Sage Journals: Discover world-class research

Abstract

Punishments are not always administered immediately after a crime is committed. Although scholars and researchers claim that third parties should normatively enact punishments proportionate to a given crime, we contend that third parties punish transgressors more severely when there is a time delay between a transgressor’s crime and when they face punishment for it. We theorize that this occurs because of a perception of unfairness, whereby third parties view the process that led to time delays as unfair. We tested our theory across eight studies, including two archival data sets of 160,772 punishment decisions and six experiments (five preregistered) across 6,029 adult participants. Our results suggest that as time delays lengthen, third parties punish transgressors more severely because of increased perceived unfairness. Importantly, perceived unfairness explained this relationship beyond other alternative mechanisms. We explore potential boundary conditions for this relationship and discuss the implications of our findings.

Keywords

cognition morality punishment open data preregistered

Governing bodies, organizations, and judges primarily aim to enact punishments proportional to a given crime (Keller et al., 2010). In determining punishment severity, extant theory indicates that people are intuitive punishment theorists and punish others in a proportional and deserved manner to give transgressors their just deserts (Carlsmith, 2006; Crockett et al., 2014). These perspectives suggest that decision-makers are primarily driven by restoring fairness by enacting an appropriate punishment vis-à-vis the crime (Mooijman & Graham, 2018).

However, there are times when individuals cannot immediately administer punishment to transgressors. Indeed, time delays—or gaps between when a crime was committed and when a transgressor experiences punishment—are frequent and common.¹ For example, government agencies often fail to immediately sanction citizens for lying on their tax returns because they lack the resources to detect fraud when it happens (Brickner et al., 2010). Similarly, overburdened police departments with limited resources often prioritize certain crimes over others, leading to time delays in convicting criminals (Martin & Sherman, 1986). Furthermore, the COVID-19 pandemic has generated a significant backlog of court cases (Witte & Berman, 2021).

Despite the ubiquity of time delays in enacting justice, normative perspectives on punishment (Treviño & Weaver, 1998) suggest that time delays should be irrelevant when determining punishment severity. This is because fundamental principles of penal justice suggest that punishments should vary only with regard to the “moral gravity” of the alleged wrongdoing, and morally similar cases should be treated alike, regardless of other factors (Hart, 2008, p. 80). Relatedly, scholars have argued that it would be normatively inappropriate to use time as an input for punishment decisions and have suggested that, if anything, it would be appropriate for negative attitudes toward transgressors to weaken over time (Coleman & Sarch, 2012). Collectively, these perspectives suggest that punishments should fit the crime, and differences in punishment for similar crimes due to external factors—such as time—would be unjust. This is especially the case when time delays are not the fault of the transgressor, and transgressors do not use the time delay to engage in additional nefarious behaviors. Thus, normative perspectives suggest that a transgressor who faces punishment immediately should be punished similarly to a transgressor who is punished long after the crime was committed.

Despite the normative prerogative that time delays should not influence punishment severity, we draw on—and advance—a descriptive view of punishment to suggest that time delays may increase punishment severity. On the basis of extensive research that portrays people as intuitive justice theorists (Bies, 1987; Mooijman & Graham, 2018), we suggest that time delays could lead to increases in punishment severity because people interpret time delays—more specifically the process that led them to occur—as unfair. Indeed, evaluators often believe that crimes should be swiftly followed by an appropriate punishment that will provide transgressors their just deserts (Kleiman et al., 2014; McDonnell & Nurmohamed, 2021). However, time delays, at least temporarily, enable transgressors to avoid facing the repercussions that typically follow a crime. That is, third parties generally believe that transgressors should receive an appropriate punishment shortly after their crime, yet time delays inhibit this process. Furthermore, third parties may view time delays as especially problematic because of the experiential nature of time—time delays are irreversible and cannot be taken back. Thus, when transgressors experience time delays between their crime and corresponding punishment, even when the delays are outside of their control and/or not their fault, they gain a period of time in which they are uniquely unpunished for their specific transgression, a process that we contend third parties will view as unfair (Cohen-Charash & Mueller, 2007).

Importantly, this perception of unfairness is likely to increase punishment severity. Existing literature underscores the role of perceptions in fairness, or subjective feelings, leading to irrational decisions or harsher judgments (Greenberg, 1983; Thompson & Loewenstein, 1992). For example, feeling wronged can lead people to treat others with anger and harsher judgments (e.g., Fetchenhauer & Huang, 2004; Zitek et al., 2010). Moreover, individuals tend to enact greater punishment for fairness violations when they are not the direct victim of the crime, as with third parties and judges (FeldmanHall et al., 2014). Thus, we argue that this perception of unfairness leads third parties to levy more severe punishments against transgressors to account for the unpunished time that the transgressor is perceived to have “unfairly” experienced.

Statement of Relevance

Transgressors are not always punished immediately after they commit a crime. Despite scholars suggesting that these time delays should not influence punishment, our research shows that transgressors who experience time delays are punished more severely because people interpret these delays as unfair, even when transgressors are not responsible for the delay. We found evidence of these effects across judges and committees responsible for administering punishment in two archival studies and replicated our findings in six additional experiments. Our research demonstrates that extraneous factors that are not supposed to influence punishment may nevertheless lead people to exact harsher punishments.

In short, our theoretical contention is that time delays result in more severe punishments for transgressors because third parties believe that time delays are unfair. Thus, we identify the role of time delays in biasing evaluators’ decision-making and highlight how delays can lead them to punish the wrongdoer more severely because of perceptions of unfairness.

Overview of Studies

We tested our hypotheses across eight studies. In two archival data sets, a sample of convicted civilian transgressors (Study 1) and another of convicted police officers (Study 2), we found support for our proposed effect of time on punishment, even when controlling for a host of alternative explanations. In the subsequent six experimental studies, we aimed to strengthen causal inferences regarding the effect of time delays on punishment severity. We found that perceptions of unfairness explain harsher punishment beyond several cognitive and affective mechanisms (Study 3). To help rule out alternative explanations, we replicated this effect even when the transgressor was not responsible for the time delay (Study 4a) and when third parties were explicitly informed that the transgressor did not commit additional crimes during the time delay (Study 4b). Moreover, we replicated our effect when the transgressor was arrested immediately (Study 5a) and when the transgressor committed a crime that provided no ongoing material benefit (Study 5b). Finally, we conducted an experiment in which we turned off the established effect of time delays on punishment severity to provide stronger evidence for unfairness as the mechanism (Study 6). Our experimental studies were reviewed and approved by the institutional review board at the University of Pennsylvania.

Open Practices Statement

Data, syntax, and output for all eight studies have been made publicly available at OSF and can be accessed at https://osf.io/w3qrn/. Five of the studies were preregistered: Study 4a: https://aspredicted.org/MPG_T3C, Study 4b: https://aspredicted.org/G68_GBZ, Study 5a: https://aspredicted.org/9SR_7BC, Study 5b: https://aspredicted.org/6D7_FVX, and Study 6: https://aspredicted.org/KD5_7LF.

Study 1

Study 1 tested the relationship between time delays and punishment in an archival data set of felony sentences from Cook County, Illinois. Starting in 2017, Cook County has released this anonymized data to increase transparency in the Cook County State Attorney’s Office. More information can be found at https://osf.io/w3qrn/. We downloaded the data in May of 2021. In this study and Study 2, we assumed that no unmeasured time-varying confounding factors would influence both time and punishment (Rohrer & Lucas, 2021). We tested alternative models designed to validate the appropriateness of this assumption and rule out potential spurious associations and confounds.

Method

Context and sample size

These data were ideal for multiple reasons. First, this data set was comprehensive, spanning felonies from 1980 to 2020 and sentencing decisions from 2010 to 2021. Second, this data set included specific sentences from various judges within the Cook County judicial system, offering a clear and objective measure of punishment. Third, this data set provided information on date of crime, arrest, and sentencing, providing multiple conceptualizations of our independent variable that allowed us to run robustness checks. Fourth, this data set included an array of measures for potential confounds, allowing us to increase confidence in our findings.

The data set includes 245,891 sentencing decisions of 209,894 transgressors. We focused only on original sentencing decisions and removed decisions where transgressors were resentenced or had their sentences changed. Moreover, Cook County (https://datacatalog.cookcountyil.gov/api/views/tg8v-tm6u/files/8597cdda-f7e1-44d1-b0ce-0a4e43f8c980?download=true&filename=CCSAO%20Data%20Glossary.pdf) notes that a case is “usually referred to by its primary charge,” and, as a result, we focused on the primary charge that each transgressor faced, providing a conservative test of our hypotheses.² Thus, we removed secondary sentencing decisions from our data (i.e., those that were not for the primary charge). However, as we discuss below, we controlled for the number of counts a transgressor was charged with. In addition, our supplementary analyses included secondary charges and nested each sentencing decision within a transgressor, finding similar results. These exclusions resulted in a sample of 169,587 sentencing decisions (i.e., one decision per person per case).

