Introduction

Prosecutors are arguably the most powerful players in the criminal justice system. They exercise the most discretion in two general areas—the decision to file charges and plea bargaining. While evidentiary factors typically drive these decisions, other considerations may also affect their discretion. Limited research suggests that external pressures, such as the widespread media coverage of a high-profile case, might also influence prosecutorial decision-making. Using unique longitudinal data, we examine the extent to which the mass shooting at Marjory Stoneman Douglas High School in Parkland, Florida, may have influenced how prosecutors contemplate and handle firearm-related cases.

Background

On February 14, 2018, 19-year-old Nikolas Cruz entered Marjory Stoneman Douglas High School in Parkland, Florida, and killed 17 people and injured 17 others with a semi-automatic rifle. A media frenzy followed the shooting, which was the deadliest mass shooting at a high school in U.S. history. Victims included students and staff members of the school. Although Cruz, on the surface, appeared to resemble any other 19-year-old high school student, he was profoundly impacted by his White supremacy and life and death outlook (Turkewitz et al., 2018). According to reports, he intended to join the military to die in the same manner as those he had murdered. Cruz’s opinions on Blacks, Muslims, Jews, and LGBTQ+ individuals were clearly radical on the question of social influence. Cruz’s contempt for these separate social categories demonstrated that he believed he was superior to others and that those he deemed unworthy of life should perish. Nevertheless, despite his prior statements about disliking specific groups of people, Cruz did not discriminate between individuals during the shooting.

Cruz fled the scene after the shooting by blending in with the fleeing students. He was apprehended by law enforcement later that day in a nearby neighborhood. In the aftermath, it was revealed that there had been multiple warning signs leading up to the attack, including troubling social media posts and specific threats made by Cruz. Cruz had also been reported to law enforcement multiple times for his disturbing behavior and violent tendencies, and his expulsion from Marjory Stoneman Douglas High School should have raised significant red flags. Yet, no action was taken. The overlooking of these clear warning signs led to widespread criticism of how law enforcement and school officials handled the situation.

The shooting also prompted discussions about the need for better coordination and communication among schools, law enforcement agencies, prosecutors, and mental health professionals to identify and address potential threats. In response to these discussions, the Broward State Attorney’s Office adopted a more aggressive stance in charging criminal defendants, particularly in cases involving threats to schools or public safety. The office also implemented new policies related to the prosecution of firearms offenses and threats to schools, reflecting a more proactive approach to gun-related cases. The Broward State Attorney’s Office endeavored to enhance transparency by providing regular updates on the status of the Parkland shooting case and engaging with the community. The public perception of the Broward State Attorney’s Office’s handling of the Parkland shooting case was influenced by media coverage and the broader national debate on gun control and school safety.

There is little doubt that the Parkland shooting had a salient impact on the national conversation regarding gun control, school safety, and the role that mental health plays in these incidents. The shooting also sparked nationwide protests and led to the founding of Never Again MSD (Marjory Stoneman Douglas), a political action committee for gun control that advocates for tighter regulations to prevent gun violence. The group was founded by students at the high school, and it organized the March for Our Lives protest on March 24, 2018.

The Parkland shooting incident also received extensive local and national media coverage. An analysis by the Washington Post found that this event received substantial media coverage and considerably more media attention than other recent mass shootings (Siegel, 2018). The shooting also engendered a significant public discourse regarding a plethora of issues, including a discussion about mental health and the prevalence of hate crimes in society because Cruz expressed contempt for minority groups. The tragedy also resulted in several policy and legislative changes. For example, more state and federal governments have hate crime laws on the books, and there has been a renewed push for stricter gun control measures (Luca et al., 2019).

The Effects of Media and Political Pressure on the Prosecution of Criminal Cases

Prosecutors are not immune to political pressures. Most lead prosecutors in the U.S. are elected to four-year terms, and prosecutorial elections are becoming more contested. A recent study by Wright et al. (2021) found that in high-population districts, the likelihood that incumbent prosecutors run unopposed has decreased significantly, and incumbents are now less likely to win reelection as well. As prosecutorial elections become battlegrounds, external factors may play a larger role in shaping prosecution. Prosecutors acknowledge that pressure from the media and advocacy groups can subtly influence high-profile case decisions (King & Kutateladze, 2023). Analyses of election cycles show that the pressure to retain one’s position often results in more punitive sanctions, such as more prison admissions and longer incarceration sentences (Okafor, 2021). Future career ambitions may also influence prosecutors when interacting with defendants (Miller & Curry, 2018); not only must prosecutors justify their handling of high-profile cases to the electorate, but being favorably viewed by the public can springboard them into more lucrative positions such as higher elected offices, judgeships, and partnerships at prestigious law firms (Green & Roiphe, 2017).

