An event study approach to evaluating the economic returns of advertising in the super bowl.

Choong, Peggy ; Filbeck, Greg ; Tompkins, Daniel L. 等

ABSTRACT

While there are several anecdotal evidence of positive outcomes from Super Bowl advertising, there have not been any credible measurements of returns on investing in this expensive strategy.

This paper evaluates the returns on management's strategy to advertise during the Super Bowl using the event study methodology. This is a useful method often used in the finance field, to capture the market's valuation of the management decision by measuring the abnormal returns associated with the announcement of that strategy. A knowledge of the outcome associated with this strategy will enable marketing managers to better allocate their advertising resources among the array of media and media events.

INTRODUCTION

Reaching the largest proportion of prospective buyers is a problem faced by many advertising executives in media planning. This problem has been exacerbated in recent years by the increased fragmentation of the market. With numerous television channels and radio stations nationwide, programs on each of these media are consistently reaching smaller defined audiences. To add to this problem of reach, advertising executives are also faced with the lack of response most consumers have toward broadcast advertising.

In response to these issues, marketers have, over the last fifteen years, gravitated toward television programs such as the Super Bowl, Olympics, Oscars and Grammy Awards. Among the handful of important media events, the Super Bowl has remained one of the most coveted and expensive advertising venues (Elliott, 1999). The annual scramble to secure strategic time slots during the game has become a major media event in itself. Many companies spend the bulk of their annual advertising budget on this single costly program. A thirty-second spot during the Super Bowl cost an average of $2.0 million dollars compared with $1.6 million for the Oscars and $0.57 million for the Grammy Award (O'Connell, 2002).

The two reasons most advertisers cite for advertising on the Super Bowl are its substantial reach and the high degree of audience receptivity to commercials during the program. For example, advertising agents for 7UP claim that their target market of 12 to 24 year-olds is so highly fragmented that Super Bowl Sunday is one of the few television venues that captures the highest percentage of this market. Similarly, when executives at Philips Electronics, North America released their new advertisement of the 64-inch hi-definition television priced just under $10,000, they claimed that the Super Bowl was their best opportunity to reach 40 percent of their 18 to 49 year-old target market Reaching "the same unduplicated number of viewers" would have required buying the equivalent of 13 or 14 prime-time spots (Elliott, C8, 1999).

Audience receptivity during the program is also reported to be higher than during normal programming. Unlike regular programming where advertisements are often viewed as intrusions, Super Bowl audiences are actually paying attention to the commercials. A research conducted by SAA/Research reported that 48 percent of respondents listed seeing the commercials as a reason they watch (Elliott, 1999). Audiences view the advertisements during Super Bowl as part of the entertainment. Interestingly, reviewing the commercials the day after have become an event in itself ranging from the informal office conversations to the annual consumer survey known as the Super Bowl Ad Meter conducted by USA Today. Marketing executives are loath to find themselves at the bottom of the well-publicized meter.

There has been extensive research in marketing about advertising and its effects. Market response models form a significant segment of these studies on how advertising works (see Demetrios et. al 1999 for an excellent discussion of the taxonomy of advertising effects). Generally, this genre of studies examines the relationship between advertising and some measure of behavioral response. Aggregate level studies typically use sales or market share as proxies for the market response (Bass & Clarke, 1972; Rao & Miller, 1975; Little, 1979; Blattberg & Jeuland, 1981; Rao 1986; Zufryden 1987; Hanssens, Parsons & Schultz, 1990; Rao & Burnkrant, 1991). However, measuring the overall effect of advertising expenditure on sales and profits is fraught with problems. One significant problem is that the duration of advertising effects uncovered in many studies clearly shows that the effects of advertising accrue over time, thus making current profits or sales less useful measures of advertising effectiveness (Winer, 1979, 1980; Dekimpe & Hanssens, 1995; Leone, 1995; Lodish, 1995;; Mela, Gupta & Lehmann, 1997). These problems are exacerbated when attempting to measure the effects of a single advertising strategy. Thus to circumvent these problems and to measure the effects of a single advertising strategy, another method of measurement is required.

The event study methodology is a feasible and useful method often used to measure the direct effects of a strategy. Commonly used in the finance and economics fields, this methodology captures the market's valuation of a management decision by measuring the abnormal returns associated with the announcement of that strategy. When applied to the measurement of advertising effectiveness, this method is able to capture the abnormal returns of advertising in a specific program such as the Super Bowl.

