow &
Principe, 1981). According to Bello, et al. (1983) and Pitts (1986), a high
involvement context produces the higher level of "need for closure" and
makes audiences emotionally more intense than a low involvement context
does. Therefore, when audiences are under a high level of need for closure,
such as in the later quarters in the Super Bowl, ads are more likely to be
considered disrupting stimulus, and consequently, negatively cause viewers'
memory of the ads.
One particularly interesting finding of this study is that the brand memory
scores of ads placed in the fourth quarter were lowest among four quarters.
This finding is inconsistent with the previous studies that found the
memory advantages of the first and last positions (the U-shaped
relationship). This finding is noteworthy. One possible explanation for
this finding may come from viewers' emotional intensity. Because viewers'
emotional arousal should reach the highest level during the fourth quarter,
their attention toward ads should be lowered, as would their memory of the
brands advertised.
   However, the present study failed to find the primacy effect of
advertising liking during the game. There were no significant differences
between ads distributed in the game. This finding indicates that liking the
ads is not significantly influenced by the quarter-based order of
advertisements. Rather, the ad liking might have been more likely to be
influenced by other factors, such as the length of ads, creativity of the
ads, or types of the products advertised.
The present study has yielded several theoretical and marketing
implications. First, this study found that the quarter-based order
significantly influences the ad performance by examining an actual
television program, the Super Bowl. Previous studies used a limited number
of stimuli in relatively short duration of time (e.g., survey questions,
paired verbal and auditory stimuli) to determine the order effects. Thus,
by examining a four hour-long, actual program, this study successfully
extended the area of order effect research. Second, the findings of the
present study yield the great implications on media planning practice
because deciding the ad position (pod) is one of few controls advertisers
have in advertising at the televised program. Moreover, the recency effect
has been pervasively adopted as an important marketing concept among
advertising practitioners and applied to numerous advertising campaigns
(Ephron, 1997, 1998; Orsini, 1998; Salvo, 1998). However, according to the
findings of this study, the primacy effect dominates throughout the Super
Bowl. Thus, current recency strategies should be reconsidered. Third, this
study used two Super Bowls, one close game (2002 Super Bowl) and one loose
game (2003 Super Bowl). Thus, this study could find the generalizable
quarter-based order effect on advertising effectiveness, regardless of
tension associated with the game.
The limitations of the present study and directions for future research are
in order. First, one of the limitations may come from the sample frame, two
Super Bowls. The Super Bowl is considered as the most important, meaningful
sports/entertainment event in American society. Thus, viewers' emotional
arousal, involvement level, physiological status and attention toward
commercials may be different from other games or programs. Second, the
telephone interviews were conducted on the following week from the Super
Bowl. Thus, this study could determine the immediate impact of order
influence on the ad performances. However, although this measurement is
frequently utilized in advertising research (e.g., DAR – day after recall),
this study might not have detected the sleeper effect of order influences
on the effectiveness of ads.
        There are several directions which future research may follow. First, this
study has paved a road for studying the order effects in advertising field
by extending the number of stimuli and increasing exposure time. Thus, one
approach is to broaden this area using more variables with various
programs. For example, more dependent variables, such as recall, ad
evaluation, and purchase intention, can be used in the future study with
other types of television programs, such as regular-season football games,
other sports games, or general television shows. Second, this study
suggests determining the context effects of television programs in
examining the order effects on advertising success. Because ads accompany
television programs, viewers may be influenced by not only the advertising
order, but by television contexts as well. Thus, this study suggests
determining viewers' involvement in the game (high versus low), mood (happy
versus sad), physiological status (aroused versus calmed), and the degree
of excitation transfer (high versus low). Finally, although previous
studies have found the first and last stimuli are generally better
memorized than those in the middle, the last ads were least memorized in
this study. Thus, it is recommended that future study thoroughly examine
this unexpected finding to have a better understanding of the order effect
on advertising effectiveness in television programs.
 Table 1
Mean recognition, recall, and liking scores for quarter positions
Recognition
Recall
Liking
Ads in the First Quarter
53.78
14.42
63.61
Ads in the Second Quarter
49.46
12.68
61.57
Ads in the Third Quarter
46.82
9.30
61.66
Ads in the Fourth Quarter
34.07
7.22
58.50
Ads in Other Game Segment (Average)
29.70
5.10
59.85


TABLE 2
MEANS OF RECOGNITION, RECALL, AND LIKING SCORES FOR ADS IN GAME
Frequency
Recognition
Recall
Liking
Before Pregame show
6.00
20.37
2.37
63.66
Pregame show
16.00
24.63
3.73
59.90
Between Pregame and 1 Quarter
2.00
30.38
3.35
59.48
1Quarter
24.00
53.78
14.42
63.61
Between 1 Quarter and 2 Quarter
7.00
37.62
5.16
58.92
2 Quarter
27.00
49.46
12.68
61.57
Between 2 Quarter and Half time show
4.00
35.32
2.66
59.76
Half time show
7.00
28.68
0.44
58.82
Between Half time show and 3 Quarter
3.00
14.00
1.56
54.87
3 Quarter
21.00
46.82
9.30
61.66
Between 3 Quarter and 4 Quarter
5.00
43.54
23.04
69.50
4 Quarter
22.00
34.07
7.22
58.50
Between 4 Quarter and postgame show
5.00
48.46
8.91
57.91
Postgame show
32.00
29.66
5.42
58.42
After Postgame show
6.00
21.41
1.10
61.98