Studies also find that media attention influences the behavioral patterns of prosecutors and other courtroom actors. For example, in a study of two California jurisdictions, Utz (1978) found that prosecutors were less likely to reach a plea agreement with a criminal defendant when the defendant’s case received widespread news coverage. In another study, Pritchard (1986) investigated the impact of newspaper coverage on a prosecutor’s decision to plea bargain a homicide case in Milwaukee County, Wisconsin. His results showed that as the newspaper coverage of a homicide case increased, prosecutors were less inclined to plea bargain the case after controlling for other factors. Survey research also finds prosecutors report being less prone to negotiate with a criminal defendant when the defendant’s case has received news coverage (Jones, 1978). Other courtroom actors can be influenced by the media as well. Juror decisions, for instance, can be shaped by the amount and type of pretrial media publicity given to their case (Ogloff & Vidmar, 1994). Lim et al. (2015) also found that greater newspaper coverage was associated with more punitive sentence lengths issued by judges in violent crime cases.

The relationship between media attention and the hesitancy of prosecutors to negotiate a plea bargain can have adverse repercussions for defendants because defendants found guilty at trial are more likely to be severely sanctioned than similarly situated defendants whose cases were settled with a plea deal (Kim, 2015). For example, a prison rather than a probation sentence can have deleterious consequences for a defendant because it increases recidivism (Smith et al., 2018), amplifies unemployment (Lindsay, 2022), decreases lifetime earnings (Garin et al., 2023), fosters adverse health outcomes (Beckett & Goldberg, 2022), and can adversely impact a defendant’s family (Gilani, 2023). As for communities, areas that experience higher levels of incarceration could experience lower levels of solidarity, social interaction, and a negative perception of the criminal justice system (Clear, 2007).

Methodological Challenges in Studying the Effects of High-Profile Crimes

Although the results generated in prior empirical research suggest that high-profile events and media coverage impact prosecutorial decision-making, we believe that further empirical evidence is warranted before accepting such a conclusion as definitive. Several methodological ambiguities must be considered when evaluating the importance of previous research on this topic. First, most prior studies are cross-sectional in design because longitudinal data regarding the prosecution of criminal defendants at specific decision points in the criminal justice system are rare. However, conclusions drawn from the analysis of cross-sectional data are only suggestive at best because an ongoing trend in the prosecution of criminal cases may have affected the results and conclusions proffered in prior studies. For example, suppose prosecutors were becoming more punitive in their dealings with criminal defendants immediately before the highly publicized case due to a rise in violent crime or some other factor. In such a situation, it is entirely plausible that any observed modification in a prosecutor’s behavior might not be due to changes in media attention but other unmeasured factors. An analysis of cross-sectional data would not be able to account for this possibility because measurements are collected at a single point in time. This view is bolstered by research showing the tendency of prosecutors to react to changes in violent crime levels. In a study of data drawn from the state court processing dataset, Fearn (2005) observed that felony defendants adjudicated in communities with a higher violent crime rate had an enhanced proclivity to receive a prison sentence. Thus, one may reasonably argue that any conclusions drawn from an analysis of cross-sectional data may be problematic because it is difficult to determine, with any degree of empirical certainty, whether the publicity given to a specific case or a preexisting trend in prosecutorial behavior was responsible for an effect evinced by researchers.

A second issue is that while a single case being tainted by media coverage, as highlighted in prior research, is problematic in the dispensation of justice, the issue of more pressing salience is whether the excessive media attention given a specific case has a far-reaching effect on the prosecution of other similar crimes. While crime incidents are unique events, one might logically surmise that the effect of publicity on prosecutorial behavior would likely encompass crimes that share similar characteristics to the highly publicized crime. A similarity among prosecuted cases may motivate prosecutors to become more punitive in handling cases that resemble the highly publicized case. Still, this possibility has received little empirical examination.

While mass shootings generate a sizable amount of media attention, they are rare events. Although definitions of mass shootings vary to some degree (Booty et al., 2019), it is estimated that there have only been 594 reported and verified mass shooting incidents of four or more individuals shot or killed (not including the shooter) in the U.S. between 2013 and November 3, 2023 (Gun Violence Archive, 2023). Besides there being more than one victim, a common denominator in mass shootings is that all these crimes involve the use of a firearm by the offender. Numerous calls for implementing gun control measures frequently occur in the aftermath of a mass shooting because of the belief that the carnage inflicted by the offender would have been lessened had the offender been armed with a non-firearm weapon such as a knife. Luca et al. (2019) reported a 15% increase in the number of firearm bills introduced in the states in the year immediately following a mass shooting incident.

