Firing advertising agencies: possible reasons and managerial implications.

This paper examines the possible reasons for and managerial implications of the firing of advertising agencies. The firing of an advertising agency is an important event for the agency and for the client. For the agency, losing just one major account can seriously reduce agency revenues and

profits, particularly if news of the loss encourages further account losses. Anecdotal evidence suggests that many ad agencies lose two-thirds of their accounts every five years (see, for instance, Adweek, May 25, 1992). Thus, ad agencies need to understand the reasons for their firing so they can establish effective programs to help reduce client turnover.

Ad agency changes are serious decisions for client firms because of the risks involved. Ad agencies provide intangible services that may substantially affect sales, profits, and market share. Yet these effects cannot be accurately predicted. Which agency will produce great ads and which will not? Manufacturing firms may routinely change suppliers of components, parts, or materials because the product purchased can be exactly specified and tested ahead of time for compliance to specifications. This is simply not possible with advertising agencies.

Another risk for clients considering an agency change involves the possible communication effect from the change itself. Agency firings--particularly those involving well-known consumer products or firms--are public events, often covered in the business press. How will investors react to the news of the firing? Will they see it as good news or will they sell the stock? In sum, ad agency firings pose important managerial issues for agencies and their clients. This research is intended to help firms manage better by providing new information about why firings occur and why investors react as they do.

In this paper we investigate factors such as changes in sales, market share, or profitability that drive a firm to terminate the client-agency relationship. Our analysis of these factors will help us understand the underlying reasons for ad agency firings and suggest some preventive steps for ad agencies to contemplate. In addition, we study a sample of ad agency firings to identify the impact of such firings on the stock prices of the client firms using event-study methodology. Basic finance theory suggests that the market value of a firm is the market's assessment of the future prospects of the firm. When an event such as the firing of the firm's ad agency occurs, the market quickly incorporates this new information into its assessment of the future prospects of the firm and bids up or down the price of the firm's stock. Thus, by studying the stock market's reaction to the event, we provide evidence on whether the market considers the firing of an ad agency to be good news or bad news. Finally, we study the relation between changes in sales, market share, or profitability and the stock market reactions to the firing.

LITERATURE REVIEW AND HYPOTHESES

A review of the academic literature reveals only one prior study, by Hozier and Schatzberg (2000), of ad agency firing that examines its stock price impact. Hozier and Schatzberg report a significant reduction in the stock price of client firms two days before the announcement of the firing, decay in firm sales growth before and after the event, deterioration in operating income after the firing, and a decline in liquidity before the firing. The present study differs from Hozier and Schatzberg's in a number of important respects. Whereas their sample consists of 16 terminations that occurred during 1986-1994, our sample consists of 49 ad agency account terminations that occurred during 1981-1999. In addition, our study differs from the Hozier and Schatzberg study because: (1) we provide evidence of the effect of the firing of the ad agency on the value of the fired and the replacement agencies, and (2) we find that one significant reason behind the firing (loss in market share) is closely related to the stock price consequence of the firing.

Mathur and Mathur (1995, 1996) investigate stock price reaction to advertising slogan changes (1995) and to the initiation of new ad agency-client relationships (1996). The latter includes placing new accounts with new agencies or existing agencies but does not include firing an agency. They find that stock prices rise in response to advertising slogan changes but show mixed reactions to agency-client changes. For example, when firms announce that a new account has been given to an agency that has an existing relationship with the company, the stock price of the company drops significantly on the announcement day but recovers ten days later.

Earlier studies by Buchanan and Michell (1991), Michell, Cataquet, and Hague (1992), and Henke (1995) report on the ad agency-client relationship but do not examine its stock price consequences. Buchanan and Michell (1991) use structural factors such as account size, past account switching behavior, age of the account, agency size, and product class to study the risk of account failure in Great Britain. They find that organizational factors such as total client advertising budget, total agency billings, and historical switching rates for the industry correlate significantly with risk of failure. Michell, Cataquet, and Hague (1992) survey client firms after an account switch. They find that dissatisfaction with agency performance is the primary reason given for the change. This broad category includes such items as "standard of creative work," "disagreement over marketing objectives," "standard of marketing advice," and "not close enough to client's business." Henke (1995) interviewed 151 "decision makers" by phone to determine what factors are critical to making an agency switch. She found that decision makers who had made a recent switch differed from those who had not in a number of ways. For example, decision makers who had made a recent switch were less satisfied with the size of their account relative to the agency's other accounts and with the agency's media skills. Paradoxically, they were more satisfied with agency creative skills but placed less importance on this variable. Finally, they were more aware of other ad agencies than decision makers who had not made a recent switch.