In addition, we also removed nine cases with death sentences, as the death penalty was outlawed in Illinois in 2011, suggesting that these cases would have been treated differently after this date. Moreover, we included only cases in which the defendant was found or pleaded guilty, removing the rare cases of dismissal or decision not to prosecute. At this point, we had 168,668 sentencing decisions. Finally, we conducted listwise deletion where there were missing data for any of our listed measures or controls, resulting in a final sample of 150,392 sentencing decisions for our primary analyses (see Table 1 for descriptive statistics and correlations).³

Table 1.

Study 1: Descriptive Statistics and Correlations

Variable	M	SD	Correlations
Variable	M	SD	1	2	3	4	5	6
1. Time delay (days)	33.16	220.99	—
2. Male	.87	.33	−.00	—
3. Number of counts	1.26	1.63	.03	.01	—
4. Pled guilty	.95	.22	−.03	−.03	−.17	—
5. Age (years)	32.78	11.96	.01	−.04	−.03	−.00	—
6. Minority	.85	.36	−.02	.08	.01	−.03	−.06	—
7. Months sentenced to prison	22.72	52.70	.13	.10	.22	−.22	−.01	.06

Note: N = 150,392. Correlations greater than |.01| are significant at p < .01. Male was coded 1 for male and 0 for female. Pled guilty was coded 1 for pled guilty and 0 for did not plead guilty. Minority was coded 1 for not White, including Hispanic, and 0 for White.

Core measures

Time delay

We operationalized time delay as the number of days from when the crime was committed to when the transgressor was arrested. Our robustness analyses included additional conceptualizations of time delays.

Punishment

We operationalized punishment as the number of months an individual was sentenced to prison.⁴ In the case of life sentences, we converted these sentences to 120 years (1,440 months). However, our results were unaffected when these transgressors were removed from our data or when we converted the sentence to various lengths (i.e., 100 years/1,200 months).

Controls

Following recent guidelines (Bernerth & Aguinis, 2016), we included control variables to increase our ability to make causal inferences about the relationship between time delays and punishment, despite the correlational nature of our data. Thus, we aimed to conduct various robustness checks to increase confidence in our findings. To be as conservative as possible, we calculated these variables before making any exclusions for listwise deletion. In our main set of analyses, we included the control variables described below. However, as shown in our supplementary analyses, we also conducted our analyses without these controls, finding similar results.

Number of counts

To rule out the alternative explanation that an increase in time delay leads to more severe punishment because additional time provides increased opportunities to engage in more crimes, we controlled for the number of counts transgressors faced. To be as conservative as possible, we used the charge with the highest number of counts, even if it was not the primary charge. However, results were consistent when we used only the number of counts for the primary charge.

Guilty plea

We controlled for instances in which the transgressor pleaded guilty or not (0 = no guilty plea, 1 = guilty plea).

Type of crime

We included fixed effects for the specific category of crime the transgressor committed. To be as conservative as possible, we used the most specific crime type available (e.g., “possession of a controlled substance with intent to deliver”), but results were consistent when we used a broader crime category (e.g., “narcotics”). The three most common crimes were possession of a controlled substance, aggravated driving under the influence of alcohol, and retail theft.

Judge

We included fixed effects for the specific judges in our data. Including judge fixed effects made this a more conservative test because they allowed us to account for any individual-level variance associated with a particular judge.

Year the crime was sentenced

We controlled for fixed effects at the year level to address the possibility that transgressors were punished differently in different years.

Demographic information

We controlled for the age, race, and gender of the transgressor. Cook County documentation indicates that these data were recorded at the time of the incident.

Results

Primary analysis

We first entered our control variables and fixed effects into our model (Table 2, Model 1). Next, we entered time delay, the focal independent variable. Results indicated that there was a positive effect of time delay on punishment (b = 0.004, SE = 0.001, p < .0001; Table 2, Model 2). Using the lm.beta package in R (Behrendt, 2014), we then estimated the standardized coefficient of time delay (b = .13, p < .0001).⁵

Table 2.

Study 1: Results of the Regression Predicting the Effect of Time Delay on Punishment

Variable	Model 1:controls	Model 2:primary analysis	Model 3:no fixed effects
Intercept	−7.813 (37.592)	−7.885 (37.583)	38.329** (0.861)
Time delay		0.004** (0.001)	0.030** (0.001)
Male	8.366** (0.301)	8.389** (0.301)	14.993** (0.386)
Number of counts	2.222** (0.061)	2.232** (0.061)	5.939** (0.080)
Pled guilty	−23.119** (0.466)	−23.124** (0.465)	−43.848** (0.604)
Age	0.362** (0.009)	0.362** (0.009)	−0.020 (0.011)
Minority	2.404** (0.292)	2.428** (0.292)	6.064** (0.363)
Fixed effects	Yes	Yes	No
R²	.520	.520	.109
Adjusted R²	.516	.516	.109

Note: N = 150,392. Except where noted, the table shows unstandardized coefficients (standard errors are reported in parentheses). Male was coded 1 for male and 0 for female. Plead guilty was coded 1 for pled guilty and 0 for did not plead guilty. Minority was coded 1 for not White, including Hispanic, and 0 for White.

p < .001.

Robustness analyses

We conducted a series of additional analyses as robustness checks to increase confidence in the interpretation of our results.

Without controls

We analyzed our results without controls, finding similar effects of time delay on punishment (b = 0.032, SE = 0.0006, p < .0001).

Alternative conceptualizations of time delays

We conducted our analyses where we conceptualized time delays as the time between the crime and sentencing, finding similar results (b = 0.013, SE = 0.0003, p < .0001). We also conceptualized time delays as the time from arrest to sentencing, finding similar results (b = 0.02, SE = 0.0004, p < .0001). Although we cannot be sure that transgressors were unpunished in these analyses—because we did not have data on whether or not they received bail—prior data support the suggestion that the majority of felony defendants are released until sentencing (T. H. Cohen & Reaves, 2007).

Curvilinear relationship

We used Simonsohn’s (2018) “two-line” test to determine whether there was a significant and meaningful U-shaped effect of time delays on punishment. Results indicated that the slope was positive and significant when the time delay was lower (b = 2.07, p < .0001) and was positive and significant when the time delay was higher (b = 0.02, p < .0001), which rejects the suggestion that there might be a curvilinear relationship.

Mixed-effects model with different sampling criteria

We also tested our hypotheses on the full data set, in which multiple sentencing decisions (N = 204,954) could be nested within each transgressor (N = 183,855). To do so, we used the lme4 (Bates et al., 2015) and lmerTest (Kuznetsova et al., 2017) packages in R (R Core Team, 2022) to fit a multilevel model with random intercepts specified at the transgressor level. Results indicated a significant positive effect of time delays on punishment (b = 0.03, SE = 0.0007, p < .0001).

Discussion

Study 1 demonstrated that Cook County judges gave harsher sentences as the time between crime and arrest increased. The effect ranged from small to medium, even after we controlled for possible confounds (Grissom & Kim, 2005). To further complement this analysis, we conducted Study 2, in which (a) the transgressors were mostly White police officers (vs. mostly minority-group member civilians in Study 1; see Table 1), (b) the misconduct consisted of violations of New York Police Department (NYPD) rules (vs. societal laws), and (c) the sentencing decisions were punishment recommendations by an independent agency (vs. a county judge).

Study 2

In Study 2, we examined violations of NYPD rules by police officers and the sentencing recommendations by an independent agency in an archival data set. As in Study 1, our analyses rested on the assumption that no unmeasured time-varying confounding factors would influence both time and punishment (Rohrer & Lucas, 2021). We tested alternative models designed to validate the appropriateness of this assumption and rule out potential spurious associations and confounds.

Method

Context and sample size

We tested our hypotheses using police misconduct archival data. The New York Civil Liberties Union (NYCLU; 2021) publicly posted data on police misconduct from 1984 to 2020. We downloaded the data in May of 2021. This data set includes information on complaints filed against NYPD police officers and punishments recommended for said complaints by the Civilian Complaint Review Board (CCRB), an independent agency. The initial data set includes 279,644 unique accusations records involving 102,121 complaints and 48,757 NYPD officers. As in Study 1, we analyzed only cases in which individuals were found guilty of committing a transgression (i.e., an accusation was substantiated) and in which punishment recommendation data were available. These data yielded 17,300 punishment recommendations.

Furthermore, as in Study 1, we limited our analyses to the primary, most severe charge an officer faced within a given complaint. By doing so, we mitigated the possibility that more complex cases (e.g., more officers involved, more charges associated with a single officer) were differentially punished. To do so, we (a) removed instances in which individual police officers faced multiple charges of equally high severity within the same complaint and/or (b) focused on the most severe charge in cases in which an officer faced multiple accusations. Thus, our final sample size was 10,380 punishment recommendations (i.e., one recommendation per person per case). As detailed in the robustness analyses section below, our results did not change when we included all accusations.

