Legally Charged: Embeddedness and Profit in Large Law Firm Legal Billings

Abstract

We examine how forms of a firm's embedding in market relationships affect the size of its spreads – i.e., the difference between the selling price and production costs of its goods and services. Building on Harrison White's work on the relational underpinning of market behavior, we argue that the embeddedness of market transactions in social structures furnishes actors with private information and informal governance benefits that shape spreads by adding unique value to transactions and by revealing the price sensitivity of clients. We propose arguments about how a firm's embedded client relationships, interlock ties, and status influence the size of its spreads. Using longitudinal data on the economic and sociological characteristics of law firms that represent the Fortune 200 corporations and top 250 financial firms in America, we find that social structure has significant effects on spreads and that the effects change in scale and direction, depending on the form of embeddedness.

Acknowledgments

We would like to thank the Russell Sage Foundation and the Department of Sociology, Princeton University, for their financial assistance and Frank Dobbin and three anonymous reviewers for the helpful comments on this research.

Notes

¹Figures underestimate total spending by excluding the large numbers of in-house counsels working in business and government.

²This intensive data collection and processing does not presume that our decision-makers are hyper-rational. Rather, we account for relatively reasonable actors who, in making a decision, search out information, categorize it, and use it to make inferences consistent with Simon's behavioral decision theory (Tversky and Kahneman 1974). While it is rare for decision makers to calculate regression-like models, they aspire to this ideal by collecting information that would make such calculations possible. They attend to both private and public information, while remaining aware that their inferences contain error.

³This methodology produces a directory of best-informed expert evaluations of other experts, which is distinguished from other rankings such as Who's Who in America and The Social Register, which conflate quality and reputation because they base inclusion on non-performance related factors such as birth position, political activities, peripheral involvement in cases, etc.

⁴We checked the validity of our results by running several different models and looking for convergence. A fixed effects time-series model estimated coefficients of similar size to our random effects model, but produced standard errors that were three to four times larger than the random effects model. These estimates suggest that the parameter point estimates of the random effects model are accurate, even if the fixed effects model lacks enough power to meaningfully test hypotheses on its own. We also ran an autoregressive function with one-time period lags to control for price stickiness. This model also yielded similar point estimates for our reported coefficients. However, because our panels were unbalanced and unevenly distributed, our data violate the assumptions of these models, so we do not report them here. Because all models converged, we report the model with the assumptions that most appropriately fit our data.