Authors: Banks Miller, Brett Curry, and Maureen Stobb
Publication: Social Science Quarterly, published online 2026
Original source: ssq_presidential_pardons_2026.pdf
Conversion note: this Markdown was generated by text extraction from the hosted file. Review equations, tables, notes, and citations before treating it as a final web copy.
| Social Science Quarterly ORIGINAL ARTICLE The Politics of Presidential Pardons: Partisan Differences in Modern Clemency Decisions Banks Miller1 Brett Curry2 Maureen Stobb2 1 University of Texas at Dallas, Richardson, Texas, USA 2 Georgia Southern University, Statesboro, Georgia, USA Correspondence: Brett Curry ( bcurry@georgiasouthern.edu) Received: 14 October 2025 Revised: 6 January 2026 Accepted: 19 January 2026 Keywords: criminal justice policy | executive power | issue ownership | political risk | presidential pardons |
Objective: We investigate how partisanship, political context, and recipient characteristics relate to presidential pardon behavior. We ask whether Democratic and Republican administrations exhibit unique patterns in the likelihood and timing of pardons across different crime categories. Methods: We analyze presidential pardons issued from 1989 to 2024. Leveraging data on the crimes for which pardons are granted, recipient characteristics, and additional factors, we estimate logit and count models controlling for demographic characteristics, crime type, and political dynamics. We also incorporate interaction terms to examine how partisan differences vary across such contexts. Results: Our results suggest that presidents utilize pardons to advance criminal justice priorities while navigating political constraints. Democratic and Republican administrations behave differently in the likelihood and timing of pardons in particular crime categories. Democrats are more likely to pardon non-Whites, women, and narcotics offenders than Republicans, but are less likely to grant pardons when facing potential electoral threat. As Democratic presidents increase their rhetorical attention to crime, they grow less likely to issue pardons. Conclusions: Our findings speak to issues surrounding executive clemency, issue ownership, criminal justice policy, and partisan governance. These patterns demonstrate that pardons can operate as policy tools shaped by ideological preferences and political constraints. 1 Introduction The past year has been a notable one for presidential pardons. When former President Biden pardoned his son Hunter in late 2024, he characterized the legal process surrounding his convic- tion as a “miscarriage of justice”(NPR 2025 ). The special counsel in the case, David Weiss, vehemently disagreed and accused the President of undermining “the very foundation of what makes America’s justice system fair and equitable” (NPR 2025 ). Less than 2 months later, President Trump defended his pardon of January 6 rioters by arguing they had spent several years in prison under “disgusting” and “inhumane” conditions Richler et al. ( 2025 ); his critics accused him of perpetrating grievous harm to the American justice system (e.g., Reilly 2025 ). These recent examples highlight the controversial nature of the presidential pardon power, a largely unlimited executive authority designed to provide mercy and correct judicial errors. While the power faces few formal legal constraints, presidents may sometimes be deterred from granting controversial pardons by the prospect of political consequences (Kassop 2004 , 73; e.g., Eckstein and Colby 2019 ). Since George Washington first exercised it in 1795 to grant amnesty to participants in the Whiskey Rebellion (Shogan 2020 ; ©2026 Southwestern Social Science Association. Social Science Quarterly , 2026; 00:e70133 https://doi.org/10.1111/ssqu.70133 1of11
Ruckman 1997 ), this executive authority has represented a pow- erful and politically charged tool presidents can wield. It has grown with the size of the federal criminal code, from just a handful of offenses in the republic’s early days to more than 4500 today (Fields and Emshwiller 2011 ). This larger scope likely incentivizes presidents and other relevant actors in their administrations to be more systematic in the pardon power’s application, using it as a tool to manifest their policy priorities. As one example, President Barack Obama ( 2017 , 836) instructed officials in his administration “to look more systematically at how clemency could be used to address particularly unjust sentences in individual cases.” In part, the former president rooted his approach in a desire to advance criminal justice reform and to “make it easier for future Presidents, governors, and other public officials” to use the pardon power for good (Obama 2017 , 838). This echoes the Hamiltonian view of the pardon as an executive tool to be exercised with an eye toward public policy objectives (Erler 2007 , 429). In many ways, the pardon is a signal of presidential preferences about criminal justice policy. It conveys priorities and philosophies in the area of criminal justice policy, where presidents differ in their approaches to law enforcement, sentencing, and criminal justice reform, and utilization of the pardon is likely to reflect those approaches. Though we focus primarily on how pardons may systematically reflect criminal justice policy priorities, we certainly acknowledge that individual presidential clemency decisions can be motivated by a variety of factors. These can include the personal (friends and family members), the partisan (political allies), the symbolic (historically wronged individuals), responses to campaigns from advocacy groups, or some combination thereof. By approaching this study with the particular goal of examining how criminal justice policy priorities, and their differential political costs across party lines, shape systematic patterns in modern pardon behavior, our aim is to identify important partisan differences in pardon patterns even as we recognize additional factors outside that analytical scope may influence individual clemency decisions. A more comprehensive understanding of presidential pardon power will ultimately require scholars to undertake additional research exploring these alternative motivating factors and how they might interact with the policy-oriented patterns we uncover here. While prior evidence indicates that the use of this power depends upon the president’s agenda and ideological positions, congres- sional attention to criminal justice issues, and broader societal conditions, we do not know if these patterns vary systematically across presidential administrations in ways that reflect different approaches to criminal justice reform. For example, President Obama emphasized “many of the most tragic failings of the justice system are disproportionately felt by communities of color” (Obama 2017 , 820). Do such demographic characteristics play a significant role in presidential decision-making across administrations? Do the types of crimes committed influence the likelihood of receiving a pardon? In addition, we need to understand the limits to presidential reform efforts through the use of the pardon power. Do presidents temper their actions based on potential backlash from the other branches and the electorate? We seek to address these questions while acknowledging that presidents may vary somewhat in their approaches to the par- don process. President Donald Trump’s first term was unique among modern presidents in that he routinely circumvented the bureaucratic channels responsible for screening applications (see Goldsmith and Gluck 2020 ; Crouch 2022 ). Some presidents have been more politically cautious than others in extending par- dons (Crouch 2022 ); and even where traditional review processes are present, they can sometimes be influenced if not overridden by personal connections—the pardons of Caspar Weinberger and Mark Rich by, respectively, Presidents Bush and Clinton, come to mind (Crouch 2009 ). Still, we anticipate that presidents systematically utilize their pardon power to advance their criminal justice policy preferences, even as they remain mindful of the political context in which they are operating. While Landes and Posner ( 2009 ) and Ruck- man ( 1997 ) document aggregate partisan differences in pardon rates, we advance understanding of presidential clemency in three additional ways. First, we disaggregate pardons by crime type and show that partisan differences are not uniform but vary systematically across offense categories—most notably, with respect to narcotics and white-collar crimes. Thus, Republican and Democratic presidents focus on different threats and areas of social concern. Second, we examine how recipient demo- graphics matter differently across administrations—Democratic presidents are considerably more likely to pardon non-Whites and women than their Republican counterparts. This demonstrates variance in systemic reform priorities. Third, we examine timing: Do certain offenders wait longer than others for pardons? The answer to this question speaks to presidents’ sense of the urgency of action on each specific policy priority. Our results indicate that partisan differences run deeper than simple disagreements about how many pardons to grant—they reflect competing visions of criminal justice priorities. Our paper proceeds as follows. First, we cover the mechanics of modern presidential pardons while also devoting attention to their constitutional foundations and historical practice. We then summarize several important studies of pardons and, in doing so, provide important context for our study and its contributions. Next, we outline our theory and hypotheses, describe the data we have compiled to test them, and the methods we utilize in doing so. After presenting our results, we offer some concluding thoughts about their implications and possible directions for future research. 2 The Pardon Power: Constitutional, Historical, and Modern Context Article II, Section 2 , of the US Constitution articulates the presi- dent’s power to grant “reprieves and pardons”for offenses against the United States. Though it cannot invalidate impeachment, the presidential pardon is an otherwise wide-ranging power that is not contingent on acquiescence from any other part of government (e.g., Whitford and Ochs 2006 ). Drawing heavily on British practice and the “prerogative of mercy,” the clause was not subject to significant debate in the Founding era (Shogan 2020 ; Ruckman 1997 , 252–253). The US Supreme Court has said that the power can be exercised before criminal charges have been brought, is not subjected to congressional modification, and must be deferred to by the courts (e.g., Ex parte Garland 2of11 Social Science Quarterly,2026
1866 ; U.S. v. Klein 1871 ; Ex parte Grossman 1925 ; Morison 2010 ). According to a 2007 estimate, more than 25,000 presidential pardons have been issued since the Washington administration and, although history’s most contentious pardons have garnered disproportionate attention, the vast majority have been routine and uncontroversial (Crouch 2022 , 694). For well over a century, presidents have been assisted by the Office of the Pardon Attorney in their consideration of clemency applications. President Cleveland’s 1893 executive order delegated the paperwork on pardons to an office in the Department of Justice (DOJ), and today’s iteration of that office contains a head pardon attorney, a deputy, and several other attorneys (Linzer and LaFleur 2011 ). It is not uncommon for presidents to bypass the standard pardon process on occasion (Crouch 2022 ). However, DOJ guidelines mandate a 5-year waiting period between an individual’s release from incarceration and their eligibility to apply for a pardon. In cases where no prison sentence was imposed, the waiting period begins on the sentencing date.1 Today, the standard pardon process vests considerable gatekeep- ing authority in the DOJ. As Erler ( 2007 , 437) describes, this authority spans “both the number of pardon applications that reach the president’s desk to be either approved or rejected as well as the quality of those applications.” Once an application is filed with the Office of the Pardon Attorney and the applicant is confirmed to be pardon-eligible, the US Probation Office, the federal prosecutor, and the sentencing judge may be consulted (Erler 2007 , 437). Officials review personal, professional, and financial histories; FBI agents interview the applicant’s close friends and associates; and it is not unusual for family members or even members of Congress to offer letters in support of individuals seeking pardons (Linzer 2011 ). Ultimately, the pardon attorney’s recommendation to the president is informed by the applicant’s post-conviction conduct, the nature of the crime, acceptance of responsibility for that crime, and whether there is a concrete need (i.e., as a condition for employment) for relief (Erler 2007 , 438).2 3 The Patterns and Politics of Presidential Pardons Although there is a lack of systematic evidence as to how applicant characteristics, crime types, and other factors may shape pardon decisions, a number of important empirical studies have considered the history and determinants of presidential pardons. Employing the probability that a pardon will be denied as their dependent variable, Whitford and Ochs ( 2006 ) conclude that a president’s decision to refuse clemency is influenced by personal policy priorities, as represented by the president’s rhetoric, agendas, and ideological positions, Congressional atten- tion to criminal justice issues, and broader societal conditions as reflected in the homicide rate. Similarly, Landes and Posner ( 2009 ) speak of “political costs” and argue that higher public concern about crime, which they proxy with a measure of the crime rate, makes it more politically costly—and thus less likely—that presidents will grant large numbers of pardons. This understanding of pardons also connects to broader theoretical questions about executive power. For example, Thelen ( 2025 ) considers the pardon a “benign”example of executive prerogative that allows presidents to act outside normal legal processes while maintaining respect for the rule of law. Several studies identify a persistent partisan divide in pardon grants, generally finding Republican presidents to be more restrictive than their Democratic counterparts. Analyzing data from 1900 to 1993, Ruckman ( 1997 ) found Democratic presidents granted an average of 241 pardons per fiscal year compared to 157 for Republicans. Landes and Posner’s ( 2009 ) study contains a similar finding, which they attribute to modern Republican presidents’ emphasis on being “tough on crime.” As Ruckman ( 1997 , 258) suggests, presidents who take a “well-advertised ‘hard line’ stance on crime may feel compelled to demonstrate fidelity to that principle in their pardon behavior.” At the very least, this partisan dynamic appears to be one factor in how presidents weigh the political costs of clemency decisions. In our view, which we flesh out more in the next section, any potential macro-level partisan divide in pardon patterns likely reflects a more nuanced reality—that there are funda- mental differences in criminal justice priorities and philosophies between the political parties. It may well be the case that, on an overarching level, there are differential