Firm Boundaries and Delegation Research Chapter 2

Table 1 The effects of input value and uncertainty on delegation IOi,j Mean Productivityi,c CV Productivityi,c Input FE Output FE Country FE Plant controls Firm controls WMS noise controls Observations (1) 0.986*** (0.350) 0.027** (0.013) 0.003 (0.0029) Yes No No No No Yes 2,889 (2) 0.973*** (0.345) 0.028** (0.014) 0.002 (0.003) Yes No Yes No No Yes 2,889 (3) 1.133** (0.511) 0.033** (0.016) 0.003 (0.003) Yes Yes Yes No No Yes 2,889 (4) 1.209** (0.501) 0.035* (0.018) 0.004 (0.002) Yes Yes Yes Yes Yes Yes 2,889 Notes: The dependent variable is Delegationf (j,c),p(i) , the degree of autonomy granted by the CEO of the parent firm f (with primary activity j, located in country c) to the senior manager of plant p (with primary activity i). IOi,j is the direct requirement coefficient for the sector pair ij. Mean Productivityi,c is the mean of supplier productivity. CV Productivityi,c is the coefficient of variation of labor productivity of the independent suppliers in input industry i located in country c. Plant-level controls are the percentage of the plant’s employees with a bachelor’s degree or higher and the plant’s share of the firm’s employment. Firm-level controls are the employment and age of the parent firm. The noise controls are related to the WMS survey (e.g., duration and day-of-the-week of the interview process, seniority, and job tenure of the plant manager interviewed, and whether the CEO is located in plant p). Output and input fixed effects are respectively the primary activities of the parent and of the plant (defined at 3-digit SIC). Standard errors clustered at the input-industry (i) level in parentheses. ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels. As expected, the coefficient of IOi,j is positive and significant across all specifications, indicating that integrated suppliers of more “important” inputs (i.e., those with larger value contributions) are granted more authority (prediction P.1). In terms of magnitude, based on the estimates reported in column 4 of Table 1, increasing the input-output coefficient by 1 standard deviation increases delegation by around 0.072 standard deviations.36 36 The standard deviation of IOi,j in the delegation sample is 0.06, so 1.2*0.06 = 0.072. This effect might appear to be a small, but can be compared to the effect of other covariates. For example, the coefficient of log firm employment, one of the firm controls included in column 4 of Table 1, is 0.101, which implies that an increase in firm employment of 40% is associated with a 0.101*0.4=0.040 standard deviation change in the delegation index. 25 The coefficient of Mean Productivityi,c is also positive and significant in all specifications. We thus find strong support for prediction P.3 of our model.37 Based on the estimates reported in column 4 of Table 1, increasing mean supplier productivity by 1 standard deviation increases delegation by around 0.034 standard deviations.38 Recall that the theory has no implications for the effect of CV Productivityi,c on delegation choice. As it happens, the coefficient of this variable is not statistically significant. Concerning the auxiliary controls included in Table 1 (coefficient estimates not reported), we find that larger and older firms are more likely to delegate decisions to their integrated suppliers and that plants that are larger and have a more educated workforce are granted more autonomy. It is interesting to compare our results with those of Bloom et al. (2012), who use the same measure of within-firm delegation (which they refer to as decentralization). They find that trust is a key driver of the internal allocation of decision rights and has larger effects than any of the other covariates in their study: when including the full set of these covariates, they find that a 1 standard deviation increase in trust is associated with a 0.07 standard deviation increase in delegation. Our results indicate that input value is also an important driver of delegation, as its effect is as large as that of trust. Notice that, in our analysis, the role of national trust is absorbed by the country fixed effects. We have carried out a series of additional estimations to verify the robustness of the results of Table 1. The results of these estimations are reported in Section A-2.3 of the Empirical Appendix. A first set of robustness checks addresses is related to the delegation variable, which measures the degree of autonomy granted by the firm’s CEO to the plant’s top manager. Recall that firms in which the plant manager is also the CEO are excluded from the dataset of Bloom et al. (2012) used in our analysis. Moreover, Table 1 includes “noise controls” (e.g., seniority and job tenure of the plant manager interviewed, and an indicator variable for whether the CEO is on site) to reduce measurement error in the delegation variable. Table A-7 restricts the analysis to multi-plant firms. Although this drastically reduces sample size to 1,663 observations, the coefficient of IOi,j is always positive and significant and the coefficient of Mean Productivityi,c is positive and significant in three of the four specifications. A second set of robustness checks concerns the uncertainty measures. In Table A-8, we verify that the results of Table 1 are robust to using US stock market data from Bloom et 37 This is notwithstanding the fact that our proxy for first-order stochastic differences in G(y) suffers from measurement error, since we cannot fully control for industry prices. As mentioned before, variation in the distribution of labor productivity across industries reflects not only differences in output, but also in the prices prevailing in those industries. To the extent that prices vary at the industry level, they are accounted by the fixed effect δi in (7). Sector-country fixed effects cannot be included, since they would be collinear with the uncertainty measures. 