Intra-Industry Contagion Effects of Layoff Announcements

Transcription

1 Intra-Industry Contagion Effects of Layoff Announcements Adam Bordeman Leeds School of Business University of Colorado Bharadwaj Kannan y Leeds School of Business University of Colorado Roberto Pinheiro z Leeds School of Business University of Colorado Abstract Using a database of 1,007 permanent mass layo announcements by US S&P 500 public rms between 1980 and 2010, we show that the market reaction to layo announcements depend on the industrial structure of the sector in which the layo occurs. The market reaction towards announcers is more negative when the announcing rm faces stronger competitors in the sector, measured in terms of sales growth and positive Returns on Asset. Moreover, we nd that the impact of the change in sector structure has become relatively stronger overtime, as indicated by an increase in the dispersion of the market reaction towards competitors through time. In addition, we also show that an overall industry e ect is still present wherein there exists a positive correlation between the direction of market reaction towards announcers and competitors. Adam.Bordeman@Colorado.EDU, address: 995 Regent Dr Boulder, CO y Bharadwaj.Kannan@Colorado.EDU, address: 995 Regent Dr Boulder, CO z Roberto.Pinheiro@colorado.edu, address: 995 Regent Dr Boulder, CO

2 1 Introduction In a market, bad news about a competitor may have an ambiguous impact on a rm s value. On one hand, the information release may signal a weakening position of the competitor s ability to attend demand. This would be seen as good news for other rms in the industry, generating a positive impact on their stock prices. The magnitude of this impact should depend on the rm s ability to take advantage of such opportunities. On the other hand, bad news for a competitor may present itself as a cautionary tale about the downturn in that particular industry, which would sooner or later a ect all companies in that sector. Even more, how the market interprets these signs will depend on the competitive structure of the sectors. For an announcing rm that is an unopposed leader in its sector, the news could signal a long due restructure that will ultimately increase the rm s pro tability. Usually, weak competitors are not in a position to take advantage of these temporary decreases in the announcer s supply, and therefore, the ultimate market reaction is negative for competitors. Di erently, for a rm with strong competitors, the short term reduction in supply will be a golden opportunity to steal hard fought market share. These interactions are so important that, even if the information release is seen as bad news for the industry overall, the net impact for any particular rm may be positive or negative depending on how the competitive structure within the industry shifts. In this paper, we try to disentangle these impacts for one important episode: permanent mass layo announcements by competitors. Layo s are particularly hard to interpret. They could signal either good or bad rm-speci c news. A company making the announcement of a large layo may be in nancial trouble, due to poor managerial decisions. Di erently, a mass layo may indicate that the company is becoming more e cient, by restructuring the productive process, closing older and less e cient plants, while shifting production to more e cient ones. Similarly, a mass layo may also say something about the industry itself, indicating a reduction in the overall demand for its products or potentially a breakthrough technological change in productive processes, that creates growth opportunities for all companies in the sector. Therefore, a layo decision can be associated with either an increase or a decline in expected future pro ts and rm values for both the laying o company as well as its competitors. Given this ambiguity about the layo s impact on rm performance, it is not surprising that the literature has found mixed results about the impacts of layo decisions on a rm s stock price performance. Farber and Hallock (2009) nd that the stock price reaction of announcing rms to layo decisions has markedly changed over time, being quite negative in earlier decades and becoming negligible in the nineties. These authors suggest that modern era layo decisions 2

3 are driven by e ciency improvement, which has a positive e ect on market price for the announcing rm. Di erently, Palmon, Sun, and Tang (1997) showed that the positive or negative impact of layo announcements depend on the reasons cited by the layo rm at the moment of announcement. Layo s have a positive impact if the announcement cites e ciency-enhancing measures, while the impact is negative when the given reason is declining demand. However, the same reasons where found having no impact by Hallock (2005). Similarly, while the usual stylized fact is that rms laying o are in dire straits, Chen, Mehrotra, Sivakumar, and Yu (2001) nd that announcing rms have better operating performance than their peers, although their performance has been in decline for the three years prior to the announcement. They also nd that the announcer s performance improves during the 3 year-window after the announcement, a nding that is in odds with Guthrie and Datta (2008) s results. Di erently, the literature on the impact of mass layo on competitors is almost nonexistent. Up to this point, studies by Madura et al. (1995), Goins and Gruca (2008), and Bhabra et al. (2011) have examined the contagion e ect of layo s. The rst two studies examine intraindustry information transfers within speci c industry; neither of these studies makes broad, cross-sectional predictions about the information transfer of layo announcements. Bhabra et al. (2011) attempts to study layo contagion e ects across all market sectors, for a sample of 403 mass layo announcements from 1984 to Our study di ers from this study signi cantly, not only because we have a larger and more in depth sample of mass layo announcements, but also because we are able to control for announcer, sector, and competitor characteristics. This allows us to avoid some misguided conclusions that they obtained, in particular about the contagion of mass layo announcements, which we will discuss in more depth in the next section. Our paper seeks to bridge this gap. We study a sample of 1,007 layo announcements by rms listed in the S&P 500, at some point between 1980 and 2010 and measure the contagion e ect of the intra-industry information transfer between the announcing rms and non-announcing rms using the cumulative abnormal return (CAR) through an event study. Di erently from the previous literature, we control for measures of nancial health of the displacing rm as well as the characteristics of its sector peers, thereby capturing not only how unexpected the layo was, but also how the announcement changes the competitive pressures in the industry. In particular, does the market react to layo s of better run companies as an indication of e ciency improvements, rm speci c idiosyncratic negative shocks, or bad news for the industry overall? Even more, how is this reaction modulated by the competitive environment in the industry? Once we look at the impact on both the announcing rms and its competitors stock returns, we are able to disentangle these di erent impacts. 3