political costs for Republican and Democratic administrations when it comes to the pardon decision. However, this oversimplifies a potentially more complex set of ideological calculations. Within the enterprise of law enforcement, Republicans and Democrats are not equally likely to prioritize enforcement of all types of criminal conduct (Miller and Curry 2019 ; Baum 2017 ). Simply put, in the language of Landes and Posner ( 2009 ), there may not only be differential political costs for the two parties generally, but those costs may vary significantly depending on the specific type of crime at issue. This study covers a period that spans important shifts in the politics of criminal justice. Our analysis begins at a point when both Republicans and many Democrats competed to be tough on crime–as an illustration, the 1994 Crime Bill passed with biparti- san support, including that of then-Democratic Senator Joe Biden (Gottschalk 2006 )—through the 2010s when reform discussions gained traction in both parties (Beckett et al. 2016 ). Even as the political environment evolved, we argue that fundamental partisan differences likely persisted in which crimes and which offenders each party prioritized. Republicans remained more focused on street crime, violent offenses, and drug enforcement, while Democrats gave more attention to white-collar crime and inequities in the criminal justice system (Miller and Curry 2019 ; Baum 2017 ). The GOP’s ownership of crime as a political issue (e.g., Petrocik 1996 ) also meant Democratic presidents faced different political risks when exercising clemency. One additional factor to consider when assessing the politics of presidential pardons is the reality that presidents can control when pardons are issued to mitigate political risk. From a purely procedural standpoint, it may take a new administration time to get up-to-speed on the clemency process (Ruckman 1997 , 260). From a political perspective, though, presidents are more politically “safe” in some circumstances than others. Crouch ( 2022 , 696) observes that W. Bush and Obama took a cautious approach to pardons during their first terms, reserving their more controversial clemency decisions for the end of their second terms when they no longer faced electoral consequences. Landes and Social Science Quarterly,2026 3of11
Posner ( 2009 ) similarly control for a president’s second term as well as final year in office and, though not statistically significant, conclude that presidents seem more likely to grant pardons in the second term. Even when not captured by multivariate analysis, anecdotal evidence on this point is strong: The aforementioned pardon of Caspar Weinberger and five others implicated in the Iran-Contra scandal occurred on Christmas Eve in 1992, following President Bush’s electoral defeat; on his final day in office, Bill Clinton issued pardons of financier Marc Rich for tax evasion, half- brother Roger Clinton for cocaine distribution, former Housing and Urban Development Secretary Henry Cisneros, and former Whitewater business partner Susan McDougal; and, after losing the 2020 election, Donald Trump pardoned former strategist Stephen Bannon, former campaign chair Paul Manafort, long- time confidant Roger Stone, and his son-in-law’s father Charles Kushner. Most recently, of course, former President Biden issued a series of pardons—to his son, Hunter, and preemptive pardons to key players in the January 6 Committee, his immediate family, and several others—in the waning days (and even hours) of his administration. 4 Hypotheses, Data, and Variables As we have noted, the pardon represents a key mechanism through which presidents can advance their criminal justice policy preferences. While presidents may set broad priorities, the actual implementation of the pardon power—like many aspects of criminal justice policy—tends to rely on bureaucratic agents working within established institutional frameworks. These institutional arrangements (see Erler 2007 ; Linzer and LaFleur 2011 , for detail) raise important questions about how presidential preferences translate into actual pardon decisions. The process typically involves multiple actors and decision points, from initial application reviews to final recommendations. Assuming, as we do, that presidents have specific preferences or priorities that influence their approach to granting pardons, and that those responsible for compiling pardon applications and making recommendations are generally aware of these prefer- ences, how might these presidential preferences be reflected in the decisions to grant pardons? Our reading of the literature suggests three broad clusters of factors that may be consequential in structuring presidential pardon behavior: demographic char- acteristics, the types of crimes committed, and an assortment of what we term “relevant political dynamics.” Further, we expect that the influence of factors in the first two of these clusters is likely to vary depending on the partisan affiliation of the president considering the pardon decision. 4.1 Demographic Characteristics A critical consideration in examining the patterns of presi- dential pardons to achieve reform goals is the potential role of demographic characteristics, particularly race and gender. Decades of empirical research have documented disparities in the criminal justice system based on race (Rehavi and Starr 2014 ; Steffensmeier and Demuth 2001 ) and gender (Mustard 2001 ). The most consistent finding related to demographic characteristics is that non-White individuals are drastically overrepresented in US prisons relative to their share of the broader population (Carson and Kluckow 2022 ). Gender dynamics are more complex. Women make up a relatively small percentage (approximately 7%) of the nation’s incarcerated population and, therefore, are dramatically underrepresented in the overall prison population. Research has generally indicated that, all else being equal, women obtain more lenient sentences than men most of the time (e.g., Doerner and Demuth 2012 ), although some studies have concluded that women may receive harsher sentences when convicted of crimes violating positive gender stereotypes (Jenkins and Davidson 1990 ; Spohn and Beichner 2000 ). While we do not have theoretical reasons to anticipate across- the-board differences in the treatment of non-White or female pardon applicants, we do anticipate that Democratic presidents will be more likely to issue pardons to non-Whites and females than their Republican counterparts. First, from an electoral standpoint, women and minorities constitute reliable parts of the Democratic Party’s coalition. It is conceivable that Demo- cratic presidents may work to be more responsive to issues affecting these groups, and this could spill over into pardon behavior. Second, Democratic administrations may be more prone to view criminal convictions through the lens of systemic biases in the justice system and/or social inequality (Gottschalk 2006 ) that could disproportionately impact non-Whites. In a similar vein, Erler concludes that Democratic presidents are particularly susceptible to pressure from interest groups seeking pardons for individuals thought to be wrongly convicted or those whose punishments exceeded the severity of their sentences (Erler 2007 , 435). Hypothesis 1. Democratic presidents will be more likely to issue pardons to non-White applicants than their Republican counterparts. Hypothesis2. Democratic presidents will be more likely to issue pardons to female applicants than their Republican counterparts. 4.2 Crime Characteristics Earlier, we referenced aggregate scholarly findings demonstrating Republican presidents were more restrained in granting pardons than their Democratic counterparts and suggested this may only represent part of the story. To be sure, the conservative preference for stricter enforcement of criminal law (Smith et al. 2011 ; Baum 2017 ) fits well with this explanation. However, the apparent partisan divide in pardon patterns may mask a somewhat more complicated reality: Republicans and Democrats have historically prioritized being “tough”on different types of crime. Republicans have typically advocated stricter approaches to street crime, violent offenses, and narcotics offenses; while Democrats have tended to take a harder line against white-collar crimes, they have generally exhibited more leniency in these other areas. To our knowledge, prior studies have not disaggregated crimes for which pardons were issued to consider these more fine- grained possibilities. Consistent with this previous scholarship on presidential priorities in criminal justice enforcement (e.g., Miller and Curry 2019 ), we test the following hypotheses3 : 4of11 Social Science Quarterly,2026
