Partisan bias in securities enforcement

Accusations that some federal agency has acted with partisan bias in enforcing the law—treating partisan allies more favorably than enemies, or enemies more harshly than allies—are commonplace in American politics today. Critics have leveled such charges at the Department of Justice, the Federal Bureau of Investigation, the Internal Revenue Service, and even the independent agency that regulates the nation’s securities markets, the Securities and Exchange Commission or SEC. Moreover, these accusations have landed on officials from both sides of the aisle. When Donald Trump was President, critics accused the Department of Justice of going easy on President Trump’s political allies, as reflected by the dropping of charges against Michael Flynn and weakening of the sentencing recommendation for Roger Stone. After Joe Biden became President and Republicans took the majority in the House, congressional Republicans formed a “Select Subcommittee on the Weaponization of the Federal Government” to investigate alleged partisan bias against Republicans.

Despite the frequency of such allegations, however, there has been limited effort to systematically test them. Although numerous studies in political science have found evidence of partisan favoritism in federal spending and contracting (e.g., Dynes and Huber 2015), the few studies that have investigated potential partisan bias in law enforcement and administration have focused on highly salient political settings, namely public corruption cases (Gordon 2009; Nyhan and Rehavi 2018; Davis and White 2021) and election administration (Porter and Rogowski 2018). It remains an open question whether partisan bias influences agency enforcement outside those overtly political settings. Leveraging data on the partisanship of the targets of over a decade of SEC investigations and enforcement actions, I seek to correct that omission.

It is not immediately obvious how partisan bias should manifest in observed enforcement outcomes. To clarify expectations about what partisan bias should look like empirically, I develop a formal model of a bargaining game between a lower-level bureaucrat and a political appointee about whether to bring an enforcement action. In the model, the penalty-maximizing bureaucrat has the exclusive ability to propose an enforcement action, and the appointee, who generally likes higher penalties but is biased in favor of copartisans, has veto authority over the action. The model suggests two potential channels for partisan bias to creep into agency enforcement: in the selection of targets for enforcement and in the penalties they are assessed. In theory, bias could manifest in either place, with predictions about where it should surface depending on the maximum penalties that could be imposed in potential cases. But, all else being equal, two things are clear: partisan bias should reduce the ex ante likelihood of enforcement and ex ante expected penalties against allies relative to enemies.

To test these predictions empirically, I marshal data on the partisanship of the targets of SEC investigations and enforcement actions across multiple presidential administrations. Connecting these data with changes in partisan control of the SEC, I use a difference-in-differences-like approach to examine three different potential channels for partisan bias: the initial selection of subjects for investigation, the final selection of targets for enforcement, and the assessment of monetary sanctions against offenders. This research design leverages variation in partisan control of the SEC over time—which during my sample period changes only following relatively close presidential elections and thus is plausibly exogenous to the SEC’s enforcement activity—and variation in firm partisanship across units to assess the impact of partisan alignment on enforcement outcomes.

My results paint a nuanced picture of partisan bias in securities enforcement. In the initial openings of investigations—where career civil servants have authority to open investigations without the approval of political appointees—I find little evidence of partisan bias. Leveraging the transition from the Bush to the Obama administration, my estimates suggest that a change in a firm’s partisan alignment with the SEC has little effect, if any, on the likelihood that the SEC investigates the firm. For the enforcement stage, by contrast, my results provide evidence consistent with substantial partisan bias in SEC enforcement. Here, relying primarily on the transition from the Obama to the Trump administration, I estimate that for a firm that would ex ante be equally likely to be targeted for enforcement or not, a typical increase in partisan alignment with the SEC following a change in party control of the agency reduces the likelihood of enforcement by approximately 19.2% or 23%, depending on the specification. A similar increase in partisan alignment reduces the expected number of enforcement actions by an estimated 22.4% or 33.5%, depending on the specification. Partisan alignment appears to reduce aggregate monetary sanctions as well, in the neighborhood of 3% for a similar increase in alignment, though this estimate is not statistically significant at conventional levels in some specifications.

To probe possible alternative explanations for these results, I run a series of additional empirical tests, focusing on explanations based on industry, geography, programmatic politics, and strategic timing by bureaucrats. My results suggest that these alternative explanations are implausible.

As I explain greater detail in the article, these findings have important implications for democratic governance. Normatively, one might worry that partisan bias in agency enforcement undermines fairness, efficiency, or both. Such concerns call for careful attention to the details of agency design and may help to inform ongoing debates in the courts and among legal scholars about how to best calibrate administrative law so as to safeguard against the arbitrary exercise of administrative power.