A very different perspective in the literature on ad agency firing involves the alleged use of agencies as scapegoats by their clients. Advertising agency executives and others familiar with the industry have long believed that agencies are often fired for reasons having little to do with their performance. Rather, it is claimed that agencies are sometimes replaced as a signal to important corporate stakeholders such as board members, current and prospective stockholders, creditors, dealers, and retailers that a troubled company's management is acting aggressively to improve its position. Joanne Lipman (1991) quotes David Ogilvy in the Wall Street Journal that it is "easier to fire your agency than admit to your stockholders that there is something wrong with your product or management." Lipman suggests that this "blame the messenger" strategy is a way of buying time during which the economy may turn up and the firm's fortunes might improve. To illustrate her point, Lipman notes that American Express fired its ad agency, Ogilvy and Mather, even though the agency's ad campaign for American Express was honored as the decade's best by Advertising Age. Other examples, she notes, of firms reviewing their ad agencies include Johnny Walker Red scotch whiskey, Compaq computer, and Mercedes Benz of America. According to Lipman, "All of these clients have one thing in common: major business problems that probably weren't caused--and almost certainly won't be cured--by advertising." Mercedes, for example, suffered a sales decline of 21% the year it fired its ad agency. Lipman writes that the client companies involved typically reject any suggestion that firing an agency is unwarranted. In the case of Compaq, for instance, a spokesperson noted that the company valued a "fresh start" and felt it was important to look at "new ideas." The spokesperson concluded, "We don't lay any blame on our agency for any of our problems" (Lipman 1991).

This brief literature review suggests several conclusions. First, the methodology of studying ad agencies has changed over time. Starting with interviews and surveys of client employees, it has evolved into empirical analyses, such as event studies. Both of these approaches are useful in their own ways and have inherent strengths and weaknesses. Empirical analyses can test hypotheses rigorously but may miss important relationships if the necessary data is not available or if empirical models are misspecified. Interviews can be flexible and informative and suggest hypotheses for future testing, but they implicitly assume that respondents know the truth and are willing to report it. In the case of the firing of ad agencies, we wonder if client or agency employees would be willing to report reasons for agency firings that might not reflect well on their employer. Second, the application of event-study methodology to advertising provides an example of how the fields of marketing and finance are growing together. Marketing activities are now studied with finance methodology and judged against financial standards, such as the impact on stock prices. Third, our study is a logical extension of earlier work; it contributes to knowledge by replicating part of the Hozier and Schatzberg study and then extends analysis into new areas, which help explain--and may even predict--agency firing.

We developed three major hypotheses that are tested in this paper. The first hypothesis is related to the reasons for the firing. We want to determine whether the client experienced a significant decline in sales, market share, or profitability in the period preceding the firing and whether this decline in sales or in profitability was isolated to the client or endemic to the industry. If the firing is preceded by declining sales or profitability but this decline is endemic to the industry, then the evidence would be consistent with David Ogilvy's claim that the ad agency is being made a scapegoat for forces that are essentially unrelated to the agency's performance. However, if the declining sales or profitability is isolated to the firm that fired its ad agency, the evidence would lead to the conclusion that there is something unique about this firm, including the possibility that its advertising strategy may be flawed in some respect. Therefore, the first hypothesis is as follows:

H1: Sales revenue for the client falls, either absolutely or in relation to the industry, in the period preceding the firing of its ad agency.

Our second hypothesis is related to the stock price consequences of the firing for all the firms that are affected. We posit that the firing of a firm's ad agency is viewed as positive news for the client by the stock market. Assuming that the firing is prompted by declining sales or profitability, the agency that is hired as a replacement will presumably be able to improve the client's fortunes by recasting its ad campaign. Therefore, the second hypothesis is as follows:

H2: The market value of a firm increases on firing its ad agency.

We would reject our second hypothesis if we found that the announcement of the firing has no impact or has a negative impact on the value of client firms.

In addition, the firing of an ad agency has significant revenue implications for the agency that is fired as well as for the agency that replaces it. Although ad agencies were traditionally organized as either partnerships or as privately held corporations, a number of ad agencies grew large in the 1990s through mergers and consolidations, and were able to tap into the capital market by going public. Therefore, we are able to examine whether being fired or being hired as a replacement has any impact on the stock price of the agency. The following two hypotheses for the fired and hired agencies are closely related to the second hypothesis above; they are therefore called H2A and H2B:

H2A: The market value of an ad agency decreases on being fired.

H2B: The market value of an ad agency increases when it is hired as a replacement for a fired agency.

In the third hypothesis we tie together the evidence on changes in sales or in profitability and the stock price consequences of ad agency firing. To test H2, we calculate the mean return on stock for the firing firms surrounding the event. However, regardless of the value of the mean return, there is likely to be considerable return variability surrounding it. Thus, investors may view the ad agency firing as good news for some firms and bad news for others, depending upon the unique circumstances of each firm. For instance, a firm might have suffered a significant decline in sales due to changing demographics that are in line with sales declines experienced by every firm in the industry. For this firm and industry, no amount of creative advertising will reverse the downward spiral in sales. Therefore, the firing of the firm's ad agency may be viewed as "no news" by the market. In another instance, the firm's sales may have declined purely due to a poor advertising strategy while the industry is thriving. Investors will view the firing of this firm's ad agency as an attempt by the managers to recognize and deal with the problems that appear to be confined to the firm. Consequently, the value of the firm will rise as investors recognize the potential for increased sales and profitability in the future. Thus, in order to tie together the evidence on changes in sales or in profitability and the stock price consequences of ad agency firing, we relate the return on stock for each client firm to its unique characteristics such as sales growth and profitability of the firm and the industry. The third hypothesis is as follows:

H3: The change in the market value of stock for a client firm that fires its ad agency is related to recent changes in its sales or profitability, either absolutely or relative to the industry.

