Technology, Information, and the Decentralization of the Firm
GAICIL Insights
Headquarters rules when it knows as much as the field.1 When it does not, the field should rule. Technology, Information, and the Decentralization of the Firm, by D. Acemoglu, P. Aghion, C. Lelarge, J. Van Reenen and F. Zilibotti, analyzes how the diffusion of new technologies relates to firm decentralization. Centralized control relies on the principal's information, equated with publicly available information. Decentralized control delegates authority to a manager with superior information, who may use that advantage to make choices that do not serve the principal's interests.
The tradeoff moves with information. As public information about a specific technology grows, centralization becomes more attractive, because the principal can decide well without relying on the manager. When technologies are new and implementation information is scarce, decentralization is more valuable. The model predicts that firms closer to the technological frontier, firms in more heterogeneous environments and younger firms are more likely to decentralize. Novelty pushes authority downward; familiarity pulls it back up.
Three datasets on French and British firms in the 1990s show robust correlations consistent with these predictions. Firms dealing with frontier technologies, facing heterogeneous environments or recently established are more likely to delegate authority. The predictions are intuitive and novel, and the authors note they had not previously been investigated empirically.
Context sharpens the contribution. Despite considerable interest in changes in firms' internal organization, there is no theoretical or empirical consensus on what determines organizational decisions or why there has been a significant move toward greater decentralization. The model offers one explanation: engaging with new technologies, firms delegate authority to managers holding superior information about implementation.
Artificial intelligence is a frontier technology whose implementation involves considerable uncertainty. The model predicts firms adopting AI will benefit from decentralizing authority to managers and teams who understand how to implement it in specific contexts. Centralized control may fail when headquarters lacks information about how AI works in practice. The team that has actually wired the model into a workflow knows things the board cannot.
Maturity changes the answer. As information about implementing AI becomes widely available, through best practices, standardized tools and accumulated experience, the model predicts a shift toward centralization. Firms may then manage AI from headquarters using public information. The right structure for AI adoption may change over the technology's life cycle. Today's empowered AI team may become tomorrow's standardized function.
Younger firms decentralize more, which fits AI startups' experimental, adaptive structures. Established firms adopting AI may need to delegate more authority than usual to capture the technology's benefits. An incumbent that runs AI adoption like a procurement exercise may be organizing for the wrong stage.
AI also rewrites the information balance itself. By giving headquarters better data and analysis, AI could make centralized control more effective, counteracting the tendency to decentralize when adopting new technologies. The net effect depends on whether AI's information benefits accrue to principals or to local managers. AI may be the first frontier technology that also strengthens the center's ability to manage it.
Organizational design matters for adoption. Firms aligning structures with new technologies' information requirements may adopt them more successfully. Policies supporting management quality and organizational flexibility can facilitate AI diffusion.
The evidence is robust: predictions tested across three datasets in two countries. Similar analyses of firms adopting AI could reveal how the technology is reshaping organizations.
Distilled to one claim: technology and information shape the allocation of authority within firms. New, frontier technologies favor decentralization to managers with superior information, while widely available information favors centralization. As AI diffuses, this framework predicts how firms will reorganize and guides how to organize to capture the technology's benefits.
One question to put to CEOs launching AI programs: decide in advance when authority will move back to the center. Decentralize while the know-how is local, and plan to recentralize once it is public; firms that never switch, in either direction, may be organized for the wrong stage.
Introduction
This paper analyzes the relationship between the diffusion of new technologies and the decentralization of firms. Centralized control relies on the information of the principal, which we equate with publicly available information. Decentralized control, on the other hand, delegates authority to a manager with superior information. However, the manager can use his informational advantage to make choices that are not in the best interest of the principal. As the available public information about the specific technology increases, the tradeoff shifts in favor of centralization. We show that firms closer to the technological frontier, firms in more heterogeneous environments, and younger firms are more likely to choose decentralization. Using three data sets on French and British firms in the 1990s, we report robust correlations consistent with these predictions.
Takeaway
Despite considerable academic and popular interest in the changes in the internal organization of the firm, we are far from a theoretical or an empirical consensus on the determinants of the organizational decisions of firms and on the reasons why there has recently been a significant move towards greater decentralization. In this paper we presented a simple model of the relationship between technology, information and decentralization, and empirically investigated the main predictions of this model using three micro-level datasets. In our model, firms delegate authority to managers, i.e., “decentralize”, in order to use the manager’s superior information about implementation of new technologies. Because the interests of the manager and the principal are not perfectly aligned, such delegation entails a costly loss of control for the principal. The model predicts that as available public information about the implementation of new technologies increases, firms should become less likely to decentralize, whereas firms dealing with new (frontier) technologies should be more likely to decentralize. We also showed that firms in more heterogeneous environments and young firms are more likely to choose decentralization. These are intuitive, but quite novel, predictions, and have, to the best of our knowledge, never been investigated empirically. We documented that in all three datasets the correlations are consistent with the predictions of our model. Firms in more heterogeneous environments and those that are closer to the frontier of their industry are more likely to choose decentralization.1
Moreover, consistent with the predictions of the theory, these results are stronger for firms in high-tech sectors. The results are robust to using UK values to instrument for French heterogeneity and proximity to frontier, alleviating some of the endogeneity concerns. They are also robust to using alternative measures of decentralization and heterogeneity. We also found that younger firms tended to be more likely to decentralize, although this result was less robust when we looked at alternative measures of decentralization. The theory and empirical results taken together suggest that learning and information accumulation may have important effects on the internal organization of firms, and may be especially important for decentralization decisions. It would be interesting to document the same issue using data from other countries, and also investigate whether the relationship between proximity to frontier or heterogeneity and decentralization correspond to the causal effects of the these variables on the internal organization of the firm. Specifying and estimating a more structural model would be a fruitful approach for this purpose. Finally, our approach suggests a natural reason for cross-country differences in the internal organization of firms; we may expect less decentralization in developing countries where most firms use well-established (rather than frontier) technologies. Theoretical and empirical analysis of cross-country patterns of organizational forms and their implications for economic performance is another promising area for future research.
Notes
- D. Acemoglu et al., Technology, Information, and the Decentralization of the Firm, 122 Q.J. Econ. 1759 (2007), https://doi.org/10.1162/qjec.2007.122.4.1759. Takeaway source: Conclusions, reproduced from a free author or repository copy (working-paper version). ↩
Cite as: D. Acemoglu et al., Technology, Information, and the Decentralization of the Firm, 122 Q.J. Econ. 1759 (2007), https://doi.org/10.1162/qjec.2007.122.4.1759
Header image: Korbel Champagne Cellars - August 2024 - Sarah Stierch 14.jpg by Missvain, CC0, via Wikimedia Commons.