This kind of measurement should also be of particular relevance nowadays as marketing managers are increasingly pressured to provide evidence of economic returns to their strategic decisions. An oft quoted remark attributed to John Wanamaker, founder of the first U.S. department store (1838-1992), "Half the money I spend on advertising is wasted, the trouble is, I don't know which half," underlines the skepticism corporate America still has for how advertising works (Lee & Lade, 2001). While there are several anecdotal evidence of positive outcomes from Super Bowl advertising, in the form of increased sales, phone inquiries and hits on Web sites of advertisers, there have not been any credible measurements of the returns on investing in this expensive programming. Just as companies seldom make major capital expenditures without estimating the likely financial returns, so too is corporate America increasingly demanding to know the returns on their advertising strategy.

The purpose of this paper is to evaluate the returns on management's strategy to advertise during the Super Bowl using the event study methodology. Knowledge of this will enable managers to better allocate their advertising resources among the array of media and media events.

EVENT STUDY METHODOLOGY

The event study methodology is a statistical procedure to examine the effect that the release of information has on the stock market returns of the firm. It has been used extensively in the economics and finance literature. In the marketing field, it has been used to examine strategies such as product innovation, change in a company's name, bad publicity associated with a product introduction and recall, announcements of green activities and introduction of e-commerce (Horsky & Swyngedouw 1987; Capon, Farley & Hoenig, 1990; Szymanski, Bhardwaj & Varadarajan, 1993; Mathur & Mathur 2000; Subramani & Walden 2001).

The theory underlying this procedure is the efficient market hypothesis, which states that a firm's stock price is the present value of the stream of future cash flows. At any point in time, this stock price reflects all known information pertaining to the firm's current and future profits. In this way, stock prices reflect the true value of the firm because they reflect the discounted value of future earnings as well as all relevant information known in the market. New information that will impact on the firm's current and future profitability will result in a change in security price. The purpose of this methodology is to determine whether the announcement of some upcoming event produces a significant stock price reaction around the time of the announcement. For example, the announcement of new information in the form of the company's strategy to advertise on the Super Bowl constitutes new signals about the present and future earnings of the company. This signal will impact its stock price. If the market views this strategy as worthwhile, the perceived value of the firm is likely to increase and be reflected in a simultaneous increase in the market price of the stock. To conduct such tests, daily stock returns are measured around the announcement or event date and are compared with the expected return. The change in the price of the stock, or abnormal returns, after the event reflects the unbiased market valuation of the economic value of the event. This methodology provides a unique opportunity to evaluate the financial effect of a corporate strategy. By examining a sample of firms over a period of years, it is possible to evaluate the overall market valuation of a particular business strategy.

The hypothesis in this paper is tested using the standard event study methodology (see e.g., Brown & Warner, 1985). The expected return is generally based on the Capital Asset Pricing Model or some other suitable market based return generating mechanism. The market model is specified as:

(1.) [R.sub.j, t] = [[alpha].sub.j] + [beta].sub.j] [R.sub.mt] + [u.sub.jt], j = 1, ... , N; t = -325, ... , -71,

where N is the number of issues in the sample, [R.sub.j, t] is the return on stock j for day t, [R.sub.m, t] is the return on market proxy m for day t, [u.sub.j, t] is the random error for stock j for day t normally distributed with E|[u.sub.j, t] | =0, [[alpha].sub.j] is the estimated intercept term for stock j, and [[beta].sub.j] is the estimated risk coefficient for stock j. The market model is estimated by using the equally-weighted market returns from CRSP files as the market proxy.

The prediction errors are calculated for each day in the test period, which begins 50 days before the announcement day and ends 50 days after the announcement day. The prediction error for stock j for day t is defined as

(2.) [PE.sub.j, t] = [R.sub.j, t] - ([[alpha].sub.j] + [[beta].sub.j][R.sub.m,t]), 1,....,N t = [T.sub.1], [T.sub.1] + 1,....,[T.sub.2]

such that E|P[E.sub.j]| =0, i.e., the prediction error is not expected in an efficient market in equilibrium. If E|P[E.sub.j]| [not equal to]0 , then we infer that some unanticipated information had come to the market and was used by well-informed individuals.