4 Our results highlight the importance of the sector s industrial organization. Even though we nd, similarly to Chen et. al. (2001), that announcing rms have a better operating performance than their peers on average, the market generally sees a mass layo as bad news. This negative e ect is particularly strong when the rm s industry peers are in a upward trend, measured by the fraction of competitors that observe sales growth and positive return on assets (ROA). Similarly, the probability that an industry peer s layo will be seen as good news for a peer - measured by a positive cumulative abnormal return (CAR) - depends on the rm s ability to seize the opportunity, i.e., by its operational conditions and ability to react to the news of other rms in the sector. In particular, in order to take advantage of potential bene ts of a mass layo, a competitor rm needs to be able to expand its operations by having enough capacity to increase production fast. In order to test the importance of these assertions, we look at the impact of potential operational and nancial constraints on the competitor s cumulative abnormal returns 1. We nd that the degree to which the market expects a competitor to bene t from the mass layo depends on the competitor s characteristics. Larger companies - measure in total lagged sales - and value companies - measured by a high market to book value - tend to bene t more from a rival s mass layo. More liquid companies seem to bene t less from a competitor s layo. These results indicate that the rms that bene t the most of a competitor s layo are the ones that are in a position to occupy the temporary supply void generated by the competitor s mass layo. As presented by Chen et. al. (2001), most layo rms go back to their original employment level within 3 years. So, if most of the rm s value comes from long-term growth opportunities and/or the rm has to initially invest in expanding capacity before being able to seize the opportunity created by the mass layo, such opportunity is most likely missed. However, the overall impact depends on the characteristics of the sector to which the announcer and competitors belong. The presence and quality of other rms in the sector, measured by the fraction of rms in the sector with positive sales growth, and the number of S&P500 competitors with positive return on assets, reduces the positive market reaction for the average competitor, particularly for the rst mass layo announcement for a given announcing rm within a year. Announcers characteristics, specially their overall performance measured by an index that aggregates nancial ratios indicate that the overall market reaction depends on how much the market expects the layo to a ect a lasting change for the announcer. This interpretation is corroborated by the negative coe cient of the interaction of the weak announcer dummy and 1 We look at the impact of 3 di erent indexes of nancial constraints: the Kaplan and Zingales (KZ) index, suggested by Lamont et. al. (2001), the Whited Wu (WW) index (Whited and Wu, 2006) and the Size and Age (SA) index (Hadlock and Pierce, 2010). Since results are similar for these measures, we focus on the SA index. 4

5 the competitor s leverage. This indicates that competitors that may have di culties raising additional capital may not have enough resources to increase productive capacity in situations in where announcers may be slower to recover. However, leverage by itself has a statistically insigni cant impact on the market reaction, given the fact that the window of opportunity usually closes fairly quickly when the announcer is not weak relative to the industry. In terms of the impacts of the layo characteristics, we conclude that larger layo s, as well as layo s of more skilled workers, using salaried worker layo s as a proxy, have a more negative impact on the announcer s stock price, in particular for announcers that are nancially constrained. In this sense, our results corroborate the results obtained by Milanez (2013) for a sample of California rms in the 2000s. Although these characteristics seem to also matter for competitors, they lose a lot of their statistical signi cance, especially for rst announcements. In contrast with prior literature, we nd that announced layo reasons have no statistically signi cant impact on the returns of announcers or competitors. Our analysis of the relationship between announced reasons and announcer characteristics based on a multinomial logit nds no clear relationship between the two, indicating a potential cheap talk by announcers, especially given that these reasons seem to correlate over time, showing some fad component. Finally, our analysis also indicates an industry-wide e ect, once there is a clear positive correlation between announcer s and competitors CAR, even after controlling for announcer, competitors, and sector characteristics. However, the impact of changes in the sector competitive structure originated by the layo seems to have become relatively more important over time, which is captured by the increase in the dispersion of reactions across competitors from the 1980s to the 2000s. In the next section, we discuss the literature on layo s announcements and market reactions. In section 3 we describe the data selection. Section 4 describes the methodology. Section 5 presents our results. Finally, section 6 concludes the paper. 2 Related Literature The literature on the impacts of mass layo s has focused mainly on the magnitude of layo announcements on the announcer s stock return. Generally, most of the earlier literature (see Worrell, Davidson, and Sharma (1991), Abowd, Milkovich, and Hannon (1990), among others) has found a negative impact of mass layo announcements to stock returns. However, the more recent work by Farber and Hallock (2009) showed that the e ect of layo s announcements on 5

