In early 2025, at the beginning of President Donald Trump’s second term, the White House issued an executive order creating the Department of Government Efficiency (DOGE) (Trump Reference Trump2025). Charged with enhancing federal efficiency and productivity, DOGE quickly launched sweeping initiatives to eliminate what the administration labeled as bureaucratic waste. This included cutting federal contracts (Nguyen and Blaeser Reference Nguyen and Blaeser2025), initiating mass layoffs of federal employees (Cameron Reference Cameron2025), and expanding oversight and surveillance of civil servants (Shear, Barnes, and Schmitt Reference Shear, Barnes and Schmitt2025). Although presented as a neutral technocratic reform, I present evidence that DOGE’s actions were politically structured, providing President Trump with an instrument to impose targeted harm on communities that had not supported him electorally.
In this research note, I argue that presidential particularism extends beyond the unequal distribution of benefits to include the imposition of costs. Existing research shows that presidents allocate federal resources to reward electorally valuable constituencies, such that politically aligned areas receive more benefits while unaligned areas receive fewer by comparison (Berry, Burden, and Howell Reference Berry, Burden and Howell2010; Ha Reference Ha2023; Kriner and Reeves Reference Kriner and Reeves2015b). In this framework, disadvantage arises passively as a byproduct of preferential allocation.
I build on this literature by showing that presidential particularism can operate through a distinct punitive dimension. Rather than merely receiving fewer benefits, disfavored constituencies may be directly targeted with costs. This distinction matters because withholding or limiting particularistic benefits produces a passive form of disadvantage, whereas the imposition of costs involves active harm that may signal punishment or retribution. This logic aligns with prospect theory, which demonstrates that individuals evaluate gains and losses asymmetrically, with losses experienced more intensely than equivalent gains (Kahneman and Tversky Reference Kahneman and Tversky1979). I argue that this punitive form of presidential particularism helps explain DOGE’s behavior, demonstrating that presidential particularism operates not only through the distribution of resources but also through the imposition of costs.
From this perspective, DOGE was not merely a technocratic agency tasked with eliminating redundancy; it functioned as an instrument of partisan discipline. Viewed through the lens of presidential particularism, DOGE provides evidence that presidents may use the machinery of federal administration to penalize political opposition rather than govern in a neutral, universalistic manner. This perspective broadens the study of distributive politics beyond the “carrot” of reward-based presidential particularism to include the “stick” of targeted harm. By incorporating punishment alongside reward, this framework expands existing theories of presidential particularism to encompass the targeted distribution of administrative harm.
Emerging work on DOGE provides additional context for understanding its political implications. Bellodi and Lee (Reference Bellodi and LeeN.d.) analyze the distribution of DOGE contract cuts and show that firms donating to the Republican Party were less likely to experience contract cuts, while those donating to the Democratic Party were more likely. Similarly, in a blog post, Bonica (Reference Bonica2025) argues that DOGE’s actions reflect political considerations rather than purely technocratic objectives, demonstrating that more liberal agencies were disproportionately affected by layoffs. Together, these findings suggest that DOGE’s activities were structured by political incentives. This study builds on this emerging literature by shifting the focus to the geographic distribution of harm, showing how presidential particularism may operate punitively across electorally defined constituencies.
If presidential particularism extends to the imposition of costs, it should be most observable in institutional settings characterized by strong presidential control such as the conditions exemplified by DOGE. Created by executive order and housed within the Executive Office of the President, the agency was granted substantial authority to determine which contracts to cut and operated in close coordination with the White House (Trump Reference Trump2025). Its structure reflects a broader pattern identified in the literature: modern presidents consolidate influence over the bureaucracy through centralization and politicization (Moe Reference Moe, Chubb and Peterson1985). As Moe (Reference Moe, Chubb and Peterson1989) argues, bureaucratic structures are crafted not to maximize administrative efficiency, but to advance political objectives. Presidents seek institutional arrangements that enhance their capacity to shape implementation and maintain control over policy outcomes (Moe Reference Moe, Chubb and Peterson1989, 280). This pattern is consistent with prior work showing that more politicized agencies distribute contracts and engage in favoritism toward electorally important constituencies (Dahlström, Fazekas, and Lewis Reference Dahlström, Fazekas and Lewis2021; Gordon Reference Gordon2011). Viewed in this light, DOGE’s institutional design was not incidental. By consolidating presidential control (Marisam Reference Marisam2025), DOGE created channels through which presidential priorities could be embedded directly within the administrative state. This structure increased the likelihood that bureaucratic action would reflect the president’s particularistic incentives rather than operate independently of them. Given DOGE’s distinctive institutional position, it represents a case in which we would expect presidential particularism to be especially pronounced, due to its high levels of centralization and politicization.