It is also important to remember that notwithstanding their use in mass shooting incidents, firearms are generally the weapon of choice by criminal offenders. Approximately 18% of State prison inmates and 15% of Federal inmates report using, carrying, or possessing a firearm during the crime for which they were sentenced (Harlow, 2001). Criminals gain several advantages by using a firearm when committing their crimes. Firearms enable offenders to initiate a crime against victims who would otherwise appear too invulnerable to challenge; they facilitate the element of surprise since offenders can engage victims from a greater distance, and they enhance an attack by persons too squeamish to come into close contact with their victims (National Research Council, 2005). Research supports the practicability of these advantages in that crime victims are more likely to acquiesce to an armed offender’s demands (Kleck, 1997), and offenders armed with a deadly weapon have a lower likelihood of being arrested by police than unarmed offenders (Stolzenberg & D'Alessio, 2004). Besides criminals using firearms to gain an advantage in plying their illegal trade, firearms serve as a means of self-protection. Survey research finds that not only citizens (Lizotte et al., 1981) but also criminals (Wright & Rossi, 1986) frequently arm themselves for self-defense rather than offensive purposes. For example, a recent study found that stand-your-ground and castle doctrine laws amplified the likelihood of gun use by criminal offenders to counter the heightened threat posed by armed targets (D’Alessio et al., 2023).

Current Study

In the current study, we analyze ten years of data calibrated in monthly intervals obtained from the Broward County State Attorney’s Office and an interrupted time series research design to ascertain whether the mass shooting in Parkland, Florida, influenced local prosecutors to modify their behavior in the prosecution of firearm-related crimes.Footnote 1 While the Parkland mass shooting received national media attention, the circuit is used as the relevant geographical unit in the time series analysis because the prosecution of most felony criminal defendants occurs at the county rather than at the city, state, or national level. Additionally, the psychological repercussions of any mass shooting are primarily felt locally because friends of the victims typically live in close physical proximity to the victims (Small & Adler, 2019). It seems likely that the physiological trauma experienced by people associated with the victims (friends and relatives) and their close physical proximity to the mass shooting provide them the motivation and the opportunity to exert political pressure on their locally elected prosecutor.

We focus our attention on several decisions made by a prosecutor. These decisions include whether to file a case, dismiss the charges against the defendant, reduce the initial arrest charges at filing, reduce the arrest charges before disposition, reduce the initial filing charges, and negotiate a plea deal with the defendant. These decision points were selected for study because a prosecutor has considerable discretion in making these decisions in criminal cases and because most criminal cases are resolved at one of these decision points (Kutateladze, 2018). For example, a prosecutor has considerable discretion in determining whether to file a criminal case. This discretion can be observed in the variability in different jurisdictions’ case filing/rejection rates. Data compiled by the Prosecutorial Performance Indicators project (2022) showed that among the 15 jurisdictions examined, case rejection rates varied from 3% (1st Judicial District of Colorado) to more than 60% (Milwaukee). The percentage of cases accepted for prosecution is typically viewed as an indicator of whether the police are conforming to the law in performing their law enforcement duties because evidentiary and legal issues influence prosecutors’ decision to proceed with a case (Pizzi, 1993). While the police must show probable cause to effectuate a valid arrest, a prosecutor requires a higher threshold of likely guilt beyond a reasonable doubt to secure a conviction. The higher threshold of evidence needed to prove a defendant’s guilt often motivates prosecutors to reject cases they believe have little chance of culminating in a successful prosecution.

The adverse repercussions faced by a prosecutor for failing to file a criminal case are also minimal because the police are the ones who failed in the performance of their duties when the individual was initially arrested. A prosecutor is simply declining to prosecute a problematic case. Additionally, these decisions tend to be hidden from the public view. Most people are unaware that many individuals arrested by police are never prosecuted for their alleged crimes. Thus, if media attention does influence the filing decision, one might theorize that prosecutors would be less apt to accept questionable firearm cases for prosecution for fear that the voting public would view their inability to secure a conviction, notwithstanding the strength of the evidence in the case, as a failure.

Like the filing decision, a prosecutor has considerable discretion in deciding whether to dismiss the charges against the defendant, modify the initial arrest charges at filing, adjust the initial filing charge, change the arrest charge after filing, and negotiate a plea deal with the defendant. However, in contrast to the filing decision, these decisions are much more transparent because the prosecution of the criminal defendant by the Prosecutor’s Office has already been initiated. Despite the increase in transparency, these charging decisions have relevance because they play a salient role in determining a criminal defendant’s outcome and can be used to scrutinize the workings of prosecutors in performing their duties (Wright, 2008).

For example, a prosecutor has the discretion to modify charges during plea negotiations to obtain a settlement. The initial charge, which is determined by police, acts as a baseline to draw inferences about the behavior patterns of prosecutors in their dealings with criminal defendants. A change to the initial charge might be mitigating because the prosecutor believes the case is weak against the defendant. A modification to the initial charge might also be aggravating to intimidate the defendant into pleading guilty. The amount of variability charging needed to secure a conviction can be employed to measure the overall performance of prosecutors, even more so than the conviction rate, because most guilty pleas result from a prosecutor manipulating the original charge. The amount of plea bargaining needed to obtain a conviction can thus be discerned from changes to the initial charge, notwithstanding when the initial charge was modified. As Wright (Wright, 2014, p. 611–12) notes, “This measure treats plea negotiations, rather than trials, as the central activity of prosecutors. It encourages more transparent charging and plea practices, allowing easier public scrutiny of case valuation.”

