Suspension of the FCPA and the Enforcement Hypothesis

President Trump’s Executive Order 14209, signed on 10 February 2025,Footnote ¹³ mandated that the DOJ cease initiation of any new investigations or enforcement actions under the FCPA for 180 days and revise its enforcement priorities.Footnote ¹⁴

Since it first passed the US Congress in 1977, the FCPA had made it illegal for US citizens or companies investing abroad to bribe foreign public officials, subjecting them to punishment in the US for acts committed overseas—a legal principle known as extraterritoriality.Footnote ¹⁵ Twenty years later, the FCPA became the cornerstone of the global regime against corporate corruption, centered on the 1997 OECD ABC.Footnote ¹⁶ FCPA punishments can include colossal fines, such as the USD 1.6 billion fee imposed on Siemens in 2008 (in coordination with German authorities) for bribing public officials in multiple countries and, in 2020, Goldman Sachs’s USD 2.9 billion assessment for facilitating bribery in Malaysia.Footnote ¹⁷ There is substantial empirical evidence that the FCPA successfully reduced bribery as well as investment in and trade with corrupt regimes by US firms and by foreign firms with activities in the US.Footnote ¹⁸

Donald Trump had long been an opponent of the bill, stating his intention to repeal it and referring to it as a “horrible law” in a television interview during his 2016 campaign.Footnote ¹⁹ Trump charged that the FCPA placed American companies at a disadvantage in international business transactions vis-à-vis competitor firms who do not face the same onerous regulations.Footnote ²⁰ Yet throughout Trump’s entire first term, the FCPA remained untouched and operational. Consequently, the decision to suspend the FCPA at the beginning of his second term was unexpected,Footnote ²¹ appearing among a wave of executive orders and actions reshaping US domestic and foreign policy priorities.Footnote ²²

Studying the immediate financial impact of Trump’s reversal of US anti-corruption priorities provides a unique window into how investors understand the threat of international anti-corruption enforcement. Consequently, we analyze the effect of the executive order on stock market evaluations of US-traded companies that have been cited as (potential) FCPA violators in the past.

Two reasons make FCPA histories a useful heuristic for investors to infer future corruption risk. First, FCPA targets are often repeat offenders involved in multiple anti-bribery cases over time. Our data (Appendix A) show that more than 40 percent of all FCPA actions are followed by another action targeting the same firm within ten years. This is explainable in terms of firms’ “type” (for example, firms operating in certain industries or countries are more likely to be involved in bribing) and also based on the cost—in terms of evidence—for prosecutors to launch new enforcement actions. Prosecutors might reduce the evidence costs of enforcement with information emerging from past actionsFootnote ²³ because investigations may uncover new information leading to new cases.Footnote ²⁴ Similarly, past targets’ adoption of compliance and internal monitoring programs often surfaces evidence regarding future FCPA violations. This makes it likely that an FCPA target will risk prosecution or investigation again.

Second, FCPA histories also inform market analysts’ measures of reputational risk exposure. In Appendix B, we show that the average reputational risk score of a past FCPA target is about 28 percent greater than that of a comparable nontainted firm. This increase in risk emerges immediately after an FCPA action and persists over several years. Because of such long-term reputational effects, investors might deem firms with an FCPA history as more risky of future corruption. These considerations lead us to believe that investors use FCPA histories as a useful heuristic to gauge future corruption risk. Given the long-term effect of FCPA histories on repeated enforcement and reputational risk, contextual factors—such as how long ago the FCPA action happened—should not be more relevant heuristics than the involvement in an FCPA case itself.Footnote ²⁵

We suggest that financial markets welcomed Trump’s executive order as a signal that corrupt business activities—such as the risky ones entertained by past FCPA targets—would be less costly in the future because the order reduced the probability of bribery investigations, punishment, and fines.Footnote ²⁶ That is:

Sample Construction and Data Collection

We expect stock market investors to interpret Donald Trump’s unexpected suspension of FCPA enforcement as a sudden lifting of law-enforcement-related corruption costs for firms. If investors consider the FCPA costly for its expected enforcement costs, but less so for the burdens of regulation and legal compliance, investing in firms that likely engaged in corrupt activities should have become less risky, and more profitable, after the announcement. If, instead, the regulatory burden of the FCPA hampered the competitiveness of all foreign-operating firms, investment in all MNCs should have increased. This expectation can be empirically tested by considering whether stock symbols for firms “at risk of corruption” appreciated after the executive order relative to the rest. This would provide evidence that suspending FCPA enforcement reduced political risk for these firms by green-lighting corrupt behavior.

