Stock Prices and Resignation of Members of the Board: The Case of the Warsaw Stock Exchange Henryk Gurgul Pawei Majdosz In this paper we provide an empirical analysis of announcements of res- ignation of board members using data which comes from the Warsaw Stock Exchange. The market reaction to this information is tested at dif- ferent time horizons by means of event study methodology. The results show that market reaction is rather positive immediately before the announcement release and negative over the following six-day-period starting on the event day. A possible explanation for this phenomenon is suggested. Besides the traditional examination of abnormal return behaviour, we also check whether or not resignation announcements induce increases in the variance of stock returns over the period un- der consideration. It turns out that a tendency towards increased stock return volatility can be observed in the whole period prior to the an- nouncement release. Key Words: managerial resignations, abnormal returns, event-induced variance, emerging stock market jEL Classification: G14; C22 Introduction Stock price reactions to announcements of managerial resignations have been investigated by many researchers. Part of this research focuses on forced resignations. Forced resignations are relatively rare and are due more often to external factors like blockholder pressure or takeover at- tempts, than to normal internal monitoring. According to economic the- ory internal control mechanisms are effective if there are more changes of top management in poorly performing firms than in firms whose perfor- mance is good. Moreover improvements can be observed in firms' per- formance after top management changes. In general, identifying forced Dr Henryk Gurgul is a Professor at the Department of Applied Mathematics, University of Science and Technology, Poland. Pawel Majdosz is an Assistant at the Department of Quantitative Methods in Economics, School of Economics and Computer Science, Poland. Managing Global Transitions 5 (2): 179-192 departures is difficult, because press reports do not describe them as such. Sometimes e. g. a departure announced as a retirement may be in reality a forced resignation. However, if a newspaper release states that a resignation is forced, or that it results from the poor performance of a company, a researcher can take it for granted that the change is really forced. In order to build a data set of forced departures, one has to identify the properties of forced resignations. Then resignations which share these properties can be clas- sified as forced resignations even if those are not announced as such. The interpretation of event study effects of a resignation is not easy; a man- agement change may signal different things: that a firm's performance is worse than expected, or that a firm's performance will improve as a re- sult of the management change but also that the firm is considered as a takeover target. In addition, top management changes can be probably partially anticipated by taking into account poor performance before the change. Based on the forced managerial turnover data from the us stock mar- ket, Furtado and Rozeff (1987) found increases in stock prices due to the event, but from a statistical point of view this result was insignificant. Unlike Furtado and Rozeff (1987), Worell, Davidson, and Glascock (1993) documented a statistically significant price increase of 2.3%. A very in- teresting work is that of Weisbach (1988), who reported that, on the one hand, there is no price impact if the managerial resignation takes place in a company whose board is dominated by executive directors. On the other hand, there is a significant positive stock price reaction if the ma- jority of the board consists of external, independent directors. Khanna and Poulsen (1995) examine whether management turnover leads to improvement in firms' performance. They argue that remov- ing poorly performing managers is an important step toward maximiz- ing shareholder wealth. A management board must identify poor man- agement and attract superior replacement managers. This is the main criterion of the effectiveness of internal monitoring. However a neg- ative correlation between prior stock price trends and managements turnover may coexist with effective internal board monitoring. Khanna and Poulsen supply two alternative explanations. The first one is that managers of poorly performing companies may voluntarily resign in order to avoid shareholder lawsuits. The second one is that company boards may replace the managers of poorly performing firms even if those managers are not responsible for the bad financial situation of a company. Under neither of these two scenarios would a change in management necessarily be expected to induce improvements in perfor- mance. In contrast to the above-mentioned results, Warner, Watts, and Wruck (1988) provide empirical evidence of negative market reaction to forced managerial resignations. A possible reason for this is, according to Warner, Watts, and Wruck, the fact that the announcement of a forced resignation is interpreted as a signal of worse current and future