38 The standard deviation of Mean Productivityi,c in the delegation sample is 0.97, so 0.035*0.97=0.034. 26 al. (2018) to construct alternative uncertainty measures. The variable Mean Stock Returnsi is the mean of stock market returns in SIC4 industry i, while SD Stock Returnsi captures the cross-section dispersion in stock market returns across firms in that industry. Unlike Mean Productivityi,c and CV Productivityi,c , these uncertainty measures vary only at the sector level. Moreover, they are only available for listed firms in some (manufacturing) industries, which reduces the size of the sample. Notwithstanding these limitations, the results are broadly in line with our model’s predictions: the coefficient of IOi,j remains positive and significant, while the coefficient of Mean Stock Returnsi is positive but not significant.39 In Table A-9, we control for contracting frictions at the industry, country, and industrycountry level. The variable Rule of Lawc used to proxy for a country’s judicial quality is taken from Nunn (2007), while the variable Contract Intensityi is taken from Alfaro et al. (2019), who use Nunn’s methodology to construct at measure of input contractibility at the SIC4 level. Our coefficients of interest remain unaffected while the interaction between a sector’s contract intensity and legal quality is mostly insignificantly associated with more delegation. The coefficients of the additional controls suggest that contracting frictions increase the incentives to delegate decisions to integrated suppliers. We have also explored non-linearities in the effects of input value on delegation choices, by including different quartiles of the variable IOi,j in equation (7). The results of Table A-10 show that the effects are concentrated in the top quartile, which captures the most important inputs. Overall, the results of Table 1 and the robustness checks in Section A-2.3 of the Empirical Appendix confirm the role of input value and supplier uncertainty on delegation choices, in line with predictions P.1 and P.3 of our model. 5.2 The Effects of Input Value and Uncertainty on Integration We next assess validity of our model predictions about the effects of input value and uncertainty in supplier markets on integration choices. To this purpose, we use the integration sample and estimate the following linear probability model:40 Integrationf (j,c),i = +α1 IOi,j + α2 Mean Productivityi,c + α3 CV Productivityi,c (8) +β4 Xf + δi + δj + δc + ϵf (j,c),i , The dependent variable is the probability that firm f (with primary activity in sector j, located in country c) integrates input i within its boundaries. IOi,j is the direct requirement coefficient 39 We have also verified that the delegation results are robust to controlling for the number of suppliers in each sector-country i, c (results available upon request). 40 The results are robust to restricting the analysis to firms in the delegation sample. 27 for the sector pair ij. Mean Productivityi,c and CV Productivityi,c are respectively the mean and coefficient of variation of labor productivity of independent suppliers in input industry i in country c. Xf is a vector of firm-level controls, and δi , δj , and δc denote input-industry, output-industry, and country fixed effects. In the most demanding specifications, we exploit only within-firm variation to identify the role of input value and supplier uncertainty. In these specifications, we replace output-sector and country fixed effects with firm fixed effects (δf ), which allow us to account for the role of unobservable firm characteristics. We cluster standard errors at the input-industry (i) level.41 According to prediction P.2 of our theoretical model, producers of good j should be more likely to integrate suppliers of more valuable inputs, implying that the coefficient of IOi,j should be positive and significant. According to prediction P.4, when controlling for Mean Productivityi,c , the coefficient of CV Productivityi,c should be positive and significant. The results of estimating (8) are reported in Table 2.42 We first regress Integrationf (j,c),i against the key controls of interest and input-industry fixed effects (column 1).43 We then add country fixed effects (column 2), output-industry fixed effects (column 3), and additional firm-level controls (column 4). In the last specification, we include firm fixed effects, exploiting only within-firm variation in integration choices (column 5). In this specification, firm controls as well as country and output-industry fixed effects are absorbed by the firm fixed effects (each firm f is associated to one location and one primary activity). The coefficient of IOi,j is positive and significant at the 1 percent level across all specifications, confirming that final good producers are more likely to produce in house more valuable inputs, in line with prediction P.2 of our model. In terms of magnitude, based on the specification of column 3, moving the input-output coefficient by 1 standard deviation increases the probability of vertical integration by 0.64 percentage points — a 64 percent increase compared to the baseline probability of one percentage point.44 The estimated coefficient for CV Productivityi,c is also positive and highly significant. This finding is robust to including additional controls and different sets of fixed effects. In particular, it continues to hold when including firm fixed effects in column 4. In this specification, the coefficient α3 in (8) is identified by exploiting only within-firm variation in the dispersion of supplier productivity across input industries. As for the economic magnitude of 41 The results of are unaffected if we cluster at the input-output (i, j) or at the industry-country (i, c) level. Recall that the WordlBase sample focuses on manufacturing firms that integrate at least one input different from their primary output. In these regressions, we do not include firms from Greece. This is because establishments in Greece only report their primary SIC codes. As a result, we cannot use within-firm variation to study the determinants of integration choices. 