These institutional dynamics operate within a broader context of high levels of elite polarization. As partisan elites have become more ideologically divided (McCarty, Poole, and Rosenthal Reference McCarty, Poole and Rosenthal2016), the presidency has faced new demands and expectations (Lowande Reference Lowande2024). One consequence is the rise of “negative representation,” where politicians prioritize attacking the opposing party over advancing their own agenda (Costa Reference Costa2025). While presidents have long used administrative tools to reward allies, elite polarization may increase the incentives of using those same tools to punish supporters of the opposing party. This alternative conceptualization of presidential particularism as punitive broadens how we conceptualize presidential representation.
Notably, a punitive dimension of presidential particularism runs against foundational theories of presidential incentives. The idea that presidents would intentionally impose material harm on specific constituencies conflicts with the logic of retrospective voting. In classic accounts, citizens evaluate incumbents largely on economic performance, rewarding good outcomes and punishing poor ones (Fiorina Reference Fiorina1981; Key Reference Key1966; Kramer Reference Kramer1971; Lewis-Beck and Stegmaier Reference Lewis-Beck and Stegmaier2000). Because voters are presumed to engage in pocketbook evaluations, presidents are expected to maximize broad economic well-being rather than selectively impose costs. Indeed, Tufte (Reference Tufte1980) argues that incumbents strategically manipulate economic conditions to improve their electoral prospects. This logic—that voters reward economic performance—has also underpinned explanations for why politicians engage in pork-barrel politics, directing targeted benefits to constituents in hopes of securing electoral support (Hudak Reference Hudak2014; Stein and Bickers Reference Stein and Bickers1994). Against this backdrop, the strategic deployment of targeted harm appears electorally risky. In an era of intense partisan polarization, however, a punitive dimension of presidential particularism may reflect a shift in incentive structures in which mobilizing and consolidating a partisan coalition outweighs the traditional imperative to deliver universalistic economic gains.
To assess this punitive dimension of presidential particularism, I analyze DOGE during the early months of Trump’s second term. I assemble a dataset of nearly 700,000 federal contracts active at the start of the second Trump administration and matched them to contracts cut by DOGE. Each contract is geocoded to the recipient’s county and linked to 2024 county-level presidential election returns, enabling fine-grained analyses at both the contract and county levels.
The results support expectations generated by an expanded understanding of presidential particularism as a punitive, rather than purely reward-based, strategy. Contracts in counties that provided weaker electoral support for Trump in 2024 were significantly more likely to be cut. Moreover, counties with lower levels of electoral support for Trump experienced a significantly greater expected number of DOGE contract cuts. I demonstrate these contract- and county-level patterns using both a continuous measure of 2024 county-level vote share and a categorical classification of counties, with results remaining consistent across model specifications.
These findings contribute to the study of executive politics by expanding the concept of presidential particularism beyond its traditional focus on rewarding allies. While prior work shows that presidents use distributive tools to benefit supporters (Ha Reference Ha2023; Kriner and Reeves Reference Kriner and Reeves2015a; Reference Kriner and Reeves2015c), this study demonstrates that the same tools can be deployed to inflict targeted harm. In certain institutional settings, such as those exemplified by DOGE, the modern presidency can administer punishment. This case highlights how the punitive form of presidential particularism can emerge in the wake of an election and highlights an instance in which presidential particularism operates as a mechanism of retributive governance.
Situated within an environment of heightened elite polarization (Enders Reference Enders2021; Hetherington Reference Hetherington2001; McCarty, Poole, and Rosenthal Reference McCarty, Poole and Rosenthal2016; Theriault Reference Theriault2008), and given the president’s increasingly central role in party politics (Alexander and Jacobs Reference Alexander and Jacobs2023; Galvin Reference Galvin2009; Jacobs and Milkis Reference Jacobs and Milkis2021; Jacobson Reference Jacobson2019), these results suggest that presidential power can become an extension of partisan conflict. As such, presidential power may function not only as a tool of governance but also as a mechanism of partisan punishment, shaping how benefits and burdens are distributed across constituencies.
PRESIDENTIAL PARTICULARISM AS PUNITIVE
Presidential particularism has traditionally been understood as a strategy of reward: presidents direct federal resources toward electorally valuable constituencies in order to consolidate support and enhance political standing (Kriner and Reeves Reference Kriner and Reeves2015a). I extend this framework by arguing that the same incentives that motivate targeted rewards can also motivate targeted punishment. This theoretical expansion builds directly on the literature on presidential particularism, which conceptualizes it as the strategic allocation of federal benefits to electorally valuable constituencies. As theorized by Kriner and Reeves (Reference Kriner and Reeves2015a), presidential particularism reflects the president’s dual role as both national executive and partisan leader, uniquely positioned to deploy federal resources for political gain. This perspective aligns with a broader literature emphasizing the presidency’s institutional centrality in advancing partisan objectives (Galvin Reference Galvin2009; Milkis, Rhodes, and Charnock Reference Milkis, Rhodes and Charnock2012; Wood Reference Wood2009). Distributive benefits in this framework serve both material and symbolic functions, shaping public perceptions and influencing presidential approval (Ang et al. Reference Ang, Reeves, Rogowski and Vishwanath2022; Cohen Reference Cohen2006; Gasper and Reeves Reference Gasper and Reeves2011).