The cross-sectional average prediction error for day t for a sample of size n is

(3.) SP[E.sub.1] = [i.summation over (j=1)] P[E.sub.j,t],

and the cumulative average prediction error is

(4.) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

The time-series of CAPEs tells us whether prediction errors would have occurred had investors bought the test portfolio in day T1 and held until day I, i=[T.sub.1], [T.sub.1]+1,...,[T.sub.2].

Following Patell (1976), tests of statistical significance are based on standardized prediction errors. The standardized prediction error for stock j in day t, is calculated as

(5.) SP[E.sub.j,t] = P[E.sub.j,t]/[S.sub.j,t]

where

(6.) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

Where [[sigma].sup.2.sub.j] is the residual from the market model estimation for stock j, T is the number of days in the period used to estimate the model and [??] is the mean market return in the estimation period.

The average standardized prediction error for day t is:

(7.) ASP[E.sub.t] = 1/N [N.summation over (j=1)] SP[E.sub.j,t].

For each day, the Z-statistic is calculated as

(8.) . ASPE N = Z t t *

The limiting distribution of [Z.sub.t] is the unit normal, under the null hypothesis that the mean normalized, standardized prediction error equals zero. Over the testing period, which begins with [T.sub.1] and ends with [T.sub.2], the cumulative normalized, average standardized prediction error is

(9.) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

Then, the Z-statistic is calculated as

(10.) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

and has a unit normal limiting distribution under the null hypothesis that the cumulative normalized, average standardized prediction error over the period from [T.sub.1] through [T.sub.2] equals zero.

DATA

Data was collected for ten years from 1990 to 1999. Announcement day is defined as the date in which the company's announcement of its intention to advertise in the upcoming Super Bowl was published in the print media. Major newspapers (e.g. The Wall Street Journal and New York Times) were consulted as well as leading advertising outlets such as Advertising Age and AdWeek. A search was also conducted on Lexis/Nexis data files to search for possible leaks to alternative sources.

A pre-requisite of the event study methodology is the availability of market data for the firm under investigation. The sample of firms with well-defined announcement dates during the ten-year period totaled 112 or about 80% of total firms. Daily stock prices were obtained from the University of Chicago's Center for Research in Security Prices (CRSP) tape.

RESULTS AND DISCUSSION

Following Brown and Warner (1985), the event day is defined as day '0' (t=0). Using the standard estimation duration of 255 days (t= -325 to t=-71), the parameters of the market model in equation (1.) is estimated. These estimated parameters are then used to calculate expected returns for each of the firms for the specified event window. Abnormal returns are simply obtained by comparing these to the actual returns observed in the event window as shown in equation (2.).

Two events were identified and analyzed. Event 1 is defined as the date that the firm announced its intention to advertise during the Super Bowl and Event 2 is defined as the first trading day after the Super Bowl.

Event 1: Announcement of the Strategy

Results for the 112 announcements for this event, for a window of -10 and +10 are exhibited in Table 1. The results indicate that a firm's announcement of their intention to advertise on the Super Bowl is associated with statistically insignificant negative excess returns. On the event day itself, firms experience an average of a 0.23 percent decline in excess returns. These results are only significant at the ten-percent level.

Cumulative abnormal returns (CAR) are calculated to capture any possibility that the effects may be spread over a few days due to the gradual spread of information and their implications on future cash flows. Results for different windows are exhibited in Table 2. The CAR associated with the announcement of the strategy, is negative and statistically significant for the (-1; 0) and (-1; +1) windows, indicating that there is a pattern of information leak before the announcement date. Cumulative abnormal stock price across the windows -1 and +1 average negative 0.67 percent. This result together with the previous result of insignificant negative abnormal returns (p=0.1) suggests that the market is somewhat pessimistic about the firm's strategy. The announcement of the firm's decision to advertise on the Super Bowl does not result in any significant positive abnormal returns. This indicates that the market believe the strategy will not result in any future flows of positive earnings. A possible reason could be that the market has at that point no credible means of measuring the probability of success in this strategy. The market has little information about the advertising campaign and has indeed not seen the broadcast of the campaign. Thus, it reacts accordingly and does not reward the firm with positive and significant abnormal returns. The significant CAR results merely indicate that the information flows are gradual and there are leaks in the market. Bearing in mind that many Super Bowl advertisers are repeat advertisers, the CAR results probably indicate that investors have simply anticipated that the firms will again participate in yet another Super Bowl.