We start by identifying the firms that investors would consider at risk of corruption (and FCPA action). Measuring foreign bribery risk is complex because corruption occurs out of the sight of stock market investors. However, as we have mentioned, past FCPA enforcement actions or investigations are a valuable heuristic for investors to infer a firm’s future corruption risk, due to the repeated exposure of past targets to future FCPA actions (Appendix A) and to the reputational risk induced by FCPA actions (Appendix B).Footnote ²⁷ We obtained the exhaustive list of publicly traded firms that faced FCPA enforcement actions or investigations in the past from Stanford’s FCPA Clearinghouse.Footnote ²⁸ We manually corrected entries due to name and corporate-structure changes since the initial FCPA action.Footnote ²⁹ Some firms were later renamed and changed their ticker symbols (for example, Norwegian oil company Statoil, formerly STO on the NYSE, is now Equinor, EQNR on the NYSE). Others underwent mergers and are now part of different corporate groups: Fiat is now part of Stellantis; Zimmer and Biomet, formerly independent (and traded as ZMH on the NYSE and BMET on the NASDAQ, respectively), are now Zimmer Biomet Holdings (ZBH on the NYSE). We kept firms in our sample that had been renamed or merged with other firms but were still publicly traded. However, we removed cases involving delisted, bankrupt, or defunct firms—such as Aegerion Pharmaceuticals, which merged with QLT in 2016 to form Novelion Therapeutics but filed for bankruptcy in 2019—or firms that have been acquired and are now entirely owned by other companies. Finally, we removed firms whose trading was suspended on US exchanges due to stock market regulations or geopolitical events—for example, trading of American depositary receipts (ADRs) of the Russian Mobile TeleSystems was suspended after the invasion of Ukraine in February 2022. The result is a list of 286 publicly traded firms that were involved in at least one FCPA action.

Our first sample is the 261 firms that traded securities on a US stock exchange and were targeted in a past FCPA case. From the list of 286 publicly traded targets with FCPA histories, this sample discards 24 firms that trade securities on foreign exchanges (as reported by Compustat).Footnote ³⁰ We also discard one company trading penny stock (Corsa Coal).

We built a second sample of firms that were not involved in FCPA actions in the past, which our statistical analysis will use to discount the effect of Trump’s FCPA executive order from general market trends. From Compustat, we gathered the Standard and Poor (S&P) 500 as of February 2025 (that is, the 500 largest companies by capitalization that trade on US markets). Excluding the 89 firms that were also involved in an FCPA action (and thus belong to the first sample of companies) brought the 500 down to 411.

Finally, we built a “placebo” sample of firms that are comparable to past FCPA targets but have never experienced an FCPA action. We use this portfolio to substantiate our claim that we are estimating the effect of the executive order on the political risk of tainted firms, as priced by investors, rather than a generic reduction of FCPA legal-compliance-related costs for similarly situated multinational firms. From the Compustat list of firms trading their securities on North American stock exchanges, we excluded past FCPA targets. From the remaining firms, we used propensity-score matching to select a portfolio of 236 firms that are as similar as possible to past FCPA targets, based on observable covariates.Footnote ³¹ Except for not having ever been involved in an FCPA action, this matched placebo sample is significantly similar to past FCPA targets in terms of foreign activity via subsidiaries, financials, and industry characteristics.Footnote ³² For instance, past FCPA targets and placebos operate, on average, in eighteen and nineteen foreign jurisdictions, respectively, and own fifty-eight and fifty-six foreign subsidiaries. They are also similar in terms of financials, with about USD 123 billion and USD 106 billion, respectively, in average assets. Finally, they are similar in terms of industry, equally representing all considered sectors, including those with high corruption exposure, such as mining, construction, heavy manufacturing, and pharmaceuticals.