DATA AND METHOD

We searched the Wall Street Journal Index to identify announcements of advertising agency firings from 1981 to 1999. We selected only those events that were related to publicly listed U.S. companies. We took great care in identifying the first public announcement of every event that appears in our sample by searching the full text of the Wall Street Journal electronically to ensure that no prior news related to the termination of the ad agency had appeared. We eliminated any announcements of ad agency terminations that were preceded by a news item suggesting the possibility of such an event. We also eliminated any ad agency terminations that were preceded by an announcement that the account had been put up for a review. This process left us with a sample of a total of 49 firings. Table 1 shows summary statistics on the annual distribution of the 49 events together with the median sales and total assets of client firms. We retrieved the income statement and balance sheet data from COMPUSTAT. It is worth noting that the agency firing events are not evenly distributed over time. Firings are somewhat more common in the period 1989-1991, a time of economic difficulty in the United States. The overall median sales for client firms of $8.97 billion and the median total assets of $11.26 billion suggest that these firms are relatively large.

We examine the impact of advertising agency firings on company stock prices by using event-study methodology. Event-study methodology can be used to study the impact of an event--such as a stock split, a merger, a strategic alliance, or an ad agency firing--on the market value of the stock of firms affected by that event. Finance theory suggests that the appropriate goal for business managers is to maximize the value of the firm's stock (see Fama and Miller 1972, chap. 2, for a classic treatment of this issue). Thus, event-study methodology was created to determine whether a particular decision is consistent with this goal. For instance, suppose a researcher wants to study whether a stock split is consistent with maximizing the value (or price) of stock. To determine whether a stock split has any impact on the price of stock, the researcher would first compile a list of firms that announced a stock split together with a list of the dates on which the split was announced. Then they would study the impact of the split on the price of the stock and draw conclusions based on the results.

Event-study methodology has been employed in numerous studies in the literature on finance, economics, management, marketing, and information systems. Indeed, based on the breadth and the number of studies that have been published using this methodology, it would be fair to state that it is a mature and well-accepted methodology. This methodology was developed and first used by Fama et al. (1969) to examine the impact of stock splits on stock returns. Subsequently, in the finance literature, event studies have been used to study corporate security issues (Smith 1986), capital expenditure decisions (McConnell and Muscarella 1985), asset sell-offs (Lang, Poulsen, and Stulz 1995), mergers and acquisitions (Jensen and Ruback 1983), spin-offs (Desai and Jain 1999), and virtually any other finance-related corporate decision.

However, the methodology can be used to analyze any event that has an impact on stock prices, not just finance-related events. Thus, event studies have been employed in the economics literature to study the impact of monetary policy (Santomero 1991) and to study the impact of regulatory changes on stock returns (Prager 1989). They have appeared in the management literature (see, e.g., Cascio, Young, and Morris 1997; Das, Sen, and Sengupta 1998; Madhavan and Prescott 1995), in the information systems literature (Chatterjee, Richardson, and Zmud 2001; Subramani and Walden 2001), and in the marketing literature. In the marketing literature, event studies have been used to study new product introductions (Chaney, Devinney, and Winer 1991), brand extension announcements (Lane and Jacobson 1995), the addition of an Internet distribution channel (Geyskens, Gielens, and Dekimpe 2002), and corporate name changes (Koku, 1997). Finally, event studies have been used in the advertising literature as well. We discussed the advertising-related event studies that are most relevant to our current study in our section on literature review. Other advertising-related event studies that are worth mentioning include those of Agrawal and Kamakura (1995) and Mathur, Mathur, and Rangan (1997), both examining the stock price effects of having a celebrity endorser. Thus, the use of event-study methodology has a number of important precedents in marketing and in advertising.

We provide the details of implementing an event study in a primer that appears in the Appendix. We estimate the market model over a 255-day period, from t = -300 to t = -46, where t = 0 is the date the announcement of an ad agency firing appeared in the Wall Street Journal. In the results that we report here, we use the CRSP Equally Weighted Portfolio to proxy for the market index; however, the results are unchanged if other indices are used to proxy for the market index. Using the estimated regression parameters and the market index return, expected returns, E([R.sub.jt]), for each firm j were estimated for t = -30 to t = +30. For the sake of brevity, and in keeping with practice followed in other studies, we report results beginning from t = -5 to t = +1.

