國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
17
4. Empirical Results 4.1 Probability of Bidding
Table 4 exhibits the results of probit analysis, where the dependent variable, Bid, denotes 1 if a firm announce an M&A deal in a firm year and 0, otherwise. We adapt the model of Harford and Uysal (2014), which is described in Section 3.1.3 as well as add our overconfident variable and recession variable to the regression. All the standard error of control variables are clustered by firm. We also run regression with year fixed and industry fixed effect.
First, we examine the effect of CEO overconfidence on acquisitiveness for the full sample.
As shown in columns (1) and (2) of Panel A, the coefficient of overconfidence is positive and the effect is strongly significant no matter we control the fixed year effect or not. This result is consistent with previous literatures (Doukas and Petmezas, 2007; Billett and Qian, 2008;
Malmendier and Tate, 2008), which evidence that overconfident CEOs are more likely to conduct M&As than non-overconfident CEOs. Second, we examine the effect of recessions on the probability of bidding. Here, we adapt NBER’s definition as our recession measurement.
Columns (2) and (5) in Panel A shows that recessions have significantly negative effect on the probability of bidding, which means that firms undertake less M&As during recessions. This result is aligned with the Figure 1, which shows that the number and value of M&A deals decline significantly during recessions. Third, we test if overconfident CEOs would undertake more M&As than non-overconfident CEOs during recessions. The results are shown in columns (3) and (6) in Panel A with the interaction of Overconfidence and Recession. However, both coefficients are not significantly different rim zero. This may due to difference in magnitude of recessions. Hence, we examine the effect of the 2001 Recession and the 2008 Recession respectively.
To examine the effects of the 2001 Recession and the 2008 Recession respectively, we divide the sample into 2 subsamples, which cover data from 1993 to 2007 and 2002 to 2016.
Also, we replace the variable Recession with Post01 or Post08, which denotes 1 if the firm year is 2001/2008 or later. Panel B exhibits the results of our first subsample, which includes data from 1993 to 2007 to study the effect of the 2001 Recession. As shown in columns (1), (4), and (7), coefficients of Overconfidence are all positively significant. These are consistent with the results if we run the regression using full sample, which show that overconfident CEOs undertake more M&As than non-overconfident CEOs overall. Then, we test if there’s
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
18
systematic difference in firms’ M&A behavior after 2001. As shown in columns (2), (5), and (8), the coefficients of Post01 is positive. This shows that firms launch more M&As after 2001.
Finally, as we can see in column (3), (6), and (9), the interaction of Overconfidence and Post01 is insignificant, which evidence that the acquisitiveness of overconfident CEOs do not change after the 2001 Recession. The insignificance may due to the smaller magnitude of the 2001 Recession, which is not sufficient to influence overconfident CEOs’ behavior.
Panel C exhibits the results of our second subsample, which includes data from 2002 to 2016 to study the effect of the 2008 Recession. Here, we find a different story from our first subsample. First, as shown in columns (1), (4), and (7), the coefficients of Overconfidence reduce or even lose their significance. This implies that overconfident CEOs merely undertake little or no more M&As than non-overconfident CEOs during the period from 2002 to 2016 overall. Second, column (2) reports that the amounts of M&A deals decrease significantly three years after the eruption of the 2008 Financial Crisis. Finally, as shown in columns (3), (6), and (9), the coefficients of interaction of Overconfidence and Post08 are significantly negative. The results evidence that the acquisitiveness of overconfident CEOs decrease sharply after the 2008 Recession. We propose that the results can support our hypothesis 1.b. that overconfident CEOs undertake less M&As relative to non-overconfident CEOs after recessions, for they tend to cut off investment rather than access fund from capital market if they do not have sufficient internal resources to finance investment projects.