 Table 3
Number of Brands advertised by year and product categories

Across: Year
Down: Product Categories
2002
2003
Total
Services
21
15
36
Auto Related Products
13
11
24
Shoes & Clothes
4
4
8
Health & Beauty
4
3
7
Household Products
4
9
13
Food & Beverages
17
17
34
Public Announce-ments
4
3
7
Entertain-ment
15
12
27
Total
82
74
156

 8
How adverting effectiveness changes
Table 4
Univariate Statistics of Major Variables (n=156)
Minimum
Maximum
Mean
Median
S.D.
Skewness
Kurtosis
Liking (0-100)
36.85
79.30
58.81
27.00
8.46
-0.02
0.24
Recognition (0-100)
1.00
85.38
30.88
59.17
18.61
0.77
0.04
Recall (%)
0.00
97.37
3.68
0.44
11.78
5.55
35.37
Advertising Frequency
0.00
7.00
1.37
1.00
1.05
2.47
8.39

Total Number of Other Ads
1.00
22.00
5.79
5.00
3.89
1.92
4.33





 8
How adverting effectiveness changes
TABLE 5
Position Effects on brand memory and advertisement Liking
Dependent Variables:
Brand Recognition (%)
Brand Recall (%)
Advertisement Liking (0-100)
1. Control Block a
     Constant
27.05***
-6.66**
59.50***
     Year 2003
-2.22 (-0.06)
-2.23 (-0.09)
-4.98 (-0.29)***
     Services
-7.14 (-0.16)*
1.50 (0.05)
-2.77 (-0.13)
     Auto Related Products
-11.88 (-0.25)**
0.46 (0.01)
-3.70 (-0.16)
     Shoe & Clothes
4.80 (0.06)
3.08 (0.05)
6.85 (0.17)*
     Health & Beauty
-16.01 (-0.20)**
3.60 (0.06)
-11.19 (-0.28)***
     Household  Products
-5.12 (-0.06)
2.70 (0.05)
0.15 (0.00)
     Food & Beverages
-1.28 (-0.03)
6.41 (0.22)*
1.66 (0.08)
     Public Service Announcement
-3.09 (-0.04)
-4.05 (-0.07)
1.41 (0.04)
     Advertising Frequency
4.23 (0.24)
6.17 (0.51)**
3.47 (0.41)
2. General Clutter Block
     The number of other ads
-0.29 (-0.06)
-0.40 (-0.13)
-0.29 (-0.13)
3. Quarter Block b
     First Quarter
19.30 (0.41)***
7.46 (0.23)**
2.50 (0.11)
     Second Quarter
12.15 (0.27)***
5.72 (0.18)*
-2.20 (-0.10)
     Third Quarter
10.34 (0.21)**
0.73 (0.02)
-1.29 (-0.05)
     Fourth Quarter
1.49 (0.03)
0.95 (0.03)
-1.86 (-0.08)
4. Total R2 of Control Block (%)
44.2***
39.0***
30.8***
5. Incremental R2  due to the number of Other Ads (%)
4.9***
2.5*
0.2
6. Incremental R2  due to the quarter block b
13.6***
5.0*
2.6
Note. Cell entries in Sections 1-4 are regression coefficients and
standardized beta coefficients (the latter are in parentheses).
a:      The brands from 2002 serve as a baseline for comparison with the brands
from the other two years.  Entertainment serves as a baseline for
comparison with the other seven product categories.
b:      The ads in other game segments serve as a baseline for comparison with
the ads from the quarters.
*:      p<.05
**:     p<.01
***:    p<.001
 8
How adverting effectiveness changes







 References

Becker, S. L. (1954). Why an order effect. Public Opinion Quarterly, 18(2),
271-278.

Bello, D. C., Pitts, R. E., & Etzel, M. J. (1983). The communication
effects of
controversial sexual content in television programs and commercials.
Journal of Advertising, 12(3) 32-42.

Blunch, N. J. (1984). Position bias in multiple-choice questions. Journal
of Marketing
Research, 21, 216-220.

Bryant, J., & Comisky, P. W. (1978). The effect of positioning a message
within
differentially involving portions of a television segment on recall of the
message. Human Communications Research, 5(1), 63-75.

Burke, R. R., & Srull, T. K. (1988). Competitive interference and consumer
memory for
advertising, Journal of Consumer Research, 15, 55-68.

Callahan, T. (1991, January 28). America's supersymbolism. U.S. News &
World Report,
110(3), 60.