RESULTS

Table 2 shows results on the first hypothesis. To test the first hypothesis, we calculated the growth rate in sales for the sample firms in the period preceding the firing of the ad agency. If the fiscal quarter in which the firing occurred was quarter 0, then we calculated the growth rate in sales from quarter 2 to quarter 1, from quarter 3 to quarter 1, and from quarter 4 to quarter 1. In Panel A of the Table we present the figures for the absolute growth rate in sales, while in Panel B we present the results for the growth rate minus the growth rate for the industry. The growth rate for a firm minus the growth rate for the industry in which the firm operates can be interpreted as an indicator of the change in the market share for the firm. For instance, if the sales of a particular firm grow by 3% over a quarter while the sales for the average firm in the industry grow by 5%, the industry-adjusted growth in sales for this firm--2%--suggests that this firm lost market share. To calculate the growth rate for the industry, we identified from COMPUSTAT all the firms in a sample firm's four-digit SIC code, retrieved the sales for all of them, calculated the growth rate in sales, and then calculated the median growth rate for the industry. This procedure was repeated for each firm that appears in our sample. For 2 out of the 49 firms in our sample, we were unable to find any other firms in the four-digit SIC code, leaving us with 47 firms for the industry-adjusted growth rate in sales.

Firms that fire their ad agencies experience a negative growth in their sales, or a decline in sales, in the quarter immediately prior to the one in which they fire their ad agency. On an absolute basis, sales decline by a statistically insignificant 2.95%. However, sales relative to peers in the firm's industry decline by a statistically significant 4%. Furthermore, although there is no evidence that in absolute terms sales decline over a longer period, the evidence suggests that, relative to peers in their industry, sales decline from quarter 3 to quarter 1 by 7.38% (statistically significant at the 10% level). In addition, the evidence suggests that sales may have declined relative to the industry over an even longer period, from quarter 4 to quarter 1 by 2.47%; however, this figure is statistically insignificant. Therefore, it is clear that firms that fire their ad agencies have lost market share over the last two quarters.

To test for declining profitability in the period preceding the firing, we calculated a number of frequently used profitability ratios, such as return on assets, return on equity, and gross and net profit margins However, in results that are not reproduced here, we found no evidence that the declining sales have any effect on the profitability of the firm. In addition, we looked for evidence for declining sales and profitability over a longer period by retrieving annual data for three years prior to the year in which the agency was fired. We found no evidence in the annual data that sales or profitability declined for the firms in our sample. We therefore conclude that firms that fire their ad agencies are (1) somewhat sensitive to a decline in market share regardless of whether such a decline affects profitability, and (2) they are quick to react to the decline in market share by dismissing their ad agency in a matter of months.

Table 3 shows results on our second set of hypotheses, which are related to the market value of the stock of the firms that are affected by the firing of an ad agency. Results for agency clients are shown in Panel A, results for agencies that were fired are shown in Panel B, and results for agencies that were hired as replacements are shown in Panel C. Our sample size for clients fell from 49 to 41 because we eliminated eight events where we found that another significant corporate incident, unrelated to the firing of the ad agency but related to the client, had occurred on either the same day that the firing was announced or within one day before or after that day. In such situations we cannot unequivocally claim that the change in the price of the stock is related to the firing of the ad agency; we therefore eliminated such events from further analysis. From these 41 events we identified all the fired and replacement agencies that were public corporations and screened them to ensure that no other incident for the agency occurred close to the time of the firing. After applying these screens, we were left with a sample of 18 fired agencies and 10 replacement agencies. Table 3 shows the value of the average abnormal return from day -5 to day +1 and its related t statistic. In addition, we provide the number of firms that have a negative return and the number of firms that have a positive return on each day. We also tested the hypothesis that the number of firms with a positive return is equal to the number of firms with a negative return by providing, in the last column of the table, the Z statistic of the nonparametric sign test for that hypothesis. Corrado and Zivney (1992) demonstrate that this sign test is better specified than the parametric t test under the hypothesis of zero abnormal return.

The average abnormal return for firms that fired their ad agencies is negative in each of the three days prior to the day the announcement was made public. On day -2, the abnormal return is -.38% and the t test is statistically significant at the 10% level. The related sign test is statistically significant at a level of less than 1%. The abnormal return cumulated over days -3 to -1 is -.87% (the related t statistic is statistically significant at 5%). Therefore, it is clear that in the three days prior to the firing of its ad agency, a firm experiences negative and statistically significant returns. On the day that the firing of the ad agency is made public, however, there appears to be no reaction in the stock market. These results are consistent with Hozier and Schatzberg's findings that two days prior to announcing a review of an ad agency account or firing of an ad agency, the firm experiences a negative return, but on the day the firing is made public, there is no reaction in the stock market. It appears that the impending firing of an ad agency becomes public knowledge before it is formally announced by the client.