We test if our proposition is true by running the regression in subsamples divided by their level of financial constraint and cash holdings. We adapt WW Index constructed by Whited and Wu (2006) to measure financial constraints. The index in constructed as follow: -0.091CF-0.062DIVPOS+0.021TLTD-0.044LNTA+0.102ISG-0.035SG, where CF is the ratio of cash flow to total assets; DIVPOS is an indicator that takes the value of one if the firm pays cash dividends; TLTD is the ratio of the long-term debt to total assets; LNTA is the natural log of total assets, ISG is the firm’s 3-digit industry sales growth; SG is firm sales growth. The value of index increase with the degree of the financial constraint a firm faces. Then we rank the observations by their level of financial constraint each year and classify the 30% highest financial constrained observations into high financial constraint subsample and the 30% lowest financial constrained observations into low financial constraint subsample. We also classify observations whose level of financial constraint lie between 30%-70% in our sample as medium financial constraint subsamples. We present the results in Panel D.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
19
As shown in Panel D, first, the coefficients of Overconfidence are significantly positive for the lowest constrained group but insignificant for groups with medium and high level of financial constraint. The results are consistent with Malmendier and Tate (2005, 2008), who show that overconfident CEOs undertake more M&As than non-overconfident CEOs only when they have enough financial resources. Also, the coefficients of interaction between Overconfidence and Post08 within highly financially constrained groups are insignificant while significantly negative within firms which are low financially constrained. This evidences that overconfident CEOs who have abundant financial resources before the 2008 Recession would reduce significantly more in undertaking M&As after recessions. Braun and Larrain (2005) and Campello, Gramham, and Harvey (2010) show that firms would face significant decrease in financial resources during recessions. On the other hand, Malmendier and Tate (2005, 2008) evidence that overconfident CEOs view external finance too costly and they would rather bypass investment project than financing it externally when they have limited internal resources. Combining the theories we mention above, we propose that overconfident CEOs react more to the reduction in financial resource and reduce more in undertaking M&As than non-overconfident CEOs after recessions. These results show that the effect of reduction in financial resources outweigh the effect of underestimating risk or overestimating synergies by overconfident CEOs, which makes overconfident CEOs to reduce more in undertaking M&As than non-overconfident CEOs after the 2008 Recession.
Next, we examine if different levels of excess cash holdings would cause overconfident CEOs to behave differently after the 2008 Recession and be the channel to explain the reduction in undertaking M&As by overconfident CEOs. Malmendier and Tate (2005, 2008) show that overconfidence CEOs with abundant internal resources would undertake more M&As than non-overconfident CEOs and underinvest relative to non-overconfident CEOs when they have few internal resources. To test this hypothesis, first, we rank our observations by their level of excess cash holdings each year. We adapt Bates (2009)’s method to construct our measurement of excess cash holding. The details of excess cash holding constructions are described in Bates (2009)’s paper. Then we divide sample into 3 different groups, which are the 30% highest excess cash holding firms, the 30 % lowest excess cash holdings firms, and firms with excess cash holdings level lying between 30%-70% among the sample. Panel E exhibits the results of regression of different levels of excess cash holdings. First, the coefficients of Overconfidence are only significant in group with medium level of excess cash. This is slightly difference from
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
20
Malmendier and Tate (2005, 2008)’s proposition that overconfident CEOs only with high level of internal financial resources would undertake more M&As than non-overconfident CEOs.
Second, coefficients of Overconfident and Post08 are significantly negative within group with medium level of excess cash. This may imply that overconfident CEOs who have enough excess cash to overinvest in M&As before the 2008 Recession find it unable to finance M&A deals after the recession due to decrease in internal resources caused by recession (Braun and Larrain, 2005). Viewing external finance too costly, they are unwilling access funds from capital market even if the M&A deal would bring positive NPV to the firm. Thus, they reduce more in undertaking M&As than non-overconfident CEOs after the 2008 Recession.
We also test if overconfident CEOs with different levels of investment opportunities would behave differently after the 2008 Recession. We measure the level of investment opportunities by Tobin’s Q, which is the sum of the market value of book assets (AT) and the market value of common equity (CSHO×PRCC), minus the sum of common equity (CEQ) and deferred taxes (TXDB), all over book value of assets (AT). The results are shown in Panel F.