Cohen, Jacob,  Patricia Cohen. Applied Multiple Regression/Correlation
Analysis for the
Behavioral Sciences, 2nd Ed.  Hillsdale, NJ: Lawrence Erlbaum, 1983.

Coney, K. A. (1977). Order-bias: The special case of letter preference.
Public Opinion
Quarterly, 41(3), 385-388.

Crano, W. D. (1977). Primacy versus recency in retention of information and
opinion
change. Journal of Social Psychology, 101, 87-96.

Day, R. L. (1969). Position bias in paired product test. Journal of
marketing Research, 6,
98-100.

Dean, M. L. (1980). Presentation order effects in product taste tests.
Journal of
Psychology, 105, 107-110.

Deveny, K. (1993, January 26). In weighing impact of Super Bowl ads,
marketers have
trouble keeping score. The Wall Street Journal, 26, B1.

Ephron, E. (1998, March 16). The new 'recency' planning. MEDIAWEEK, 8(11),
14-15.

Ephron, E. (1997). Recency planning. Journal of Advertising Research,
37(4), 61-65.

Friedman, L. W., & Friedman, H. H. (March 1994). A comparison of vertical and
horizontal rating scales. Mid-Atlantic Journal of Business, 30(1), 107-111.

Goldman, K. (1994, February 1). Super Bowl ads looked worse than bills. The
Wall
Street Journal, B12.

Gunter, B., Furnham, A., & Beeson, C. (1997). Recall of television
advertisements as a
function of program evaluation. The Journal of Psychology, 131(5), 541-553.

Haugtvedt, C. P., & Wegener, D. T. (1994). Message order effects in
persuasion: An
attitude strength perspective. Journal of Consumer Research, 21(1), 205-218.

Hugick, L. (January 1992). Super Bowl now an American institution. The
Gallup Poll
Monthly, 316, 36-37.

Kanner, B. (2004). The Super Bowl of Advertising: How the Commercial Won
the Game.
Princeton, NJ: Bloomberg Press.

Krosnick, J. A., & Alwin, D. F. (1987). An evaluation of a cognitive theory
of response-
order effects in survey measurement. Public Opinion Quarterly, 51, 201-219.

Lamb, K. (February 1992). An American holiday: There's New Year's,
Thanksgiving,
the Fourth of July, and Super Sunday. Sport, 83(2), 52-56.

Landon, E. L. Jr. (1971). Order bias, the ideal rating, and the semantic
differential.
Journal of Marketing Research, 8(3), 375-378.

Linnett, R., Halliday, J., & Johnson, B. (2004, January 12). Super Bowl
busts records.
Advertising Age, 75(2), 1-2.

Lord, K. R., & Burnkrant, R. E. (1988). Television program elaboration
effects on
commercial processing. In Advances in Consumer Research. Ed. Michael J.
Houston. Provo, UT: Association for Consumer Research, 15, 213-218.

McAdams, D. D. (November 1999). Women's net eyes goal line. Broadcasting &
Cable,
129(45), 10.

Moore, M. T. (1993, February 2). Super ad! what's it for? USA Today, 1A.

Norris, C. E.., & Colman, A. M. (1992). Context effects on recall and
recognition of
magazine advertisements. Journal of Advertising, 21(3), 37-46.

Orsini, P. (1998, April 20). Upfront: A matter of timing media buyers like
recency, but
which medium do they like? ADWEEK Eastern Edition, 39(16), S4.

Park, C. W., & McClung, G. W. (1986). The effect of TV program involvement of
involvement with commercials. Advances in Consumer Research, 13, 544-548.

Pieters, R. G. M., & Bijmolt, T. H. A. (1997). Consumer memory for television
advertising: A field study of duration, serial position, and competition
effects. Journal of Consumer Research, 23(4), 362-372.

Pitts, R. E. (1986). Commercials in context. Marketing & Media Decisions,
21, 76.

Quindt, F. (February 2003). Vital stats. The Sporting News, 227(8). 6.

Salvo, C. (November 1998). Get 'em while they're hot: Are hotline names the
most
responsive names on the market? Target Marketing, 21(11), 46-47.

Sekely, W. S., & Blakney, V. L. (1994). The effectiveness of response
position on trade
magazine readership and usage. Journal of Advertising Research, 34(6), 53-60.

Sharps, M. J.,Price, J. L., & Bence, V. M. (1996). Visual and auditory
information as
determinants of primacy effects. The Journal of General Psychology, 123(2),
123-136.

Simon, H. A. (1957). Models of man. New York, NY: Wiley.

Soldow, G. F., & Principe, V. (1981). Response to commercials as a function
of program
context. Journal of Advertising Research, 21(2), 59-65.

Whipple, T. W., & McManamon, M. K. (1992). Primacy order effects in the
measurement of trade magazine receipt and readership. Journal of
Advertising Research, 32(5), 24-29.

Zhao, X. (1997).  Clutter and serial order redefined and retested. Journal
of Advertising
Research, 37(5), 57-73.

Zimmerman, P. (1995). Day one. Sports Illustrated, 83(15), 74-81.