firm's performance. This finding was later confirmed by Mahajan and Lum- mer (1993), who also documented a significant negative reaction over a two-day-period, starting one day before the announcement release. The second very important topic of research is non-conflictual resig- nations and their impact on stock prices (see e. g. Mahajan and Lummer 1993). The conclusion that can be drawn from these studies is that the announcements of resignations on a non-conflictual basis are accompa- nied by a decline in stock prices. This means that such announcements are interpreted by market participants as a loss of valuable human cap- ital by a firm. Resignations due to the retirement by managers need to be analysed separately from other non-conflictual resignations. The lat- ter can usually be well anticipated, and as a consequence, no stock price reaction should be observed. The empirical work of Weisbach (1988) and Mahajan and Lummer (1993) provides support for this statement. In ad- dition, forced resignations and normal retirements also exhibit a signifi- cant amount of post turnover corporate asset-restructuring sales, layoffs, cost-cutting measures and so on. In this paper we provide an empirical analysis of announcements about the resignation of board members which took place in the compa- nies listed on the primary market of the Warsaw Stock Exchange (wse). Poland is a representative case for event study in an emerging stock mar- ket due to the Polish experience in the establishment and development of a stock market. The stock market in Poland did not exist, practically, un- til the beginning of 1990s. The wse, the only stock exchange in Poland, became operational in April 1991. Despite the fact that the period under consideration is relatively short and comprises only five years, our results reveal a statistically significant stock price reaction to announcements of resignations. To be more precise, market reaction is rather positive im- mediately before the announcement release, and negative over the fol- lowing six-day-period, starting on the event day. Besides a traditional examination of abnormal return behaviour, we also check whether or not resignation announcements induce increases in the variance of stock returns over the period under consideration. It turns out that an increas- ing tendency towards stock return volatility can be observed during the period prior to an announcement release. The rest of the paper proceeds as follows. The second section outlines the methodology that aims at uncovering the anomalous behaviour of stock prices induced by an event. The third contains a brief description of the data and the rules underlying sample selection procedure. In the fourth section we start with some basic descriptive analysis of the ab- normal return series, and then the test results for the significance of an event effect over the period under consideration are presented. The last Section provides a summary of the main findings, comments and some guidelines for future research. Methodology Over thirty years ago Fama, Fisher, Jensen, and Roll (1969) introduced event study methodology which still seems an unbeatable tool for un- covering stock price as well as trading volume reactions to the arrival of new information. Obviously, the methods that are used now under event study differ from those of Fama et al. but the main idea has remained the same over the whole period since this methodology was introduced. Let n be the set of day indices t belonging to an event window, and O be a set of day indices t which are attributed to a pre-event or observation window. As a first step, stock prices (Pt) are transformed into returns (Rt) by means of a discreet or continuous formula. The latter, which is given by is especially popular due to the well-known fact that return series (1) is better approximated by normal. In addition, the use of a continuous formula usually improves the stationarity properties of the return series (stabilizing the stock return variance with respect to time). As a second step, the abnormal return series (ARt) is obtained by sub- tracting the actual return from the expected return Note that the expected return in (2) is conditional on the returns ob- served over the pre-event window. The most popular model for gener- ating expected returns is the market model (mm) introduced by Sharpe Rt = log , Vf e n U n, (1) ARt = Rt - E[RtjRkeol, Vt e n. (2) (1963). This model shows the expected return as the sum of two com- ponents. The former is a constant (a). The latter, is a product of the systematic-risk parameter and the market-portfolio return (ßRm). With mm serving as an expected return model, (2) may be rewritten as ARt = Rt - a - ßRm,t, Vt e n, (3) where a and ß mean the estimators of the corresponding model param- eters applied over the pre-event window. In order to check whether the average abnormal return on a given day t e n statistically differs from zero the t-statistic is employed, which is given by N-1! N=1 AR1>t tstat — CAAR (4) where N stands for the number of firms included in the sample and the denominator (the standard deviation of the average abnormal returns) can be