43 Recall that, for each industry j, we focus on its top 100 inputs in terms of (strictly positive) IO coefficients. 44 The standard deviation of IOi,j is 0.036. Thus, 0.179*0.036*100 = 0.644. 42 28 the effects, based on the specification in column 3, a one-standard-deviation increase in CV Productivityi,c increases the probability of integrating a supplier by around 0.34 percentage points.45 This corresponds to a 34% increase relative to the baseline integration probability of one percentage point. Table 2 The effects of input value and uncertainty on integration IOi,j Mean Productivityi,c CV Productivityi,c Input FE Output FE Country FE Firm FE Firm controls Observations (1) 0.150*** (0.013) -0.000 (0.000) 0.001*** (0.000) Yes No Yes No No 6,644,884 (2) 0.179*** (0.015) -0.000 (0.000) 0.001*** (0.000) Yes Yes Yes No No 6,644,884 (3) 0.179*** (0.015) -0.000 (0.000) 0.001*** (0.000) Yes Yes Yes No Yes 6,644,884 (4) 0.203*** (0.016) -0.000 (0.000) 0.001*** (0.000) Yes Yes No 6,644,884 Notes: The dependent variable is Integrationf (j,c),i , a dummy variable equal to 1 if firm f (with primary activity in sector j, located in country c) integrates input i within its boundaries. IOi,j is the direct requirement coefficient for the sector pair ij. Mean Productivityi,c and CV Productivityi,c are respectively the mean and the coefficient of variation of labor productivity of the independent suppliers in input industry i located in country c. Firm-level controls are the employment and age of the parent firm. Output and input fixed effects defined at 3-digit SIC. Standard errors clustered at the input-industry (i) level in parentheses. ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels. These results provide strong support for prediction P.4. Increases in risk increase the option value of integration: when there is more uncertainty about the problems that will arise during production HQ attaches greater value to the option of being able to centralize or delegate production decisions. When productivity is distributed lognormally, as it is the case in our data, this effect dominates the (ambiguous) effect on the cost of integration, so firms should be more likely to integrate suppliers in riskier input industries. Concerning the additional controls, the negative coefficient of Mean Productivityi,c , although insignificant, is consistent with our theoretical model: recall from Section 3.3.2 that 45 The standard deviation of CV Productivityi,c is 4.63, thus 0.0007*4.63*100 = 0.342. 29 stochastic increases in the supplier’s ability to solve problems reduce the option value of integration, but also reduce its cost, and it is an empirical question whether one effect dominates the other. In our data, the benefit effect appears to dominate the opposing cost effect. The coefficients of the firm controls (not reported), indicate that larger and older firms are more likely to integrate inputs within their boundaries. In Section A-2.4 in the Empirical Appendix, we report the results of series of additional estimations to verify the robustness of the results of Table 2. We first consider different samples of firms. Some concerns have been raised in the literature about the measurement of vertical integration, particularly for multi-plant firms (Atalay et al., 2014). Measurement error should work against us, making it harder to find a significant effect of input value and supplier uncertainty on vertical integration. Nevertheless, we have verified that the results of Table 2 continue to hold when we restrict the analysis to single-plant firms (see Table A11). The coefficients of input value and input risk also remain positive and significant when restricting the analysis to multi-plant firms (see Table A-12).46 Another set of estimations is related to the uncertainty variables. In Table A-13, we restrict the analysis to input industries in which there are at least 50 suppliers in sectorcountry i, c, to rule out that the results of Table 2 are due to demand-driven supply assurance motives for integration (see discussion in Section 6).47 In Table A-14, we use data from Bloom et al. (2018) on stock market returns of US firms to measure uncertainty in input industries. As mentioned before, these alternative uncertainty measures vary only at the sector level and can only be constructed for some (manufacturing) industries. The results confirm that firms are more likely to integrate suppliers that produce more valuable inputs and operate in riskier industries. In Table A-15, we control for contracting frictions at the industry, country, and industrycountry level, as in Nunn (2007). In these regressions, the sample is restricted to manufacturing inputs, for which the variable Contract Intensityi , measuring the fraction of inputs that are differentiated, can be constructed. Following Nunn (2007), we interact this variable with, Rule of Lawc , a measure of each country’s judicial quality. The coefficient of IOi,j remains highly significant and increases substantially in magnitude. This is because measurement error in IO coefficients is substantially smaller for manufacturing inputs compared to others (see Section 4.3) and thus the attenuation bias is less severe. Note that the coefficient on the interaction of Contract Intensityi with Rule of Lawc is negative and highly significant, in line 46 Table A-12 includes multi-plant firms that have establishments in different countries. The results are robust to excluding multinational firms. 