Core measures

Time delay

Our data did not provide a date of when the transgressor was apprehended, as in Study 1. Accordingly, we conceptualized time delays as the number of days from when the incident was reported to when the case was closed by the CCRB. A “closed date” represents when the CCRB makes a punishment recommendation and thus conceptually falls between the arrest date and the sentencing date used in Study 1. However, there were three dates recorded during CCRB investigations that we further explored as robustness checks.

Punishment

We operationalized punishment using the CCRB’s internal classification system for disciplinary actions. This system differentiates between three levels of punishment (for full details on punishment categories, see CCRB, 2023b). First, and most lenient, the CCRB can recommend punishing police officer wrongdoing using “instructions or formalized training” (e.g., mandatory training at the NYPD police academy or a warning). Second, and more punitive, the CCRB can recommend punishing police officers using “command disciplines” (e.g., losing up to 10 vacation days). Third, and most punitive, the CCRB can recommend punishing police officers using “charges and specifications” (e.g., losing more than 10 vacation days, being suspended, being terminated).

In line with this framework, we coded punishment using a 3-point scale—providing us with a measure of punishment in which higher values indicated more punishment severity (1 = instructions/formalized training, 2 = command disciplines, and 3 = charges and specifications). Of the 10,380 observations that the CCRB recommended punishments for, 1,876 were instructions/formal training (low punishment), 3,111 were command disciplines (medium punishment), and 5,393 were charges and specifications (high punishment).

Controls

We included a number of control variables in our focal analyses to help rule out alternative explanations. As in Study 1, and to be as conservative as possible, we calculated control variables before making any exclusions via listwise deletion. Our results did not substantively change when we conducted our analyses without these controls.

Year crime occurred

We controlled for fixed effects at the year level to address the possibility that officers were punished differently in different years.

Number of charges officer faced

We controlled for the number of accusations an officer faced in a given complaint. This variable addressed the possibility that a given officer would be punished more (vs. less) harshly when they were accused of multiple transgressions (vs. a single transgression) in the same complaint (e.g., a victim accused the officer of abusing force, using racial epithets, and attempting a bribe).

Number of officers associated with a given complaint

We controlled for the number of officers accused in a given complaint. This accounted for the possibility that accusations involving multiple officers within a single complaint (e.g., multiple officers jointly assaulting a single victim) would be punished more (vs. less) harshly than accusations involving one officer.

Type of accusation officer faced

We controlled for the type of accusation officers faced by coding the type of accusation using a classification system for misconduct outlined by the CCRB and NYPD. Both organizations differentiate misconduct into one of four categories based on its scope: (a) force, (b) abuse of authority, (c) discourtesy, and (d) offensive language. Force refers to the “use of excessive or unnecessary force” (e.g., punching or deadly force), abuse of authority refers to “abuse of police powers to intimidate or mistreat a civilian” (e.g., improper stop and frisk), discourtesy refers to “cursing and using other foul language or gestures,” and offensive language refers to “slurs or derogatory remarks or gestures based upon race, ethnicity, religion, gender, sexual orientation, or physical disability” (CCRB, 2023a).

Demographic information

We controlled for the accused officer’s gender, race, and years worked at the NYPD.

Results

Means, standard deviations, and correlations are displayed in Table 3. Results from our regression analyses are shown in Table 4.

Table 3.

Study 2: Descriptive Statistics and Correlations

Variable	M	SD	Correlations
Variable	M	SD	1	2	3	4	5	6	7	8	9	10
1. Time delay (days)	251.24	184.91	—
2. Punishment severity	2.34	0.77	.08	—
3. Number of accusations	2.72	1.78	.11	.03	—
4. Number of officers	2.22	1.24	.19	−.02	.01	—
5. Accusation type: abuse of authority	0.62	0.49	.26	−.12	.03	.13	—
6. Accusation type: discourtesy	0.14	0.34	−.05	−.15	−.04	−.10	−.51	—
7. Accusation type: use of force	0.23	0.42	−.25	.26	.01	−.05	−.69	−.22	—
8. Accusation type: offensive language	0.02	0.12	−.01	.00	−.02	−.03	−.16	−.05	−.07	—
9. Female	0.08	0.27	.01	−.06	−.08	−.01	.03	.02	−.06	.02	—
10. Minority	0.42	0.49	.09	−.05	.01	−.00	.06	−.02	−.05	−.01	.15	—
11. Years at NYPD	8.31	5.48	.14	.03	.04	−.01	.11	.01	−.13	.00	−.07	−.05

Note: N = 10,380. Correlations greater than |.02| are significant at p < .05. Female was coded 1 for female and 0 for male. Minority was coded 1 for not White and 0 for White. NYPD = New York Police Department.

Table 4.

Study 2: Results of the Ordered Logistic Regression Predicting the Effect of Time Delay on Punishment

Variable	Model 1:controls			Model 2:primary analysis			Model 3:no fixed effects			Model 4:no controls
Variable	b	SE	OR	b	SE	OR	b	SE	OR	b	SE	OR
Time delay^a				0.001**	0.000	1.001	0.002**	0.000	1.002	0.002**	0.000	1.002
Number of accusations	0.101**	0.012	1.11	0.096**	0.012	1.100	−0.004	0.011	0.996
Number of officers	0.123**	0.018	1.13	0.106**	0.018	1.111	−0.077**	0.016	0.926
Accusation type: abuse of authority	−0.128	0.168	0.880	−0.115	0.168	0.892	−0.235	0.154	0.791
Accusation type: discourtesy	−0.702**	0.174	0.496	−0.685**	0.174	0.504	−0.670**	0.160	0.512
Accusation type: use of force	1.406**	0.174	4.08	1.400**	0.174	4.055	1.250**	0.160	3.490
Female	−0.340**	0.076	0.712	−0.333**	0.076	0.716	−0.261**	0.070	0.770
Minority	0.026	0.044	1.03	0.024	0.044	1.024	−0.188**	0.040	0.829
Years at NYPD	0.034**	0.004	1.03	0.033**	0.004	1.033	0.015**	0.004	1.015
Fixed effects	Yes			Yes			No			Yes
Pseudo R²	.408			.410			.124			.329

Note: N = 10,380. Offensive language was excluded as a predictor from the analyses because including it caused the model to be rank deficient (because of its rarity in the data set; see Table 3). Female was coded 1 for female and 0 for male. Minority was coded 1 for not White and 0 for White. OR = odds ratio; NYPD = New York Police Department.

Time delay was measured in days.

p < .001.

Primary analysis

Because our dependent variable was ordinal rather than continuous (i.e., the three tiers of punishment were not necessarily equally different in magnitude), we used ordered logistic regression to test our hypothesis (e.g., Lerner et al., 2013; Packard & Berger, 2020). Support for our hypothesis did not change when we used ordinary least squares (OLS) regression instead.

We first entered our control variables and fixed effects into the model. Next, we entered our independent variable of the time delay between crime and punishment. As in Study 1, there was a significant positive effect of time delay on punishment (b = 0.0010, SE = 0.0001, p < .0001).

Robustness analyses

We summarize a series of additional robustness analyses below, which we conducted to increase our confidence in interpreting these results and to rule out possible alternative explanations.

Results without controls

Even when we excluded the above control variables, time delays were significantly positively associated with punishment (b = 0.0019, SE = 0.0001, p < .0001).

Alternative conceptualizations of time delay

In our primary analysis, we conceptualized time delay as the number of days from when the incident was reported to when the case was closed by the CCRB. Given the typical CCRB investigation timeline, these two dates were most pertinent to our theory. However, we also incorporated the received date to conduct robustness checks. Our results were similar regardless of which operationalization of time delay we used. When we controlled for all the variables above, the time between when the CCRB received the complaint and when the CCRB recommended punishment was significantly positively related to punishment severity (b = 0.0009, SE = 0.0002, p < .0001). We also explored whether the time between the incident and received dates was positively related to punishment. It is important to note that this length of time is likely too early in an incident’s timeline to constitute a true delay because it is before the CCRB is aware of a crime and can investigate the incident. Still, when we included the controls and fixed effects outlined above, this variable was positively related to punishment (b = 0.0011, SE = 0.0005, p = .0305). Overall, these results were substantively similar to those of our focal hypothesis tests.

Curvilinear relationship

As in Study 1, we tested for a potential U-shaped effect of time delays on punishment. Simonsohn’s (2018) two-line test, using OLS regression, revealed that the slope was positive and significant when the time delay was lower (b = 0.00003, p = .0173) and still positive and marginally significant when the time delay was higher (b = 0.00000, p = .0605). This fails to support the possibility that there might be a curvilinear relationship.