For the ad agencies that are fired, the abnormal returns are negative from day -2 to day 0. On day 0, the average abnormal return is -.84%, and both the t statistic as well as the sign test are significant at the 5% level. For the ad agencies that are hired as a replacement, the return on the announcement day is 3.71% and the t statistic is significant at a level of less than 1%. On the next day, however, these agencies appear to give up some of the increase in the stock price, because the return on day + 1 is -1.21% (statistically significant at the 5 % level). Overall, investors view being hired as a replacement to be good news and being fired as bad news for an agency.

In Table 4 we bring together the results on the changes in sales or in profitability and the stock price consequences of ad agency firing to examine the relation between them. The abnormal return for each client firm is an estimate of the market's consensus on the effect that the firing will have on the firm. Although the average abnormal return for clients on day -2 is -.38%, there is substantial variability around that average. For instance, it ranges from a low of -4.56% to a high of 3.73%. We relate the variability in abnormal return to the circumstances of every firm by regressing it against firm characteristics such as the absolute and industry-adjusted growth rate in sales. For client firms, it appears that the announcement effect of firing the ad agency begins on day -3. Therefore, we calculate the cumulative abnormal return for every firm over day -3 to day 0 (our results are qualitatively unchanged if we use other periods to cumulate the abnormal returns) and regress it against several measures of sales growth and profitability.

In regressions that are not reproduced here, we found that there is no relation between abnormal return and absolute growth in sales. In addition, from regressions 1 and 2 shown in Table 4, we conclude that there is no relation between abnormal return and the indicator for change in market share from quarter 2 to quarter 1 or from quarter 3 to quarter 1. However, from regression 3 we conclude that there is a strong positive relation between abnormal return and the indicator for change in market share from quarter 4 to quarter 1. From regressions 4 and 5 we see that this relation is robust to other specifications. In particular, we allow the industry-adjusted return on equity to enter into the regression as an independent variable, and we even allow it to interact with the growth rate in order to test for any interaction effects between the growth rate and return on equity. Yet we find that the only variable that has the power to explain the variability in abnormal return is the industry-adjusted growth rate in sales from quarter 4 to quarter 1.

We interpret the lack of any relation between abnormal return and the short-term indicator of the change in market share as evidence that investors do not pay much attention to the firing of ad agencies by firms that have transitory problems with growth in revenue and change in market share. In addition, we believe that the strong relation between abnormal return and the medium-term change in market share suggests that investors pay close attention to firms that appear to be losing market share over a medium term. Most important, the strong positive relation between abnormal return and the indicator for the change in market share from quarter 4 to quarter 1 suggests that investors take the firing of the ad agency to be a confirmation by the management of these clients of a loss in market share. This is because a positive relation between abnormal return and the indicator for change in market share implies not only that firms that had an increase in market share had high abnormal returns but also that firms with a decrease in market share had a low or negative abnormal return. Investors do not view the firing of the ad agency as management taking corrective action that will remedy the sales growth situation.

CONCLUSIONS AND DISCUSSION

In this study we show that the firing of an ad agency is preceded by a loss in the client's market share in the two immediate quarters prior to the quarter in which the ad agency is fired. This finding highlights some important managerial implications for ad agencies as well as for client firms.

Market share is an important indicator of relative positioning of a firm vis-a-vis its competitors in the firm's product market. In the long run, a growing market share will help establish a firm's dominant position relative to its competitors in a particular product market. A number of factors, such as pricing, product quality, distribution, and competitors' marketing strategies, play a role in securing market share. Ad agencies, in their role as creators of advertising strategies, are instrumental in helping a firm to secure market share. As our results indicate, however, ad agencies appear to get a disproportionate share of the blame--as evidenced in their being fired--for a falling market share even though ad campaigns are generally developed with mutual consultation between the client and the ad agency.

Given the above finding, it is important for ad agencies to pay particular attention to sustaining market share for their clients. Ad agencies may be able to anticipate their contract review and termination by keeping a close tab on their clients' market share. It is possible for ad agencies to be complacent about the effectiveness of their ad campaigns if they merely pay attention to increases in sales and not on the market share of their clients. However, we recognize that advertising is only one component of a company's marketing mix and that ad agencies have control over only a limited set of parameters that have any impact on their clients' market share.

In addition, we replicate part of the work of Hozier and Schatzberg (2000) and confirm one of their findings using substantially more events spread over a longer time period. We report that the market value of the common stock of the firms that fire their ad agencies declines during the three days before a public announcement is made. Most important to note is that we extend the analysis by showing from our cross-sectional regressions--for the first time--that the decline in the stock prices of client firms is strongly related to a decline in market share over two consecutive quarters but unrelated to a reduction in market share over one quarter. We interpret this result to mean that investors are concerned about firms that have lost market share over a medium term, but brush aside as unimportant, losses in market share over a short term.