We hypothesize that mass shootings will make prosecutors less apt to modify charges for firearm crimes, particularly when these decisions are much more transparent to the public than filings. Additionally, we expect fewer plea deals offered to defendants in firearm crimes following the mass shooting because charge modification is used primarily in the negotiation of plea deals.

Data and Methods

The data used in this study were obtained from the case management system of the Broward County Prosecutor’s Office. Data are captured at the case-defendant level, and the dataset includes information capturing the progression of each case from police referral to punishment, along with other charging and case processing characteristics. Given the study’s focus on prosecuting firearm crimes, the study’s primary sample is limited to cases referred for prosecution by law enforcement with at least one felony firearms charge. Felony firearms charges include offenses such as aggravated assault with a firearm, robbery with a firearm, possession of a firearm by a convicted felon, possession of a firearm in the commission of a felony, and armed burglary of a dwelling with a firearm.

Our study period encompasses cases resolved between January 2011 and December 2020, representing 120 monthly measurement periods. We selected January 2011 as the starting point for data collection based on data availability. This beginning date also furnishes suitable measurement periods needed to model the preintervention period adequately. The type of data and the analytic strategy employed here are advantageous for several reasons. First, while no research design fully guarantees correct causal inferences, the interrupted time-series design is considered an effective quasi-experimental design for drawing causal inferences (McCleary et al., 2017). A depiction of an interrupted time-series design is presented below:

$${\textrm{O}}_1{\textrm{O}}_2{\textrm{O}}_3{\textrm{O}}_4{\textrm{O}}_5\left(\textrm{X}\right){\textrm{O}}_6{\textrm{O}}_7{\textrm{O}}_8{\textrm{O}}_9{\textrm{O}}_{10}$$

The design involves modeling a series of observations (O) over time that are expected to be affected by an intervention (X). Although the above diagram depicts one series, we evaluate the effect of the mass shooting on several decisions made by a prosecutor to help reduce the possibility of erroneous conclusions. If the difference between the preintervention and postintervention series is greater than expected by chance, it can be deduced that the mass shooting significantly influenced the decision-making process of prosecutors.

Monthly rather than yearly data is also beneficial because causality is more readily discernable when calibrated into fine temporal units (Tiao & Wei, 1976). For there to be an external threat to validity in the time series analysis, the extraneous causal factor must arise in the same month that the mass shooting occurred. This is an important but often overlooked point because cause and effect can be distinguished by using fine units of time since simultaneous effects do not occur in nature (Einstein, 1920). As Granger (Granger, 1969, p. 430) points out: “in many economic situations an apparent instantaneous causality would disappear if the economic variables were recorded at more frequent time intervals.” It is important to recognize that using a small unit of time, like in the current study, is far superior for inferring causal effects than using yearly data with various control variables.

Monthly data also permits greater flexibility in applying more sophisticated and efficient statistical procedures because of the increased number of observations. Lastly, because this analysis compares changes in prosecutorial decision-making regarding firearm offenses crimes within a single circuit over time rather than across different circuits, potential biases resulting from unaccounted dissimilarities between counties are minimized.

Six dependent variables pertaining to firearm offenses are analyzed in this study. Firearm cases are identified using Florida’s “10–20-Life” law. This law was enacted in 1999 and imposes mandatory minimum sentences for certain felony offenses involving using or possessing a firearm. The following are the key provisions of this law: (1) 10 Years: If a person is convicted of certain felonies and possessed a firearm during the commission of the crime, they face a mandatory minimum sentence of 10 years; (2) 20 Years: If the firearm was discharged during the commission of certain felonies, the mandatory minimum sentence increases to 20 years; and (3) 25 Years to Life: If the discharge of the firearm resulted in death or great bodily harm, the mandatory minimum sentence ranges from 25 years to life in prison.

The first dependent variable, firearm cases filed, measures the monthly firearm crimes accepted for prosecution. Prosecutors evaluate a case following an arrest and then determine whether it is appropriate to charge the defendant. Cases can be dropped at initial screening for various reasons, including insufficient evidence, credibility issues with the arresting officers, and constitutional problems. The second variable, nolle prossed firearm crimes, is operationalized as the monthly number of firearm cases dropped by the prosecution after the cases were initially accepted for prosecution. We measure the third dependent variable, arrest to filing reduction, as the monthly number of firearm cases that experienced a charge reduction between arrest and filing. The fourth dependent variable, arrest to disposition reduction, is measured as the monthly number of firearm cases undergoing a reduction in charges between arrest and disposition. The fifth dependent variable, filing to disposition reduction, is the monthly number of firearm cases with reduced charges between filing and disposition. The final dependent variable, plea negotiation, is operationalized as the monthly number of firearm cases resolved by a plea deal with the defendant. All these dependent variables provide us with some indication of whether media attention impacts the processing of criminal defendants.