Hypothesis 3. Democratic presidents will be more likely to issue pardons for narcotics crimes than will their Republican counterparts. Hypothesis 4. Democratic presidents will be less likely to issue pardons for white-collar crimes than will their Republican counterparts. With respect to firearms crimes, we anticipate divergent partisan approaches between administrations. While Republican rhetoric often emphasizes support for Second Amendment rights, Repub- lican administrations typically maintain support for prosecuting improper weapons possession—particularly when connected to other criminal activity (Smith et al. 2011 ). Democrats, though less supportive of Second Amendment rights, may be more inclined to view certain firearms offenses as examples of excessive criminalization. Consequently: Hypothesis 5. Democratic presidents will be more likely to issue pardons for firearms crimes than will their Republican counterparts. 4.3 Relevant Political Limits The remaining variables we analyze consist of environmental elements that could operate to influence the president’s propen- sity to grant pardons at a given point in time. Most of these factors relate to the expected reactions of the public and Congress to a pardon. Just as self-proclaimed “law and order” presidents could feel constrained to utilize the pardon, prior studies have considered the possibility that presidents may be less likely to issue pardons when the crime rate is elevated, as the probability of a negative reaction from the electorate is increased (Landes and Posner 2009 ). Likewise, the political calendar could be relevant to a president’s propensity to pardon offenders. Prior studies have operationalized this notion of political risk in different ways, though controlling for a president’s first term or their last months in office are common approaches. Whatever the specific operationalization, the general point is that the political costs (or benefits) to exercising the pardon power “may be asymmetric with regard to the stage in the president’s term in office”(Landes and Posner 2009 , 69). Presidents may also be hesitant to pardon relatively recent crimes, particularly if the crime (or the offender) is viewed as violent. Older crimes may be more politically “safe.” The effect is likely greater if the president made crime a central focus of his time in office. An administration may also feel it is less politically risky to pardon when it enjoys what we term “trifecta control,”or control of the House, Senate, and presidency, as the president does not need to fear noisy backlash from opponents in Congress. We also consider the extent to which the effects of the political dynamics are contingent upon the president’s ideology, as pardons vary in compatibility with the positions of the two parties. We test these expectations below. Hypothesis6. Regardless of partisanship, presidents will be less likely to issue pardons in their first term. Hypothesis7. Presidents will be more likely to issue pardons to older crimes in their first term. 4.4 Time-to-Pardon Up to this point, our expectations have focused on dichotomous outcomes—the likelihood of a pardon being granted to specific types of individuals or for particular types of crimes or under certain circumstances. But an equally important dimension to understanding presidential pardon behavior is not just whether a pardon is granted, but how long it takes for that grant to occur. The theoretical expectations we have outlined are also relevant to this question because they suggest pardons are strategic political decisions that are shaped by policy priorities and perceptions of risk. If certain types of pardons are considered more politically risky or align with lower policy priorities—due to factors like the nature of the crime, the recipient’s demographic characteristics, or the current political climate—we would expect those pardons to take longer for presidents to approve compared to less risky ones. To account for this, we incorporate a time-to-pardon variable, described in the next section, as an independent variable in our first set of models and as the dependent variable in the second. Hypothesis 8. Non-Whites will take longer to receive pardons than whites, and men will take longer to receive pardons than women. Hypothesis9. The time-to-pardon will be influenced by political dynamics, such that: 9a: Presidents facing political electoral consequences will take longer to grant pardons than those who do not . 9b: As a president’s rhetorical focus on crime increases, the time-to-pardon will increase . 9c: The murder rate will negatively impact pardon timing, with higher crime rates leading to longer delays in pardons . 5 Measurement and Operationalization We obtained data on all presidential pardons granted from 1989 through 2024 from the Office of the Pardon Attorney’s website.4 We begin our analysis in 1989 because, prior to the H.W. Bush administration, the Office did not provide data on the type of crime being pardoned. We then collected data to operationalize several variables that can function as either dependent or inde- pendent variables, depending on the specific statistical models we employ. In our analysis, we utilize both traditional logit models and negative binomial count models. To begin, we created a dummy variable, Democratic President, to control for presidential partisanship. As we have noted, we have a priori partisan expec- tations about pardon grants. To explore potential differential effects across presidential partisanship, we include interaction terms in a supplementary analysis to allow for the possibility that certain contextual variables may have party-specific impacts on pardon behavior. We then coded two dichotomous variables to capture the demographic characteristics of pardon recipients— Social Science Quarterly,2026 5of11