For each firm, we gathered stock prices at closing on each trading day from 3 June 2024 (180 trading days before Trump’s order) to the end of the trading week of the announcement (14 February 2025). We use the price at closing on a trading day to compute our outcome variable of interest: returns, defined as the percentage change in stock price at closing between two trading days.Footnote ³³

Research Design

We apply the standard two-window event study from corporate finance. Here, we describe our procedure for the sample of past FCPA targets, but we apply the same design to the matched placebo sample. The goal of the design is to estimate each firm’s counterfactual returns after Trump’s FCPA executive order.Footnote ³⁴ We compare observed and counterfactual returns to estimate whether (and in what direction) the event affected risk as market participants priced it. To this aim, we divide returns to each firm $i$ into two time windows: an “estimation window” before the event and an “event window” after it (Figure 1). The estimation window spans the days $t \in \left[ {{t_0},{t_1}} \right)$ ; the event window covers the days $t \in \left[ {{t_1},{t_2}} \right]$ and includes Trump’s FCPA order ( ${t_{\rm{e}}}$ , 10 February 2025).

Figure 1. Research design: estimation and event windows

For each firm in our sample, we estimate counterfactual returns in the event window using information from the estimation window, which entirely predates the event. Using data from $t \in \left[ {{t_0},{t_1}} \right)$ , we estimate a model of daily returns (“market model”) for each firm in the sample of past FCPA targets, as a function of predictors ( ${\bf {\it X}}$ ). Once a market model is estimated for each firm, we use it to predict firms’ returns over the full period in consideration, $t \in \left[ {{t_0},{t_2}} \right]$ . Equations 1 and 2 represent these two steps. When looking at the estimation window— $t \in \left[ {{t_0},{t_1}} \right)$ —the prediction is useful to assess the quality of our models (e.g., in terms of ${R^2}$ ). When stretched out-of-sample and into the event window— $t \in \left[ {{t_1},{t_2}} \right]$ —this prediction offers returns to a firm $i$ that are counterfactual to the event, under the assumption that no other unrelated shock impacts only the subset of firms of interest.

(1)

(2)

Traditional applications use aggregated market indexes—such as the S&P 500 or the Dow Jones—to build a matrix ${\bf {\it X}}$ . However, these indices themselves are composed of market movements of companies that might be impacted by Trump’s FCPA announcement. For instance, eighty-nine (18 percent) of the firms in the S&P 500 composite index are themselves past FCPA targets. Therefore, using aggregate indices could bias counterfactual returns. To overcome this, we follow the suggestion of Meredith Wilf.Footnote ³⁵ We construct a matrix ${\bf {\it X}}$ using individual returns to each of the 411 S&P 500 firms that have never been targets of an FCPA action. Using these 411 stock movements to construct counterfactuals allows us to discount broader market trends that affected traded companies in February 2025. Because our estimation has more predictors than observations (we have at most 180 trading days for each market model), we use the least absolute shrinkage and selection operator (LASSO)Footnote ³⁶ to select which S&P 500 firms are the most predictive for each past FCPA target. LASSO excludes nonpredictive S&P 500 firms from each market model.Footnote ³⁷ Effectively, for each past FCPA target, we generate a specific index that is the most predictive of stock returns by weighing the S&P 500 constituents that yield the best fit in the estimation window.

When performing this estimation, we choose to employ LASSO (and the 411 individual firms’ returns as predictors) as opposed to estimating Equation 1 using ordinary least squares (OLS) and an aggregated market index (such as the S&P 500 composite). We arbitrarily choose the length of the estimation window $\left[ {{t_0};{t_1}} \right)$ . Also, we choose an arbitrary number of LASSO cross-validation (CV) folds. We tested fifteen possible sets of market models resulting from all combinations of these arbitrary choices—we considered OLS and LASSO, estimation windows of 180, 90, or 30 days, and LASSO CV folds of 3, 5, 10, or 15—and selected the single combination that yields the best prediction (in terms of average ${R^2}$ ): the LASSO implemented on an estimation window starting thirty days and ending five days before the event, with 15-folds CV (average ${R^2}$ of these market models is 0.673 for past FCPA targets, 0.669 for matched placebos). In Appendix D, we fully describe our estimation results and show the ${R^2}$ distribution yielded by these fifteen combinations. In Appendix H we demonstrate that our results are robust to varying these arbitrary choices.