First, the coefficients of Overconfidence are only significant in group of low Tobin’s Q. This may results from the fact that overconfident CEOs overestimate the low investment opportunities which are not captured by non-overconfident CEOs. As a results, overconfident CEOs are more acquisitive relative to non-overconfident CEOs within firms with low investment opportunities. However, the coefficients of interaction between Overconfidence and
Post08 are all insignificant. This show that level of investment opportunities is not the channel
for recession to influence overconfident CEOs’ investment decision.In summary, our empirical results are consistent with Malmendier and Tate (2005, 2008), who show that overconfident CEOs undertake more M&As than non-overconfident CEOs overall due to overestimation of return and underestimation of risk and the phenomenon is significant within firms within low level of financial constraints. However, overconfident CEOs reduce significantly more in undertaking M&As after the 2008 Recession, especially for firms with low level of financial constraint, which overinvest before the recession. We show that the effect of reduction in financial resources available outweighs the effect of underestimation of risk and overestimation of synergies by overconfident CEOs and cause them to reduce more in undertaking M&As.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
21
4.2 Announcement Effect
Table 5 exhibits the summary statistics of the cumulative abnormal return (CAR) and Table 6 exhibits the OLS regression of CAR. The construction of CAR is described in Section3.1.4. As shown Panel A, the average CAR for the full sample is significantly positive, which is consistent with Harford and Uysal (2014) and Masulis, Wang, and Xie (2007), who also find positive average CAR in their whole samples. Next, we check difference in average CAR between deals made during recessions and non-recessions.
The second column in Panel A exhibits that average CAR of M&As undertaken during recessions is significantly negative while the average CAR of M&As undertaken during non-recessions periods is significantly positive. This result evidences that firms are more likely to undertake value-decreasing M&As during recessions than during non-recession periods.
Then we examine the difference in CAR of M&As undertaken by overconfident and overconfident CEOs. As shown in the third column, average CAR for overconfident and non-overconfident CEOs are both positive and significant. However, average CAR for overconfident CEOs is significantly higher than non-overconfident CEOs’. This result contradicts to Malmendier and Tate (2008) and Billet and Qian (2008)’s findings, which show that overconfident CEOs have negative impact on the announcement effect measured by CAR.
We believe that this inconsistence may result from different in sample periods between our study and previous literatures. For example, the data covered by Malmendier and Tate (2008) range only from 1980 to 1994, and the average CAR for their sample is slightly significantly negative. While the average CAR of our study is significantly positive. Furthermore, recent literatures shows that overconfident CEOs do not always destroy firm value but create value
Next, we present the difference in average CAR before 2001/2008 and after 2001/2008 in Panel B and Panel C respectively to check if there’s systematic difference after the recession.
The third column in Panel B, which include data ranging from 1993 to 2007 shows that M&As undertaken after 2001 have significantly higher average CAR than M&As undertaken before 2001. The results can somewhat explain why our average CAR for the whole sample is positive and is contrary to older literatures (e.g. Andrade et al. (2001) and Moeller et el. (2004)), which show that acquirers do not gain from the deals. There might be systematic difference in announcement effect of M&As between difference periods. Panel C exhibits that average CAR
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
22
for M&As undertaken after 2008 is lower than those undertaken before 2008 but the difference is insignificant. The result show that the reaction of investors to the announcement of M&A does not change after the 2008 Recession. This means that M&As undertaken after 2008 do not create less value than those undertaken before 2008.
Table 6 shows the regression results of CAR. Panel A exhibits the results when we examine the difference in CAR between recession and non-recession periods. First, as shown in column (1) and (4), the coefficients of overconfidence are significantly positive. This result is consistent with the statistics we present in Table 5, which shows that M&As undertaken by overconfident CEOs have higher announcement effect. Columns (2) reports that recessions have negative impact on the announcement effect. This implies that M&As undertaken during recessions create less value for acquirers and is consistent with Lambrecht (2004), who show that synergy created by M&As would reduce during recessions. Next, we examine if the announcement effect of M&As undertaken by overconfident CEOs would change during recessions. As shown in column (3) and column (6), the coefficients of overconfidence and recessions are positive but insignificant, which do not support our hypothesis that overconfident CEOs are more likely to undertake value-decreasing M&As during recessions.
This may imply that overconfident CEOs become more cautious when undertaking M&As during recessions.
We further test if the the 2001 and 2008 Recession have different impact on overconfident CEOs and the CAR of M&As undertaken by them. The results are shown in Panel B and Panel C with independent variables Post01 and Post08, which denote 1 after 2001/2008 and 0, otherwise. In Panel B, columns (1), (4), and (7) exhibit that positive effect of overconfidence on CAR weakens in periods before 2008. Also, we do not see significant difference in CAR after the 2001 Recession nor do the effect of overconfident on CAR change after the 2001 Recession, which is shown in columns (3), (6), and (9). In Panel C, the positive effect of overconfidence on M&As is still strongly significant, as shown in columns (1), (4), and (7).
