Capítulo 2. El sonido como generador de identidad
3.3. Elementos curriculares y vinculación con el curriculum
The first part of the empirical analysis employs the hazards regression model proposed by Cox (1972) to answer the question whether increased firm size can indeed act as takeover protection. Takeover data can be interpreted as failure time data, when we assume that all firms are at risk to be taken over and the survival time until takeover is observed. I construct such a data set by identifying all firms on the SDC database where the majority of shares out-standing were acquired and matching these observations with the Worldscope database. The year of the takeover is defined as the time of failure. The resulting sample consists of firm-year observations with two types of firms. For one set of firms the firm-year of failure is observed within the sample period until 2006 but for the other set of firms it is not observed because these firms still existed at that point in time. Therefore, a right censoring problem is present, which cannot be treated properly in standard logit or probit models. Those models would
117 As robustness check also the 11-day event window (-5, 5) is used. Results (not tabulated) are consistent for both event windows.
sider firms that are not taken over until the end of the sample period as non-targets.118 There-fore, estimation outcomes crucially depend on the end point of the study. The Cox model treats these firms as not taken over yet, which means that for these firms the variable takeover year is correctly specified as right-censored at the end year of the study. Therefore, the de-pendent variable in this case is duration (from entering the sample until takeover) and not just a dummy variable, which distinguishes targets and non-targets. Closely related to the right censoring problem of the takeover year is left truncation, which means that not all firms enter the sample at the same time. The problem of delayed entry is present in this study because accounting databases such as Worldscope add firms at different points in time. This happens either because they expand the universe of firms covered only stepwise or because over time firms newly founded or listed are added to the database. The Cox model also accommodates left truncation. Another advantage of the Cox model is that it allows including time-varying covariates. This is crucial because most firm and industry data change annually and it is likely that the hazard for takeover depends more on recent values of the covariates than on values at the beginning of the sample.119
The prior literature on predicting takeover targets has already identified a number of variables affecting the likelihood to become a target. Palepu (1986) finds that smaller firms, less effi-cient firms, firms with low growth but large financial resources and firms with high growth but low financial resources are likely takeover targets. Hasbrouck (1985) observes a negative influence of size and market-to-book ratio on the takeover probability. Trimbath, Frydman, and Frydman (2001) also find that smaller and more inefficient firms face a greater risk of takeover. Due to these results, I control for a number of firm characteristics when testing Hypothesis 1, which says that the risk of takeover is decreasing in firm size. Besides firm
118 Shumway (2001) was the first, who proposed to use a hazard model to avoid the deficiencies of standard logit and probit models in the related context of bankruptcies.
119 See Kalbfleisch and Prentice (2002), chapter 4, or Cameron and Trivedi (2005), chapter 17, for a thorough statistical discussion of the Cox model. A more applied discussion of the Cox model can be found in Hos-mer, Lemeshow, and May (2008). Trimbath, Frydman, and Frydman (2001) discuss the econometric issues of the Cox model important to the context of takeovers.
characteristics, I also include a few industry level variables, which proxy for the competitive situation and the merger activity in the firm’s major industry. In total, there are 210,224 firm-year observations of 24,563 sample firms (Number of subjects) and 9,004 takeovers (Number of failures).
Insert Table 4.4 here
The results in Table 4.4 clearly show that the risk of a takeover is significantly lower for lar-ger firms. Across six different measures for size, including sales, assets, and market capitali-zation this finding persists. The effect is not only statistically but also economically signifi-cant. The coefficients on Large and Small in model (1) imply, for example, that firms with a market capitalization in the highest quartile (Large=1) of their three-digit SIC industry have an 18% lower probability of being taken over than the firms in the two middle quartiles. In contrast firms in the lowest quartile (Small =1) have a 10% higher chance of becoming a tar-get than firms in the two middle quartiles. These results strongly support Hypothesis 1 and I conclude that firm size indeed can act as a takeover defense. However, not only firm size has a negative impact on the risk of takeover, also firm age (FirmAge), recent acquisition activity (Log#Deals), and higher profitability (NetProfitMargin) of the firm reduce the likelihood of being acquired.120 Log#Deals is the natural logarithm of the number of acquisitions (plus one) undertaken by the firm over the last two years. The result supports the notion that own (poten-tially defensive) acquisitions can prevent a possible takeover in the future. The effect is also economically quite large: a one standard deviation increase in Log#Deals decreases the prob-ability of a takeover by about 25%. The negative effect of firm age on the likelihood to be-come a target is consistent with the widespread belief that old corporations usually do not disappear. Firm age is also a measure of unobserved firm characteristics, which influenced its prior survival and predict its survival in the future. More profitable firms (measured by Net-ProfitMargin) are also less like to be taken over; however, the economic effect is rather small.
120 Comparable results (not tabulated) are obtained when using LogAcqVolume instead of Log#Deals.
In line with the results of Palepu (1986), I find that firms with financial resource constraints or high growth potential are more likely to be taken over. Leverage (Leverage) defined as total debt over total debt plus common equity and liquidity (PercentCash) defined as cash plus short-term investments over total assets are used as proxies for financial resource avail-ability. Growth potential is proxied by research and development expenses over total asset (R&D).121 Firms with higher leverage are significantly more likely to be taken over, whereas the coefficient for PercentCash has the expected negative sign but is insignificant. Firms with high R&D expenditures are significantly more likely to become a target. Both results together imply that in particular firms with high growth potential but small financial recourses have a higher probability to be acquired. These are possibly smaller and younger firms, which are acquired by older and larger firms to augment their product portfolio. The coefficients of market-to-book ratio and capital expenditures (CapEx) are insignificant.
Besides firm specific factors, industry characteristics also have an influence on the risk of takeover. I focus here on industry rather than country characteristics, because as neoclassical merger theories and also the model of Gorton et al. predict, merger waves mostly cluster by industry (see, for example, Mitchell and Mulherin, 1996). In times of high takeover activity in an industry (Log#DealsIndustry), the probability of becoming a target increases as well.122 Moreover, it seems that firms from homogeneous and heterogeneous firm size industries have a lower risk to become a target, however, only the coefficient of HomogeneousIndustry is significant. This result is in line with the prediction of Gorton et al. that most merger waves should be expected in mixed firm size industries, because in these industries inefficient (de-fensive) as well as efficient merger waves should occur. The competitive situation in an in-dustry also affects the likelihood of takeover. Firms in less competitive industries (with higher
121 Since research and development expenses often need not to be reported, I set this item equal to zero if it is missing.
122 Comparable results (not tabulated) are obtained when using LogAcqVolumeIndustry instead of Log#DealsIndustry.
net profit margins) have a higher likelihood to become a target. However, the coefficient of MediumNetProfitMargin is in most of the models only marginally significant.