Finally, the fact that the stock market does not react positively to the firing of an ad agency leads us to conclude that investors do not believe that firing the existing ad agency or hiring a new ad agency will immediately alleviate the reasons for the decline in the market share. This may be because several factors besides ad campaigns help determine a firm's market share. This suggests that client firms need to pay careful attention to the generally negative impact on their firm's stock prices as a result of ad agency firings. For investors to react positively, the firms probably also need to make other appropriate changes.

We also studied the impact on the stock prices of ad agencies that get fired as well as those hired as replacements. We found that agencies that are fired experience a decline in the price of stock, while agencies that are hired as replacements experience an increase in the price of stock. Further study of the hired firms and their clients may be of interest, particularly to investigate if these ad agencies were able to reverse the declining market share for their clients.

A few words regarding our analysis and methodology are in order. Much of the literature on the topic of client-agency relationships relies on responses received through surveys and questionnaires. Respondents--consisting of employees of client firms--are asked about their relationship with their advertising agency. It is possible that respondents may be unwilling to reveal the underlying reasons for the firing of the ad agency. In fact, the existing literature does not report declining market share as a reason for the firing of ad agencies. Our results are based on publicly available financial information; thus, they do not suffer from the sort of bias that can be problematic in information obtained from surveys and questionnaires. Therefore, our results provide new insights on this topic and our methodology is likely to be of interest to other marketing and advertising researchers.

APPENDIX

A Primer on Event-Study Methodology

In theory, event studies examine the stock price consequences of an event. In practice, however, the methodology deals with stock returns. The return on stock j at time t, [R.sub.j,t] is

(A1) [R.sub.j,t] = [P.sub.j,t] - [P.sub.j,t-1] + [D.sub.j,t]/[P.sub.j,t-1].

where [P.sub.j,t] is the price of the firm's stock at time t, [P.sub.j,t-1] is the price of firm's stock at time t - 1, and [D.sub.j,t] is the dividend paid by the firm at time t. By using returns in lieu of prices it is possible to avoid any issues that may arise due to the different magnitudes of the price of stock for the firms that appear in the study while being able to account for any dividend paid by a firm during the scope of the study. To determine the impact of an event on the returns of the firms that are affected, an abnormal return is computed for every firm:

(A2) A[R.sub.j,t] = [R.sub.j,t] - E([R.sub.j,t]),

where A[R.sub.j,t] is the abnormal return and E([R.sub.j,t]) is the expected return on that stock. Thus, the abnormal return is the actual return on the stock less the expected return on the stock in the absence of the event. If the event is unanticipated by the market, then the abnormal return captures the impact of the event on the value of the firm's stock. To ensure that the event is unanticipated, it is vital to gather, for every firm that appears in the sample, the first public announcement of the event. There are three different models for expected return that have appeared in the literature: (1) the mean-adjusted return model, (2) the market-adjusted return model, and (3) the market model. Brown and Warner (1985) demonstrate that the power and specification of the tests that emerge from the three models are similar; therefore, it does not matter which model is employed. Most event-study papers employ the market model where the expected return for firm j on day t is given by

(A3) E([R.sub.j,t]) = [[alpha].sub.j] + [[beta].sub.j][R.sub.m,t],

where [R.sub.m,t] is the return on the market portfolio, and [[alpha].sub.j] and [[beta].sub.j] are the parameters of the market model. The CRSP Value-Weighted Portfolio or the CRSP Equally Weighted Portfolio is used as a proxy for the market portfolio in event studies based on U.S. data. The parameters of the market model are estimated, typically, from data prior to the event date. If the public announcement day of an event is denoted as t = 0, then the estimates may be based on, for instance, data from t = -250 to t = -60. After estimating the parameters of the market model for every firm in the sample, abnormal returns are calculated for each firm during an event window. The length of the event window varies from study to study, but days that are close to the event day are always included in the analysis. Thus, for example, most event studies report results for days t = -2, -1, 0, and + 1. To test hypotheses regarding the abnormal return for firms that are affected by the event, a daily average abnormal return, AA[R.sub.t] is calculated:

(A3) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

In (A3), n is the number of firms that are affected by the event. The t statistic for testing the null hypothesis of zero abnormal return,

(A4) t = AA[R.sub.t]/s(AA[R.sub.t]),

is Student t distributed with n-2 degrees of freedom and s(AA[R.sub.t]) is the standard error of the average abnormal return. In addition to the t statistic above, frequently the Z statistic of the nonparametric sign test that the number of firms with a positive abnormal return is equal to the number of firms with a negative abnormal return is also reported. Corrado and Zivney (1992) demonstrate that this sign test is better specified than the parametric t test under the hypothesis of zero abnormal return.