Empirical evidence demonstrates that presidential particularism is a routine feature of executive governance. Kriner and Reeves (Reference Kriner and Reeves2015b) show that presidents systematically direct domestic grants toward politically valuable counties, with similar patterns documented in trade protection (Lowande, Jenkins, and Clarke Reference Lowande, Jenkins and Clarke2018) and COVID-19 relief allocation (Ha Reference Ha2023). While this literature overwhelmingly focuses on the rewarding of political allies, the same institutional tools can plausibly be used to punish political opponents.
Existing work has only indirectly engaged with this possibility. Studies of military base closures note instances in which presidential particularism could have operated punitively (Kriner and Reeves Reference Kriner and Reeves2015a; Reference Kriner and Reeves2015c). However, as Kriner and Reeves (Reference Kriner and Reeves2015a) emphasize, the most explicit example—Dick Cheney’s proposal to target partisan opponents through base closures—was never implemented. As a result, the punitive dimension of presidential particularism has received limited direct empirical attention, particularly with respect to its real-world operation and geographic consequences. Insights from prospect theory highlight why the administration of punishment is especially consequential: individuals are loss averse and evaluate losses more negatively than comparable gains (Kahneman and Tversky Reference Kahneman and Tversky1979). Consequently, policies that impose costs are likely to be highly salient and politically meaningful, underscoring the importance of examining how presidents may deploy punishment as a particularistic strategy.
This study extends the logic of presidential particularism by examining its punitive form: how presidents may use executive authority to impose costs on electorally oppositional constituencies. Underlying this behavior is a dual reality—presidents are both uniquely empowered to shape the flow of federal resources and politically incentivized to wield that authority for partisan advantage.
A growing body of scholarship underscores the structural dominance of the presidency in both federal policymaking and partisan politics. Presidents have expansive unilateral powers that allow them to shape public policy with minimal legislative constraint (Childree Reference Childree2025; Howell Reference Howell2003). In addition to formal unilateralism, the president leads the federal bureaucracy, a domain where the vast majority of major laws delegate implementation authority, thus amplifying executive discretion (McCann, Clouser, and Shipan Reference McCann, Clouser and Shipan2022; VanSickle-Ward Reference VanSickle-Ward2014). The centralized and politicized bureaucracy over which presidents exercise authority (Moe Reference Moe, Chubb and Peterson1985) underscores their substantial institutional power within the federal government.
Presidents are also central to partisan strategy and representation. As leaders of their parties, presidents shape electoral strategies, party branding, and legislative positioning (Galvin Reference Galvin2009; Jacobs and Milkis Reference Jacobs and Milkis2021; Milkis and Jacobs Reference Milkis and Jacobs2017). Public perception of the president often extends to their party more broadly, influencing evaluations of co-partisans and shaping down-ballot electoral outcomes (Jacobson Reference Jacobson2019). Moreover, presidents have become so central to partisan politics that legislators increasingly look to them for cues on policy positions and legislative strategy (Alexander and Jacobs Reference Alexander and Jacobs2023). The nationalization of American politics has only magnified the importance of presidents, making presidential support increasingly synonymous with party performance (Hopkins Reference Hopkins2018; Rogers Reference Rogers2023).
While institutional and partisan centrality has led scholars to focus on the president’s capacity to reward allies, the same logic extends to punishment. The very mechanisms that sustain presidential particularism also enable presidents to engage in punitive behavior. At the elite level, polarization has intensified the stakes of partisan conflict. Over the past several decades, party leaders have diverged not only in policy preferences but also in governing styles and institutional strategies (McCarty, Poole, and Rosenthal Reference McCarty, Poole and Rosenthal2016; Theriault Reference Theriault2008). As elites have grown more polarized, cooperation in Congress has eroded, and institutional gridlock has deepened, placing greater weight on presidents (Lowande Reference Lowande2024). In this environment, presidents face both heightened incentives and fewer constraints to use administrative power not only to cultivate allies, but to penalize opponents.
In this context, presidents—who serve both as national executives and party leaders—may come to see the exercise of administrative power as an extension of partisan conflict. Rather than governing in a universalistic manner, they may advance the interests of their partisan coalition while actively disadvantaging the opposing side. This perspective aligns with emerging research on negative representation, which highlights how politicians increasingly define their actions not by responsiveness to constituents but by hostility toward the out-party (Costa Reference Costa2025). Negative representation has been most visible in rhetoric and symbolic acts, but it also has potential material implications.
