In the fall of 2018, Supreme Court nominee Brett Kavanaugh suddenly faced allegations that he had committed a rape in the 1980s. When his confirmation appeared in peril, Kavanaugh’s supporters accused his detractors of infidelity to a cherished American ideal - the presumption of innocence (“POI”). President Trump thus lamented, “My whole life I’ve heard you’re innocent until proven guilty, but now you’re guilty until proven innocent.”1 RNC chairwomen Ronna Romney likewise tweeted, “A Democrat on the Judiciary Committee (who graduated from law school) doesn’t believe in the presumption of innocence for conservatives… That’s terrifying and goes against the bedrock of our entire justice system.”2
For advocates and practitioners working in the trenches of the criminal justice system, these sudden high-profile pleas to honor the innocence presumption were grimly ironic. First, Kavanaugh himself faced no criminal charges and his personal liberty was never in jeopardy.3 What Kavanaugh actually stood to lose was a lifetime job - a golden opportunity he seemed destined to seize no matter the allegations against him. Second, and more importantly, the implicit assumption that the system was only just now ignoring the POI was deeply flawed. Every day, hundreds of thousands of legally innocent criminal defendants are deprived of their liberty and caged in jails despite the bedrock principle.4 These anonymous defendants languish behind bars, guilty until proven innocent, and no high-profile chorus protests.
This Essay aims to set the record straight. Our system is in crisis because the POI is indeed in peril. However, this crisis did not begin with the Kavanaugh controversy and it did not end with the swearing-in of the 114th Justice of the Supreme Court. Instead, the real danger to the presumption lies in the day-to-day operation of our criminal justice system. The true affront to our cherished POI tradition is in the phenomenon of mass pretrial incarceration.5
How bad is the problem of mass pretrial incarceration? Why isn’t the POI more effective in shielding criminal defendants from unjustified pretrial incarceration? The short answer is that our mass pretrial incarceration problem is profound, and that doctrinal misunderstanding has rendered the POI an impotent safeguard to pretrial liberty. The long answer - based on new empirical data and back-to-basics doctrinal analysis - is provided in this remainder of this Essay. Here’s the roadmap:
Part II of the Essay harnesses empirical evidence to show that our pretrial incarceration problem is far worse than most realize. After considering national data that suggests the scale of the phenomenon, I focus on an original dataset of over 150,000 Maryland District Court cases that reveals the disturbing nature of the injustice. Specifically, this original data shows how every year thousands of accused persons are routinely jailed for extended periods on charges that are ultimately dropped. In other words, presumed innocent individuals are deprived of their liberty for weeks and months at a time and then are unceremoniously released because the accusations against them are withdrawn. This utterly unjustified pretrial incarceration devastates lives and exacerbates systemic inequalities.
In Part III, I lay partial blame for the POI’s blaring impotence at abating this problem on a doctrinal position I call “the argument that cries Wolfish.” This doctrinal position, most prominently advanced in dicta by Justice Rehnquist in Bell v. Wolfish, maintains that the POI merely restates of the prosecution’s burden of proof at trial and therefore has no relevance to pretrial matters.6 Since courts largely accept the argument that cries Wolfish, the POI provides no effective legal barrier to unnecessary and unwarranted pretrial incarceration. Yet I maintain that the argument that cries Wolfish is flawed. While the POI does indeed require proof beyond a reasonable doubt at trial, it also requires - at a minimum - proof by clear and convincing evidence before detention can be imposed pretrial.7 This commonsense interpretation of the POI’s imperative both squares with due process doctrine and forecloses aggressive readings of Rehnquist’s Wolfish dicta.
In this Essay’s final Part, I conclude that we must reject the appeal to Wolfish because it is every bit as misguided and dangerous as the fabled boy who cried wolf. As the novel data presented in this Essay demonstrates, the theoretical presumption that criminal defendants will not be proven guilty is empirically warranted. In other words, the POI accurately describes reality in our era of police and prosecutorial overreach.8 People who are accused of crimes but will never be convicted should not spend months in jail except in very rare circumstances. To confront the alarming reality of mass pretrial incarceration, the notion that the POI does not apply pretrial must be rejected.
At any given moment, approximately 20% of the 2.3 million people incarcerated in United States prisons and jails - 462,000 individuals - are detained pretrial.9 These individuals face criminal charges but stand unconvicted; they are thus legally innocent. Though wrapped in the POI, they languish behind bars.
Formally, pretrial incarceration of the presumed innocent occurs through established legal processes. Every pretrial criminal defendant has seen a judicial officer and only ends up in jail after either (a) she has been assigned a monetary bail but fails to pay it; or (b) she has been deemed dangerous to her community or to pose a high risk of non-appearance at trial.10 Yet the existence of formal legal processes only accentuates the POI's utter failure, in pretrial hearing after pretrial hearing, to prevent unconvicted persons from suffering imprisonment.