TABLE 1
Summary Statistics on Firms That Fired
Their Ad Agencies

                 Median sales    Median assets
Year             (in millions)   (in millions)    n

1981                 $5,312         $2,702        1
1983                 $9,736          5,986        1
1984                 $1,272          1,346        1
1986                 $4,044          2,681        6
1988                $59,681         73,037        1
1989                 $8,730          8,460        5
1990                 $6,206          8,249        9
1991                $17,295         24,170        5
1992                 $1,855          2,282        1
1993                $42,260         43,830        1
1994                 $5,626         11,045        6
1995                 $8,647         14,499        4
1996                $48,603         44,322        2
1997                $34,697         27,544        3
1998                $24,152         28,141        2
1999                $61,751         61,381        1

Overall median       $8,965         11,261       49

TABLE 2
Quarterly Growth Rate in Sales Revenue for Firms
That Fired Their Ad Agencies

Panel A: Absolute   growth rate in sales

                                Mean
                               growth
                                rate    t statistic   n

From  quarter 2 to quarter I   -2.95%     -1.27       49
From  quarter 3 to quarter I     .43%       .25       49
From  quarter 4 to quarter I    5.09%      2.33 **    48

Panel B: Growth rate in sales relative to the industry

                                Mean
                               growth
                                rate    t statistic   n

From  quarter 2 to quarter I   -4.00%     -2.13 **    47
From  quarter 3 to quarter I   -7.38%     -1.66 *     47
From  quarter 4 to quarter I   -2.47%      -.47       46

* Significant at the 10% level.

** Significant at the 5% level.

*** Significant at the 1% level.

TABLE 3
Daily Average Abnormal Returns for Firms Affected by the Firing of Ad
Agencies

Panel A: Results for 41 client firms that fired their ad agency

                           Positive:   Z statistic for
Day   AAR    t statistic   Negative       sign test

-5    -.18      -.65        17:24          -.76
-4     .17       .62        20:21           .18
-3    -.31     -1.13        20:21           .18
-2    -.38     -1.40 *      11:30         -2.64 ***
-1    -.19      -.72        21:20           .49
0      .22       .82        18:23          -.45
1      .19       .70        19:22          -.14

Panel B: Results for 18 ad agencies that were fired

                           Positive:   Z statistic for
Day   AAR    t statistic   Negative       sign test

-5     .54      1.16         8:10          -.12
-4    -.21      -.45         7:11         -1.07
-3     .35       .75          9:9           .35
-2    -.08     -0.17          9:9          -.12
-1    -.35     -0.75         8:10          -.12
0     -.84     -1.80 **      4:14         -2.49 **
1      .27       .57         11:7          1.30

Panel C: Results for 10 ad agencies that were hired as replacements

                           Positive:   Z statistic for
Day   AAR    t statistic   Negative       sign test

-5      .52      .90          6:4           .72
-4     -.60    -1.02          3:7          -.55
-3      .32      .55          6:4           .72
-2      .30      .51          4:6         -1.18
-1      .02      .03          4:6           .08
0      3.71     6.37 ***      6:4           .72
1     -1.21    -2.07          3:7          -.55

* Significant at the 10% level.

** Significant at the 5% level.

*** Significant at the 1% level.

TABLE 4
Cross-Sectional Regressions for Clients
That Fired Their Ad Agencies

              1      2      3         4      5

Intercept   -.01     0    -.01      -.01   -.01
SG2-IA        0
SG3-IA              .03
SG4-IA                     .04 **           .09 ***
ROE-IA                              -.02    .02
ROESG4                                     -.03
[R.sup.2]    .06    .06    .15       .02    .21
N             38     38     37        37     36
F           2.35   2.35   6.34       .70   2.90

Note.: SG2-IA is the industry-adjusted growth in sales from quarter 2
to quarter 1; SG3-IA is the industry-adjusted growth in sales from
quarter 3 to quarter 1; SG4-IA is the industry-adjusted growth in
sales from quarter 4 to quarter 1; ROE-IA is the industry-adjusted
return on equity; ROESG4 is the interaction between ROE-IA and
SG44A.

* Significant at the 10% level.

** Significant at the 5% level.

*** Significant at the 17% level.

The authors gratefully acknowledge the comments received from the Editor and three anonymous reviewers. Any remaining errors are the authors' own.

REFERENCES

Agrawal, Jagdish, and Wagner A. Kamakura (1995), "The Economic Worth of Celebrity Endorsers: An Event Study Analysis," Journal of Marketing, 59 (July), 56-70.

Brown, Stephen, and Jerold Warner (1985), "Using Daily Stock Returns: The Case of Event Studies," Journal of Financial Economics, 14 (March), 3-31.

Buchanan, Bruce, and Paul C. Michell (1991), "Using Structural Factors to Assess the Risk of Failure in Agency-Client Relations," Journal of Advertising Research, 31 (August/September), 68-75.

Cascio, Wayne F., Clifford E. Young, and James R. Morris (1997), "Financial Consequences of Employment-Change Decisions in Major U.S. Corporations," Academy of Management Journal, 40 (October), 1175-1189.

Chaney, Paul K., Timothy M. Devinney, and Russell S. Winer (1991), "The Impact of New Product Introductions on the Market Value of Firms," Journal of Business, 64 (October), 573-610.