Mixed-effects model with different sampling criteria

We also tested our hypotheses on the full data set, in which multiple accusations could be nested within a single complaint (N = 17,300). To do so, we fitted a multilevel model with random intercepts specified at the complaint level. Results indicated a significant positive effect of time delay on punishment (b = 0.0004, SE = 0.0001, p < .0001). Thus, we found substantively similar results.

Discussion

In Study 2, we found support for our hypothesis using a different context (police officer misconduct) and operationalizations of our focal variables. In two archival studies, we took multiple steps to better understand the effect of time delays on punishment (e.g., sampling different contexts, operationalizing time delays in different ways, including control variables to prevent spurious associations). Despite these steps, the archival and correlational nature of Studies 1 and 2 limited our ability to draw causal inferences or identify explanatory mechanisms. Therefore, we conducted a series of experiments to strengthen causal inferences.

Study 3

In Study 3, we sought to constructively replicate our results in an online experiment and test our proposed mediating mechanism. Across Studies 3 to 5, we tested whether time influences punishment because of unfairness and assumed that unfairness temporally precedes punishment. In Study 3, we ruled out alternative mechanisms, including specific forms of moral outrage, competence, and deterrence.

Method

Sample size and exclusions

Our power analysis indicated that we would require 191 participants to detect an effect of small to medium size (J. Cohen, 1992). We treated this as a minimum and recruited 500 participants to ensure that our study was appropriately powered and to account for potential exclusions. Because TurkPrime can oversample by a small margin, we recruited 505 participants from Mechanical Turk (MTurk) to complete the study. We removed 33 participants who incorrectly reported a key detail of our manipulation (the time between the crime and the arrest), five participants who provided a nonsense response to an open-ended question (“What did you have for breakfast today?”), six participants who had suspicious Internet protocol (IP) addresses (Prims et al., 2018), and 18 participants who used automatic form fillers (Buchanan & Scofield, 2018). Given that some participants failed multiple checks, this resulted in a final sample of 450 participants (age: M = 38.28 years, SD = 12.35; 46% female; 72% White). We used these same exclusion criteria across our remaining studies.

Procedure

Each participant was randomly presented with one of two different conditions, in which we manipulated the time that elapsed between the crime and the arrest. All participants received the same information about the crime and how the transgressor was identified in the two conditions: “Jamie shoplifted $1698 worth of merchandise from an electronics store on a Saturday night. Police identified Jamie using security cameras.”

To manipulate time delays, we informed participants in the short-time-delay condition that Jamie was arrested less than 24 hr later, and we informed participants in the long-time-delay condition that Jamie was arrested 30 days later. We chose 30 days as the time interval for the long-time-delay condition because it represented the average amount of time in Study 1 (i.e., the time between the crime and the arrest) and thus provided a conservative test of our hypotheses.

Punishment

To measure punishment, we asked participants how many months the transgressor should be sentenced to prison (0–40 months, in 2-month increments).

Perceived unfairness of process

To test whether perceived unfairness⁶ was the explanatory mechanism, we adapted a three-item measure from Cohen-Charash and Mueller (2007; originally used by Smith et al., 1994) to fit our scenario: “[Jamie] gained an unfair advantage,” “[Jamie] achieved an advantage through unjust procedures,” “[Jamie] unfairly escaped justice” (α = .87). Anchors were 1 (not at all) and 7 (very); higher scores indicate greater perceptions of unfairness.

Alternative mechanisms

We also investigated alternative explanations based on prior theory and research. Specifically, time delays may spur moral outrage at the transgressor’s behavior, driving more severe punishment. We assessed moral outrage directed at the transgressor’s behavior with three items from Batson et al. (2007; e.g., “Jamie’s behavior makes me” . . . “outraged,” “angry,” “offended”; α = .91). Moreover, time delays may result in transgressors being perceived as more competent, warranting more severe punishment. We assessed competence of the transgressor with three items from Fiske et al. (2007; e.g., “Jamie is” . . . “intelligent,” “skillful,” “competent”; α = .94). Finally, some scholars contend that punishment is used to deter other people from committing crimes or refusing to turn themselves in (Carlsmith et al., 2002). To address this, we created a deterrence scale based on language from Carlsmith et al. (2002; e.g., “It needs to be made clear to others that Jamie’s behavior is unacceptable,” “Jamie should be made an example of wrongdoing to others,” “It is important for others to know that Jamie’s behavior is not tolerated”; α = .88). Thus, in addition to testing the central relationship between time delay and punishment (and perceptions of unfairness as the mediating mechanism), we explored moral outrage, competence, and deterrence as alternative mechanisms.

Results

Manipulation check

We asked participants to indicate whether “a long time” had passed before Jamie was arrested (1, strongly disagree, to 5, strongly agree). As expected, participants in the long-time-delay condition (M = 3.71, SD = 0.96) were more likely to agree than participants in the short-time-delay condition (M = 1.50, SD = 0.85), F(1, 448) = 662.36, p < .001, η_p² = .60.

Punishment

A one-way analysis of variance revealed a significant effect of our manipulation on punishment as participants in the long-time-delay condition (M = 12.72, SD = 9.65) levied more severe punishment than participants in the short-time-delay condition (M = 10.09, SD = 7.77), F(1, 448) = 10.08, p = .002, η_p² = .02.

Indirect effect of perceived unfairness

Next, we tested the mediating role of perceived unfairness. We found that the time delay increased perceptions of unfairness (b = 1.42, SE = 0.16, p < .0001, 95% confidence interval [CI] = [1.10, 1.73]). Moreover, unfairness was positively related to punishment (b = 1.33, SE = 0.24, p < .0001, 95% CI = [0.86, 1.80]), and controlling for unfairness eliminated the effect of time delays on punishment (b = 0.75, SE = 0.87, p = .39, 95% CI = [–0.96, 2.46]). Most notably, there was a significant indirect effect of time delay on punishment through unfairness (b = 1.89, SE = 0.42, 95% CI = [1.10, 2.79]).

Alternative mechanisms

We conducted additional analyses in which we modeled all mechanisms (unfairness, moral outrage, competence, deterrence) as parallel mediators. We found that the indirect effect through unfairness remained significant (b = 0.97, SE = 0.39, 95% CI = [0.23, 1.80]), and there was no significant indirect effect through moral outrage (b = 0.34, SE = 0.21, 95% CI = [–0.04, 0.83]), competence (b = 0.29, SE = 0.25, 95% CI = [–0.20, 0.82]), or deterrence (b = 0.08, SE = 0.14, 95% CI = [–0.20, 0.37]; see Fig. 1).

Fig. 1.

Study 3: influence of time delay on punishment, as mediated by competence, unfairness, deterrence, and moral outrage. Green arrows indicate significant paths and red arrows indicate nonsignificant paths. We modeled but do not visually depict the nonsignificant direct effect of time delay on punishment (b = 0.95, SE = 0.83, p = .26).

Discussion

Study 3 replicated our results from Studies 1 and 2 and provided evidence for perceptions of unfairness as the underlying mechanism. Although Study 3 ruled out several alternative explanations, it might be the case that third parties infer that the transgressors (a) were responsible for the time delays or (b) committed additional crimes during the time between crime and punishment, both of which could warrant greater punishment. As a result, we conducted follow-up experiments to investigate and rule out these possibilities.

Study 4a

In Study 4a, we sought to constructively replicate our results from Study 3 and rule out the possible explanation that transgressors incurring longer time delays are punished more severely because they are viewed as responsible for the time delay.⁷

Method

We preregistered the study design, sample size, hypothesis, and analysis plan in advance (https://aspredicted.org/MPG_T3C).

Sample size and exclusions

We sought to recruit 500 participants. As discussed in Study 3, our power calculation suggested that detecting a small to medium effect required approximately 200 participants. Study 3 confirmed that the effect size of time on punishment was indeed small to medium. To account for the extra manipulated factor, we doubled this and added a buffer to account for exclusions, recruiting 501 participants from MTurk. Following the preregistered exclusion criteria, we removed participants who did not correctly identify the time between crime and arrest (n = 4), provided a nonsense response to an open-ended question (“What did you have for breakfast today?”; n = 15), had suspicious IP addresses (n = 5), or used automatic form fillers (n = 10). Given that some participants failed multiple checks, our final sample was 470 (age: M = 40.46 years, SD = 12.03; 43% female; 81% White).

Procedure

Using a between-subjects design, we randomly assigned each participant to one of two time-delay conditions (same conditions as in Study 3). To rule out the possibility that observers punish those in the long-time-delay condition because they see them as responsible for creating the time delay, we included an additional factor: Participants were informed that defective security cameras resulted in the time delay (a time delay of either 24 hr or 30 days). Participants in the security-camera-information condition received the following additional details: “Because there were technical issues with the security cameras used to identify Jamie, it took police [24 hours/30 days] to locate and arrest Jamie.” Thus, we had a 2 (time delay: short vs. long) × 2 (security camera information: present vs. control) factorial design.

