Firm Boundaries Matter

The following post comes to us from Amit Seru, Professor of Finance at the University of Chicago.

Do firm boundaries affect the allocation of resources? This question had spawned significant research in economics since it was raised in Coase (1937). A large body of work has focused on comparing the resource allocation in conglomerates relative to stand-alone firms to shed light on this issue. Theoretically, there are competing views on this aspect. On the one hand, Alchian (1969), Wiliamson (1985), and Stein (1997), among others, have put forth the view that conglomerates, by virtue of exerting centralized control over the capital allocation process, may do a better job in directing investments than the external capital markets. On the other hand, the “dark side” view of internal capital markets argues that problems of corporate socialism are more prevalent in conglomerates making them less efficient in resource allocation (Rajan, Servaes, and Zingales, 2000; Scharfstein and Stein, 2000).

Estimating the effects predicted by these theories has proven challenging. On the one hand, there is a broad brush approach that argues that efficiency of conglomerates can be compared to stand-alone firms by examining their relative market values. This approach has, however, been criticized as being indirect and tainted by endogeneity bias which is hard to account for. The other, more direct approach, has been to examine the productivity differences across organizational forms to make assessment about resource allocation (Maksimovic and Philips, 2002; Schoar, 2002). In the paper, Firm Boundaries Matter: Evidence from Conglomerates and R&D Activity, forthcoming in the Journal of Financial Economics, I extend the latter by focusing on one activity and demonstrating that a link exists between R&D productivity differences and organizational form. By doing so, I hope to provide evidence that firm boundaries may matter for allocation of resources.

I choose to focus on innovative activity following the argument made in Wiliamson (1985) that “… in the presence of asset specificity, uncertainty, and opportunistic behavior—differences in internal organization may impact innovative behavior …” The intuition behind this idea is simple. Novel research projects are especially characterized by significant informational asymmetries between researchers and outside evaluators. This may provide researchers in divisions leeway to manipulate the information they transmit to corporate bosses, especially if they are faced with the possible threat of reallocation of resources by corporate headquarters. Recognizing this problem, high-level managers may be reluctant to embark on novel projects in the first place. Thus, it is precisely those organizations that attempt to exploit the efficiencies of a centralized resource allocation process that may end up fostering mediocrity in their divisional R&D activities.

I use information in the Compustat files and from the patents granted by the United States Patent and Trademark Office (USPTO) to shed light on this question. I measure the scale of a company’s R&D output by the number of patents its research generates. In addition, I measure the novelty of its research program by the average number of citations its patents receive in subsequent patent applications. I start by providing some suggestive evidence by evaluating these measures and showing that conglomerates with more active internal capital markets and higher implied competition for R&D resources do, on average, conduct less-novel research.

These results, however, only show an association between internal capital markets and research output. There may be a concern that these effects are driven by endogenous selection rather than the impact of organizational form on R&D activity. For instance, many conglomerates may have grown by acquiring firms that have the potential to come up with novel ideas in the future. Alternatively, they may acquire firms with one big idea, which has already been developed. Both these arguments would lead to different biases in estimates that compare the average R&D productivity of conglomerate firms. The main identification strategy of the paper accounts for these selection concerns by exploiting a quasi-experiment.

The experiment constructs two groups of firms: a “treatment group” comprised of firms taken over in a friendly merger and a “control group” that is assembled from a sample of targets whose mergers failed to go through. The important consideration for empirical design is that the reasons for failure of the friendly merger of the control group be unrelated to R&D policy of the target. I read news articles for each of the failed mergers in my sample and select only those to be a part of the control group where one can argue this to be the case (e.g., deals around 1987 crash). The two groups then comprise a sample where I claim that the assignment of a firm into an acquirer is random. Under this assumption, I can difference out any selection concerns by comparing the R&D productivity of the firms in the treatment group pre- and post-merger with those of the control group.

This research design allows for two tests. The identification of the main estimate comes from the unsuccessful targets that were going to conglomerate acting as a counterfactual for how the successful targets would have performed R&D after the merger, had they not been acquired by conglomerates. In addition, the research design allows me to conduct a placebo test that involves targets in non-conglomerating mergers.

I employ a difference-in-differences specification, which exploits within-firm variation and find that, relative to the control group, firms in the treatment group suffer a significant decline (about 60%) in novelty of their research output after the merger. This drop is driven by diversifying mergers with targets involved in non-conglomerating mergers not exhibiting such change in their R&D output. What is more, I find that the drop in novelty is significantly more in treatment firms that were acquired by diversified firms, which already had an active capital market in operation.

These findings also alleviate concerns that my results are driven by firms in the control group being more productive after the event, due to elevated market pressure after the unsuccessful merger. If it was the case, I would have also found similar effects for firms that were involved in unrelated mergers. As well, it would not immediately follow that market pressure would intensify for firms where I find the strongest results—i.e., in firms that are involved in mergers where acquirers operated a conglomerate with an active ICM.

I further investigate the drivers of the treatment effect by examining the R&D productivity of inventors around the merger event. There are two margins which could be responsible for a decline in the R&D productivity of the treatment group: on the extensive margin, individuals with ‘entrepreneurial spirit’ may leave the diversified firm; on the intensive margin, individuals may choose to stay in the firm but become less productive on the R&D dimension—both because the combined firm might be reluctant to fund their entrepreneurial ideas (Bhide, 2000; Gompers, Lerner and Scharfstein, 2005). I hand-collect information on all the inventors responsible for patents in the sample and exploit within-inventor variation in the data. The results suggest that the treatment effect is largely driven on the intensive margin. In particular, the impact of invention of an average inventor in the treatment group falls more than 50% post-merger. While there is an exodus of inventors after the merger event, the rate of exit is similar for both the control and treatment groups.

Next, I examine whether conglomerates that operate active capital markets, possibly aware of a consequent drop in R&D productivity, might be taking steps to counteract diminished research incentives. I examine this issue first by broadly evaluating if there is any change in value around the merger event and find no difference in either the market or the accounting performance of the two groups. Since novel innovations have been shown to generate firm value (see Hall, Jaffe and Trajtenberg, 2005), observing no change in value despite a reduction in R&D output suggests that acquirers in the treatment group did take steps to mitigate the value loss.

I investigate this issue further by assessing if there are margins on which acquirers of firms in the treatment group might have changed their behavior after the merger. Indeed, I find that conglomerate acquirers in the treatment group significantly increase their intensity of strategic alliances and joint ventures after the merger relative to those in the control group. Moreover, about 40% of the treatment effect is in cases when the acquirer in the treatment group subsequently engages in a joint venture or strategic alliance. This is consistent with recent work which argues that organizational forms such as strategic alliances may be more conducive for innovative and uncertain research (Robinson, 2008). While there is still some drop in R&D productivity in the treatment group that cannot be accounted for, I suggest possible value-enhancing measures that acquirers might be undertaking to compensate for this reduction. Overall, the evidence suggests that firm boundaries matter for the type of research activity that is done inside versus outside the firm.

The full paper is available for download here.

Both comments and trackbacks are currently closed.