This research note argues that presidents can institutionalize partisan hostility by embedding it within the administrative apparatus of government. Through the selective imposition of harm—such as cutting contracts—presidents may convert the federal bureaucracy into a tool of partisan punishment. Crucially, unlike traditional discretionary spending programs where presidents aim to maximize distribution in politically advantageous ways, DOGE faced no fixed threshold for required cuts. This lack of constraint made its decisions over where and how to impose harm particularly susceptible to political considerations.
To provide preliminary descriptive support for the argument that presidential particularism can operate in a punitive form, I present a bivariate scatterplot illustrating the relationship between a county’s Trump two-party vote share and the logged number of contracts cut by DOGE (plus one). As shown in Figure 1, the scatterplot includes a smoothed LOESS line to visualize the trend between electoral support and the geographic concentration of DOGE contract cuts.
Scatterplot of Relationship between Trump Two-Party Vote Share and Logged (Plus One) Number of Contract Cuts
Note: LOESS (black) line visualizes relationship in scatterplot.
Figure 1. Long description
The x-axis is labeled County Trump Two-Party Vote Share, ranging from 0.25 to 1.00. The y-axis is labeled Logged Number of Contract Cuts, ranging from 0 to 6. The data consists of numerous semi-transparent grey circular points. A high concentration of points is clustered along the y equals 0 baseline across the entire x-axis. Above this baseline, the points are more dispersed, showing a higher density in the lower-left quadrant (low vote share, higher contract cuts) and thinning out significantly toward the bottom-right quadrant (high vote share, low contract cuts). A thick black L O E S S curve begins at approximately x equals 0.1, y equals 3.2, and follows a smooth, non-linear downward trajectory, flattening out as it approaches y equals 0 near the x equals 0.85 mark, indicating a negative correlation between the two variables.
The relationship is distinctly negative: counties that provided stronger electoral support for President Trump experienced fewer contract cuts, whereas those with lower levels of support faced more cuts. The Pearson correlation coefficient of
$ -0.48 $
further underscores this inverse association—indicating that as Trump’s county-level vote share increases, the number of DOGE-imposed cuts declines. This pattern is consistent with a punitive dimension of presidential particularism, suggesting that DOGE’s administrative actions were disproportionately concentrated in electorally oppositional counties. Although descriptive in nature, this initial evidence provides a visual and intuitive foundation for the empirical analyses that follow. It signals that DOGE’s discretionary authority was not deployed uniformly across jurisdictions, but may have been conditioned by partisan electoral considerations.
DATA AND METHODS
While the bivariate scatterplot provides preliminary evidence of an association between Trump’s county-level electoral support and DOGE’s impact, the following section outlines the data and empirical strategy used to more rigorously evaluate the presence of a punitive dimension of presidential particularism.
To begin, I assemble a dataset of federal contracts that were plausibly eligible to be cut by DOGE. I draw contract-level data from the Federal Procurement Data System (2026), collecting all contracts entered between fiscal year 2004 and January 20, 2025—the inauguration date of Trump’s second term.Footnote 1 To be included, contracts must have originated within this period and have a scheduled completion date after inauguration day, ensuring that they were active at the start of the second Trump administration. These contracts constitute the federal contracts that could have theoretically been subject to cuts by DOGE, as they were initiated before the start of the administration and remained in progress.Footnote 2 In total, the resulting dataset contains nearly 700,000 federal contracts.
I next incorporate a dataset of federal contracts cut by DOGE during the first seven months of President Trump’s second administration. As of September 9, 2025, the DOGE website listed 13,231 contract cuts (Department of Government Efficiency 2025). With respect to data completeness, DOGE notes that reported contract cuts may be subject to delays due to lags in FPDS reporting. DOGE also indicates that the contract, grant, and lease cancellation information displayed on its website represents a subset of DOGE’s total reported savings, which also include additional components, such as workforce reductions and asset sales.Footnote 3 Every possible DOGE contract cut was cross-referenced with the Federal Procurement Data System (2026). In total, the dataset comprises nearly 700,000 federal contracts eligible to be cut and includes a binary indicator identifying whether each contract was cut by DOGE.Footnote 4
I next incorporate a series of control variables. First, I assign each contract to a county based on its reported ZIP Code Tabulation Area (ZCTA). For each ZCTA, I calculate a centroid and match it to the county in which it is located. This procedure enables the linkage of each contract to a county-level dataset containing results from the 2024 U.S. presidential election.Footnote 5 In addition to using Trump’s two-party vote share as a continuous independent variable, I also construct a three-category classification of counties as core, swing, or hostile, following prior work on presidential particularism (Kriner and Reeves Reference Kriner and Reeves2015a; Reference Kriner and Reeves2015b).Footnote 6
I then add county-level demographic characteristics drawn from the 2024 U.S. Census Bureau’s American Community Survey, including logged population, logged median household income, the percentage of adults without a high school diploma, and population density (U.S. Census Bureau 2026). Next, each county is associated with a congressional district by assigning it to the district covering the largest share of the county’s land. I include information on the county’s representative in the 119th Congress, focusing on the representative’s partisan relationship to the president.Footnote 7
Finally, I construct measures capturing the logged number of federal contracts awarded to each county between fiscal year 2004 and January 20, 2025, as well as the logged number of contracts awarded during the Biden administration.Footnote 8
To construct the county-level dataset, I aggregate the total number of DOGE-related contract cuts within each county. Counties with no reported DOGE contract cuts are assigned a value of zero. Counties that received no federal contracts between fiscal year 2004 and January 20, 2025 are assigned a value of zero for total contracts, and counties that received no federal contracts during the Biden administration are assigned a value of zero for Biden-era contracts.
