Nihat Aktas, Eric de Bodt, and Richard Roll
Learning from repetitive acquisitions: Evidence from the time between deals
Journal of Financial Economics | Volume 108, Issue 1 (Mar 2013), 99–117

Our JFE paper studies whether firms gain valuable experience from repetitive acquisitions in the mergers and acquisitions (M&A) market. This is an important question, because if repetition translates into learning gains, the effort to structure learning processes within the organization could be a key driver of value creation. If instead repetition does not lead to positive gains, then stricter supervision of acquisitions programs would be warranted.

By undertaking successive deals, repetitive acquirers have the potential to build experience and acquire knowledge about the takeover process. But repetitive acquisitions increase firm size and diversity, thus generating additional integration costs. If repetition is value creating for shareholders, the increase in acquisition benefits must dominate the increase in integration costs.

1: M&A activity by year
figabr1
The number of M&A transactions peaked first in 2000 and then again in 2008.

 

The sample includes 321,610 completed M&A transactions announced during 1992–2009, and in which the acquirer owned less than 50% of the target prior to the purchase and more than 50% after the transaction. The figure looks similar if one plots the dollar value of acquisitions instead (which we do in our paper).

Using a global sample of more than 320,000 deals announced during the period 1992–2009, we test empirically whether repetitive acquirers gain from learning. We focus on M&A activity because the potential for learning possibly represents sizeable economic effects. Figure 1 shows the yearly pattern in aggregate M&A activity. The average yearly total deal value by all acquirers during 1992–2009 was $1,789 billion, with a peak value in 1999 of $3,597 billion (in 2008 constant dollars). The figure here just plots the number of deals. The sample displays a first peak in the number of transactions between 1997 and 2000, consistent with the well-known friendly M&A wave of the end of the 1990s, and then a second peak between 2005 and 2007.

When you learn, time between deals (TBD) decreases

Assessing whether learning gains exist is challenging because neither acquisition benefits nor integration costs are directly observable from outside the firm. We rely on a simple theoretical model to overcome this problem. Our model relates the time elapsed between successive deals (TBD) to the ratio of changes in acquisition benefits to changes in integrations costs. Our theoretical framework is based on Hayward (SMJ 2002)'s paper. Figure 2 illustrates the theoretical premises used to derive our key proposition. Learning gains for a given deal attempt is a function of the time elapsed since the previous deal (TBD). TBD* denotes some TBD threshold around which the experience building phase gives way to the memory loss phase. Learning first increases with the TBD up to a threshold value, before which acquisitions too quickly in succession cannot provide enough time to build experience (experience building phase). Then, learning decreases with the TBD because, after the threshold, acquisition expertise decays when successive takeovers are too far apart (memory loss phase). Under these premises, we derive our testable proposition: as long as the firm is in the experience building phase (i.e., learning gains are increasing in the TBD), a decrease in the TBD through the acquisitions sequence implies positive learning gains.

2: Learning gains as a function of the TBD
figabr2

 

Repetitive acquisitions are common

We depict the acquisitions sequence of a given acquirer by using the deal order number (DON), which refers to the number of transactions already completed by the firm. To highlight that repetitive acquisition is common in the M&A market, we report the distribution of the DON variable in Table 1. More than 55% of all deals are a first deal for a particular acquirer, though multiple acquisitions are common: 67,387 transactions are associated with a DON of at least 5, 36,835 transactions with a DON of at least 10, and 19,341 transactions with a DON of at least 21. Therefore, Table 1 clearly highlights the presence of serial acquirers into the M&A market. The average deal size by DON in Table 1 is clearly increasing throughout the acquisitions sequence. For example, the fifth deal in a sequence has an average deal size three times higher than the first deal, and later deals in the sequence have an average deal size approximately four times higher than the first deal. This substantial increase in deal size through the acquisitions sequence appears consistent with learning: firms begin with smaller deals to learn the basics, then gain more knowledge and start to risk bigger acquisitions. This is a practice which is emphasized by consulting firms advising acquirers (see Harding-Rovit (HBR 2004)).