Measures

We used the same measures of punishment and unfairness (α = .91) as in Study 3.

Results

Manipulation check

We used a similar manipulation check as in Study 3. Consistent with our manipulation, results showed that participants in the long-time-delay condition (M = 3.52, SD = 1.10) were more likely than participants in the short-time-delay condition (M = 1.28, SD = 0.64) to agree that a long time had passed before Jamie was arrested, F(1, 467) = 715.65, p < .0001, η_p² = .61. Moreover, we asked participants to indicate whether there were issues with the security cameras. As expected, those in the security-camera-information condition (M = 4.34, SD = 1.17) were more likely to agree than participants in the control condition (M = 1.51, SD = 0.94), F(1, 467) = 871.57, p < .0001, η_p² = .65.

Punishment

We began by testing the effects of time delays on punishment, which revealed a significant effect: Participants in the long-time-delay condition (M = 12.03, SD = 8.57) levied more severe punishment than participants in the short-time-delay condition (M = 10.43, SD = 7.63), F(1, 467) = 4.51, p = .03, η_p² = .01. There was no significant main effect of the security-camera-information condition, F(1, 467) = 0.02, p = .88, η_p² = .00, and no interaction with the time-delay condition, F(1, 466) = 0.01, p = .92, η_p² = .00.

Indirect effect of perceived unfairness

Next, we tested the mediating role of perceptions of unfairness. We found that time delays increased perceptions of unfairness (b = 1.50, SE = 0.14, p < .0001, 95% CI = [1.22, 1.79]). Moreover, unfairness was positively related to punishment (b = 1.40, SE = 0.24, p < .0001, 95% CI = [0.94, 1.85]), and controlling for unfairness eliminated the effect of the time delay on punishment (b = −0.50, SE = 0.80, p = .53, 95% CI = [–2.08, 1.08]). Importantly, there was a significant indirect effect of time delays on punishment through unfairness (b = 2.10, SE = 0.42, 95% CI = [1.33, 2.94]).

Study 4b

In Study 4b, we sought to rule out the argument that transgressors incurring longer time delays receive more severe punishment because they are assumed to have committed more crimes during the period between the crime and the arrest.

Method

We preregistered the study design, sample size, exclusions, and analysis plan in advance (https://aspredicted.org/G68_GBZ).

Sample size and exclusions

We used the same power calculation as in Study 4a. Five hundred two participants from MTurk completed the study. Following our preregistered exclusion criteria, we removed participants who did not correctly identify the time between crime and arrest (n = 12), provided a nonsense response to an open-ended question (“What did you have for breakfast today?”; n = 15), had suspicious IP addresses (n = 9), or used automatic form fillers (n = 7). Some participants failed multiple checks, leaving a final sample of 466 (age: M = 39.93 years, SD = 11.84; 42% female; 76% White).

Procedure

We randomly assigned each participant to one of the same two time-delay conditions as in the prior experimental studies. To rule out the argument that additional crimes were driving this effect, we included a second factor: no crime information. In one condition participants were explicitly informed that the transgressor did not commit any additional crimes: “Jamie did not commit any additional crimes between the time of the shoplifting and their arrest.” Participants in the other condition did not receive this information. Thus, we used a 2 (time delay: short vs. long) × 2 (no crime information: present vs. control) factorial design.

Measures

We used the same measures of punishment and unfairness (α = .90) as in the prior experimental studies.

Results

Manipulation check

Using a similar manipulation check as in the previous studies, we found that participants in the long-time-delay condition (M = 3.16, SD = 1.12) were more likely than those in the short-time-delay condition (M = 1.30, SD = 0.69) to agree that a long time had passed between crime and arrest, F(1, 463) = 466.57, p < .001, η_p² = .50. Moreover, we asked participants to indicate their agreement that Jamie had committed additional crimes during the time delay. Those in the control condition (M = 2.02, SD = 1.02) were more likely to agree than participants in the no-crime-information condition (M = 1.25, SD = 0.70), F(1, 463) = 94.70, p < .001, η_p² = .17.

Punishment

We began by testing the effects of time delays on punishment, which revealed a significant effect: Participants in the long-time-delay condition (M = 12.20, SD = 8.86) levied more severe punishment than participants in the short-time-delay condition (M = 10.39, SD = 8.17), F(1, 463) = 5.19, p = .023, η_p² = .01. There was no main effect of the no-crime-information condition, F(1, 463) = 0.61, p = .436, η_p² = .00, and no interaction with the time-delay condition, F(1, 462) = 0.05, p = .816, η_p² = .00.

Indirect effect of perceived unfairness

Next, we tested the mediating role of unfairness. We found that time delays increased perceptions of unfairness (b = 1.12, SE = 0.13, p < .0001, 95% CI = [0.85, 1.38]), unfairness was positively related to punishment (b = 1.50, SE = 0.27, p < .0001, 95% CI = [0.98, 2.03]), and controlling for unfairness eliminated the effect of time delays on punishment (b = 0.12, SE = 0.82, p = .88, 95% CI = [–1.49, 1.73]). As hypothesized, there was a significant indirect effect of time delays on punishment through unfairness (b = 1.68, SE = 0.40, 95% CI = [0.95, 2.51]).

Discussion

Studies 4a and 4b extended our results by ruling out the possibility that the severe punishment for time delays was due to third parties inferring that transgressors were responsible for the delay or committed additional crimes during the time delay between the crime and the punishment. In Study 5a and 5b, we used complementary designs to rule out two more possibilities: that transgressors are punished more severely because they are evading capture (i.e., “at large”), which in and of itself is seen as a violation (Study 5a), and that transgressors are punished more severely because they have additional time to enjoy the material benefits of their crime (Study 5b). Moreover, we used a new measure of punishment to show that our results map onto additional forms of punishment.

Study 5a

We conducted Study 5a to rule out the possibility that transgressors in the long-time-delay condition are punished more severely because they are at large for longer. We preregistered the study design, sample size, exclusions, and analysis plan in advance (https://aspredicted.org/9SR_7BC).

Method

Sample size and exclusions

Given that Studies 4a and 4b revealed a smaller effect size than Study 3, we performed a new power analysis in G*Power (Faul et al., 2009), averaging the effect sizes from Studies 3, 4a, and 4b, which indicated that a two-condition study required 779 participants for 80% power. Following our preregistration, we recruited 900 participants on Prolific to account for potential exclusions. Following the preregistered exclusion criteria, we removed participants who did not correctly identify the time between crime and punishment (n = 65), provided a nonsense response to an open-ended question (“What did you have for breakfast today?”; n = 11), had suspicious IP addresses (n = 2), or used automatic form fillers (n = 7). Some participants failed multiple checks, leaving a final sample of 815 (age: M = 40.18 years, SD = 13.05; 50% female; 77% White).

Procedure

In this study, participants were informed that the transgressor, Jamie, worked at a charity and used $1,000 of charity funds as partial payment for a new car for personal use. Participants were informed that Jamie was arrested shortly after. Moreover, participants were informed that Jamie was then brought before a judge, who determined that an appropriate punishment was the full seizure of the car.

As in the prior experimental studies, we randomly assigned each participant to one of two time-delay conditions. However, we modified our paradigm to rule out the argument that participants in the long-time-delay condition assigned more severe punishment because they were penalizing Jamie for being at large. In the long-time-delay condition, we informed participants that because of “a backlog in the court, Jamie’s car cannot be seized for an extra 6 months,” whereas in the short-time-delay condition, we informed participants that Jamie’s car was seized shortly after the sentence was announced. Thus, in both conditions, Jamie was arrested immediately, and the only difference between conditions was whether there was a short or long time delay before the seizure of Jamie’s car.

Measures

To measure punishment, we informed participants that they could supplement the judge’s punishment (i.e., prison time) with a fine, ranging from $0 to $10,000 in $1,000 increments, if there was justification for doing so. To measure our mechanism, we used the same measure of unfairness (α = .93) as in the prior experimental studies.

Results

Manipulation check

Using an adapted manipulation check from the previous studies, we found that participants in the long-time-delay condition (M = 4.53, SD = 0.71) were more likely than those in the short-time-delay condition (M = 1.19, SD = 0.55), F(1, 813) = 5,509.53, p < .001, η_p² = .87, to agree that a long time had passed between crime and punishment.

Punishment

We tested the effect of time delay on punishment (reported in thousands of dollars), which revealed a significant effect: Participants in the long-time-delay condition (M = 4.91, SD = 3.54) levied more severe punishment than participants in the short-time-delay condition (M = 3.15, SD = 3.40), F(1, 813) = 52.16, p < .001, η_p² = .06.