6 stock returns has varied over time, being extremely negative during the 1970s while becoming close to insigni cant in later periods. This pattern seems to be partially reverted from 2000 on. Hallock, Strain, and Webber (2011) extended the database used by Farber and Hallock (2009) until They showed that stock price reaction to job loss announcements were less negative in the 1980s and 1990s compared to the 1970s, but the 2000s were not statistically di erent from the 1970s. Although the authors emphasized the reduction in the market s reactions to mass layo s, it seems that there is a clear indication of the reversal of this pattern in the last decade. Another topic that has been discussed at length in the literature is the impact of the announced reasons of the layo s. The SEC only requires disclosure of layo s in relation to restructuring behavior through an 8k ling (SEC 2012) so most disclosures of layo s in the popular press are discretionary, which means that the story spun by the rm may also be discretionary. According to Palmon, Sun, and Tang (1997), the negative impact of stock returns from mass layo announcements comes from announcements in which the rm cites adverse market conditions as the main reason for the layo. Layo s in which the announced reason is related to e ciency improvements actually presented positive abnormal returns. As there is no cost but reputational costs for misleading announcements, it seems reasonable to ask if the announced reasons are actually related to future performance. Palmon et. al. provide an analysis that corroborate the relationship between the announced layo reason and measures of future pro tability and sales. Similarly, Farber and Hallock (2009) nd that the market reacts less negatively to mass layo s in which the announced reason is other than demand slump. However, in interview data with senior investors and managers, Hallock (2003) shows that some investors seem quite skeptical about the actual validity of these announced reasons. In terms of the time of the announcement, Aggarwal and Kolev (2012) show that public companies tend to cluster their mass layo announcements at the time that other large public companies announce their mass layo s, particularly during recessions. This clustering behavior is particularly important for rms with new or short tenure CEOs. As a result, rms that follow the layo of another Fortune 500 company see a less negative CAR, while the rst mover in the mass layo wave sees a small positive CAR once the followers start announcing their layo s. Similarly, Farber and Hallock (2009) discuss the possibility that layo announcements occur systematically around the time of other announcements that also a ect stock price. In order to isolate the e ect of mass layo announcements, they generate a sub-sample of clean announcements, i.e., announcements in temporal isolation from other important announcements 2. They showed that, 2 They consider layo announcements that were not within 10 days of the rm s dividend or earnings announcements, as well as not within 100 days of layo announcements by the same rm. 6

7 although the size of the market reaction to the news was bigger with the "clean" data, the results were qualitatively similar. The literature has also examined the characteristics of the mass layo s. Using data from the WARN act for California, Milanez (2013) shows that nancially constrained rms tend to announce larger mass layo s and they tend to layo more workers with more human capital, onthe-job training, and experience then unconstrained rms. She also shows that the rm reaction is worse for nancially constrained rms. Similarly, Farber and Hallock (2009) show that the market reaction to a layo is higher when salaried workers are dismissed and when the mass layo represents a larger fraction of the rm s total labor force. Finally, the literature that looks at the impact of mass layo s on the announcing rm s competitors is quite small. Studying a sample of 403 layo announcements in the period of 1984 to 1994, Bhabra, Bhabra, and Boyle (2011) showed evidence of intra-industry contagion as well as competitive e ects. They nd a positive and small e ect of layo announcements on competitors, but they show that this is in fact an indication of the dispersion of e ects across di erent competitors, given their characteristics. Even though our paper is related to theirs, their analysis is somewhat incomplete. Not only do they have a reduced sample, but they also do not control for most of the sector characteristics and interactions with the announcers and competitors characteristics, which induces their analysis to reach potentially awed conclusions. For example, their analysis of the contagion hypothesis focusing on the Tobin q seems misleading. First, because mass layo s are a short term phenomena, since most rms announcing a layo return to their original employment levels during the next 3 years. Second, because the negative impact of Tobin s q is present even after we control for the sign of the announcers CAR. This indicates that the result of a larger Tobin s q not only goes beyond the industry-wide contagion e ect that they propose, but also it is mostly circumscribed to layo s that are expected to last longer than the short-term restructures. 3 Data and Sample Selection 3.1 Sample Construction Our sample consists of layo announcements between 1980 and 2010 for all rms ever to be listed in the S&P 500 during that time period; 1,269 unique rms were listed in the S&P 500 in those years. We use Factiva to search for mass layo announcements in the Wall Street Journal for these sample rms. We focus on Wall Street Journal announcements since we believe that 7