non-White and woman. Non-White respondents were coded as 1, and women were coded as 1, with zero representing White and male recipients, respectively. We determined the former by relying on the Namsor name identification software, which uses statistical probabilities derived from census data to estimate likely racial and ethnic background. We simply examined the names to code for gender. To address concern that there might be measurement error in this algorithmic race classification (e.g., Shah and Davis 2017 ), we assessed the confidence scores (scaled 0 to 1, with higher values reflecting greater certainty) that accompany each Namsor classification. The median confidence score in our data is 0.72, and 75% of observations exceeded 0.55. As a robustness check, we re-estimated our main models using different confidence thresholds; as Appendix Table 1A in the Supporting Information indicates, our results are substantively unchanged. To capture crime characteristics, we coded whether (1) or not (0) the pardon involved narcotics, firearms, or white-collar offenses—together these three categories comprise almost two- thirds of all pardons issued, with the majority in narcotics and white-collar crimes. However, as noted in Footnote 3, we have few firearms cases and so anticipate some difficulty in discerning results for this category of crime. The remaining cases involved other offenses not captured by these categories (e.g. being Absent Without Leave from the military); they were coded as zero on all three variables. To provide a rough proxy for the pool of potential pardon recipients, Appendix Table 2A in the Supporting Information summarizes relevant aspects of the federal prison population at several intervals in our study period. While presidential pardons are typically granted to those who have completed their sentences, the population of federal prisoners offers a reasonable approximation of the pool of offenders eligible for clemency. As Table 2A in the Supporting Information shows, there was some shift in the offense composition of federal prisoners over our period of study, with drug offenses declining some, weapons offense more than doubling, and a slight decrease in the percentage of prisoners whose most serious offenses were fraud. Evidence indicates that the percentages of female and non- White federal prisoners remained stable during the period of our study. A number of our variables tap into political dynamics that may be relevant to structuring presidential pardon behavior. To operationalize first term, we simply code whether a president has the potential to face the electorate again. Thus, presidents in their first term are coded 1, and those who have secured reelection are coded zero.5 When same-party or trifecta control exists across the presidency, House, and Senate, we code that as 1, with zero representing all other circumstances.6 As a proxy for the president’s rhetorical attention to crime (e.g., Whitford and Ochs 2006 ), we create a measure—SOTU Crime—to capture the percentage of the speech devoted to crime in the State of the Union Address. Though admittedly just one important sign of administration priorities, the SOTU measure has several advantages for our purposes: It provides consistent data across administrations, directly captures presidential priorities rather than bureaucratic implementation, and has been validated by others (e.g., Whitford and Yates 2009 ; Miller and Curry 2019 ) as an indicator of presidential issue attention. We also include the murder rate to control for variations in the crime rate. Finally, as noted, we include time-to-pardon in our models, first as an independent variable and then as a dependent variable in our count model. In general, we anticipate that presidents who face political risk, give greater rhetorical attention to crime, and are operating in environments where the murder rate is higher will be more reluctant to grant pardons (and take longer to grant them). On the other hand, presidents enjoying trifecta control will be more likely to grant pardons (and do so more quickly) than those who do not. We present basic descriptive statistics and a breakdown by presidential administration in Appendix Tables 3A and 4A in the Supporting Information. The data show meaningful variation across administrations in who receives pardons and for what types of offenses. These descriptive patterns help motivate our multivariate efforts to isolate more systematic effects. 6 Results We estimate a series of logit models for our dichotomous depen- dent variables and a negative binomial model for our count dependent variable (time-to-pardon). Table 1 displays the results for the logit regressions, which we will discuss first. In these models, we are essentially asking how our variable of interest affects the likelihood that a pardon is granted to someone with a particular characteristic. Each model includes standard errors that are robust to heteroskedasticity. With the exception of the model for firearms, our independent variables of theoretical interest predict variation in the pardon rate. Our model for firearms is presented for completeness, but there are few pardons for firearms offenses in our data (only 32), making inference here difficult. The modest PRE values in Table 1 are not inherently surprising, given the highly discretionary nature of the pardon power— and the reality that, as we have noted, individual clemency decisions can be motivated by factors ranging from personal relationships to symbolic efforts to right historical wrongs (see, e.g., President Trump’s pardons of boxer Jack Johnson, 2018, and suffragist Susan B. Anthony, 2020). Of course, our present goal is explaining variation—identifying systematic patterns that reveal how demographic, crime, and political factors sys- tematically associate with this executive power—rather than predicting outcomes. As the statistically significant coefficients demonstrate, meaningful differences do exist even within this difficult-to-model outcome. To better summarize the results, we present them grouped by variables of interest and display average marginal effects and 95% confidence intervals (Tables 2 through 4). In Table 2 , we display outcomes for presidential partisanship. Consistent with Hypothe- ses 1 and 2, Democrats are significantly more likely to pardon non-White and female convicts. They are also significantly more likely to pardon narcotics crimes as compared to a Republican President—a finding that echoes Hypothesis 3. Most strikingly, Democratic presidents appear highly reluctant to grant pardons 6of11 Social Science Quarterly,2026
Logit Regressions Non-White Women Firearms Narcotics White-collar First term Dem. pres. 0.57 (0.16)* * 1.02 (0.24)* − 0.45 (0.41) 0.52 (0.19)* * 0.10 (0.16) − 0.28 (0.14)* Woman 0.05 (0.18) — − 1.75 (1.02) 0.50 (0.19)* * 0.09 (0.18) 0.28 (0.18) Non-White — 0.03 (0.18) − 0.40 (0.40) − 0.23 (0.16) 0.30 (0.14)* 0.19 (0.14) Time-to-pardon 0.02 (0.01)* * − 0.02 (0.02) 0.00 (0.01) − 0.00 (0.01) − 0.03 (0.01)* * − 0.01 (0.01) Firearms − 0.34 (0.41) − 1.38 (1.01) — — — − 0.05 (0.36) Narcotics − 0.13 (0.18) 0.71 (0.23)* * — — — 0.09 (0.18) White-collar 0.23 (0.16) 0.40 (0.22) — — — 0.57 (0.16)* * First Term 0.08 (0.17) − 0.07 (0.25) − 0.18 (0.41) − 0.62 (0.22)* * 0.59 (0.16)* * — Trifecta 0.24 (0.22) 0.64 (0.31)* − 0.23 (0.55) 1.01 (0.27)* * − 0.17 (0.21) — SOTU Crime − 5.46 (1.97)* * − 6.38 (2.44)* * 0.45 (5.55) − 3.96 (2.47) 0.94 (1.86) 5.22 (1.67)* * Murder Rate 0.01 (0.06) 0.14 (0.08) 0.15 (0.10) − 0.28 (0.09)* * 0.06 (0.06) 0.37 (0.07)* * Constant − 1.07 (0.41) − 2.83 (0.58) − 3.89 (0.90) 0.59 (0.51) − 0.88 (0.39) − 2.49 (0.46) N 994 994 994 994 994 994 Wald 33.41 (0.00) 44.01 (0.00) 14.38 (0.07) 54.99 (0.00) 52.29 (0.00) 94.32 (0.00) PRE 0.01 0.00 0.00 0.00 0.03 0.28 Pseudo R2 0.03 0.06 0.04 0.06 0.05 0.09 * is significant at p < 0.05; ** is significant at p < 0.01 (both two-tailed).
Democratic president effects Non-White 13 [6, 20] Woman 13 [7, 18] Narcotics 9 [3, 15] First term − 6 [ − 12, 0]
Time-to-pardon effects (10 to 90) Non-White 10 [2, 17] White-collar − 18 [ − 25, − 11] when they must potentially face the electorate again (first term). The chance that a Democratic administration grants a pardon when facing some political risk is 6 percentage points lower than in a Republican administration. We take this as evidence of the GOP issue ownership of criminal justice policy, and we explore these effects in more nuanced ways below.