Chatterjee, Debabroto, Vernon J. Richardson, and Robert W. Zmud (2001), "Examining the Shareholder Wealth Effects of Announcements of Newly Created CIO Positions," MIS Quarterly, 25 (March), 43-70.

Corrado, Charles J., and Terry L. Zivney (1992), "The Specification and Power of the Sign Test in Event Study," Journal of Financial and Quantitative Analysis, 27 (September), 465-478.

Das, Somnath, Pradyot K. Sen, and Sanjit Sengupta (1998), "Impact of Strategic Alliances on Firm Valuation," Academy of Management JournaL 41 (February), 27-41.

Desai, Hemang, and Prem C. Jain (1999), "Firm Performance and Focus: Long-Run Stock Market Performance Following Spinoffs," Journal of Financial Economics, 54 (October), 75-101.

Fama, Eugene F., Lawrence Fisher, Michael Jensen, and Richard Roll (1969), "The Adjustment of Stock Prices to New Information," International Economic Review 10 (January), 1-21.

--, and Merton H. Miller (1972), The Theory of Finance New York: Holt, Rinehart and Winston.

Geyskens, Inge, Katrijn Gielens, and Marnik G. Dekimpe (2002), "The Market Valuation of Internet Channel Additions," Journal of Marketing, 66 (April), 102-119.

Henke, Lucy L. (1995), "A Longitudinal Analysis of the Ad Agency-Client Relationship: Predictors of an Agency Switch," Journal of Advertising Research, 35 (March), 24-30.

Hozier, George C., and John Schatzberg (2000), "Advertising Agency Terminations and Reviews: Stock Returns and Firm Performance," Journal of Business Research, 50 (November), 169-176.

Jensen, Michael C., and Richard S. Ruback (1983), "The Market for Corporate Control: The Scientific Evidence," Journal of Financial Economics, 11 (April), 5-50.

Koku, Paul Sergius (1997), "Corporate Name Change Signaling in the Services Industry," Journal of Services Marketing, 11 (December), 392-402.

Lane, Vicki, and Robert Jacobson (1995), "Stock Market Reactions to Brand Extension Announcements: The Effects of Brand Attitude and Familiarity," Journal of Marketing, 59 (January), 63-72.

Lang, Larry H. P., Annette Poulsen, and Rene M. Stulz (1995), "Asset Sales, Firm Performance, and the Agency Costs of Managerial Discretion," Journal of Financial Economics, 37 (January), 3-37.

Lipman, Joanne (1991),"Blame-the-Messenger Mentality Leaves Scars on Madison Avenue," Wall Street Journal (November 20), B4.

Madhavan, Ravindranath, and John E. Prescott (1995), "Market Value Impact of Joint Ventures: The Effect of Industry Information-Processing Load," Academy of Management Journal, 38 (June), 900-915.

Mathur, Lynette Knowles (1996), "Is Value Associated with Initiating New Advertising Agency-Client Relations," Journal of Advertising, 25 (Fall), 1-12.

--, and Ike Mathur (1995), "The Effect of Advertising Slogan Changes on the Market Values of Firms," Journal of Advertising Research, 35 (January/February), 59-65

--, Ike Mathur, and Nanda Rangan (1997), "The Wealth Effects Associated with a Celebrity Endorser: The Michael Jordan Phenomenon," Journal of Services Marketing, 37 (May/ June), 67-73.

McConnell, John J., and Chris J. Muscarella (1985), "Corporate Capital Expenditure Decisions and the Market Value of the Firm," Journal of Financial Economics, 14 (September), 399-422.

Michell, Paul C., Harold Cataquet, and Stephen Hague (1992), "Establishing the Causes of Disaffection in Agency-Client Relations," Journal of Advertising Research, 32 (March/April), 41-48.

Prager, Robin A. (1989), "Using Stock Price Data to Measure the Effects of Regulation: The Interstate Commerce Act and the Railroad Industry," Rand Journal of Economics, 20 (Summer), 280-290.

Santomero, Anthony M. (1991), "Money Supply Announcements: A Retrospective," Journal of Economics and Business, 43 (February), 1-23.

Smith, Clifford W., Jr. (1986), "Investment Banking and the Capital Acquisition Process," Journal of Financial Economics, 15 (January/February), 3-29.

Subramani, Mani, and Eric Walden (2001), "The Impact of E-Commerce Announcements on the Market Value of Firms," Information Systems Research, 12 (June), 135-154.

Mukund S. Kulkarni (Ph.D., University of Kentucky) is an associate professor of finance, Pennsylvania State University-Harrisburg, School of Business Administration.

Premal P. Vora (Ph.D., Pennsylvania State University) is an assistant professor of finance, Pennsylvania Stare University-Harrisburg, School of Business Administration.

Terence A. Brown (D.B.A., University of Maryland) is an associate professor of transportation and marketing, Pennsylvania State University-Harrisburg, School of Business Administration.