1: Sample distribution by deal order number (DON)
DON N % Average deal size
1 179,057 55.68% 103
2 40,012 12.44% 133
3 21,402 6.65% 165
4 13,752 4.28% 241
5 9,706 3.02% 325
6 7,237 2.25% 234
7 5,564 1.73% 274
8 4,428 1.38% 339
9 3,617 1.12% 381
10 and higher 36,835 11.45% 437
Total 321,610 100% 197
The sample includes 321,610 completed M&A transactions announced during 1992–2009. Respectively, N and % denote the number of acquisitions and the percentage of the sample for each deal order number. We also report the average deal size by deal order number (in millions of dollars)

Less frequent deals in industries with high concentration, large firms, large ROA, and tighter financing

To test our proposition, we regress the TBD on the DON. Our main measure of the TBD is the number of days between the most recent completed deal and the announcement date of the current deal. Given that the TBD is potentially affected by industry factors that are not necessarily related to the trade-off between acquisition-related benefits and integration costs, we control for industry determinants of the TBD in our empirical analysis.

2: Repetitive acquirers and learning: Explaining time between deals with TBD
Variable All deals Acq Pgms
DON –3.841*** –27.805***
DON2 0.015*** 0.227***
Firm fixed effects Yes Yes
Fisher statistic 82.73 199.97
Number of observations 129,223 17,590
One, two, and three stars mark statistical significance at the 10%, 5%, and 1% level, respectively.
This table shows that TBD decreases with successive acquisitions.

 

The estimated coefficient of the deal order number variable (DON) is from a fixed-effects panel regression, in which the dependent variable is the industry-adjusted time between successive deals (TBD). Column 1 reports on the “all deals” sample. In column 2, the sample includes deals embedded in acquisitions programs, defined as an acquisitions program that starts after a dormant period of 24 months and includes successive acquisitions separated by at most 12 months.

Among the industry factors, we document that the TBD increases with industry concentration, industry median firm size, and industry median return on assets. These three variables seem to characterize industries with fewer transaction opportunities. The TBD is shorter in growing industries, suggesting that growing industries offer more opportunities for acquirers. We also find that tighter financing conditions increase the TBD, which is consistent with the observation of lower M&A market activity during economic recessions and financial downturns.

Table 2 summarizes our main results. Column 1 reports on the main sample of 38,875 unique firms that have completed at least two deals during 1992–2009.

TBD decreases with successive acquisitions

The DON variable relates negatively to the TBD, with a statistically significant coefficient estimate of –3.841. This indicates that the TBD is significantly decreasing through acquisitions sequences. From the second to the fifth deal, the decrease in TBD averages 15 days. This corresponds to a 5% reduction with respect to the sample average TBD. This result provides strong and robust evidence that net gains from learning increase over the course of sequential deals.

The presence of several acquisitions over a period as long as 1992–2009 does not necessarily mean that the firm actually implemented an acquisitions program. To assess whether firms implementing acquisitions programs develop learning-by-doing, we create another sample that includes only “program acquirers”—firms that did not announce any acquisitions during a period of at least 24 months, and then engaged in successive acquisitions separated by at most 12 months. These results, which are reported in column 2 of Table 2, are even more supportive of the learning hypothesis: the decrease in the TBD from deal to deal is substantial, with an average reduction of almost 30 days from the second deal to the third deal, or a 9% reduction with respect to the sample average TBD. This finding is consistent with the view that program acquirers are more focused on transferring their positive experience from one deal to the next as emphasized by consulting firms and the business press (see, e.g., Ashkenas-DeMonaco-Francis (HBR 1998)). For example, General Electric, a well-known repetitive acquirer, has managed to specialize its acquisition process in order to effectively integrate most of its acquisitions within 100 days.

We examine also whether the relation between the TBD and DON is a function of variables known a priori to be correlated with learning. In this respect, we consider three potential factors: heterogeneity in acquisition experience, CEO continuity, and integration capability.

Learning is impeded when deals are different...