In both the contract- and county-level analyses, the inclusion of the control variables described above helps isolate the relationship between Trump’s electoral support and DOGE’s contract cuts. For example, counties with lower levels of support for Trump may have historically received a greater volume of federal contracts, while more populous counties may mechanically experience larger numbers of cuts due to their larger contracting base. By accounting for these structural and contextual differences, the analysis provides an estimate of the relationship between electoral support for President Trump and the geographic distribution of DOGE-imposed contract cuts.
To evaluate this relationship, I employ complementary modeling strategies.
The first uses the contract-level dataset to examine the factors shaping the likelihood that an individual federal contract was cut by DOGE. I estimate logistic regression models appropriate for a binary outcome indicating whether a contract was cut by DOGE (1) or not (0). At the contract level, I estimate two sets of models that differ in how electoral support is operationalized. In the first set, the key independent variable is the county-level Trump two-party vote share in the 2024 election. These models include covariates capturing county-level socioeconomic and political context: logged population, logged median household income, the percentage of the population without a high school diploma, population density, an indicator for whether the county is represented by a co-partisan member of Congress, the logged total number of contracts awarded to the county between fiscal year 2004 and January 20, 2025, and the logged number of contracts awarded during the Biden administration. All models report standard errors clustered at the state level, and specifications are estimated both with and without state fixed effects.
In the second set of contract-level models, electoral support is measured categorically using county type—core, swing, or hostile—based on the 2024 Trump two-party vote share. Core counties serve as the reference category, with indicator variables for swing and hostile counties. The covariates, estimation strategy, and standard error clustering remain identical to those in the continuous vote-share models.
The next set of analyses parallels the contract-level approach but shifts the unit of analysis to the county level. Using the county-level dataset, I model the total number of DOGE contract cuts experienced by each county. Because the dependent variable is a count of the number of contract cuts, I estimate negative binomial regression models.
In the first set of county-level models, the key independent variable is the county’s Trump two-party vote share in the 2024 election, with the same set of control variables, state fixed effects, and state-clustered standard errors as in the contract-level analyses. In the second set, electoral support is operationalized using the county-type classification (core, swing, and hostile), again mirroring the controls and estimation strategy employed at the contract level.
By analyzing how presidential particularism may have manifested through the punitive use of DOGE contract cuts across the United States, this study leverages both contract- and county-level data to investigate the political logic underlying these administrative actions. Employing multiple levels of analysis and operationalizations of electoral support provides insight into the geographic and political contours of DOGE’s contract cuts. Together, these analyses provide a systematic test of whether presidential particularism extends beyond the distribution of benefits to include the targeted imposition of costs.Footnote 9
RESULTS
To begin assessing the political contours of DOGE’s contract cuts through the lens of a punitive dimension of presidential particularism, I first examine how Trump’s 2024 two-party vote share shaped the distribution of contract cuts. Table 1 presents the results. Columns 1 and 2 report contract-level analyses, while columns 3 and 4 present county-level models, offering two complementary approaches to evaluating the relationship.
Regression Analysis of the Relationship between 2024 Trump Presidential Two-Party Vote Share and DOGE Contract Cuts at the Contract (Logit) and County (Negative Binomial) Levels
Table 1. Long description
The table presents regression coefficients with standard errors in parentheses for four models. Models 1 and 2 are contract-level logit models (671,775 observations), while Models 3 and 4 are county-level negative binomial models (3,114 observations).
* Trump two-party vote share: Shows a consistent negative correlation across all models. Model 1: -4.60. Model 2: -4.08. Model 3: -2.12. Model 4: -2.53. All are statistically significant at p < 0.05.
* Log (population): Coefficients range from -0.14 to 0.31, with significant positive results in Models 2 and 4.
* Log (Med. household income): Consistently positive and significant across all models, ranging from 0.80 to 2.31.
* % Less than HS: Small, non-significant coefficients ranging from 0.00 to 0.05.