Indirect effect of perceived unfairness

Next, we tested the mediating role of unfairness. We found that time delays increased perceptions of unfairness (b = 2.85, SE = 0.11, p < .0001, 95% CI = [2.63, 3.06]), unfairness was positively related to punishment (b = 0.98, SE = 0.07, p < .0001, 95% CI = [0.84, 1.12]), and controlling for unfairness eliminated the positive effect of time delays on punishment (b = −1.04, SE = 0.30, p < .001, 95% CI = [–1.62, –0.45]). As hypothesized, there was a significant indirect effect of time delays on punishment through unfairness (b = 2.79, SE = 0.23, 95% CI = [2.37, 3.24]).

Study 5b

Using an adapted paradigm from Study 5a, in Study 5b, we sought to rule out the possibility that transgressors incurring longer time delays receive more severe punishment because they can enjoy the material benefits of their crime for longer. We preregistered the study design, sample size, exclusions, and analysis plan in advance (https://aspredicted.org/6D7_FVX).

Method

Sample size and exclusions

We used the same power analysis as in Study 5a. Following our preregistration, we recruited 900 participants on Prolific to account for potential exclusions. In line with the preregistered exclusion criteria, we removed participants who did not correctly identify the time between crime and punishment (n = 119), provided a nonsense response to an open-ended question (“What did you have for breakfast today?”; n = 8), had suspicious IP addresses (n = 1), or used automatic form fillers (n = 10). Some participants failed multiple checks, leaving a final sample of 766 (age: M = 37.98 years, SD = 13.79; 49% female; 75% White).

Procedure

We randomly assigned each participant to one of two time-delay conditions as in the prior experimental studies. To rule out the possibility that participants in the long-time-delay condition assigned more severe punishment because they were penalizing Jamie for enjoying the material benefits of the crime, we slightly altered the Study 5a paradigm. Instead of a crime that entailed a potential future material benefit (e.g., stolen goods, car purchased with stolen funds), participants were informed that Jamie vandalized local property. As in Study 5a, in both conditions, participants were informed that Jamie was arrested shortly after the transgression. Moreover, participants were informed that Jamie was shortly thereafter brought before a judge, who determined that an appropriate punishment was 30 days in prison. In the long-time-delay condition, participants were informed that because of “overcrowding, Jamie will not be sent to prison until 6 months from now.” In contrast, participants in the short-time-delay condition were informed that Jamie would be sent to prison the next day.

Measures

To measure punishment, we informed participants that they could supplement the judge’s punishment (i.e., prison time) with a fine, ranging from $0 to $10,000 in $1,000 increments, if there was justification. To measure our mechanism, we used the same measure of unfairness (α = .93) as in the prior experimental studies.

Results

Manipulation check

Using an adapted manipulation check from the previous studies, we found that participants in the long-time-delay condition (M = 4.19, SD = 0.88) were more likely than those in the short-time-delay condition (M = 1.13, SD = 0.42) to agree that a long time had passed between the crime and the punishment, F(1, 764) = 3,711.91, p < .001, η_p² = .83.

Punishment

We tested the effects of time delay on punishment, which revealed a significant effect: Participants in the long-time-delay condition (M = 2.58, SD = 3.20) levied more severe punishment than participants in the short-time-delay condition (M = 2.01, SD = 2.63), F(1, 764) = 7.13, p = .008, η_p² = .01.

Indirect effect of perceived unfairness

Next, we tested the mediating role of unfairness. We found that time delays increased perceptions of unfairness (b = 1.82, SE = 0.10, p < .0001, 95% CI = [1.61, 2.02]), unfairness was positively related to punishment (b = 0.68, SE = 0.07, p < .0001, 95% CI = [0.55, 0.82]), and controlling for unfairness eliminated the positive effect of time delays on punishment (b = −0.68, SE = 0.24, p = .004, 95% CI = [–1.14, –0.21]). As hypothesized, there was a significant indirect effect of time delay on punishment through unfairness (b = 1.24, SE = 0.16, 95% CI = [0.95, 1.57]).

Discussion

Studies 5a and 5b extended our results by ruling out the possibility that being at large and receiving material benefits from one’s crime drive the effects of time on punishment. Although our effect held in Study 5b, it is important to note that the effect size was smaller than in Study 5a, which suggests that some variance in prior studies may be attributed to enjoying benefits from the crime. In Study 6, we sought to test an intervention explicitly designed to mitigate the effect of time on punishment severity by reducing perceptions of unfairness.

Study 6

Study 6 manipulated an additional experimental factor to counteract the main effect of time delays on punishment by reducing perceptions of unfairness. To that end, we adapted our Study 5b design and manipulated unfairness in terms of whether or not the judge accounted for the time delay when determining the punishment. We reasoned that taking the time delay into account would reduce perceptions of unfairness, resulting in participants assigning less severe fines. Our goal with this design was thus to establish that fairness temporally precedes punishment (Spencer et al., 2005).

Method

We again preregistered the study design, key hypothesis, sample size, exclusion criteria, and statistical analysis plan before running the study (https://aspredicted.org/KD5_7LF).

Sample size and exclusions

Because we hypothesized a “knock-out” interaction, we followed recent recommendations (Anderson et al., 2017) for powering an interaction effect and quadrupled our sample size from Studies 5a and 5b (Giner-Sorolla, 2018), aiming to recruit 3,300 participants on MTurk. Because CloudResearch can oversample by a small margin, we ended up with an initial sample of 3,316. Following our preregistered exclusion criteria, we removed participants who did not correctly identify the time between crime and arrest (n = 181), provided a nonsense response to an open-ended question (“What did you have for breakfast today?”; n = 29), had suspicious IP addresses (n = 12), or used automatic form fillers (n = 39). Because some participants failed multiple checks, we had a final sample of 3,062 (age: M = 42.10 years, SD = 12.80; 53% female; 75% White).

Procedure

Our study used a 2 (time delay: long vs. short) × 2 (unfairness: unfair vs. fair) between-subjects design. We adapted the experimental approach from Study 5b and randomly assigned each participant to similar short- and long-time-delay conditions. In both conditions, participants were informed that there were backlogs in the system, which meant that the transgressor would not be sent to prison for 1 more day (short time delay) or 6 more months (long time delay). Additionally, we included a second factor, which manipulated unfairness. At its core, our argument is that time delays are perceived as unfair because transgressors experienced the benefit of unpunished time, which warrants more severe punishment (Colquitt & Rodell, 2015). Building on this, we informed participants in the fair condition that the judge was aware of the backlogs and accounted for the length of the time delay when deciding the severity of their punishment. This created a sense of justice and fairness because the judge accounted for the time delay when making their punishment decision. In the unfair condition, we informed participants that the judge was unaware of the backlogs and did not account for this when making their punishment decision. Our core preregistered prediction based on this design was that the effect of time on punishment would emerge in the unfair condition but not in the fair condition.

Measures

We used the same measure of unfairness (α = .92) as in the prior experimental studies and adapted the punishment measure from Study 5b.

Results

Manipulation check

In this study, we used anchors of 1, strongly disagree, and 7, strongly agree, for our manipulation checks. As expected, participants in the long-time-delay condition (M = 5.69, SD = 1.28) were more likely than participants in the short-time-delay condition (M = 1.42, SD = 0.96) to agree that a long time had passed before Jamie would be sent to prison, F(1, 3059) = 10,923.12, p < .001, η_p² = .78. Moreover, we asked participants whether they believed that the judge was aware of the time delay when making their decision. As expected, participants in the fair condition (M = 6.53, SD = 1.06) were more likely than those in the unfair condition (M = 1.72, SD = 1.56) to agree with this, F(1, 3059) = 9,972.39, p < .001, η_p² = .77.

Punishment

As predicted, we found that participants in the long-time-delay condition (M = 1.68, SD = 2.85) levied more severe punishment than participants in the short-time-delay condition (M = 0.90, SD = 2.23), F(1, 3059) = 72.05, p < .001, η_p² = .02. We also found a main effect of the unfairness condition on punishment, F(1, 3059) = 88.41, p < .001, η_p² = .03; participants in the fair condition levied less severe punishment than those in the unfair condition.

As predicted, we found a time-by-unfairness interaction on punishment, F(1, 3058) = 96.87, p < .001, η_p² = .03. For participants in the unfair condition, there was a significant effect of time delay on punishment, F(1, 3058) = 170.34, p < .001, η_p² = .05; participants in the long-time-delay condition (M = 2.55, SD = 3.15) administered more severe punishment than participants in the short-time-delay condition (M = 0.90, SD = 2.21). However, there was not a significant effect of time delay on punishment in the fair condition, F(1, 3058) = 0.84, p = .36, η_p² = .00, as participants in the long-time-delay condition (M = 0.79, SD = 2.16) levied equally severe punishment as those in the short-time-delay condition (M = 0.91, SD = 2.25; see Fig. 2).

Fig. 2.

Study 6: mean fine in the unfair and fair conditions, separately for the short- and long-time-delay conditions. Error bars show standard errors.