calculated as follows C AAR = N -1 1 ( N 1 N ä-TZ n 1 ten V i—1 n teA i—1 (5) where A means a cardinal number of set A. With the widely-documented fact that financial time series exhibit heteroscedasticity of variance, the use of a market model as in (3) does not seem to be fully justified. The estimator of the standard deviation of the average abnormal returns defined by (5) is not able to capture vari- ance changes which may occur over the event window. As a consequence, the value of statistic (4) is no longer sufficient for the purpose of infer- ence needs. To relax the assumption that stock return variance remains the same on each day of the event window, while improving the statistical infer- ence used under event study, has resulted in the development of several helpful techniques. One of them is that of Hilliard and Savickas (2000). An original test for abnormal performance is proposed by Hilliard and Savickas with the market model and the garch(i,i) error term. Under this study we, however, decided to use the generalized ARMA(r,m)-MM- garch(p,q) model given by Ri,t — Otf + °',)Rt-j + ß'Rm,t + Si,t + 2 0ySi,t-i, j=1 j=1 Si,t - (0, hit). 2 2 184 Henryk Gurgul and Pawel Majdosz q p hi,t = aifi + ^ aijs2t-j + ^ Mijht-j. (6) j=i j=i The proper length of time-lags in the model is identified using the Akaike Information Criterion. The model parameters are estimated by means of the ML-method from observations included within the pre- event window, i. e. for t e O. The test statistic (lt) can be expressed as h= . ASRt =(N- 1), (7) yllNL{SRi,t - ASRt)2 where SR1>t = AR1>t and ASRt = N'1 ^ SR1>t. In order to test the implications of announcements over any sub- period of the event window whose boundaries are set as m and s (m < s), the standardized cumulative abnormal return can be calculated SCARi (8) s ■\Jlit=m hi,t The corresponding test statistics are given by %?=ASCARmsJ—-N(N 1}-, (9) y T*f=i(SCARi>m>s - ASCARm>s)2 where ASCARm>s = N-1 E N=1 SCARhm>s. With the help of the methodology proposed by Hilliard and Savickas (2002) we are also able to study the event effect on the unsystematic volatility of stock returns. The multiplicative abnormal volatility param- eter (i), introduced by the above-mentioned authors, measures the scale of the increase in unsystematic volatility, caused by an event. This pa- rameter is defined as TtN-'in-l^t + N-^tMt Note that if parameter (10) is equal to unity, the event has no impact on unsystematic volatility. A value of the parameter greater than one im- plies a volatility increase due to the event. To test it more formally, one can use a statistic expressed as St = (N - 1)it, (11) which is a chi-squared distribution with N - 1 degree of freedom. , \—l V"vi,t Z-ik=1 k,t I At = (N-iy1) —i-j-———. (10) ^- 0). The shape of the line representing cumulative abnormal return fully supports this finding. As regards the conditional volatility of stock returns, one can 1 0 0 3 5 4 table 2 Test results for event effect within sub-periods of the event window Time interval {m,s} {-5,-1} {-3,-1} {-2,-1} {0,+2} {0,+3} {o,+5} {-5,+5} ASCARm,s 0.033 0.729 1.564* -1.781 -3.126** -3.229** -3.196 C 0.017 0.602 1.932 -I.598 -2.347 -2.093 -1.170 notes ** Statistically significant value at 5% level, * stat. significant value at 10% level. notice a slow rise in variance in the second half of the event window. This observation, however, is not equivalent to saying that an event- induced shift in variance can be identified. We address this issue in the next section. Table 2 reports the test results for the event effect in the seven differ- ent periods of the even window. Statistically significant values of average standardized cumulative abnormal returns (ascar) can be found in the case of three sub-periods of the event window, including the period from day t - 2 to day t - 1, from the event day to day t + 3, and finally from the event day to the last day within the window. In the whole event window the ascar is negative (-3.196), but its value does not differ from zero from a statistical point of view. What can be concluded from the figures in table 2? Firstly, the negative valuation effect in the second half of the event window clearly indicates that companies, on average, lost valuable human capital because of the resignation of board members. This finding is consistent with our in- tuition and corroborates other empirical evidence, e. g. that of Mahajan and Lummer (1993). Secondly, the positive valuation effect before the official announcement of managerial resignations may be, on the other hand, interpreted as evidence that a resignation results in conflict reduc- tion inside the company, as a consequence of which stock prices start to increase. One possible explanation for this phenomenon is that prior to an of- ficial announcement about a resignation there is trading by insiders. In- siders are well informed and probably know the true circumstances of a resignation decision. They may be convinced that the resignation of a given person will ultimately lead to improved performance. After the of- ficial announcement other investors start to trade. They are not as well informed as insiders and