Once LASSO market models are estimated, we focus on event-window data— $t \in \left[ {{t_1},{t_2}} \right]$ —and obtain two measures for each firm: abnormal returns (AR) and cumulative abnormal returns (CAR), as shown in Equation 3. AR are daily differences between observed and expected returns, representing the residuals between observations and counterfactuals. car are the running daily sum of AR for a firm after the event. They help assess whether a positive (negative) event effect accumulates to a significant gain (loss).

(3)

The final step is to study the daily evolution of average AR and CAR. We do so by calculating event-window daily average AR and CAR and testing whether these averages are statistically distinguishable from zero. In Appendix J, we complement this approach with firm fixed-effect linear regressions and parametric and non-parametric tests proposed by the corporate finance literature. Across our statistical tests, we impose a level of significance of 0.05.

Robustness Tests

We perform several robustness tests. In Appendix H, we show similar results when changing the arbitrary choices adopted in our estimation phase (estimation window lengths, number of cross-validation folds, and the choice of LASSO over OLS). In Appendix I, we find similar results when excluding firms with poorly estimated counterfactuals; one firm at a time from the samples of past FCPA targets and placebo firms; the ten past FCPA targets who realized outlier-sized gains; firms trading ADRs over the counter; and past FCPA targets whose ownership history we manually reconstructed; and when we limit the sample of past FCPA targets to those that are matched to the placebo firms (that is, those for whom we have covariate information). In Appendix J, we show similar findings when adopting linear regression models with firm fixed effects, industry fixed effects, or other parametric and nonparametric tests proposed by corporate finance to account for event-induced changes in the variance of returns. Appendix K demonstrates that we obtain comparable results if we select a placebo group of non-FCPA targets with matching techniques other than propensity scores (coarsened exact matching and entropy balancing). In Appendix L, we show that our findings do not even hinge on the chosen research design: we find similar effects with a difference-in-differences and with a generalized synthetic control method,Footnote ⁴⁵ where we use matched placebo firms as a control group. Finally, in Appendix M we explore heterogeneous effects for the sample of past FCPA targets. We distinguish between whether the past FCPA targets were subject to an enforcement action or just to an investigation; we distinguish past FCPA targets based on whether they changed their names since the FCPA action; and we study how estimated effects varied based on the last time a firm experienced an FCPA action. We find no evidence of heterogeneous effects according to these features, a null effect which indicates that investors largely use FCPA histories, regardless of contextual factors, as a heuristic for future corruption risk.

We dedicate a set of final considerations to the significant effect detected on 4 and 5 February, in conjunction with Pam Bondi taking office and circulating a memo to DOJ employees redirecting enforcement priorities. We argue that this memo, although it provided price-relevant information to investors (hence the early effect detected for past FCPA targets and matched placebo firms on those days), did not clearly anticipate the decision to suspend FCPA enforcement in the upcoming executive order. The FCPA is the subject of only two paragraphs of the five-page memo, which rather aimed at reshaping all DOJ priorities (not just FCPA enforcement) to “pursue [the] total elimination of Cartels and Transnational Criminal Organizations (TCOs)” (emphasis in the original). In the paragraphs concerning the FCPA,Footnote ⁴⁶ the memo never alludes to the possibility of halting enforcement. On the contrary, Bondi orders the DOJ to expedite enforcement of the FCPA for cases of bribery associated with cartels and TCOs, which would no longer require authorization for investigation by the Criminal Division or the involvement of the Fraud Section. That is, the memo indicates a shift in priorities (from corporate crime in and of itself to corporate crime linked with TCOs) but does not suggest a halt to enforcement. The positive effect detected on those two days for both samples should thus be understood as a result of this broader deregulatory shift at the DOJ in conjunction with Bondi taking office, not as an anticipation of the suspension of FCPA enforcement. Confirming this interpretation, and in contrast to the effect of the FCPA halt, we detect a comparable positive effect for both past FCPA targets and untainted placebo firms on the day the memo was issued.