The independent variable of theoretical interest is the occurrence of the mass shooting in Parkland. This variable is coded 0 before February 2018 and 1 thereafter. We also incorporated an additional dummy coded control variable in the analysis to represent the beginning of the COVID-19 epidemic in the U.S. in 2020. This variable is coded 0 before March 2020 and 1 thereafter. The means, standard deviations, and definitions for all the variables used in this study are reported in Table 1.

Table 1 Description of variables used in the analysis (January 2011–December 2020)
Full size table

Descriptive Analysis

Figure 1 compares the mean changes for the six decision points between the pre-and postintervention periods. For these comparisons, we use the 85 months preceding the mass shooting (January 2011 to January 2018) and the 35 months following the shooting (February 2018 to December 2020). A visual examination of Fig. 1 indicates some support for the claim that the mass shooting may have influenced the filing of cases by the prosecutor for firearm crimes. The mean change between the preintervention and postintervention periods for the frequency of firearm crimes accepted for prosecution was statistically substantive. Approximately 33 firearm crimes were accepted for prosecution before the mass shooting. However, after the mass shooting, the mean acceptance dropped to 15. Thus, based on a simple comparison of pre-and post-test means, prosecutors were less likely to accept firearm cases for prosecution following the mass shooting.

Fig. 1
figure 1

Average monthly 10–20-life cases before and after the Marjory Stoneman Douglas High School shooting in Broward County, Florida (January 2011–December 2020). Notes: N = 120 months. *p ≤ .01 (with Bonferroni correction)

Full size image

Figure 1 also depicts the remaining decision points. A visual examination of the figure suggests that prosecutors became more punitive in their dealings with firearm crimes after the mass shooting. Firearm defendants were less likely to have their cases nolle prossed, their arrest to filing charges reduced, their arrest to disposition charges reduced, their filing to disposition charges reduced, and be afforded a plea deal. For example, the mean level of the preintervention plea bargain series was 19, indicating that 19 firearm defendants received a plea bargain before the mass shooting. However, after the mass shooting, the mean was 8. Thus, after the mass shooting, prosecutors were less likely to make plea deals with defendants charged with a firearm offense.

One plausible explanation for the difference observed between filings and the other decisions made by prosecutors is that the decrease in firearm case filings made it easier for prosecutors to secure a conviction without negotiating with defendants because weak or legally problematic cases were already rejected. Prosecutors could also spend more time, effort, and resources on the fewer firearm cases forwarded for prosecution. Lastly, because ethical prosecutors would not prosecute individuals believed to be innocent, questions about a defendant’s guilt are less of an issue after the dropping of dubious cases.

However, while it seems reasonable to speculate that media attention might have motivated prosecutors to amplify their dropping of firearm cases where a conviction might be difficult, it is also possible that changes in police practices over time may be responsible for the observed drop in the filing of cases. Unlike the other decision points analyzed in this study, police activity directly influences the filing of cases. Prior to the mass shooting, police averaged approximately 34 arrests per month for firearm crimes. However, after the shooting, police averaged only 14 arrests per month for these same offenses. Even more interesting is that while arrests for firearm offenses decreased, reported firearm-related crimes showed a rise during the postintervention period.

One plausible explanation for this contradiction is that after the Parkland mass shooting, law enforcement agencies reallocated their limited resources to focus on preventing future school shootings rather than actively preventing the commission of individual firearm-related crimes generally, thereby leading to fewer arrests for firearm-related offenses. For example, after the Parkland mass shooting, Broward County increased the presence of school resource officers in schools to provide security and build relationships with students. Florida Governor Rick Scott also signed the Marjory Stoneman Douglas High School Public Safety Act into law in March 2018. This law mandated that every school in Florida must have at least one armed school resource officer. Although likely relevant in addressing the problem of school shootings, these types of policies tend to have a negative effect on traditional measures of police performance, like the arrest sanction, because organizational efficiency is related to workload. Policing is considered a labor-intensive activity, and police resources tend to be relatively inelastic, at least in the short term (Majmundar & Weisburd, 2018). Thus, based on studies suggesting that increased police workload tends to decrease the certainty of arrest (Ratcliffe, 2016), it seems possible that the shifting of police resources from traditional law enforcement operations to school-based operations because of the Parkland mass shooting may have impacted traditional law enforcement activities adversely.

That said, it is important to recognize that these descriptive findings are only preliminary. Preexisting trends may be responsible for the relationships observed in Fig. 1. The intervention analyses are conducted to investigate this issue more thoroughly.