47 We have also verified that the integration results are robust to controlling for the number of suppliers in each sector-country i, c (results available upon request). 30 with previous studies (e.g., Acemoglu al., 2009). A final set of robustness checks concern the input-output coefficients. We have explored non-linearities in the effects of input value on integration choices, by including different quartiles of the variable IOi,j in equation (8). The results of Table A-16 confirm that firms are more likely to integrate suppliers of more valuable inputs. Overall, the robustness checks of Table 2 confirm the role of input value and supplier uncertainty on integration choices, in line with our model’s predictions. 6 Discussion: Alternative Mechanisms Our empirical analysis establishes the following regularities: (i) firms delegate more decisions to integrated suppliers that produce more valuable inputs; (ii) firms are more likely to integrate suppliers of more valuable inputs; (iii) firms delegate more decisions to integrated suppliers in more productive input industries; (iv) firms are more likely to integrate inputs in industries in which supplier productivity is more dispersed. These results can be rationalized by our theoretical model, in which integration creates a real option for HQ to centralize or delegate decisions according to comparative advantage, and the value of this option increases with the degree of uncertainty about the severity of problems that may arise during the production process. Existing alternative models could provide a rationale for some — but not all — of our empirical findings. For instance, as suggested earlier, the model in Legros and Newman (2013) could explain the finding that the likelihood that a supplier is integrated grows with the input’s contribution to total firm value. However, that model does not consider the possibility of delegation and thus cannot rationalize the findings about its determinants; nor does it account for the empirical effects of risk on integration. The finding that the likelihood that a supplier is integrated grows with the riskiness of the input industry is related to the literature on supply assurance motives for integration (e.g., Carlton, 1979; Bolton and Whinston, 1993). In these models, the assurance motive is driven by uncertainty resolved after input production (e.g., product demand), possibly augmented by the supplier’s hold-up behavior. Broadly speaking, one would expect less integration when there is less of a risk of suppliers coming up short, for technological or behavioral reasons. These might also account for the positive coefficient of CV Productivityi,c in Table 2. However, we should expect these demand-driven mechanisms, particularly the variants in which suppliers opportunistically sell to other buyers, to be less relevant when firms can source inputs from many suppliers. Against this hypothesis, the coefficient of CV Productivityi,c 31 remains positive and highly significant when focusing on input industries with large numbers of suppliers (see Table A-13). The result also holds when we include output industry fixed effects, which account for product market uncertainty (see columns 2 and 3 of Table 2 and the corresponding specifications in Section A-2.4 of the Empirical Appendix), and firm fixed effects, which account for demand for inputs by other firms in the same country-output sector (see column 5 of Table 2 and the corresponding specifications in Section A-2.4 of the Empirical Appendix). Moreover, while existing models of supply assurance could in principle explain why suppliers in riskier industries are more likely to be integrated, they do not rationalize the other empirical findings, since they neither have anything to say about delegation, nor address the role of input value in integration decisions. In a similar vein, transaction-cost reasoning might account for the risk-integration finding.48 Along these lines, one may argue that more uncertainty could lead to more contractual incompleteness, which in turn could lead to more integration.49 As discussed above, our results are robust to including controls for the degree of contractibility (as in Nunn, 2007) to account for this alternative mechanism (see Table A-15). Finally, like classical supply assurance, this approach appears to be silent on the drivers of delegation and the organizational effects of input value. As mentioned in the Introduction and documented in Section A-2.2 of the Empirical Appendix, our data also show that more vertically integrated firms tend to delegate more. Following the discussion in Section 3.2, this correlation is consistent with our theoretical model: higher value enterprises (e.g., that have more productive HQs or can sell their goods at higher prices) should be more vertically integrated while simultaneously granting more autonomy to their subordinates. The covariation of delegation and integration could potentially be