8 any news relating to S&P 500 rms are reported in the Wall Street Journal. Further, since we are only concerned with e ect of layo announcements on competitors, we are not interested in unannounced layo s. We search Factiva for keywords including layo, layo s, laid o, restructure, restructuring, and downsize. Of our 1,269 rms, 294 unique rms had at least one layo announcement for a total sample of 1,027 layo announcements. From each of the Wall Street Journal layo announcements, we collect information on the date of the layo announcement, the size of the layo, the reason given for the layo, and the type of worker laid o. We next gather rm speci c nancial data from Compustat. We lag nancial data so that we can control for the nancial position of our rms prior to the layo announcement. Then, the announcers and non-announcers are grouped by 3-digit SIC codes. Within each industry and year, we construct an industry average for our nancial variables and calculate the di erence between the industry averages and the values at the individual observation level. This is done to measure the relative position of each rm within the industry. Within a given year, a competitor rm could be a competitor rm for multiple announcers if multiple rms within the same 3-digit SIC code announced layo s. Also, an announcing rm can also be a competitor rm for another announcer within the same 3-digit SIC code. We next merge our sample with Eventus to get our cumulative abnormal returns data. Our sample size is now 81,492 observations with 1,007 layo announcements. We focus on short-run event windows of 3 days, 5 days, and 11 days around the layo event date. Since results are qualitatively similar, we focus on the 3 day window in our analysis below. Following similar research by Farber and Hallock (2009), we use market-adjusted returns to calculate cumulative abnormal returns. We obtain actual and forecasted GDP data to control for macroeconomic changes that could impact all rms within a sector. The change in GDP in current year dollars was obtained from the Bureau of Economic Analysis. Finally, to get earnings forecast error data, we collect analyst forecasts from I/B/E/S. To minimize any confounding e ects of I/B/E/S split adjusting (Payne and Thomas 2003), we used the unadjusted detail le within the database. We construct the forecast error for the consensus analyst forecast for the quarter prior to the layo announcement and create an indicator variable equal to one if the rm had a positive forecast error; rms with a negative forecast error might be looking for ways to improve nancial performance. The nal sample of rms with su cient data to run the regressions is 1,007 layo announcing rms and 62,347 competitor rms when using all layo announcements. The sample is reduced to 673 announcers and 43,960 non-announcing competitors when we limit our layo announcements 8

9 to only those that have at least a one year gap between same- rm layo s. 4 Methodology In order to study the market s response to the announcement of a layo, we employ an event study methodology. This is consistent with much of the prior literature on layo announcements (i.e. Farber and Hallock 2009) and intra-industry information transfers (Madura et al. 1995, Goins and Gruca 2008, Bhabra et al. 2011). We speci cally focus on 3-day cumulative abnormal returns around the layo announcement date, as reported by the Wall Street Journal. Farber and Hallock (2009) also use Wall Street Journal announcements as the event date and acknowledge the possibility of information leakage prior to the WSJ release; if anything, leakage should bias against nding robust results. We use the event study methodology to measure the returns for both rms announcing layo s as well as their non-layo -announcing competitors. Before we begin the event study, we rst take a closer look at underlying rm characteristics to generate prediction models regarding the likelihood of laying o employees as well as the likelihood of the reasons cited for the layo. First, using a probit model, we measure the probability of a S&P 500 rm to announce a layo given the rm s nancial and productive characteristics, conditional on industry xed e ects. Both announcers and competitors are used in this analysis such that the dependent variable is dichotomous with a value of 1 if the rm announced a layo and 0 otherwise. The independent variables in the probit are meant to capture the nancial condition of the rm in the period prior to the layo. In other words, we seek to determine if layo s are a function of poor nancial health or constraints and if future layo announcements can be reasonably predicted. It is clear that the market does not accurately predict layo announcements since we observe a market response around the announcement date; the E cient Market Hypothesis (Fama 1970) suggests that the market would already incorporate the impact of a layo if it were predictable at some earlier point in time. In subsequent analysis we will use this modeled outcome, the probability of announcing a layo, as an explanatory variable in our event study (results are presented in table 2). Prior research has also placed some importance on the reasons cited by rms making layo s (i.e. Farber and Hallock 2009). These studies suggest that the reasons given can have an interpretable meaning about the future nancial prospects of the rm and, thus, the reasons in uence the market response. However, it is also possible that the reason given is cheap talk due to the voluntary nature of most layo announcements. We run a multinomial logit model to 9

10 create a predictive model of cited reasons based on nancial variables related to the announcing rm. If nancial position can predict the layo reason, then one can conclude that the reasons are accurately revealed. If reasons are unpredictable, then the market (and academic studies) should consider them noise. To begin our event study of the competitors, we rst use a probit to determine the likelihood of competitors having a positive or negative reaction to a layo announcement. We create dummy variables to partition the size of the competitors market response and run a probit based on characteristics of the announcer, competitor, layo, industry, and economy. Too often, event studies focusing on intra-industry information transfer lump all competitors into one large pool. However, companies can react di erently to the same piece of news depending on their competitive position. We intend to explore these di erentials and this probit model shows that di erent rm characteristics of non-announcers can lead to di erent market responses to a layo announcement. We then use a quantile regression to compare these nancial characteristics of non-announcers based on their market response. While the rst probits determine if certain rm characteristics predict higher or lower returns, the quantile regressions gives us a way to directly compare sub-samples within our data to determine whether these characteristics are signi cantly di erent for di erent levels of market response (Cameron and Trivedi, 2009). 4.1 Data Summary Announcer s Characteristics Table 1 presents the summary statistics for rms laying o employees and their competitors. The above table is broken down in four parts. Part A of the table displays the summary statistics for all rms laying o employees in comparison to their competitors. Part B describes summary statistics for rst time announcers in comparison to their competitors. The variable rst time announcer is calculated in such a manner that only those layo announcements that are not preceded by another layo announcement within 365 days are included. Finally, parts C and D narrows the sample of competitors to S&P 500 companies The daily cumulative abnormal returns, CAR, around a 3-day event window centered on the layo announcement date, is on average positive for competitors and negative for announcers. Based on the median measures for CAR, we see that announcers continue to do to worse than competitors. However the median CAR for competitors is negative as well. This indicates a large variability of the impact of layo announcements on competitors that will be discussed in detail below. 10