Event 2: First trading day after Super Bowl Sunday

Event 2 was specified to investigate the market reaction subsequent to the actual broadcast of the advertising campaign. It is defined as the first trading day after Super Bowl Sunday. Results of this analysis exhibited in Table 1 indicates that on the event date, firms experience on average a 0.16 percent positive excess returns. This is statistically significant at the five-percent level. The results indicate that on average firms who advertise on the Super Bowl experienced a gain of 0.16 percent in excess returns on the first trading day following the broadcast. The size of excess returns is in line with the magnitudes obtained in other event studies investigating the impact of marketing strategies on share price responses. Agrawal and Kamakura (1995) found that the announcement of new celebrity endorsers resulted in a 0.44 percent positive abnormal returns, Chaney, DeVinney and Winer (1991) showed that the announcement of new products resulted in a positive abnormal return totally 0.25 percent.

Looking at the event subsequent to the airing of the Super Bowl, CARs, exhibited in table 2, are seen to be not significant between days -1 and +1. This together with the positive average abnormal returns obtained on the first trading day after Super Bowl Sunday suggest that the market reacts in a positive direction to the advertisements broadcast during the game. Having seen the actual execution of the strategy and the reaction of the audience to the advertisement, the market reacts in a positive manner and rewards the firms with an increase in stock prices on average.

CONCLUSION

Two interesting results have emerged from the analysis. The first is that the market considers the announcement of the firm's intentions to advertise on the Super Bowl as a non-news event with insignificant effects on the future cash flows and profits of the firm. The second and more interesting finding is the positive and statistically significant average abnormal returns recorded on the first trading day after the airing of the advertisement on the Super Bowl. The market obviously rewards firms adopting this strategy at this point. The message as well as the reaction of the audience to the advertisement is viewed to positively impact on the firm's future cash flows and profits and stock prices rise accordingly.

These results suggest that the market does not reward the firm on its announcement of its strategy to launch an advertising campaign during the Super Bowl due probably to investors' lack of information about the campaign's probability of success. Therefore, it is in the marketing manager's interest at the announcement date to offer information that will help the market assess the outcome of their forthcoming campaign. Managers should also provide investors with "reminders" of the success of similar campaigns they launched in the past. These additional sets of information will enable investors to more effectively gauge the potential for success of their current campaign.

Future research should include measuring the returns on advertising in other major programs such as the Academy Awards and Grammy awards. Results from these studies can be pooled together to obtain a clearer picture of whether it makes economic sense to target advertising dollars to these specific programs.

REFERENCES

Agrawal, J. & Kamakura, W.A. (1995). The economic worth of celebrity endorsers: an event study analysis. Journal of Marketing, 59, 56-62.

Bass, F.M. & Clark, D.G. (1972). Testing distributed lag models of advertising effects. Journal of Marketing Research, 9, 298-308.

Blattberg, R.C. & Jeuland, A.P. (1981). A micro-modeling approach to investigate the advertising-sales relationship. Management Science, 27, 988-1005.

Brown, S. J. & J. B. Warner. (1985). Using daily stock returns: The case of event studies. Journal of Financial Economics, 14, 205-258.

Capon, N., Farley, J. & Hoenig, S. (1990). Determinants of financial performance: A meta-analysis, Management Science, 36, 1143-59.

Chaney, P.K., Divinney, T. & Winer, R.S. (1991). The impact of new product introduction on the market value. The Journal of Business, 64(4), 573-611.

Dekimpe, M.G. & Hanssens, D.M. (1995). The persistence of marketing effects on sales. Marketing Science, 14(1), 1-21.

Demetrios, V. & Ambler, T. (1999). How advertising works: What do we really know. Journal of Marketing, 63(1), 26-43.

Elliott, S. (1999, January 28). Trying to score big in 'ad bowl' companies jostle to reach consumers on super Sunday. The New York Times, C1/C8.

Hanssens, D.M., Parsons, L.J. & Schultz, R.L. (1990). Market Response Models: Econometric and Time Series Analysis. Boston, MA: Kluwar Academic Publishers.

Horsky, D. & Swyngedouw, P. (1987). Does it pay to change your company's name? A stock market perspective. Marketing Science, 6, 32-34.

