Peer Effects and Corporate Corruption

The following post comes to us from Christopher Parsons of the Finance Area at the University of California, San Diego; Johan Sulaeman of the Department of Finance at Southern Methodist University; and Sheridan Titman, Professor of Finance at the University of Texas at Austin.

Traditional models of crime frame the choice to engage in misbehavior like any other economic decision involving cost and benefit tradeoffs. Though somewhat successful when taken to the data, perhaps the theory’s largest embarrassment is its failure to account for the enormous variation in crime rates observed across both time and space. Indeed, as Glaeser, Sacerdote, and Scheinkman (1996) argue, regional variation in demographics, enforcement, and other observables are simply not large enough to explain why, for example, two seemingly identical neighborhoods in the same city have such drastically different crime rates. The answer they propose is simple: social interactions induce positive correlations in the tendency to break rules.

Our paper, The Geography of Financial Misconduct, extends this literature by examining geographical patterns in white-collar crime. We begin by documenting a strong geographic effect: the average rates of financial misconduct vary substantially across U.S. cities, and over time within these cities. Then, in the second part of the paper, we seek to better understand the mechanism. In particular, we try to distinguish between managers being subject to common local influences like enforcement, and managers influencing each other’s behavior through peer effects.

Our benchmark analysis shows that among the largest twenty cities in the U.S., financial misconduct is exposed at dramatically different rates. For example, averaged over 1970-2010, about 1 in 190 firms headquartered in Indianapolis, Seattle, and Minneapolis are prosecuted for misconduct in a typical year, whereas firms based in Dallas (1:62), St. Louis (1:61), and Miami (1:60) are investigated nearly three times as often. Virtually none of these differences is due to industry clustering within cities.

With these basic patterns established, we then attempt to shed light on the mechanism, drawing heavily on Manski (1993). The first possibility is that, for lack of a better term, cities differ in terms of the “types” of their inhabitants, owing to long-standing factors like cultural origin (e.g., Minnesota being home to many Scandinavian descendants), wealth, or religion. In the second alternative, what differs across cities is not so much the people, as much as their local environment, such as economic conditions or enforcement. Finally, the propensity for financial misconduct may spread within a region via peer effects, or more specifically, through interpersonal interactions.

Our attempt to distinguish between these mechanisms starts by arguing that exogenous differences among city cultures is, at best, an incomplete explanation. Specifically, we extend our analysis to include time-series variation within cities and show that time series movements in the tendency to be prosecuted for financial fraud have strong regional patterns. That is, after accounting for (say) the fact that firms headquartered in Seattle have lower than average fraud rates over our sample, we find that Seattle’s food and beverage providers (Starbucks), online retailers (Amazon), senior living providers (Emeritus), and software firms (F5 Networks) tend to commit financial misconduct during the same times, despite operating in very different lines of business. Static or slow-moving regional factors provide a poor account of such dynamics.

We next examine the potential impact of environmental factors. One possibility is that a city’s prospects—think Detroit versus San Francisco—may influence a manager’s incentives to invest in reputation or social capital. Yet, we find little relation between financial misconduct and measures of city health like population or income growth, and more importantly, their inclusion does not attenuate the effect of a firm’s local peers. A second possibility is regional fluctuation in enforcement efforts. Although this provides a good explanation for why financial misconduct may be exposed simultaneously within a region (which we also observe), it does not account for coordinated initiations of financial misconduct. Indeed, the first years of financial misconduct occurrences are highly clustered within regions, even for those that are later detected at different times.

Our final approach involves tests that rule out any generic regional factor by construction. We begin by identifying metropolitan areas containing a single, dominant industry, such as Houston (energy) or San Francisco (software). Then, we use industry-level variation as an instrument for the fraud rates of firms in these dominant industries, e.g., using the fraud rates of Oklahoma’s Chesapeake Energy to instrument for Houston’s Apache. The final step is to relate fraud rates of firms outside the dominant sector (e.g., a pharmaceutical firm in Houston) to the instrumented fraud rates of the city’s dominant industry players. That we find a strong relation here is difficult to reconcile with local environmental effects of any form.

Further analysis lends direct support to the idea that peer effects between managers are, at least in part, responsible for the observed regional correlations in financial misconduct. First, we divide a firm’s local neighbors into groups, based on the: 1) similarity of market capitalization, and 2) similarity of the CEO’s age. The idea of each is to identify proxies for the strength of local interaction. Matching on size seems intuitive (even across industries), given that the largest firms in an area—think Google, Wells Fargo, and Genentech in the Bay Area—are likely to share linkages on corporate boards, civic organizations, and so forth. The intuition behind age matching is similar, i.e., a CEO in his 40s is more likely to socially interact with CEOs in the same age range, versus those thirty years his senior.

Confirming this intuition, we find striking results. Large firms are sensitive only to the financial misconduct of other large local firms; likewise, small firms are sensitive only to the behavior of other small local firms. The results are somewhat weaker for the age-matching results, though young CEOs are roughly twice as sensitive to the behavior of other young (local) CEOs, versus their more mature counterparts.

Our second test considers the possibility that CEOs—particularly those of large public firms—likely interact regularly with elected officials. If so, and if ethical norms are transmitted through social interactions, then we might expect misbehavior in the corporate arena to correlate with misbehavior of public servants. Not only is this true on average across cities, but also over time within each city. This relation is particularly strong for large firms, whose executives are most likely to interact with elected officials. As before, such regional ebbs and flows are difficult to explain with static regional factors, moreover, because there is little (though not zero) overlap between the relevant enforcement bodies, we interpret this as additional evidence against regionally correlated enforcement driving the results.

The full paper is available for download here.

 

Both comments and trackbacks are currently closed.