have to guess the true reasons for a resignation. From their view-point a resignation means a loss of the firm's human capital. table 3 Test results for cumulative abnormal volatility within sub-periods of the event window {-5,-1} {-3,-1} Time interval |m,s| {-2,-1} {0,+2} {0,+3} {o,+5} {-5,+5} %t=m A t CSm,sx 13.485* 68.776 5.229* 26.670 2.934* 14.964 0.931 0.968 4-779 4-934 0.977 4-983 14.462* 73-758 notes * Statistically significant value at 10% level. testing for the event effect on the unsystematic volatility Finally, for the same periods of the event window as previously, we cal- culate the multiplicative abnormal volatility parameter (10) and the cor- responding test statistic (12). The results are summarized in table 3. We found an increasing tendency towards volatility in the cumulative abnormal returns over the first half of the event window (i. e. for t < 0). In the second half of the window (i. e. for t > 0) volatility is not changed. This can be seen as evidence that before the information about a resig- nation becomes public the market reacts more nervously. The volatility increase here may be a result of uncertainty about the possible resigna- tion. Conclusions The purpose of this paper is to analyse whether the announcement of res- ignations of board members conveys valuable information in an emerg- ing stock market like the Polish one. Using a variant of event study methodology, we provide empirical evidence supporting the hypothesis of market reaction to managerial resignations. Before the announcement release there is a tendency towards an increase in stock prices. When the firm announces the resignation of members of the board, this tendency is reversed, and stock prices start to fall. In order to explain this phenomenon we have referred to differences in the interpretation of a resignation announcement by insiders and other investors operating on the wse. Insiders, who know the true motives behind a resignation decision, are prone to buy shares. It may be they expect that a resignation, by reducing conflict and/or improving man- agement, will result in better firm performance. With the limited infor- mation in an official announcement, other market participants have to guess what the resignation means for the current and future position of Stock Prices and Resignation of Members of the Board 191 table 4 Companies included in the sample and the number of identified events Name of company n Name of company n 4media 2 Naftobudowa 3 7bulls.com 1 Netia 1 Agora 1 Optimus 1 Agros 1 Orfe 2 Amica Wronki 1 Pekao 2 Apexim 1 Pfleiderer Grajewo 1 Bank Millennium 3 Pollena Ewa 1 Bre Bank 2 ppwk 1 Centrozap 2 Projprzem 1 Comarch 2 Prokom Software 3 Elektrim 1 Redan 1 Energomontaz-P. 1 Softbank 2 Espebepe 1 Ster-Projekt 3 Ferrum 1 Szeptel 1 Fortis Bank Polska 1 tim 1 Getin 1 Tonsil 1 Impel 1 Tras Tychy 1 Interia.pl 1 Wolczanka 2 Kruszwica 1 zm Duda 1 Leta 1 zp c Mieszko 1 Mostostal Zabrze 2 zpue 1 Mostostal-Export 1 notes n - the number of events. a firm. As our results show, resignations are regarded as a loss of valu- able human capital. Hence, stock prices tend to go down over the period following the announcement. It would be very interesting to check whether stock prices react dif- ferently to forced and non-conflictual resignations (e. g. normal retire- ment). It would be also interesting to find out the relative importance of different factors which cause forced resignations, such as blockholder pressure, takeover attempts, financial distress, shareholder lawsuits or normal board monitoring. With the meagre sample, we cannot do so under this study. Therefore, we must leave this problem for future re- search. Acknowledgments The authors thank Anna Gruszka for helping to complete the event database. References Fama, E. F., L. Fisher, M. Jensen M., and R. Roll. 1969. The adjustment of stock prices to new information. International Economic Review 10 (1): 1-21. Furtado, E. P. H., and M. S. Rozeff. 1987. The wealth effects of company initiated management changes. Journal of Financial Economics 18 (1): 147-160. Hilliard, J., and R. Savickas. 2000. On stochastic volatility and more pow- erful parametric tests of event effects on unsystematic returns. Work- ing paper, University of Georgia. -. 2002. On the statistical significance of event effects on unsystem- atic volatility. Journal of Financial Research 25 (4): 447-62. Khanna, N., and A. Poulsen. 1995. Managers of financially distressed firms: Villains or scapegoats? The Journal of Finance 50 (3): 919-40. Mahajan, A., and S. Lummer. 1993. Shareholder wealth effects of manage- ment changes. Journal of Business, Finance and Accounting 20 (3)393- 410. Sharpe, W. 1963. A simplified model for portfolio analysis. Management Science 9 (2): 277-93. Warner, J., R. Watts, and K. Wruck. 1988. Stock prices and top manage- ment changes. Journal ofFinancial Economics 20:461-92. Weisbach, M. 1988. Outside directors and ceo turnover. Journal of Finan- cial Economics 20:431-60. Worrell, D., W. Davidson, and J. Glascock. 1993. Stockholder reactions to departures and appointments of key executives attributable to firings. Academy of Management Journal 36 (2): 387-401.