Heterogeneity in acquisition experience.
Intuitively, learning-by-doing should relate to the degree of similarity between current and past acquisitions. Because, too much heterogeneity in acquisition experience may impede firm learning in early stages of capability building, whereas experience in similar settings should facilitate learning and improve acquisition performance. To measure heterogeneity in successive acquisitions, we construct an index corresponding to the variance of a portfolio of deal characteristics. This is analogous to the variance of a portfolio of assets, the deal characteristics corresponding to the different assets in the portfolio. We consider four deal characteristics: the extent of diversification of the deal at the three-digit SIC code level, the target status (whether the target is listed, private, or subsidiary), whether the deal is domestic, and the relative deal size (i.e., deal size divided by the acquirer's equity market value). The four deal characteristics are equally weighted in the index. We compute the heterogeneity index starting from the third deal of the program, and the index is updated for each deal thereafter. The results indicate that heterogeneity across successive deals reduces the speed at which deals are undertaken. This result suggests that slower or more difficult learning (because of deal heterogeneity) gives pause to acquisition speed.

...when the CEO changes

CEO continuity.
So far, we perform the analysis at the firm level, and therefore focus on organizational learning. However, substantial learning may depend on the persons involved in deal-making, such as the CEO for large deals, and learning might be more salient in cases of CEO continuity, while being negatively affected following a CEO change. To test the effect of CEO continuity on the speed at which successive deals are undertaken in an acquisitions program, we start by building an acquisitions program sample at the CEO-firm level. We use the ExecuComp database to identify the CEOs. We only consider full year CEOs, and match these CEOs with our initial M&A sample. An indicator variable is set to one when the current deal is undertaken by a CEO who is new since the previous deal of the firm. This variable is 0 in case of CEO continuity. The proportion of deals undertaken by new CEOs is around 9% in our sample. We find that the decrease in the TBD from deal to deal is stronger under CEO continuity, while a CEO change retards the TBD decrease from deal to deal. These results indicate that the learning effect is stronger under CEO continuity and is negatively affected by a CEO change. Learning is therefore not only organizational but is also related to individuals that are involved in deal-making.

...and when integration is costly

Integration capability.
A third result bears on the integration capability of the firm. Integration costs should increase with deal size, which in turn should reduce the integration capability of the firm for subsequent deals and negatively affect the speed at which subsequent deals are undertaken. To test this idea, we use the deal size of the previous deal as a proxy for the saturation of the firm's integration capability (i.e., large acquisitions require more resources to merge with existing activities). We show that the larger the previous deal, the more the acquirer reduces its acquisition speed and takes more time to undertake its acquisition attempts.

Conclusion

Our empirical evidence uncovers a clear and significant decrease in the TBD during acquisitions sequences. This negative trend is consistent with positive learning gains associated with repetitive deal-making. We also document that the negative trend in the TBD relates to known determinants of learning, such as heterogeneity in acquisition experience, CEO continuity, and integration capability. This brings further support to the notion that TBD relates to the tradeoff between acquisition benefits and integration costs. Our results should have important managerial implications because they reveal the importance of learning-by-doing through repetitive acquisitions. An organizational structure flexible enough to encourage learning thus appears highly desirable.

Advertising
FAMe thanks the editors and publishers of
The Review of Financial Studies, The Review of Asset Pricing Studies, and The Review of Corporate Finance Studies.
Advertising Inquiries Welcome. Please contact fame-jagazine@gmail.com

03-dongyuan
Dong Yuan: Mountain Hall. China, 10th century. Landscapes have been the subject of Chinese paintings from ancient to modern times. They are often considered as the highest form of Chinese paintings. This mountain was painted during the “great age of Chinese landscape.” It beautifully exemplifies the style in which monochromatic landscapes were not mere reproductions but meant to grasp the rhythm of the nature. Dong Yuan followed the literati painting style, spanning the Tang and early Ming dynasties. Literati refers to the lifestyle of a scholar-artist (just like those of our authors in FAMe). The traditional Chinese painting styles used calligraphy techniques, with a brush dipped in ink (no oil). Different types of brush strokes were developed to paint landscapes, from the meticulous to the freehand. The canvas material was usually paper or silk mounted on a hand scroll or a wall scroll.