However, the coefficients of Overconfident×Post08 are insignificant, which do not provide evidence that the effect of overconfidence on announcement effect would change after recessions, either.
To summarize, first, our results are more close to past literatures such as Goel and Thakor (2008), Gervais, Heaton, and Odean (2011), who posit that overconfident CEOs would take the
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
23
necessary risk that rational CEOs don’t want to take. Thus, overconfidence can help CEO make first-best investment decisions. However, investors’ better reaction toward M&As undertaking by overconfident CEOs do not change during and after recessions. This indicates that the overestimation of synergies or underestimation of uncertainty do not lead overconfident CEOs to undertake value decreasing M&As during or after recessions.
4.3 Long-term Performance
Table 7 shows the summary statistics of 24-month buy-and-hold abnormal return (BHAR) after the M&A announcement date and Table8 presents the regression results for BHAR. We adapt Duchin and Schmidt (2013)’s model to run the regression with adjustment in the benchmark portfolio. The details of model are shown in Section 3.1.5.
Panel A in Table 7 reports that the average BHAR for our full sample is 0.0283 and is significantly positive. This result implies that M&As undertaken during our sample period can create value for firms overall. We then check the difference in BHAR between M&As undertaken during recessions and non-recessions periods. As shown in the second column, M&As undertaken during non-recession periods have higher BHAR. This result is also consistent with Lambrecht (2004), who shows that M&As undertaken during expansions can capture higher synergies than those undertaken during recessions. Next we divide the sample into M&As undertaken by overconfident CEOs and non-overconfident CEOs. As shown in the third column, average BHAR for overconfident CEOs is 0.0909 and significant difference from zero while the average BHAR for non-overconfident CEOs is -0.241 but insignificant. The difference in average BHAR between overconfident and non-overconfident CEOs is aligned with our CAR results, which show that overconfident CEOs have positive impact on the value of firm. Finally, Panel B and Panel C show that there’s no significant difference in BHAR before and after 2001/2008. We then examine the regression results of BHAR in Table 8. In Panel A, columns (1) and (4) exhibit that the coefficients of overconfidence positively significant. This again show that overconfident CEOs have positive impact on M&As, which is consistent with Goel and Thakor (2008) and Gervais et al. (2011), who propose that overconfident CEOs are more willing than non-overconfident CEOs to bear necessary risk to undertake investment which can create value for firms. Coefficients of Recession shown in columns (2) and (5) are both negative but insignificant. Hence, we could not find evidence that recessions have negative impact on M&A’s long-run performance. In columns (3) and (6), the
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
24
coefficients of interaction between Overconfidence and Recession are both insignificant. This shows that the long-term performance of overconfident CEOs in undertaking M&As do not worsen during recessions. This may imply that the overestimation of synergies and underestimation of risk do not lead them to undertake more value-decreasing M&As during recessions.
In Panel B and Panel C, we present the regression results using independent variables
Post08 and Post01. Columns (1), (4), and (7) in Panel B show that the coefficients of
overconfidence are positively significant, which are consistent with the results if we run the regression in full data range. However, the interaction between Overconfidence and Recession are all negatively significant. This somewhat supports our hypothesis 2 that overconfident CEOs are more likely to undertake value-decreasing M&As during recession than recessions than non-recession periods, for there are less profitable M&A deals during recessions. On the other hand, in Panel C, the coefficients of overconfidence become insignificant as shown in columns (1), (4), and (7). This may result from the effect of the 2008 Recession that make it difficult for overconfident CEOs to create value for firms. Also, the effect of interaction between Overconfidence and Post08 is insignificant, which means that long-term performance of overconfident CEOs in undertaking M&As does not deteriorate after the 2008 Recession.Combined with the results shown in Panel B, we infer that the failure experiences after the 2001 Recession cause overconfident CEOs to be more cautious when encountering the shock of the 2008 Recession. As a result, overconfident CEOs’ performance after 2008 is no significantly poorer than performance before 2008.
In summary, overconfident CEOs do not undertake value-decreasing M&As during recessions in our full sample. Nevertheless, the positive effect of overconfidence on long-term performance decrease significantly after the 2001 Recession. This finding is consistent with our hypothesis that overconfident CEOs are more tend to undertake value-decreasing M&As during or after recessions.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
25