11 We construct Tobin s Q to measure the market to book value relationship for all rms. We see that the mean and median values of Tobin s Q is greater for competitors when compared to rms announcing layo s, which implies that competitors are seen by the market as adding more value to its capital than announcers. This suggests that announcing rms are more in need of a restructure than competitors. We measure rm liquidity using the quick ratio (current assets/ current liabilities). We see that, on average, rm liquidity levels are higher for competitors when compared to announcing rms. Further, leverage levels are higher for announcing rms on average. This may imply that competitors may be in a better position to take advantage of any competitive gains created by a layo. Similarly, these results indicate that layo s may be a source of liquidity for the announcers. However, in terms of nancial constraints, we see that the SA index is lower for announcers implying that they may be less likely to be nancially constrained when compared to their competitors. This implies that announcing rms are larger and older which suggests that they have more access to external sources of funds, for example, external commercial paper markets. Finally, we notice that announcers have higher levels of return on assets and sales growth for the year preceding the announcement. This is in agreement with what was found by Chen et. al. (2001). In this sense, announcers are not rms under distress, but maybe rms with decreasing pro tability and therefore, in need of a restructure. The direction of these sample statistics hold robust even while comparing rst announcers to their competitors in panel B of table 1. Further we also present the summary statistics of all announcers and rst announcers compared to only competitors listed in the S&P 500 index in the appendix. This is done to avoid any bias resulting in table 1 parts A and B from comparing rms of di erent sizes. We nd that the descriptive statistics are similar to what we nd previously for all variables except return on assets. In order to further compare the characteristics of announcers and competitors, we run a probit on the likelihood of a given rm in the S&P500 to become a mass layo announcer, based on the rm characteristics. We control for industry, macroeconomic conditions, and year dummies. The results are presented in table 2. As you can see, announcers are older, with more employees, more pro table, and less liquid than their competitors. Even though from summary statistics, Tobin s Q is higher for competitors and leverage levels higher for announcers, these variables are not signi cant once we take into account other rm, industry, and economy-level variables. 11

12 4.1.2 Announcement Characteristics Layo Reasons Usually layo announcements are accompanied with reasons for the layo s. The literature has ambiguous results about the importance of such released reasons, but few papers related the announcer, competitors, and economy characteristics. In order to take these features into account, we run a multinomial logit, presented in Table 3 3. The baseline reason is demand slump. From the table, we can observe that few variables are signi cantly associated to the reason announced. Firms with more employees are more likely to announce Cost and Pro tability or Other as the reasons for a layo, while Plant Closing and M&A and Restructuring are more likely to be related to high Tobin s Q and nancial constraints, respectively. In terms of the economywide situation, we see that the 1990s are a period M&A and Restructuring are more likely than Demand Slump as an announced reason, while the other reasons don t seem to be signi cantly a ected by the time period. These results, paired with results presented below that indicate minimal impact of the announced layo reasons on the market reaction, corroborate Hallock (2003) s survey results that point to announced reasons having little informational content. Changes in Market Reactions Over Time Hallock (2003) reports that the average CAR for announcing rms became less negative in the 1980s and 1990s compared to the 1970s, to the point that it was not signi cantly di erent from zero. From Graph 1, we can observe that this tendency has stabilized overtime. Even more, there is a di erence in the pattern of changes in the 3-day CARs among rst and follow-up layo announcements. This di erence is even more striking once we compare outlier robust measures, like the median CAR, as shown in graph 2. However, a more important change in pattern is presented by the increase in the standard deviation of CARs, as presented in graphs 3 and 4. This pattern is robust for industry controls. Graphs 5 to 8 present the results for competitors. First of all, notice that the average 3- day CAR evolved from negative to positive over the period. However, more importantly, the median CAR has been negative through the entire period, i.e., for the median competitor rm, a mass layo announcement by a S&P 500 competitor is seen as bad news. However, the dispersion is quite high, as presented in graph 7, at least 45% of the competitors experienced positive 3-day CARs. Similarly to what happened to announcers, we observe in graph 8 that the 3 Multinomial logit includes industry and decade dummies. 12

13 standard deviation of competitors 3-day CAR increased over time. As in the previous case, this increase in variability is not explained by changes in the industry composition among competitors. Similarly, the fraction of competitors with positive CAR is not explained by being an S&P 500 company. In order to control for industry, announcement, and announcement characteristics, we ran a regression on the standard deviation of competitor s CAR per announcement. The results presented on table 4 show that the dummies for the 1990s and 2000s are positive and statistically signi cant, even after controlling for all the mentioned characteristics, which corroborates the idea of competitive e ects becoming more important over time. From the results from tables 5 and 6, we can see that the probability that a competitor has large positive or negative 3-day CAR is related not only to the announced layo characteristics, but also competitors characteristics. Competitors with more employees or larger ROA are less likely to have extreme CARs, while less liquid competitors are more likely to have large positive CARs, and competitors with large Tobin s Q are more likely to have large negative CARs. In terms of the characteristics of the layo announcement, mass layo s that represent a larger fraction of the labor force increase the probability that competitors will experience a large positive CAR, while announcer s with higher SA index increase the chance that competitors will experience large negative CARs. As expected from our previous results, dummies for the 1990s and 2000s increase the probability that competitors experience extreme CARs, which further corroborates the results that the standard deviation of market reactions went up through time. 5 Main Results 5.1 Announcer s Stock Price Reaction As we can see in table 7, the magnitude of the market reaction to a rm s announcement on its stock price depends not only on the characteristics of the layo and the announcer, but also the sector structure. The higher the fraction of competitors with positive sales growth, the more negative the market reaction to a mass layo announcement. Interestingly, the fact of having more competitors that are also in the S&P 500 and with positive ROA does not signi cantly impact the announcers CAR. In contrast with previous literature, the announced reasons do not seem to have an impact on the market reaction, apart from the M&A and restructuring, which reduces the negative impact of mass layo s. Focusing on announcer characteristics, the fact that a rm announced another mass layo in the past year does not signi cantly changes 13