* Population density: Positive coefficients, significant only in Model 2 (0.25).
* Co-partisan House Rep: Non-significant coefficients ranging from 0.03 to 0.38.
* Log (County Contracts): Positive and significant in Models 1, 3, and 4.
* Log (County Biden-Era Contracts): Negative and significant in Models 1 and 2.
* Intercept: Highly negative and significant across all models, from -12.51 to -29.62.
* Model Statistics: A I C values are 77,065.9 (Model 1), 74,790.8 (Model 2), 4,059.3 (Model 3), and 4,045.1 (Model 4). State Fixed Effects (F E) are included in Models 2 and 4.
Note: Standard errors are clustered by state. Models 1 and 2 are contract-level logit models of contract cuts. Models 3 and 4 are county-level negative binomial models of the number of cut contracts.
$ {}^{*}\hskip0.3em p\hskip0.3em < $
0.05
Focusing first on the contract-level results (columns 1 and 2), the findings indicate that contracts located in counties with higher Trump vote shares were significantly less likely to be cut by DOGE. The estimated coefficients on Trump’s two-party vote share are
$ -4.60 $
and
$ -4.08 $
, respectively. Substantively, the coefficient from Model 2 indicates that a 10-percentage-point increase in Trump’s vote share is associated with approximately a 33% decrease in the odds of a contract being cut. This is a sizable relationship and suggests that DOGE’s cuts were structured along partisan lines, disproportionately impacting counties that were less supportive of Trump in 2024.
These findings provide evidence consistent with there being a punitive dimension of presidential particularism: the strategic use of administrative authority to impose costs on electorally oppositional constituencies. In the Supplementary Material, I present additional models that assess the relationships in Table 1 using alternative specifications. For instance, one noteworthy robustness check examines whether the geographic distribution of punishment reflects where contract work was performed rather than the location of the vendor receiving the contract. This addresses the possibility that the punitive dimension of presidential particularism is tied to where the activity associated with a contract occurs, rather than where the contract recipient is formally located. Re-estimating the models using contracts’ place of performance instead of vendor location yields substantively similar results.
The control variables also yield informative patterns. Counties that had a larger number of active federal contracts were suggestively more likely to experience DOGE-related cuts, which could reflect an effort to reduce contracts in areas with substantial federal spending. However, the results do not support the interpretation that DOGE primarily targeted active Biden-era contracts. In fact, contracts in counties that had more active contracts from the Biden administration were less likely to experience DOGE contract cuts, suggesting that the pattern of cuts cannot be explained solely as a rollback of prior administration spending. As a supplementary analysis, I estimate models restricted to contracts most recently signed prior to the Biden administration. The results remain substantively similar. These patterns suggest that the punitive dimension of presidential particularism operates through the targeting of geographic constituencies rather than through the selective reversal of contracts associated with a prior administration.
Turning to Models 3 and 4 in Table 1, I present the county-level analyses. Consistent with the contract-level findings, the coefficient on Trump’s 2024 two-party vote share is negative and statistically significant. Counties in which Trump performed better electorally experienced fewer DOGE contract cuts overall. Substantively, the coefficient of
$ -2.53 $
on Trump’s 2024 two-party vote share in Model 4 suggests that a 10-percentage-point increase in Trump’s two-party vote share is associated with approximately a 22% decrease in the expected number of DOGE contract cuts in a county. These results reinforce the earlier findings: the geographic distribution and intensity of DOGE contract cuts closely track Trump’s 2024 electoral performance, consistent with a punitive logic of presidential particularism.
Notably, across all four models in Table 1, the coefficient on co-partisan House representation is not statistically significant. This suggests that DOGE contract cuts were not systematically mitigated in counties represented by Republican members of Congress. As such, the pattern of DOGE contract cuts appears more closely tied to the president’s electoral incentives than to the broader well-being of the party, implying that presidential particularism may operate primarily in service of a president’s political goals.
A potential concern is that the results may be driven by contracts associated with Democratic policy areas. To address this, I provide supplementary analyses showing that the substantive findings hold across both Democratic- and Republican-associated policy areas.
Conceptualizing electoral support using Trump’s continuous two-party vote share is one approach, but it may also be useful to consider a categorical measure of political alignment. Following Kriner and Reeves (Reference Kriner and Reeves2015b), I classify counties as core, swing, or hostile based on Trump’s 2024 two-party vote share.
This categorization allows for a more flexible, potentially nonlinear assessment of how electoral support relates to DOGE contract cuts. Under a punitive logic of presidential particularism, we would expect hostile counties—those providing the weakest electoral support—to experience the highest levels of targeting. Meanwhile, core counties—Trump’s strongest electoral base—should face the fewest DOGE contract cuts.