Indirect effect of perceived unfairness

Next, we tested unfairness as a mediator. As expected, unfairness was positively related to punishment severity (b = 0.62, SE = 0.03, p < .001, 95% CI = [0.56, 0.67]). Results revealed a stronger indirect effect of time delays on punishment through unfairness in the unfair condition (b = 1.01, SE = 0.08, 95% CI = [0.87, 1.17]) than in the fair condition (b = 0.23, SE = 0.05, 95% CI = [0.14, 0.32]) and a significant difference between these effects (Δb = −0.78, SE = 0.08, 95% CI = [–0.95, –0.63]). Thus, our second manipulation reduced the indirect effect of time delays on punishment severity via unfairness (in addition to directly reducing the effect on punishment; see Fig. 3).

Fig. 3.

Study 6: mean unfairness rating in the unfair and fair conditions, separately for the short- and long-time-delay conditions. Error bars show standard errors.

Discussion

Study 6 replicated the results from the prior studies and provided additional evidence for unfairness as our key mechanism. Indeed, we found that the effect of time delays on punishment was dampened when the time delay was accounted for explicitly. Moreover, we followed recent recommendations for sampling and powering interaction effects (Giner-Sorolla, 2018), providing strong confidence in our findings.

General Discussion

Across eight studies (two archival, six experimental) balancing external and internal validity, we found that time delays increased the severity of third-party punishment. Our archival studies advanced the external validity of our research by demonstrating that time delays in two different contexts—Study 1 featured the time between a crime and an arrest, and Study 2 featured the time delay between police misconduct and an investigation being completed—increased punishment severity while accounting for various alternative explanations and robustness tests. In our experimental studies, we tested a variety of mechanisms and found consistent evidence that perceptions of unfairness explained why time delays contribute to more severe punishment.

Our research makes important theoretical and empirical contributions. Most notably, we theorized and tested the influence of time—an ever-present factor in every transgression—on punishment severity. We found that temporal elements related to the crime influence the severity of punishment, even when accounting for other factors that play a role in determining punishment, such as characteristics of the transgressor (Kakkar et al., 2020), punisher, or crime (Kundro & Nurmohamed, 2021; Mitchell et al., 2015). Thus, our research contributed to perspectives on punishment by examining the mechanisms and boundary conditions that pertain to the effects of time delays on punishment severity. In addition, we augment perspectives that have implied that time delays may decrease punishment severity. Indeed, research taking a victim-centered approach has implied that victims may lose their desire or motivation for severe punishment as time passes (Wang et al., 2011). In contrast, by taking a third-party perspective—in which judges are not the victims of a crime—we found that third parties view time delays in a markedly different light and instead punish transgressors more severely. Future research should explicitly compare how victims—versus third parties—perceive and respond to time delays, as this distinction likely has implications for whether people believe justice is adequately served (Aquino et al., 2006; Fehr & Gelfand, 2010; Lind & Tyler, 1988).

Our research, notwithstanding its strengths, has important limitations. First, we were unable to test the mediating mechanism of perceptions of unfairness in Studies 1 and 2. However, we found consistent evidence for this mechanism in our experimental studies, even when accounting for a number of alternative explanations. Second, we investigated the effects of time on punishment in contexts with formal punishment proceedings, but it is unclear to what extent our results would generalize outside of these contexts. In this vein, our experiments were all conducted on online platforms (MTurk, Prolific), and future research should replicate our direct and mediation effects with different samples and scenarios. Third, we took steps to understand the source of unfairness perceptions, but future research should continue to explore factors that increase or decrease perceptions of fairness following time delays. For example, future researchers may consider the underlying causes of time delays, such as delayed accusations (Raj & Wiltermuth, 2022). Alternatively, researchers may investigate differences in punishment as a function of individual versus corporate punishment (Caulfield & Laufer, 2018) or across organizational boundaries (Frey et al., 2023).

In conclusion, our studies reveal that time delays increase punishment severity. This research illuminates when and why this effect emerges, offering new contributions and directions for future research on punishment.

Footnotes

Acknowledgements

For feedback on a prior version of this article, we thank Jason Colquitt, Kurt Gray, and Ann Tenbrunsel. For helpful conversations, we thank Karl Aquino, Adam Grant, Matthew Caulfield, Beth Anne Helgason, Brian Lucas, David Mayer, Sean McDonnell, Zoe Schwingel-Sauer, and Kristin Smith-Crowe. We also thank participants at the M&O Brown Bag Series at Mendoza College of Business, University of Notre Dame. For research assistance, we thank Colleen Lipa, Michelle Cho, Christopher Li, and Logan Balfantz.

Transparency

Action Editor: Yoel Inbar

Editor: Patricia J. Bauer

Author Contributions

Timothy G. Kundro: Conceptualization; Data curation; Formal analysis; Funding acquisition; Investigation; Methodology; Project administration; Software; Supervision; Writing – original draft; Writing – review & editing.

Samir Nurmohamed: Conceptualization; Data curation; Formal analysis; Methodology; Writing – original draft; Writing – review & editing.

Hemant Kakkar: Data curation; Methodology; Writing – original draft; Writing – review & editing.

Salvatore J. Affinito: Data curation; Formal analysis; Methodology; Writing – original draft; Writing – review & editing.

ORCID iDs

Timothy G. Kundro

Hemant Kakkar

Salvatore J. Affinito

Notes

References

Anderson

S. F.

Kelley

Maxwell

S. E.

(2017). Sample-size planning for more accurate statistical power: A method adjusting sample effect sizes for publication bias and uncertainty. Psychological Science, 28(11), 1547–1562.

Aquino

Tripp

T. M.

Bies

R. J.

(2006). Getting even or moving on? Power, procedural justice, and types of offense as predictors of revenge, forgiveness, reconciliation, and avoidance in organizations. Journal of Applied Psychology, 91(3), 653–668.

Bates

Mächler

Bolker

Walker

(2015). Fitting linear mixed-effects models using lme4. Journal of Statistical Software, 67(1). https://doi.org/10.18637/jss.v067.i01

Batson

C. D.

Kennedy

C. L.

Nord

L. A.

Stocks

E. L.

Fleming

D. Y. A.

Marzette

C. M.

Zerger

(2007). Anger at unfairness: Is it moral outrage?. European Journal of Social Psychology, 37(6), 1272–1285.

Behrendt

(2014). Package lm.beta [Computer software]. https://cran.r-project.org/web/packages/lm.beta/index.html

Bernerth

J. B.

Aguinis

(2016). A critical review and best-practice recommendations for control variable usage. Personnel Psychology, 69(1), 229–283. https://doi.org/10.1111/peps.12103

Bies

R. J.

(1987). The predicament of injustice: The management of moral outrage. Research in Organizational Behavior, 9, 289–319.

Brickner

D. R.

Mahoney

L. S.

Moore

S. J.

(2010). Providing an applied-learning exercise in teaching fraud detection: A case of academic partnering with IRS criminal investigation. Issues in Accounting Education, 25(4), 695–708.

Buchanan

E. M.

Scofield

J. E.

(2018). Methods to detect low quality data and its implication for psychological research. Behavior Research Methods, 50(6), 2586–2596.

10.

Carlsmith

K. M.

(2006). The roles of retribution and utility in determining punishment. Journal of Experimental Social Psychology, 42(4), 437–451.

11.

Carlsmith

K. M.

Darley

J. M.

Robinson

P. H.

(2002). Why do we punish? Deterrence and just deserts as motives for punishment. Journal of Personality and Social Psychology, 83(2), 284–299.

12.

Cassidy

Rydberg

(2020). Does sentence type and length matter? Interactions of age, race, ethnicity, and gender on jail and prison sentences. Criminal Justice and Behavior, 47(1), 61–79.

13.

Caulfield

Laufer

W. S.

(2018). The promise of corporate character theory. Iowa Law Review, 103, 101–122.

14.

Civilian Complaint Review Board. (2023a). Allegations. City of New York. https://www.nyc.gov/site/ccrb/policy/data-transparency-initiative-allegations.page

15.

Civilian Complaint Review Board. (2023b). Police discipline. City of New York. https://www.nyc.gov/site/ccrb/prosecution/police-discipline.page

16.

Cohen

(1992). A power primer. Psychological Bulletin, 112(1), 155–159. https://doi.org/10.1037/0033-2909.112.1.155

17.

Cohen

T. H.

Reaves

B. A.

(2007). State court processing statistics, 1990-2004: Pretrial release of felony defendants in state courts (NCJ Report No. 214994). U.S. Department of Justice. https://bjs.ojp.gov/content/pub/pdf/prfdsc.pdf

18.

Cohen-Charash

Mueller

J. S.

(2007). Does perceived unfairness exacerbate or mitigate interpersonal counterproductive work behaviors related to envy? Journal of Applied Psychology, 92(3), 666–680.

19.