10th to the 90th percentile (6 years to 35 years post conviction). Consistent with our expectations (Hypothesis 8), non-White convicts get pardons about 10 years later than White defendants. We noted that white-collar crimes, being non-violent, may be less politically risky than other kinds of pardons. Here, we see that white-collar convicts are pardoned on average 18 years earlier than others.
Narcotics effects Dem. pres. 9 [3, 15] Woman 9 [2, 15] Trifecta 17 [8, 26] First term − 11 [ − 18, − 4] Murder rate − 15 [ − 23, − 8] White-collar effects Non-White 6 [0, 12] Time-to-pardon − 18 [ − 25, − 11] First term 13 [6, 19] We can also divide the results based on the type of crime being pardoned. Here, we will focus on narcotics and white- collar crimes because we observe comparatively few pardons for firearms crimes (only 32 in total in our data). Table 4 displays results for narcotics and white-collar cases. We noted the presidential partisanship results above, but women who commit drug crimes are about 9 percentage points more likely to be pardoned than a man. Interestingly, we find no statistically significant differences between White and non-White defendants who commit narcotics crimes. When an administration enjoys trifecta control, they are more likely to pardon narcotics crimes, perhaps seeing less overall risk in doing so. Finally, the murder rate has a substantial influence on the rate at which drug crimes are pardoned: shifting the murder rate from 5.3/100,000 (10th percentile) to 9/100,000 (90th percentile) decreases the likelihood of a pardon for a narcotics crime by 15 percentage points. Less is notable about white-collar pardons, aside from the speed with Social Science Quarterly,2026 7of11
Negative binomial Time-to-pardon Dem. pres. − 0.05 (0.05) Woman − 0.10 (0.06) Non-White 0.10 (0.04)* * Firearms − 0.07 (0.08) Narcotics − 0.15 (0.05)* * White-collar − 0.27 (0.05)* * First term − 0.18 (0.05)* * Trifecta 0.27 (0.07)* * SOTU Crime − 1.30 (0.54)* Murder Rate − 0.02 (0.02) Constant 3.33 (0.11) N 994 Wald 99.28 (0.00) Adj. R2 0.09 Pseudo R2 0.01 * Is significant at p < 0.05; ** is significant at p < 0.01 (both two-tailed).
Effects on time-to-pardon (NB Regs) Narcotics − 2.9 [ − 4.8, − 1] Non-White 2.1 [0.6, 3.6] White-collar − 5.3 [ − 7.3, − 3.3] Trifecta 5.3 [2.8, 7.9] SOTU crime − 2.3 [ − 4.1, − 0.4] which they occur (as noted above). We do observe that non- White white-collar criminals are more likely to be pardoned (by 6 percentage points) and that first-term pardons are more likely for white-collar cases than other types, again likely because they pose relatively little political risk in terms of the violent crime that most voters are most likely to notice. Next, we present the results of our count model for time- to-pardon, in Table 5 . Because these results are not directly interpretable, we present average marginal effects from this model for statistically significant effects in Table 6 . Note, these results differ from those presented in Table 3 because there we fix the difference in time-to-pardon to display outcomes across variables of interest. In Table 6 , we are predicting directly how each variable affects how long it takes to get a pardon. Women receive pardons 2.6 years sooner than men, while non-White convicts receive pardons 2 years later than White convicts. Both results fit with our expectations. Notably, white-collar crimes are pardoned about 6 years sooner than are non-white-collar crimes. Two variables capturing the political context are also important in determining the timing of pardons. First, when an administration has trifecta control, pardons come more slowly by 4 years. This is an unexpected result because normally we expect there to be less political risk for a president making pardons in an environment of political control. In turn, this should mean that there is less risk in pardoning more recent crimes, but this is not supported by the data. This result is likely simply an artifact of modern American politics, when presidents only enjoy trifecta control in the early years on an initial term. Second, and in line with Hypothesis 9b, as the president focuses more on crime as an issue (as measured by mentions in the State of the Union address, varying from the 10th percentile to the 90th percentile), we see that the age of crimes pardoned increases by about 2.5 years. As crime becomes a more central focus for an administration, presidents appear reluctant to pardon more recent crimes. To summarize our results, we find significant differences in the pardoning behavior of Democratic and Republican admin- istrations. Some of these differences are likely attributable to ownership of crime as an issue. Demographic factors of the convicted are also predictive, with women being more likely to receive pardons quickly compared to men, and non-White convicts more likely to take longer than Whites to receive one. The substantive criminal issue, here measured as either a narcotics, white-collar, or firearms violation, is also predictive of pardon behavior. For instance, women who commit drug offenses are more likely to be pardoned than men who commit these offenses. 6.1 Interactive Regressions and Issue Ownership Below, we present the results of several interactive regression models, focusing on differences across Democratic and Repub- lican presidential administrations. Our concern here is on the various ways that GOP ownership of crime as an issue affects pardoning behavior. Note the regression results upon which the results below are based are available in Appendix Table 5A of the Supporting Information. In Figure 1 , we plot the interaction of mentions of crime in a SOTU and presidential partisanship for pardons when a president must potentially face the electorate after the pardon. A striking pattern emerges. As Democrats become more focused on crime, they also become less likely to issue pardons when they may face electoral consequences. Conversely, as Republican administrations become more focused on crime, they are significantly more likely to issue pardons when they may face the electorate again. For perspective, about half of all State of the Union addresses by Republican presidents mention crime more than 4% of the time. When there is a heavy focus on crime (the 75% percentile is 7.5%), we predict that a GOP administration is 37 percentage points more likely to pardon someone compared to a Democratic administration with a similar focus on crime.