Intervention Analyses

We began the intervention analyses by constructing the univariate autoregressive integrated moving average (ARIMA) model for the filing and charging series using the 85-month period preceding the mass shooting. The univariate model is developed through an iterative model-building strategy that accounts for the stochastic processes associated with a series. Several factors are assessed when selecting an appropriate univariate ARIMA model. First, it is essential that each series have a single constant variance throughout its course. A nonstationary variance is engendered by dramatic fluctuations between observations in a series. To determine whether each series was stationary in variance, we consulted a rule-based expert system in the statistical software program SPSS 28 that uses a goodness-of-fit measure to compare competing models (IBM Corp, 2021). This goodness-of-fit measure indicated that all the series were stationary in variance.

A second issue is whether a series has a single constant level throughout its course. In other words, a series should not trend or drift upward or downward over time. We used the augmented Dickey-Fuller test to evaluate whether either series had a unit root (Dickey et al., 1986). Series that indicated a unit root were first-order differenced. A third consideration is whether a series has any cyclical or periodic fluctuation that repeats itself each time at the same phase of the cycle or period. This repetitive variation is commonly known as seasonality. Seasonality is most apt to occur at yearly intervals with monthly data. Our examination of each series autocorrelation function at lags of 12 months, 24 months, 36 months, and 48 months suggested that none of the series required seasonal differencing.

Once the series was stationary, we examined the autocorrelation function (ACF) and partial autocorrelation function (PACF) plots for autoregressive and moving-average processes. An exponentially weighted sum of one or more previous values in an autoregressive process influences the current value in a series. The effect of one or more prior observations (i.e., the order of the autoregressive parameter) on the current observation diminishes over time (Yt = ϕ1 Yt − 1 + … + ϕp Yt − p + at). In contrast, each value in a moving-average process is determined by the average of the current disturbance and one or more previous disturbances. The effect of a moving-average process lasts for a finite number of periods (i.e., the order of the moving-average parameter) and then disappears abruptly (Yt = at − θ1 at − 1 − … − θq at − q).

After constructing the univariate ARIMA models, we assessed the impact of the mass shooting on each decision point. We evaluated three distinct intervention models for the mass shooting (McCleary et al., 2017). First, we considered the possibility that the prosecution of firearm crimes changed abruptly (i.e., increased or decreased) after the mass shooting and remained at this new level over time (ω0It). This type of intervention is typically called a zero-order transfer function. For example, if the intervention coefficient for the mass shooting were negative and statistically significant for the filing series, it would indicate support for the proposition that the mass shooting influenced prosecutors to accept fewer firearm cases for prosecution.

Second, we investigated the possibility that the mass shooting had a small initial impact on each series that grew larger over time. In this transfer function model, two effect parameters are estimated. The omega parameter (ω) measures the degree of change in the level of a series, and the delta parameter (δ) estimates the amount of time required for this change to be actualized. A higher value for the delta parameter would indicate a more gradual impact of the mass shootings on a series. In contrast, a small delta coefficient would denote a more rapid effect.

Finally, we tested a pulse model (ω0[(1 − B)It]This model implies an initial abrupt increase or decrease due to the intervention, which then quickly decays without permanently changing the mean of the series. These three intervention analyses indicated that the zero-order transfer model best fits the data for each series.

Table 2 presents the mass shooting and COVID-19 coefficients and statistical significance values for each outcome. It is important to note that because we performed multiple statistical tests, the chance of obtaining at least one statistically significant result due to random chance is amplified. Consequently, we employed the Bonferroni correction method to address the problem of multiple comparisons, which helps to mitigate the probability of making one or more Type I errors (false positives) when conducting multiple significance tests (Bland & Altman, 1995).

Table 2 ARIMA model parameters for firearm-related cases (January 2011–December 2020)
Full size table

The Ljung-Box Q statistic indicated that the residuals for all models were uncorrelated (Ljung & Box, 1978). Visual examination of this table shows that in contrast to the pre-post data reported in Fig. 1, the mass shooting had little direct effect on the number of monthly firearm crimes accepted for prosecution after accounting for the preexisting trend in the data. This finding suggests that fewer arrests for firearm crimes likely engendered the reduction in firearm cases accepted for prosecution that was initially noted in Fig. 1. Arrest data buttresses this assertion in that monthly reported firearm crime incidents resulting in at least one arrest for firearm-involved crimes declined steadily throughout the 10-year study period.

Failing to establish the importance of the mass shooting in affecting the filing decision for firearm crimes, we turn our attention to the other decisions made by prosecutors. These results are also reported in Table 2. Appraisal of this table reveals that the mass shooting impacted all the remaining decision points except for a negotiated plea bargain. Prosecutors were more likely to deal harshly with firearm defendants following the mass shooting. Firearm defendants were less likely to have their case nolle prossed, their arrest to filing charge reduced, their arrest to disposition filing charge reduced, and their filing to dispositional charge reduced. The plea bargaining of firearm cases remained unchanged pre- and post-Parkland, which itself indicates a reduction in leniency after Parkland because charge reductions are typically used by prosecutors during plea negotiations with the defendant to secure a conviction. With fewer charge reductions occurring after Parkland, many firearm defendants were forced to accept the original charge if they desired to make a deal with the prosecutor. None of the COVID-19 coefficients were substantive in any of the equations.