rationalized by managerial capacity (MC) models in which an HQ’s attention is a scarce corporate resource (e.g., Geanakoplos and Milgrom, 1991; Aghion and Tirole, 1995): if vertical integration increases the scope of decisions in a firm, HQ may simply need to cede control to lower-level managers. However, there are grounds for skepticism. According to MC models, delegation should be negatively correlated with management capacity because higher MC reduces HQ’s requirement to delegate decision making power. By contrast, all proxies for managerial capacity 48 Lafontaine and Slade (2007) summarize the evidence from a few single-industry studies. Some of the steps in this argument are rather tenuous. One can think of mechanisms through which uncertainty could enhance rather than reduce contractibility, e.g., if greater risk encourages more investment in verification technologies. In our data, input risk (as measured by CV Productivityi,c ) is uncorrelated with contracting frictions (as measured by the interaction between Contract intensityi and Rule of Lawc ). 49 32 are positively correlated with delegation in the data.50 Moreover, capacity-constrained HQs would tend to keep control of the decisions regarding more important inputs and delegate decisions concerning less important ones, which goes against one our key empirical findings: more autonomy is granted to suppliers who contribute more, rather than less value to the enterprise. 7 Conclusion Organizations are complicated. Understanding them entails simplification, and a lot has been learned by isolating distinct organizational design elements. But there are costs to isolation. Formally similar models that focus only on one dimension of the organization can mislead when embedded in two dimensions. For example, based on models that study separately firm boundaries and the allocation or control within firms, one may expect integration and centralization to covary positively at the firm level, while the opposite is true in the data. In this paper, we have brought integration and delegation together, both theoretically and empirically. A number of insights emerge from the exercise. First, the analysis reveals a novel mechanism by which supply assurance motivates integration: the ability to redeploy control to the party most competent to solve problems that may arise during production ensures a minimal level of productivity. This places the evaluation of the costs and benefits of owning suppliers under the rubric of real options. Our findings suggest that integration may increase managerial flexibility, because it allows re-allocating decision-making across different parts of the organization. Outsourcing can therefore be very costly for firms, which lose the ability to redeploy control in light of new information.51 Thus, in a world in which input risk is rising, we would expect waves of vertical mergers, as the option value of being able to reallocate decision-making within firm boundaries increases. Our framework also helps to uncover unifying themes in the analysis of organizational design. In particular, it extends the value principle, already applied to integration in previous work, to delegation: the value of an input or the profitability of a firm affect how much autonomy will be granted to the input’s supplier. 50 The (unreported) coefficients of the plant’s size and workforce education in columns 4 and 5 of Table A-6 are both positive and significant; so is the coefficient of the plant’s management score in column 5. 51 This is illustrated by Boeing’s infamous 787 Dreamliner fiasco. In a departure from its practices with earlier models of aircraft, Boeing outsourced the design and manufacture of major components of the 787 (e.g., fuselage, wings, stabilizers) to independent suppliers. As it turned out, some of these suppliers were not competent to solve various design and procurement problems that arose during the course of production. The need for remedial fixes led to years of delivery delays and tens of billions of dollars in cost overruns. Part of Boeing’s remedial reorganization for the Dreamliner was to acquire some of the major suppliers in order to have more direct control on the production of its inputs (e.g., Tang et al., 2009; McDonald and Kotha, 2015). 33 An important consequence of our findings concerns the connection between delegation (or decentralization within firms) and productivity. Delegation is often associated with better performance, i.e., higher generated value, though the direction of causality remains an open question. A frequent claim in the business press is that delegation increases value, for instance by improving incentives and morale, or freeing up top managerial resources. A few studies find evidence for links in this direction (e.g., Bresnahan et al., 2002; Aghion et al., 2021). The present model offers two selection mechanisms through which productivity could be driving delegation, rather than the other way around: exogenously more productive firms should be more willing to delegate; and subordinates with more autonomy may be the ones who happen to have encountered problems they are good at solving. As evidence mounts that organization matters for the performance of individual firms, industries, and aggregate economies, it is becoming ever more imperative to understand the functioning of organizations as a whole rather than just their parts. Studying firm boundaries together with other aspects of the firm’s internal organization helps to illuminate interdependencies that are crucial for understanding the functioning and guiding the design and regulation of organizations.