role of partisanship and focus on crime for pardons of narcotics offenses. We tested similar models for firearms and white-collar cases but found no interactive effects. Figure 2 shows that as the focus on crime goes up, as measured by mentions of crime in the State of the Union, Democratic administrations are decreasingly likely to grant a pardon for a narcotics crime, while Republican administrations become slightly more likely to grant a narcotics 8of11 Social Science Quarterly,2026
pardon. This illustrates again that issue ownership of crime by GOP administrations makes Democratic presidents hesitant to grant pardons for these drug cases when there is also a high degree of focus on crime as an issue. 7 Conclusion In this examination of presidential pardons from 1989 through 2024, we focused on how demographic characteristics, crime types, and political considerations can influence the incidence and timing of presidential clemency decisions. Our results reveal important dynamics about the ways in which presidents navigate between policy goals and political constraints. These findings demonstrate that pardons are more than isolated acts of mercy—they can operate as policy tools to fulfill priorities and cultivate perceptions of issue ownership. Specifically, Democratic and Republican administrations show distinct patterns in their pardon-granting behavior, with Democrats being more likely to pardon non-White and female offenders as well as narcotics offenses. Both Democrats and Republicans pardon more women than their share of the federal prison population, but Demo- cratic presidents use their power to help women far more than Republicans. Republicans tend to pardon non-Whites much less frequently than their share of prisoners, while Democrats are closer to parity with their representation in the prison population. We also observed differences across these and several additional dimensions when examining the time elapsed between the imposition of sentence and granting of a pardon. The existence of such systematic differences in pardon behavior stands in contrast to the high-profile, politically charged pardons that dominate media coverage. While controversial pardons of family members or political allies tend to occur toward the end of presidential terms, most pardons follow predictable patterns shaped by philosophies toward criminal justice policy and the appreciation of political risk. The Republican Party’s traditional “ownership”of crime as a political issue is reflected in our finding Social Science Quarterly,2026 9of11
that Democratic presidents are significantly less likely to grant pardons during their first terms, when electoral considerations are most pressing—and the party’s lack of ownership on the crime issue probably makes them especially vulnerable when granting pardons during this period. However, the relationship between presidential rhetoric and pardoning behavior reveals an intriguing pattern: Democratic presidents who emphasize crime in their messaging actually become less likely to issue pardons, particularly for narcotics offenses. In contrast, Republican presi- dents show the opposite trend—as they increase their rhetorical focus on crime, they become significantly more likely to issue pardons, especially during their first terms when electoral conse- quences are at stake. This suggests that “crime issue ownership” provides Republicans with political cover to exercise clemency, while Democrats become more constrained in their pardoning when they emphasize crime issues. Our findings may have important implications for understanding executive clemency at the state level, as well. Just as presidential pardons reflect broader goals and political constraints, governors likely face similar pressures in wielding their pardon powers. At the same time, institutional variation across the states adds a potentially important dimension to understanding executive clemency. First, there is variation across states when it comes to the frequency with which the pardon power is employed. According to a leading source, approximately one-third of states grant a significant number of pardon applications, and in about half, the process ranges from uneven to irregular to rare. This heterogeneity is accompanied by different models of pardon administration: putting aside some technical variation, six states utilize independent review boards, 22 see governors sharing power with a board, another 19 states permit but generally do not require governors to consult with the board, and the five remaining states possess no statutory advisory process.7 Future research might examine whether partisan patterns in state-level pardons mirror those we observe at the federal level, and how variations in state institutional arrangements affect gubernatorial pardoning behavior. As the criminal justice landscape continues to evolve, presidential pardons remain a complex instrument of executive power. The patterns we have identified suggest that pardons are fundamen- tally communicative acts—ways in which presidents articulate their understanding of criminal justice, equity, and political risk. This aligns with theoretical work showing that pardons serve as a mechanism for preserving public confidence in the justice system while allowing for individualized mercy (see Thelen 2025 ). Moving forward, understanding these dynamics becomes increasingly important as the United States grapples with debates about mass incarceration, sentencing reform, and the long- term social consequences of criminal conviction. Notably, these debates have intensified of late with discussions about a pro- posed constitutional amendment designed to curtail the pardon power, reflecting bipartisan concerns about the way Presidents Biden and Trump have exercised this constitutional prerogative (e.g., Korecki 2025 ). Irrespective of constitutional debates about the power’s scope, future research should continue to probe the intersections between executive clemency, political ideol- ogy, demographic representation, and evolving conceptions of justice. Endnotes 1 Office of the Pardon Attorney. https://www.justice.gov/pardon/help- me-choose . 2 The Office of the Pardon Attorney and the White House can sometimes interpret pardon standards differently. Linzer ( 2011 ) describes a reveal- ing disconnect during the George W. Bush administration: an associate White House counsel learned that the pardon attorney’s evaluations of “remorse” went beyond the crimes applicants had pleaded guilty to, also considering charges from their original indictments that had been dropped. The counsel noted that had the White House known about this practice, “our decision making may have been different.” This is consistent with Love’s ( 2010 ) conclusion that the DOJ’s bureaucratic gatekeeping role can create tension between prosecutorial interests and presidential discretion in clemency matters. 3 We include firearms cases in our modeling but do not draw conclusions about them due to the limited sample size—just 32 in our period of study. 4 https://www.justice.gov/pardon/clemency-recipients . The DOJ also publishes aggregate counts by fiscal year of petitions received, denied, and closed without action. Unfortunately, these aggregate statistics contain no corresponding names (from which we could glean demographics) or offense characteristics and, as such, cannot be used to test our expectations about differential individual-level treatment. 5 A small number of pardons are made by presidents after they have lost an election in November. These are not coded as having occurred in the first term, since a president faces no political risk from such pardons. 6 In our period of study, all instances of trifecta control occur in the first term. So, in some models we must omit this variable due to collinearity. 7 https://ccresourcecenter.org/state-restoration- profiles/50- state- comparisoncharacteristics- of- pardon- authorities- 2/ .