The results are presented in Table 2. As in the previous table, Models 1 and 2 report contract-level analyses, while Models 3 and 4 present county-level results.
Regression Analysis of the Relationship between County Type and DOGE Contract Cuts at the Contract (Logit) and County (Negative Binomial) Levels
Table 2. Long description
The table presents regression results for four models. Models 1 and 2 are contract-level logit models, while Models 3 and 4 are county-level negative binomial models.
Key variables and their coefficients (standard errors in parentheses):
* Hostile County: Model 1 (1.10*), Model 2 (1.13*), Model 3 (0.70*), Model 4 (0.73*).
* Swing County: Model 1 (0.15), Model 2 (0.29), Model 3 (0.14), Model 4 (0.17).
* Log (population): Model 1 (-0.14), Model 2 (0.27*), Model 3 (0.02), Model 4 (0.22*).
* Log (Med. household income): Model 1 (2.24*), Model 2 (1.91*), Model 3 (0.99*), Model 4 (0.70*).
* % Less than H S: Model 1 (0.03), Model 2 (0.03), Model 3 (0.01), Model 4 (-0.01).
* Population density: Model 1 (0.26*), Model 2 (0.54*), Model 3 (0.17), Model 4 (0.24).
* Co-partisan House Rep: Model 1 (0.16), Model 2 (0.03), Model 3 (0.13), Model 4 (-0.05).
* Log (County Contracts): Model 1 (2.04*), Model 2 (1.80*), Model 3 (1.37*), Model 4 (1.14*).
* Log (County Biden-Era Contracts): Model 1 (-2.18*), Model 2 (-2.21*), Model 3 (-0.47*), Model 4 (-0.35).
* Intercept: Model 1 (-28.59*), Model 2 (-27.75*), Model 3 (-16.03*), Model 4 (-13.57*).
Model Statistics:
* No. of obs.: Models 1 and 2 (671,775); Models 3 and 4 (3,114).
* A I C: Model 1 (77,648.8), Model 2 (75,048.6), Model 3 (4,063.4), Model 4 (4,054.7).
* State F E: Models 1 and 3 (No); Models 2 and 4 (Yes).
Asterisks denote p < 0.05. Standard errors are clustered by state.
Note: Standard errors are clustered by state. Models 1 and 2 are contract-level logit models of contract cuts. Models 3 and 4 are county-level negative binomial models of the number of cut contracts.
$ {}^{*}\hskip0.3em p\hskip0.3em < $
0.05
In Models 1 and 2, contracts located in hostile counties are significantly more likely to be cut by DOGE relative to those in core counties (the reference category). Substantively, this indicates that contracts in counties that provided the least electoral support to Trump in 2024 were more likely to be cut following the election.
Contracts in swing counties—those that were electorally competitive—also exhibit a positive coefficient, suggesting a higher likelihood of being cut relative to core counties; however, this estimate does not reach conventional levels of statistical significance. As such, the evidence for differential targeting of swing counties is inconclusive.
Taken together, the results indicate that contracts in hostile counties were more likely to be cut by DOGE than those in core counties, highlighting a clear pattern of targeting concentrated in electorally oppositional locations.
Turning to Models 3 and 4, which present the county-level results, we observe a pattern consistent with the contract-level analyses. Hostile counties exhibit a positive and statistically significant coefficient, indicating that they experienced a greater number of DOGE contract cuts relative to core counties. This suggests that electorally oppositional counties bore a disproportionate share of DOGE contract cuts.
Taken together, these findings—using an alternative operationalization of electoral support—reinforce the central result: counties that provided weaker electoral support to Trump in 2024 experienced more DOGE contract cuts. These patterns lend further support to the argument that presidential particularism can operate in a punitive form.
DISCUSSION
Drawing on evidence from DOGE and the geographic distribution of its contract cuts, this study extends existing research on presidential particularism by demonstrating that it can operate through the targeted distribution of harm. Across multiple operationalizations of electoral support and at both the contract and county levels, a consistent pattern emerges: contracts in counties that provided weaker support for Trump in 2024 were more likely to be cut, and these counties were more likely to experience a greater number of contract cuts overall. Taken together, these findings provide evidence of a systematic relationship between electoral opposition and the allocation of administrative harm.
These findings extend the literature on presidential particularism in important ways. While prior scholarship has largely emphasized the distributive logic of reward (Hudak Reference Hudak2014; Kriner and Reeves Reference Kriner and Reeves2015a; Lowande, Jenkins, and Clarke Reference Lowande, Jenkins and Clarke2018), this study provides evidence that presidents may also engage in targeted punishment. In doing so, it broadens the concept of presidential particularism beyond the allocation of benefits to include the strategic imposition of harm—revealing how modern presidents can leverage the administrative state to penalize electorally unsupportive communities. Presidential particularism, therefore, operates not only through the provision of material benefits but also through the targeted imposition of costs.