Supplemental Analysis

We conducted a supplemental analysis to help ensure that our original findings remained robust across different specifications. Our initial ARIMA analyses were rooted in the assumption that the Parkland mass shooting might influence prosecutors’ handling of firearm crimes because of political pressure. Nevertheless, we still felt it prudent to determine whether our results would vary if an outcome measure unrelated to firearm crime were analyzed. Consequently, we estimated several ARIMA models using motor vehicle theft as the dependent variable. We speculated that the Parkland mass shooting would have little impact on the prosecution of a property crime like motor vehicle theft.

We initially constructed a figure depicting the average monthly number of motor vehicle theft cases filed, nolle prossed, receiving an arrest to filing reduction, receiving an arrest to disposition reduction, given a filing to disposition reduction, and resolved by a plea negotiation before and after the mass shooting. Figure 2 shows that the mass shooting had a differential effect depending on the dispositional outcome. The number of monthly filings, nolle prossed cases, and filing to disposition reductions for motor vehicle theft remained unchanged between the pre-and post-mass shooting periods. In contrast, the effect of the mass shooting on the other outcomes varied. While motor vehicle theft defendants experienced an increase in arrest-to-filing and arrest-to-disposition reductions post-Parkland, negotiated plea deals showed a substantive decrease after the mass shooting.

Fig. 2
figure 2

Average monthly motor vehicle theft cases before and after the Marjory Stoneman Douglas High School shooting in Broward County, Florida (January 2011–December 2020). Notes: N = 120 months. *p ≤ .01 (with Bonferroni correction)

Full size image

While the descriptive data presented in Fig. 2 are informative, we undertook several ARIMA analyses to better establish whether the Parkland mass shooting influenced the prosecution of motor vehicle theft cases. The results of these analyses, which are reported in Table 3, show that the mass shooting had little substantive effect on how prosecutors handled motor vehicle theft cases because the coefficient for the mass shooting intervention variable was not substantive in any of the equations estimated.

Table 3 ARIMA model parameters for motor vehicle theft cases (January 2011–December 2020)
Full size table

However, while the mass shooting had little impact on the processing of motor vehicle theft cases, the coefficient for the COVID-19 variable was statistically substantive in the negative direction for two of the six equations estimated. Prosecutors were less likely to file and prosecute auto theft cases following the outbreak of COVID-19. None of the COVID-19 coefficients estimated in the other four equations were of salience.

Discussion

It is proffered in the extant literature that prosecutors are influenced unduly by the media attention given to specific criminal cases. This reasoning hinges on the assumption that media attention amplifies the likelihood that the voting public will scrutinize how their elected prosecutor handles their criminal cases. Prior research finds that a prosecutor’s behavior in handling a criminal case is impacted by the media attention focused on the case. Less understood is whether this effect is limited solely to the publicized criminal case. The cognitive nexus in the mind of a prosecutor may be between criminal cases involving similar circumstances rather than just the specific case receiving media attention. We made the logical assumption in this study that the media attention given to a particular case will not only influence the prosecutor in that case but also affect the prosecutor in less publicized cases that are analogous to the publicized case because of the amplification in public scrutiny.

Criminal defendants encounter several points as they progress through the criminal justice system. These common decision points include whether the defendant was released on bail, whether the defendant received an incarceration sentence and the length of the imposed criminal sentence. We focused on decisions made by the prosecutor because a prosecutor has unimpeded discretion in determining what criminal cases to pursue and what charges to levy against a defendant.

The results generated in several ARIMA analyses showed that while the Parkland mass shooting had little effect on the filing decision and plea bargaining for firearm cases, firearm defendants were less apt to have their case nolle prossed, their arrest to filing charge decreased, their arrest to disposition filing charge reduced, and have their filing to dispositional charge reduced. Taken in their totality, these findings suggest that prosecutors had an enhanced proclivity to deal harshly with firearm defendants following the Parkland mass shooting.

Although the data in this study gave us an excellent opportunity to evaluate the effect of a highly publicized criminal event on prosecutor decision-making in Broward County, a question remains about the applicability of our findings to other locations. It is essential to point out that the elected State Attorney holding office during the study period was a Democrat. Because political affiliation likely plays a role in how the prosecutor handles criminal cases, researchers must undertake similar analyses in counties with an elected prosecutor of a different political affiliation because there may be a substantial difference in how the media impacts Republican prosecutors. When only 14% of Republican voters indicate that they have a great deal or fair amount of confidence in the media, compared to 70% of Democrats (Brenan, 2022), there might be a noteworthy difference in how the media influences prosecutors of different political parties. Regrettably, we are unable to address this issue. Future research should focus on how media attention affects prosecutors of different political parties. Should these studies produce similar results, then confidence in the generalizability of our findings will be enhanced. Conversely, if additional analyses generate divergent results, our findings may only apply to Democratic prosecutors. Only through further probing can we be assured of the generalizability of our findings.