14 the magnitude of the market reaction. Another interesting feature is that if the announcing rm has more capital than the average competitor, measured by total assets, the market reaction to the layo is more negative. In terms of the quality of the employees laid o, the result corroborates what Milanez (2013) obtained, since the market reacts more negatively to the lay o of salaried workers, which is a proxy for human capital in our sample. Even more, we also nd that rms that are more nancially constrained - higher SA index - have a more negative market reaction to a mass layo announcement, a result also found by Milanez (2013) in a more restricted sample of rms from California. Finally, rms that layo a larger fraction of their labor force also see a more negative reaction from the market, as it may be an impact the rm s ability to operate in the short run. Turning to macroeconomic indicators, we can observe that the average layo during the 1990s and 2000s was accompanied by a more negative reaction by the market, although this does not seem to be related with the state of the overall economy, measured by the GDP. 5.2 Competitors Stock Price Reaction Turning our attention to competitors, the fact that a mass layo announcement is the rst or a follow-up announcement has a clear impact on the market reaction, as is shown in table 8. Because of this, we will present the results for both rst announcements and announcements overall, controlling for the fact that a given announcement is the rst. In table 9, we focus on rst announcements. As we can see, the sector structure matters a lot - the larger the fraction of competitors with positive sales growth, as well as the larger the number of S&P 500 competitors with positive ROA, the lower the stock reaction to a mass layo announcement for the average competitor. In terms of the rms own characteristics, nancially constrained rms and rms in the S&P 500 seem to bene t more, while more liquid rms and rms with larger Tobin s Q seem to bene t less. This is related to the fact that the value of rms with larger Q is more weighted on growth opportunities, so the short-term decreases in competitiveness by a given competitors has a minor positive impact, while the restructure of a major competitor that may lead to a stronger future competitor has a larger negative impact. Similarly, if in order to take advantage of this short term impact a rm needs to invest in projects that may take time to mature, the opportunity may be altogether missed, which may indicate the negative impact of liquidity. Looking at layo and announcer characteristics, none of the layo announced reasons are signi cant, while announcer s that beat expectations generate a smaller or more negative reaction 14

15 in competitors stock price movement. Announcers that have been in relative nancial di culty, as measured by the weak announcer dummy, generate a more positive impact on competitors. This seems surprising, as the weak condition was already known by the market, since the measure is based on pre-announced nancial information 4. However, this may imply a more lasting negative trend that can be exploited by a wider group of competitors. In this sense, the negative coe cient on the interaction with the leverage level seems to con rm this explanation, as more leveraged rms are less able to raise the funds to invest in growing its supply, competing for the void left by the troubled announcer. Another interesting feature is that the likelihood that the average competitor has a positive stock price reaction increases if the announcer also experienced a positive CAR. This indicates an industry wide positive or negative e ect. Therefore, even though we focus on the dispersion in the competitor s market reaction, we can still nd an overall industry e ect. In terms of the overall economy, higher GDP is considered good news for competitors. Given GDP growth autocorrelation, a layo in periods of good economic performance implies a larger short run demand that will now be absorbed by competitors, characterizing overall good news for the average competitors. Examining the trend over time, the decade dummies re ect that competitors stock price reaction became more positive over time. Notice that this goes in the opposite direction of the fact that the coe cient of the announcer s positive CAR dummy is positive and announcers CARs have become on average more negative over time. We interpret these results as a sign that the industry competitive rearrangements e ects have become more important overtime, compared to the industry-wide information content obtained through mass layo announcements. This is corroborated by the increased dispersion in competitors stock price reaction throughout the sample. Once we look at the entire sample of layo announcements, including follow-up announcements by the same rm, a somewhat di erent pattern emerges. While most of the announcer and competitors characteristics have qualitatively identical results - apart from competitors liquidity becoming statistically insigni cant - the industry variables show a di erent dynamic. The dummy for rst announcements is positive and signi cant, showing that the impact on competitors is more positive at rst announcements, and similarly the interaction terms of this dummy with measures of better competitors in the sector is negative, indicating a similar result that we obtained before. Di erently, the level variables that indicate a more competitive environment - fraction of competitors with sales growth, fraction of competitors with positive return on assets, fraction of competitors in S&P 500 with positive return on assets - have positive and signi cant 4 See appendix for details on the construction of the dummy variable. 15