Kim, P. (1992). How advertising works: A review of the evidence. The Journal of Consumer Marketing, 9(4), 5-9.

Lee, M. & Lade, D.S. (2001). Demystifying advertising investments. The Journal of Business Strategy, 22(6), 18-22.

Leone, R.P. (1995). Generalizing what is known about temporal aggregation and advertising carryover. Marketing Science, 14(3), 141-50.

Little, J. (1979). Aggregate advertising models: The state of the art. Operations Research, 27.

Lodish, L.M. (1995). A summary of fifty-five in-market experimental estimates of the long-term effects of advertising. Marketing Science, 14(3), 133-40.

Mathur, L.K. & Mathur, I. (2000). An analysis of the wealth effects of green marketing strategies. Journal of Business Research, 50(2), 193-200.

Mela, C.F., Gupta, S. & Lehman, D.R. (1997). The Long-term impact of promotion and advertising on consumer brand choice. Journal of Marketing Research, 34, 248-61.

O'Connell, V. (2002, January 28). Super bowl gets competition. The Wall Street Journal, B1.

Patell, J. M., (1976) Corporate forecasts of earnings per share and stock price behavior: Empirical tests. Journal of Accounting Research, 14(2), 246-276.

Rao, R.C. (1986). Estimating continuous time advertising sales model. Marketing Science, 5(2), 125-142.

Rao, U.H. & Burnkrant, R.E. (1991). Effects of repeating varied ad executions on brand name memory. Journal of Marketing Research, 28, 406-416.

Subramani, M. & Walden, E. (2001). The impact of e-commerce announcements on the market value of firms. Information Systems Research, 12(2), 135-154.

Szymanski, D.M., Bharadwaj, S.G. & Varadarajan, P.R. (1993). An analysis of the market share-profitability relationship. Journal of marketing, 57, 1-18.

Winer, R.S. (1979). An Analysis of the time-varying effects of advertising: The case of Lydia Pinkham. Journal of Business, 52(4), 563-76.

Winer, R.S. (1980). Estimation of a longitudinal model to decompose the effects of an advertising stimulus on family consumption. Journal of Consumer Research, 21, 708-718.

Zufryden, F.S. (1987). A model for relating advertising media exposures to purchase incidence behavior patterns. Management Science.

Peggy Choong, Niagara University Greg Filbeck, Schweser Study Program Daniel L. Tompkins, Niagara University Thomas D. Ashman, Union College

Table 1: Abnormal Returns of Announcements

 Average Abnormal Returns (%)

 (A.) Event 1: Announcement (B.) Event 2: After Super Bowl

-10 -0.03 0.03
 -9 -0.22 * 0.33 **
 -8 -0.55 ** -0.20 *
 -7 -0.22 (+) -0.14 (+)
 -6 -0.50 ** -0.39 **
 -5 -0.05 -0.14
 -4 -0.19 0.01
 -3 0.11 0.32
 -2 -0.29 * -0.51 **
 -1 -0.41 (+) -0.33 **
 0 -0.23 (+) 0.16 *
 +1 -0.05 -0.29 *
 +2 -0.19 -0.10
 +3 -0.39 * 0.01
 +4 0.04 -0.63 **
 +5 -0.14 0.03
 +6 -0.13 -0.35 **
 +7 0.13 -0.25
 +8 -0.15 -0.12
 +9 0.06 -0.31
+10 -0.34 0.26

** Significant at the p=0.01 level

* Significant at the p=0.05 level

(+) Significant at the p=0.10 level

Table 2: Cumulative Abnormal Returns for Announcement Event Dates

Windows Cumulative Abnormal Returns (CAR) (%)

 Event 1: Announcement Event 2: After Super Bowl

-10; -6 -1.51 ** -0.37
-10; -1 -2.35 ** -1.02 *
 -5; -1 -0.83 * -0.65 (+)
 -2; 0 -0.93 ** -0.67 *
 -1; 0 -0.64 * -0.16
 -1; +1 -0.69 * -0.46 (+)
 0; +2 -0.46 * -0.22
 +1; +5 -0.72 (+) -0.96 **
 +1; +10 -1.15 * -1.73 **
 +6; +10 -0.43 -0.76 **

** Significant at the p=0.01 level

* Significant at the p=0.05 level

(+) Significant at the p=0.1 level