Baum, L. 2017. Ideology in the Supreme Court . Princeton University Press. Beckett, K., A. Reosti, and E. Knaphus. 2016. “The End of an Era? Understanding the Contradictions of Criminal Justice Reform.”Annals of the American Academy of Political and Social Science 664: 238–259. Carson, E. A., and R. Kluckow. 2022. “Prisoners in 2022 – Statistical Tables.” U.S. Bureau of Justice Statistics. https://bjs.ojp.gov/document/ p22st.pdf. Crouch, J. 2009. The Presidential Pardon Power . University Press of Kansas. Crouch, J. 2022. “President Trump’s Clemency Record: Extraordinary or Just Ordinary?”Presidential Studies Quarterly 52: 692–708. Doerner, J. K., and S. Demuth. 2012. “Gender and Sentencing in the Federal Courts: Are Women Treated More Leniently?”Criminal Justice Policy Review 25, no. 2: 242–269. Eckstein, P. F., and M. Colby. 2019. “Presidential Pardon Power: Are There Limits and, If Not, Should There Be?”Arizona State Law Journal 51: 71– 108. Erler, H. A. 2007. “Executive Clemency or Bureaucratic Discretion? Two Models of the Pardons Process.”Presidential Studies Quarterly 37, no. 3: 427–448. Ex parte Garland . 71: U.S. 333 1866. Ex parte Grossman . 267: U.S. 87 1925. Fields, G., and J. R. Emshwiller. 2011. “As Federal Crime List Grows, Threshold of Guilt Declines.”Wall Street Journal , September 27. Goldsmith, J., and M. Gluck. 2020. “Trump’s Circumvention of the Justice Department Clemency Process.” Lawfare , December 29. https://www.lawfaremedia.org/article/trumps-circumvention-justice- department- clemency- process . 10of11 Social Science Quarterly,2026
Gottschalk, M. 2006. The Prison and the Gallows: The Politics of Mass Incarceration . Cambridge University Press. Jenkins, P., and B. Davidson. 1990. “Battered Women in the Criminal Justice System: An Analysis of Gender Stereotypes.”Behavioral Sciences & the Law 8, no. 2: 161–170. Kassop, N. 2004. “The View from the President.”In Making Policy, Making Law: An Interbranch Perspective , edited by M. C. Miller and J. Barnes, 72– 88. Georgetown University Press. Korecki, N. 2025. “Biden and Trump Pardons Come Under Scrutiny, Renewing Calls for Reform.”https://www.nbcnews.com/politics/donald- trump/biden- trump- pardons- scrutiny- renewing- calls- reform- rcna193056 . Landes, W. M., and R. A. Posner. 2009. “The Economics of Presidential Pardons and Commutations.”Journal of Legal Studies 38, no. 1: 61–88. Linzer, D. 2011. “Pardon Applicants Benefit from Friends in High Places.”ProPublica , December 4. https://www.propublica.org/article/ pardon- applicants- benefit- from- friends- in- high- places . Linzer, D., and J. LaFleur. 2011. “Presidential Pardons Heavily Favor Whites.”ProPublica , December 3. https://www.propublica.org/article/ shades- of- mercy- presidential- forgiveness- heavily- favors- whites . Love, M. C. 2010. “The Twilight of the Pardon Power.”Journal of Criminal Law and Criminology 100: 1169–1212. Miller, B., and B. Curry. 2019. U.S. Attorneys, Political Control, and Career Ambition . Oxford University Press. Morison, S. T. 2010. “Presidential Pardons and Immigration Law.” Stanford Journal of Civil Rights & Civil Liberties 6: 253–296. Mustard, D. B. 2001. “Racial, Ethnic, and Gender Disparities in Sen- tencing: Evidence From the US Federal Courts.”Journal of Law and Economics 44, no. 1: 285–314. NPR. 2025. “Special Counsel in Hunter Biden Case Denounces President for Criticism of the Probe.”NPR. Published January 13. https://www.npr. org/2025/01/13/nx- s1- 5258953/hunter- biden- special- counsel- report . Obama, B. 2017. “The President’s Role in Advancing Criminal Justice Reform.”Harvard Law Review 130: 811–866. Petrocik, J. R. 1996. “Issue Ownership in Presidential Elections, With a 1980 Case Study.”American Journal of Political Science 40: 825–850. Rehavi, M. M., and S. B. Starr. 2014. “Racial Disparity in Federal Criminal Sentences.”Journal of Political Economy 122, no. 6: 1320–1354. Reilly, R. J. 2025. “Trump Pardons Roughly 1,500 Criminal Defen- dants Charged in the Jan. 6 Capitol Attack.” NBC News, Jan- uary 21. https://www.nbcnews.com/politics/justice-department/trump- set- pardon- defendants- stormed- capitol- jan- 6- 2021- rcna187735 . Richler, A. D., M. Kunzelman, and A. Khalil. 2025. “Trump Pardons Upend Massive Jan. 6 Prosecution by Freeing Rioters and Dismissing Cases.” AP News, January 21. https://apnews.com/article/capitol-riot- trump- pardons- jan- 6- f6e23bcd84eaed672318c88f05286767 . Ruckman, P. S., Jr. 1997. “Executive Clemency in the United States: Origins, Development, and Analysis (1900-1993).”Presidential Studies Quarterly 27, no. 2: 251–271. Shah, P. R., and N. R. Davis. 2017. “Comparing Three Methods of Measuring Race/Ethnicity.”Journal of Race, Ethnicity, and Politics 2: 124–139. Shogan, C. 2020. “The History of the Pardon Power: Executive Unilater- alism in the Constitution.”Rubenstein Center Scholarship. https://www. whitehousehistory.org/the-history-of-the-pardon-power . Smith, K. B., D. R. Oxley, M. V. Hibbing, J. R. Alford, and J. R. Hibbing. 2011. “Linking Genetics and Political Attitudes: Reconceptualizing Politi- cal Ideology.”Political Psychology 32, no. 3: 369–397. Spohn, C., and D. Beichner. 2000. “Is Preferential Treatment of Female Offenders a Thing of the Past? A Multisite Study of Gender, Race, and Imprisonment.”Criminal Justice Policy Review 11, no. 2: 149–184. Steffensmeier, D., and S. Demuth. 2001. “Ethnicity and Judges’ Sentenc- ing Decisions: Hispanic-Black-White Comparisons.”Criminology 39, no. 1: 145–178. Thelen, I. 2025. “The ‘Benign Prerogative’: Political Theory and Executive Pardoning.”Presidential Studies Quarterly 55: 19–33. U.S. v. Klein . 80: U.S. 128 1871. Whitford, A. B., and H. L. Ochs. 2006. “The Political Roots of Executive Clemency.”American Politics Research 3, no. 6: 825–846. Whitford, A. B., and J. Yates. 2009. Presidential Rhetoric and the Public Agenda: Constructing the War on Drugs . Johns Hopkins University Press. Supporting Information Additional supporting information can be found online in the Supporting Information section. Supplementary Table 1A : Main results across NAMSOR confidence thresholds. Supplementary Table 2A : Federal prison population. Supplementary Table 3A : Descriptive statistics. Supplementary Table 4A : Descriptive statistics by presidential administration. Supplementary
Social Science Quarterly,2026 11of11