Second, while our longitudinal analyses present empirical evidence that the mass shooting influenced many of the decisions made by prosecutors, our findings are tentative because we analyze only firearm crimes and motor vehicle theft. We cannot determine what the effects might be for other offenses. Further insight into the nature and strength of the relationship between the media and prosecutorial behavior for other crimes without an identifiable victim, such as drug offenses, must await additional research.

Another issue pertains to variability. Some prosecutorial outcomes analyzed in this study have low variability because of our monthly data use. That said, it is important to point out that four of the six intervention coefficients reach statistical significance in their respective equations. This observation is noteworthy because low variability in the dependent variable generally engenders decreased statistical power (Cohen et al., 2003). The effect of the mass shooting on the different prosecutorial outcomes reported in Table 2 and graphically depicted in Fig. 1 are also all in the negative direction, including the nonsignificant coefficients reported in Table 2. These findings lead us to believe that the lack of variability in some prosecutorial outcomes analyzed did not impact our results adversely.

Fourth, contextual analyses are warranted because it is plausible that other local political considerations besides the mass shooting may also influence a prosecutor’s decision-making. For example, the racial composition of a circuit may have a noteworthy effect on how prosecutors handle their caseload. Surveys consistently show that many Black citizens lack confidence in the police (Morin & Stepler, 2016). Black citizens also frequently report feelings of injustice and believe they have been the targets of excessive police use of force. The publicized killings of unarmed black citizens by police have further exacerbated this mistrust. Based on these observations, it seems plausible that Black citizens’ widespread distrust of the police may put enhanced pressure on a prosecutor to remedy perceived police injustice in areas where the Black vote is more influential.

To illustrate, in his seminal work, Blalock Jr. (1967) argued that as the percentage of Blacks in the population grew progressively larger, the state apparatus increasingly viewed Black citizens as a threat to the political ascendancy of whites. State actors respond to this threat by relying more heavily on social control, such as the arrest sanction and the harsh prosecution of Black criminal defendants to maintain the dominant position of Whites in society.

However, the relationship between the percentage of Blacks in the population and social control is not linear. A curvilinear relationship exists because once the Black population nears or eclipses the size of the White population, the amount of social control Black citizens experience declines because of the political mobilization of the Black population. This presumed threshold reflects the ability of Black citizens to gain greater political power and influence under a situation in which their relatively large population size affords them the opportunity to mobilize their resources for collective action (Leighley, 2021). It is also asserted that Whites are more apt to seek economic accommodation with Blacks when the Black population in the community is larger (Cook et al., 2023). Based on this perspective, one could make a strong theoretical argument that contextual factors, such as a community’s racial composition, probably influence how a prosecutor handles criminal defendants.

While previous studies have investigated the issue of racial threat for more visible outcomes such as pretrial detention, jail incarceration, and prison incarceration, little research has considered the impact of racial threat on prosecutor decision-making. This oversight is surprising when one considers prosecutors to be the “kings of the courtroom” because of their power in prosecuting cases (The Economist, 2014).

A paramount concern in the administration of justice pertains to the unequal treatment of similarly situated criminal defendants. Our findings have profound implications since they bear directly on the current debate as to whether outside influences, such as media attention, impact the prosecution of criminal cases. The results of this study suggest that the mass shooting in Parkland influenced the charging of firearm crimes.

Our findings also have relevance to the ongoing debate regarding elected versus appointed prosecutors. Some argue that elected prosecutors are beneficial because of their responsiveness to community needs and concerns, their promotion of democratic legitimacy through the election process, and their transparency resulting from election campaigns. Elected prosecutors perceived as doing a poor job can also be replaced by a dissatisfied electorate. As Pizzi (Pizzi, 1993, p. 1339) notes, “If someone is to decide which laws will be aggressively enforced, which laws will be enforced occasionally, and which laws will never be enforced, it makes sense that the person who has to answer to the voters will make those determinations.”

However, as highlighted in previous research and the current study, a major potential problem with elected prosecutors is that political pressure provoked by excessive media attention may motivate prosecutors to make decisions that may not necessarily be rooted in objective legal analysis. Based on the data analyzed here, we cannot determine whether the observed changes in prosecutor decision-making following the Parkland mass shooting were legally warranted. Just because prosecutors modified their behavior after the mass shooting does not necessarily mean the changes implemented were not legally justified. While the supplanting of locally elected prosecutors with appointed prosecutors is probably a “nonstarter” (Wright, 2014), regular reviews and public access to prosecution data will go a long way in helping to ensure that elected prosecutors serve the public interest.