16 coe cients. This indicates that the information released by a follow-up announcement and its impact on the sector s competitive structure is fundamentally di erent from the rst announcement. The di erent informational content seems clear both for the industry-wide and the sector dynamic shift information, since the impact of the announcer s CAR direction is only signi cant for the rst announcement as well. In this sense, the presence of stronger competitors may be seen as good news, once it indicates a stronger overall sector condition, which could be put in check by successive layo announcements by a large rm in that particular industry. Finally, in order to evaluate the impact of di erent variables across winners and losers from the layo announcement, we look at a quantile regression with respect to the competitors stock price reaction. We run quantile regressions for the 25th, 50th, and 75th quartiles, with bootstrapped standard deviations for rst announcements. This analysis allows us to distinguish the di erent impact of explanatory variables on the reactions by winners and losers among the competitors of the announcer rm. However, we want to make clear the caveat that is common for quantile regression analysis and particularly here, that changes in the explanatory variables may push the observation outside the quartile bin, making results harder to interpret. Results are presented in Table 10. First of all, we observe the di erence in the impact of the decades on the CAR for di erent quartiles. While the 1990s increased the dispersion by both making CARs at the lower quartile more negative and higher in the highest quartile, the 2000s increased the dispersion by making median and higher quartile CARs bigger. Another important point to highlight is that the impact of the announcer having a positive CAR is positive and signi cant for all quartiles, which indicates an industry-wide e ect, as expected. The announcer s characteristics are signi cant only for the rms at the median and lower quartiles, in particular showing that layo s by weaker announcers are good news for this subset of competitors. This corroborates the more lasting impact of the mass layo that can then be also exploited by weaker competitors. In terms of the sector s competitive environment, having stronger competitors is bad news for rms that see the layo as good news, while results are mixed for rms at the median and lower quartiles. This implies that for weaker rms, having stronger competitors may be both good news, as it reduces the potential industry-wide negative shock from the layo, but at the same time reduces the chances that weak competitors can bene t from the supply reduction generated by the mass layo. 16

17 6 Concluding Remarks Mass layo announcements release information not only about industry-wide trends, but also of changes in the competitive structure in the sector. We are able to distinguish these e ects in this paper by looking at the stock price reaction to mass layo announcements by S&P 500 rms in both announcers and competitors stock prices. First of all, we notice that announcers are not on average underperforming rms, as their operating returns on assets are better than the competitors average and median, even after we adjust the competitors pool to rms with similar characteristics. This not only corroborates the results obtained by Chen et. al. (2001) that showed that announcers are rms with decreasing performance that restructure and tend to improve their position in the 3-year window after the announcement, but also allows us to interpret the median negative impact of mass layo s announcements to competitors abnormal stock returns. We observe a wide variation in competitors stock price reaction to a layo announcement, based on not only the announcer s and competitors characteristics, but also the sector s structure, in terms of the degree of competitiveness and the presence of other major competitors and their ability to exploit the opportunity presented by the announcer. Firms in a better position, as measured by sales growth and size (S&P 500 rms), are more likely to see a positive stock market reaction, particularly if they don t face strong opposition by other competitors and if the mass layo seems to be a lasting episode, given the poor condition of the announcer. This e ect is more prevalent in the announcer s rst mass layo announcement within a year. In terms of other layo characteristics, the layo of salaried workers, a proxy for human capital, seems to be seen as positive news for the average competitor, while bad news for the announcer. This corroborates the results from previous literature. However, we have not found signi cant impact of announced reasons on the market reactions for both announcer and competitors. In terms of industry-wide impact, we see a high correlation between the direction of announcer s and competitors stock reaction, even after controlling for announcers, competitors, layo, and industry characteristics. However, the impact of changes in the sector competitive structure originated by the layo seems to have become relatively more important over time, which is captured by the increase in the dispersion of reactions across competitors from the 1980s to the 2000s. 17

18 References [1] Agarwal, R., and J. Kolev, (2012). Strategic Corporate Layo s, mimeo; [2] Bhabra, G., Bhabra, H., Boyle, G., (2011). Competitive and Contagion E ects in Corporate Layo Announcements, Journal of Money, Investment, and Banking, Issue 19, pp ; [3] Cameron, C., and P. Trivedi, (2009). Microeconometrics Using Stata, STATA Press; [4] Chen, P., V. Mehrotra, R. Sivakumar, and W. Wu, (2001). Layo s, shareholders wealth, and corporate performance, Journal of Empirical Finance, Vol. 8, pp ; [5] Fama, E., (1970). E cient Capital Markets: A Review of Theory and Empirical Work, The Journal of Finance, Vol. 25, No. 2, pp ; [6] Farber, H., Hallock, K., (2009). The Changing Relationship between Job Loss Announcements and Stock Prices: , Labour Economics, Vol. 16, Issue 1, pp. 1-11; [7] Gleason, C., Jenkins, N., Johnson, W.B., (2008). The Contagion E ect of Accounting Restatements. 2008, The Accounting Review, Vol. 83, Issue 1, pp ; [8] Goins, S., Gruca, T., (2008). Understanding Competitive and Contagion E ects of Layo Announcements, Corporate Reputation Review, Vol. 11, pp ; [9] Hallock, K., (2003). A Descriptive Analysis of Layo s in Large U.S. Firms using archival and interview data: , mimeo; [10] Hoberg, G., Phillips, G., (2010). Product Market Synergies and Competition in Mergers and Acquisitions: A Text-Based Analysis, Review of Financial Studies, Vol. 23, Issue 10, pp ; [11] Hoberg, G., Phillips, G., (2010). Text-Based Network Industries and Endogenous Product Di erentiation, mimeo; [12] Madura. J., Akhigbe, A., Bartunek, K., (1995). Intra-Industry E ects of Bank Layo Announcements, Review of Financial Economics, Vol. 4, Issue 2, pp ; [13] Milanez, A., (2013). The Human Capital Costs of Financial Constraint, mimeo; [14] Payne, J. & W. Thomas, (2003). The Implications of Using Stock-Split Adjusted I/B/E/S Data in Empirical Research, The Accounting Review, Vol. 78, No. 4, pp ; 18