More broadly, these findings reinforce the view of presidents as particularistic political actors and challenge accounts that characterize the presidency as primarily universalistic in orientation (Howell Reference Howell2015; Howell and Moe Reference Howell and Moe2016). Rather than governing in an even-handed, national manner, presidents structure administrative decisions in ways that reflect electoral incentives and partisan boundaries.
Second, these findings complicate how we should understand DOGE’s contract cuts. Although the initiative was framed as a technocratic effort to eliminate government waste, the results suggest a more politically structured pattern. Counties with higher numbers of active federal contracts did experience more cuts, consistent with an efficiency-based rationale. However, the most consistent predictor of where cuts occurred was presidential electoral support. Importantly, this pattern does not appear to reflect a simple rollback of prior administration policies. Rather than disproportionately targeting active Biden-era contracts, the evidence suggests that the punitive dimension of presidential particularism operates less as inter-administration policy reversal and more as geographically targeted electoral sanctioning. The distribution of harm is tied more closely to presidential electoral incentives than to partisan undoing of a predecessor’s agenda. This distinction is theoretically important. It indicates that presidential particularism is driven primarily by electoral considerations—targeting places that failed to support the president—rather than by broader partisan or ideological conflict between administrations.
Finally, this analysis contributes to the literature on elite polarization and partisanship. As partisan divisions among political elites have intensified (McCarty, Poole, and Rosenthal Reference McCarty, Poole and Rosenthal2016), presidents have faced stronger incentives to behave as explicitly partisan actors. This study shows that polarization extends into the realm of administrative decision-making, with presidents using federal resources to institutionalize partisan conflict. While prior work on negative representation emphasizes rhetorical strategies for demonizing the opposition (Costa Reference Costa2025), this study demonstrates how such dynamics may manifest materially—through actions that harm electorally oppositional constituencies.
The DOGE case illustrates a clear instance of presidential particularism as punitive, yet further research is needed to determine whether similar dynamics extend across other domains of presidential governance—such as regulatory enforcement, disaster relief, or agency staffing. More broadly, these dynamics raise the possibility that retributive governance—the strategic use of administrative authority to impose material costs on political opponents—may be emerging as a recurring feature of contemporary executive politics. For example, prior research has emphasized the president’s capacity to impound previously appropriated funds when opposing their allocation (Wlezien Reference Wlezien1994). Presidential impoundment—or the strategic withholding of funds—may thus represent another punitive form of presidential particularism. More recently, the 2025 federal government shutdown granted Office of Management and Budget Director Russell Vought substantial discretion over funding decisions. Notably, he ordered the suspension of $2.1 billion allocated for a Chicago infrastructure project he characterized as race-based (Shepardson Reference Shepardson2025) and halted nearly $8 billion in climate-related spending directed primarily toward Democratic-leaning states (Frazin Reference Frazin2025). These episodes underscore the importance of assessing whether the logic of presidential particularism as punitive constitutes an institutionalized feature of contemporary presidential governance.
Additionally, future research should further examine the scope conditions of this punitive dimension of presidential particularism. The case of DOGE is characterized by high levels of centralization and politicization, given its location within the Executive Office of the President (Moe Reference Moe, Chubb and Peterson1985). As Dahlström, Fazekas, and Lewis (Reference Dahlström, Fazekas and Lewis2021) note, more politicized agencies exhibit partisan favoritism in the distribution of contracts. Using DOGE data, I examine a segment of the federal bureaucracy characterized by high levels of centralization and politicization. The results reveal substantively large and statistically significant relationships between presidential electoral outcomes and DOGE contract cuts. These findings raise the possibility that the punitive dimension of presidential particularism may be especially pronounced in institutional contexts marked by strong centralization and politicization. Future research should therefore assess whether, and to what extent, similar patterns emerge in less centralized and less politicized components of the federal bureaucracy, where presidential influence is more attenuated.
The findings presented here serve as a cautionary tale about the evolving nature of presidential power. As federal debt rises and political pressure mounts to cut government spending, this study highlights the danger that presidential electoral considerations may shape where and how those cuts occur. In an era of deep polarization and weakened institutional constraints, the presidency has become not only a source of distributive rewards but also a mechanism of punishment, capable of reshaping communities not on the basis of need or equity, but according to electoral loyalty.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/S0003055426102020.
DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available at the American Political Science Review Dataverse: https://doi.org/10.7910/DVN/SCKFS6.
ACKNOWLEDGEMENTS
I thank Federico Acosta y Lara, Taylor Carlson, Mike Olson, and Andrew Reeves for their feedback and support throughout the manuscript’s development.
CONFLICT OF INTEREST
The author declares no ethical issues or conflicts of interest in this research.
ETHICAL STANDARDS
The author affirms this research did not involve human participants.
Open access
Comments
No Comments have been published for this article.