Exploring the effects of allegations of sexual misconduct on political careers

ABSTRACT

Recently, many prominent public figures have been accused of sexual misconduct, with American politicians being no exception. In light of the prominence and deeply personal nature of sexual misconduct allegations, we explore the effects of these allegations on political careers. Pairing original, externally validated data with rich existing data, we explore the effects of allegations of sexual misconduct on political careers for politicians at a variety of levels of politics, across political parties and branches of government, and over a long period of time. Through a variety of statistical tests and simulations, we demonstrate that as time post-allegations increases, the likelihood of politicians losing their seats increases exponentially, regardless of surrounding context. This ultimately suggests that the effects of sexual misconduct allegations on political careers are pronounced and often terminal.

KEYWORDS:

• Sexual misconduct
• political careers
• American politics

Notes

¹ E.g., Sen. Al Franken’s (D–MN) Senate press conference on C-SPAN on 27 November 2017 (Franken, 2017).

² E.g., Rep. Vance McAllister’s (R–LA) personal appeal for forgiveness from family, friends, and God for reports of sexual misconduct (Hilburn, 2014).

³ The Cox model is preferable to other common duration models, such as the Weibull or Gompertz, for several reasons. Most notably, it is semi-parametric, where a functional form is specified for the regressors (which here is exponential), but there are no assumptions made about the baseline hazard, ${\gamma }_0\left(\mathit{t}\right)$. This is assumed to be unknown, as opposed to the fully parametric approaches, which parameterize the baseline hazard. Such a parametric approach requires numerous additional assumptions, which are avoided when estimating the Cox model. Further, the semi-parametric nature of the Cox model allows for comparisons of model fit with and without regressors included to assess the shift in relative hazard (and survival probability) for each specification. This is shown more clearly below in Figure 1. Also, the Cox model is frequently used in social and political research, where extensions such as the one leveraged in this analysis from Katz and Sala (1996) allow for greater flexibility in addressing a breadth of questions.

Figure 1. Model results without and with regressors.

⁴ E.g., E.J. Dionne’s coverage of Rep. Mark Souder’s (R–IN) allegations of sexual misconduct (Dionne, 2010).