19 [15] Piotroski, J., Roulstone, B., (2004). The In uence of Analysts, Institutional Investors, and Insiders on the Incorporation of Market, Industry, and Firm-Speci c Information into Stock Prices, The Accounting Review, Vol. 79, Issue 4, pp ; [16] Ramnath, S., (2002). Investor and Analyst Reactions to Earnings Announcements of Related Firms: An Empirical Analysis, Journal of Accounting Research, Vol. 40, Issue 5, pp ; [17] Szewczyk, S., (1992). The Intra-Industry Transfer of Information Inferred from Announcements of Corporate Security O erings, Journal of Finance, Vol. 47, pp ; 19

20 7 Appendix Weak Announcer Dummy De nition The Weak Announcer Dummy represents an indicator of whether the announcer appears to have weaker nancial ratios relative to its industry peers. Speci cally, we focus on the following nancial ratios: Tobin s Q, liquidity, leverage, ROA, sales growth, and gross margin. In order to determine the weak announcer dummy indicator, we proceeded as follows. First, we calculated the average winsorized and lagged nancial ratio within each layo announcement grouping. Then we calculated the di erence of each observation s unique value from this calculated mean to determine a deviation of the observation from the mean. Once we generated these deviations, we then broke these deviations into quantiles, with quantile one being the biggest negative di erence and quantile four being the biggest positive di erence from the industry mean; our variables are constructed such that positive deviations from the mean signal stronger nancial position. Next, we created dummies for each nancial ratio equal to one if the deviation from the mean was positive for the layo -announcing rm. Then we summed all of these dummies. By construction, an announcer could have a sum of six if this rm exceeded the group average for all ratios or a value of zero if the announcer was below the group average for all six ratios. We nally created the weak announcer dummy equal to one if this sum was equal to three or less. In other words, an announcer is weak relative to its industry peers if, at the time of the layo, it falls short of the competitive mean of at least three of the six nancial ratios of interest. 20

21 Graph 1: Announcer s average 3-day CAR over time Average CAR over time All announcers year of announcement average of announcer's 3 day CAR per year Fitted values Average CAR over time 1st. announcements year of announcement 1st announcement annual CAR fitted Average CAR over time follow up announcements year of announcement announcer annual CAR fitted Graph 2: Announcer s median 3-day CAR over time Median CAR over time All announcers year of announcement median of announcer's 3 day CAR per year Fitted values Median CAR over time 1st. announcements year of announcement 1st announcement annual CAR fitted Median CAR over time follow up announcements year of announcement announcer annual CAR fitted 21

22 Graph 3: Announcer s 3-day CAR std. Dev. over time Evolution of announc. CAR's std. dev. all announcements Evolution of announc. CAR's std. dev. 1st announcements year of announcement std. CAR Fitted values year of announcement std. CAR fitted Evolution of announc. CAR's std. dev. follow up announcements year of announcement std. CAR fitted Graph 4: CAR Distribution over time - Announcers CAR Distribution All announcers CAR Distribution 1st. announcements CAR density CAR density day CAR day CAR 80s 2000s 90s 80s 2000s 90s CAR Distribution follow up announcements CAR density day CAR 80s 2000s 90s 22

23 Graph 5: Competitor s average 3-day CAR over time Average competitors' CAR over time All announcers Average competitors' CAR over time 1st. announcements year of announcement year of announcement average of competitors' 3 day CAR per year Fitted values competitors' annual CAR fitted Average competitor's CAR over time follow up announcements year of announcement competitors' annual CAR fitted Graph 6: Competitor s median 3-day CAR over time Median competitors' CAR over time All announcers year of announcement median of competitors' 3 day CAR per year Fitted values Median competitors' CAR over time 1st. announcements year of announcement competitors' annual CAR fitted Median competitors' CAR over time follow up announcements year of announcement competitors' annual CAR fitted 23

24 Graph 7: Fraction of Competitors with positive 3-day CAR Graph 8: Competitors 3-day CAR std. Dev. over time Evolution of compet. CAR's std. dev. all announcements year of announcement std. CAR fitted Evolution of compet. CAR's std. dev. 1st announcements year of announcement std. CAR fitted Evolution of compet. CAR's std. dev. follow up announcements year of announcement std. CAR fitted 24