Coleman

Sarch

(2012). Blameworthiness and time. Legal Theory, 18(2), 101–137.

20.

Colquitt

J. A.

Rodell

J. B.

(2015). Measuring justice and fairness. In Cropanzano

R. S.

Ambrose

M. L.

(Eds.), The Oxford handbook of justice in the workplace (pp. 187–202). Oxford University Press.

21.

Crockett

M. J.

Kurth-Nelson

Siegel

J. Z.

Dayan

Dolan

R. J.

(2014). Harm to others outweighs harm to self in moral decision making. Proceedings of the National Academy of Sciences, USA, 111(48), 17320–17325.

22.

Faul

Erdfelder

Buchner

Lang

A.-G.

(2009). Statistical power analyses using G*Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods, 41(4), 1149–1160.

23.

Fehr

Gelfand

M. J.

(2010). When apologies work: How matching apology components to victims’ self-construals facilitates forgiveness. Organizational Behavior and Human Decision Processes, 113(1), 37–50.

24.

FeldmanHall

Sokol-Hessner

Van Bavel

J. J.

Phelps

E. A.

(2014). Fairness violations elicit greater punishment on behalf of another than for oneself. Nature Communications, 5(1), Article 5306. https://doi.org/10.1038/ncomms6306

25.

Fetchenhauer

Huang

(2004). Justice sensitivity and distributive decisions in experimental games. Personality and Individual Differences, 36(5), 1015–1029.

26.

Frey

Adams

G. S.

Pfeffer

Belmi

(2023). What we (do not) know about punishment across organizational boundaries. Journal of Management, 49(1), 196–236.

27.

Fiske

S. T.

Cuddy

A. J.

Glick

(2007). Universal dimensions of social cognition: Warmth and competence. Trends in Cognitive Sciences, 11(2), 77–83.

28.

Giner-Sorolla

(2018, January 24). Powering your interaction. Approaching Significance. https://approachingblog.wordpress.com/2018/01/24/powering-your-interaction-2

29.

Greenberg

(1983). Overcoming egocentric bias in perceived fairness through self-awareness. Social Psychology Quarterly, 46(2), 152–156.

30.

Grissom

R. J.

Kim

J. J.

(2005). Effect sizes for research: A broad practical approach. Lawrence Erlbaum Associates Publishers.

31.

Hart

H. L. A.

(2008). Punishment and responsibility: Essays in the philosophy of law (2nd ed.). Oxford University Press. https://doi.org/10.1093/acprof:oso/9780199534777.001.0001

32.

Kakkar

Sivanathan

Gobel

M. S.

(2020). Fall from grace: The role of dominance and prestige in the punishment of high-status actors. Academy of Management Journal, 63(2), 530–553.

33.

Keller

L. B.

Oswald

M. E.

Stucki

Gollwitzer

(2010). A closer look at an eye for an eye: Laypersons’ punishment decisions are primarily driven by retributive motives. Social Justice Research, 23(2), 99–116.

34.

Kleiman

M. A.

Kilmer

Fisher

D. T.

(2014). Theory and evidence on the swift-certain-fair approach to enforcing conditions of community supervision. Federal Probation Journal, 78, 71–74.

35.

Kundro

T. G.

Nurmohamed

(2021). Understanding when and why cover-ups are punished less severely. Academy of Management Journal, 64(3), 873–900.

36.

Kuznetsova

Brockhoff

P. B.

Christensen

R. H. B.

(2017). lmertest package: Tests in linear mixed effects models. Journal of Statistical Software, 82(13). https://doi.org/10.18637/jss.v082.i13

37.

Lerner

J. S.

Weber

E. U.

(2013). The financial costs of sadness. Psychological Science, 24(1), 72–79.

38.

Lind

E. A.

Tyler

T. R.

(1988). The social psychology of procedural justice. Springer Science & Business Media.

39.

Martin

S. E.

Sherman

L. W.

(1986). Selective apprehension: A police strategy for repeat offenders. Criminology, 24(1), 155–174.

40.

McDonnell

M. H.

Nurmohamed

(2021). When are organizations punished for organizational misconduct? A review and research agenda. Research in Organizational Behavior, 41, Article 100150. https://doi.org/10.1016/j.riob.2021.100150

41.

Mitchell

M. S.

Vogel

R. M.

Folger

(2015). Third parties’ reactions to the abusive supervision of coworkers. Journal of Applied Psychology, 100(4), 1040–1055.

42.

Mooijman

Graham

(2018). Unjust punishment in organizations. Research in Organizational Behavior, 38, 95–106.

43.

New York Civil Liberties Union. (2021, May 6). NYPD-misconduct-database-updated. GitHub. https://github.com/new-york-civil-liberties-union/NYPD-Misconduct-Complaint-Database-Updated

44.

Packard

Berger

(2020). Thinking of you: How second-person pronouns shape cultural success. Psychological Science, 31(4), 397–407.

45.

Prims

Sisso

Bai

(2018). Suspicious IP online flagging tool. https://itaysisso.shinyapps.io/Bots

46.

Raj

Wiltermuth

S. S.

(2022). Better now than later: The social cost of victims’ delayed accusations. Organizational Behavior and Human Decision Processes, 168, Article 104110. https://doi.org/10.1016/j.obhdp.2021.104110

47.

R Core Team. (2022). R: A language and environment for statistical computing [Computer software]. https://www.R-project.org/

48.

Rohrer

J. M.

Lucas

R. E.

(2021). Causal effects of well-being on health: It’s complicated. PsyArXiv. https://doi.org/10.31234/osf.io/wgbe4

49.

Simonsohn

(2018). Two lines: A valid alternative to the invalid testing of U-shaped relationships with quadratic regressions. Advances in Methods and Practices in Psychological Science, 1(4), 538–555.

50.

Smith

R. H.

Parrott

W. G.

Ozer

Moniz

(1994). Subjective injustice and inferiority as predictors of hostile and depressive feelings in envy. Personality and Social Psychology Bulletin, 20, 705–711.

51.

Spencer

S. J.

Zanna

M. P.

Fong

G. T.

(2005). Establishing a causal chain: Why experiments are often more effective than mediational analyses in examining psychological processes. Journal of Personality and Social Psychology, 89(6), 845.

52.

Thompson

Loewenstein

(1992). Egocentric interpretations of fairness and interpersonal conflict. Organizational Behavior and Human Decision Processes, 51(2), 176–197.

53.

Treviño

L. K.

Weaver

G. R.

(1998). Punishment in organizations: Descriptive and normative perspectives. In Schminke

(Ed.), Managerial ethics: Moral management of people and processes (pp. 99–114). Erlbaum.

54.

Wang

C. S.

Sivanathan

Narayanan

Ganegoda

D. B.

Bauer

Bodenhausen

G. V.

Murnighan

(2011). Retribution and emotional regulation: The effects of time delay in angry economic interactions. Organizational Behavior and Human Decision Processes, 116(1), 46–54.

55.

Witte

Berman

(2021, December 19). Long after the courts shut down for covid, the pain of delayed justice lingers. The Washington Post. https://www.washingtonpost.com/national/covid-court-backlog-justice-delayed/2021/12/18/212c16bc-5948-11ec-a219-9b4ae96da3b7_story.html

56.

Zitek

E. M.

Jordan

A. H.

Monin

Leach

F. R.

(2010). Victim entitlement to behave selfishly. Journal of Personality and Social Psychology, 98(2), 245–255.

Time and Punishment: Time Delays Exacerbate the Severity of Third-Party Punishment

Abstract

Keywords

Statement of Relevance

Overview of Studies

Open Practices Statement

Study 1

Method

Context and sample size

Core measures

Time delay

Punishment

Controls

Number of counts

Guilty plea

Type of crime

Judge

Year the crime was sentenced

Demographic information

Results

Primary analysis

Robustness analyses

Without controls

Alternative conceptualizations of time delays

Curvilinear relationship

Mixed-effects model with different sampling criteria

Discussion

Study 2

Method

Context and sample size

Core measures

Time delay

Punishment

Controls

Year crime occurred

Number of charges officer faced

Number of officers associated with a given complaint

Type of accusation officer faced

Demographic information

Results

Primary analysis

Robustness analyses

Results without controls

Alternative conceptualizations of time delay

Curvilinear relationship

Mixed-effects model with different sampling criteria

Discussion

Study 3

Method

Sample size and exclusions

Procedure

Punishment

Perceived unfairness of process

Alternative mechanisms

Results

Manipulation check

Punishment

Indirect effect of perceived unfairness

Alternative mechanisms

Discussion

Study 4a

Method

Sample size and exclusions

Procedure

Measures

Results

Manipulation check

Punishment

Indirect effect of perceived unfairness

Study 4b

Method

Sample size and exclusions

Procedure

Measures

Results

Manipulation check

Punishment

Indirect effect of perceived unfairness

Discussion

Study 5a