The mass scale of the pretrial incarceration phenomenon deserves emphasis. As large as a number as 462,000 cited above is, it only represents a daily snapshot. Annually, over 10 million people cycle through America’s local jails.11 Since majority of those who cycle though are under pretrial detention, the reality is that every year, the POI fails to prevent millions of legally innocent individuals from spending days, weeks, and months in jail.
Despite their implications, these national pretrial detention statistics do not trigger alarm bells in most circles. Perhaps this is because, in a sense, the increase in pretrial detention of the unconvicted is old news. As Figure 1 shows, the trend originates from the mid 1990s. Quite frankly, the reality of large-scale incarceration of the presumptively innocent would not surprise most policy makers or practitioners.
Yet the mere fact that the issue is not new does not adequately explain the lack of outrage. Why isn’t there a high-profile chorus protesting the assault on the bedrock POI principle?12
Without question, prejudice and apathy play a significant role. The losers in this POI equation are, after all, mostly poor and disproportionately of color. And class bias and structural racism inhibit justice now just as they did when the mass pretrial incarceration trend began.
At the same time, I suspect a different attitude exerts undue and unrecognized influence on the conversation. This attitude usually goes unstated and is not rooted in traditional American racial and class biases. Specifically, I suspect many people of good faith regard the POI as an abstract legal technicality and it does not bother these people that accused criminals are not released from jail on a technicality.
The logic animating this attitude has intuitive appeal and deserves serious consideration. While a criminal defendant is presumed innocent before trial, she may well be found guilty at trial and then sentenced to jail or prison. What does it matter if she serves some of that sentence up front? After all, she will get credit for all time served. Perhaps there are 462,000 legally innocent defendants locked up at any given moment, but surely (per this intuitive appeal) just about all of these folks are actually guilty and bound to lose their liberty anyway.
One response to this line of practical reasoning condemns its misunderstanding of relevant law. After all, Supreme Court doctrine plainly prohibits pretrial incarceration as punishment for unconvicted crimes.13 Though perfectly sound, this response will hardly persuade those who already perceive “relevant law” as an abstract technicality. Without conceding that it would ever be justified to preemptively imprison a defendant on the promise of future conviction, I suggest a different answer is needed if we wish to move the opinions of those who suspect pretrial detainees to be presumed innocent but actually guilty.14
Of course, many of those who have crossed paths with the criminal justice system - the accused, their families, and their lawyers alike - will not share the intuition that most defendants detained pretrial are actually guilty and thus “belong” in jail. Their experiences support a contrary intuition. Infuriating anecdotes abound of criminal defendants spending serious pretrial time behind bars only to have flimsy charges dismissed. Anger only mounts when former defendants are unceremoniously released from days, weeks, or months jail without so much as an apology. The presumption of legal innocence seems warranted for these defendants since the accusations against them fail.
So, who is right? Are most of the 462,000 legally innocent pretrial defendants incarcerated on any given day actually guilty or is their legal presumption of innocence empirically warranted? Up until now, there has been a dearth of good data directly speaking to the question.15
Before considering new data that does speak directly, I want to briefly speculate on perceived reality. I imagine that the dominant intuition in policy-making circles is that detained pretrial defendants are generally guilty. My own sense of this dominant intuition proceeds from this premise: without hard numbers to tell a clear story, intuition will emerge from lived experience. Since the lived experience of typical policy makers involves privilege rather than criminal-justice-system entanglement, they assume the system usually gets it right.16 Thus, policy makers typically do not raise a hue and cry about locking up presumed innocent defendants because they sincerely believe those detained pretrial are simply “banking” timed-served credits that they will subsequently claim after their inevitable conviction.17
Yet the hard numbers presented in this Essay tell a different story. This story, based on five years of District Court data from the four largest counties in Maryland, starts with the observation that most people who are charged with crimes will have all their charges dropped before trial - a procedure known in Maryland by the Latin phrase nolle prosequi, or colloquially, nolle pross. Those who have all charges against them dropped have, by definition, had their presumption of legal innocence vindicated. Remarkably, the nolle pross rate observed in this dataset is approximately 60%. Conviction on any charge is thus far less likely than nolle pross on all charges.
The story told by the numbers gets worse. Despite the high District Court nolle pross rate, large numbers of defendants whose legal innocence will be vindicated end up incarcerated pretrial anyway. Indeed, over the five-year period studied, just under 12,000 defendants spent a combined total of over 1485 years incarcerated pretrial before all the charges they faced were dropped, an average of 46 days in jail per person. Because these defendants were presumed innocent when initially subjected to pretrial lock-up, and because their legal innocence was subsequently vindicated by complete nolle pross, I call the time they spent behind bars unjustified incarceration.18
Absent reason to believe that the criminal justice systems in Maryland’s four largest counties wildly differ from those in the rest of the nation’s urban jurisdictions - and evidence suggests otherwise - the data support the inference that unnecessary pretrial incarceration is widespread and common in the United States. The mass pretrial incarceration situation is dire and so we lament: the POI is dead.
Appendix A details the method by which the data for this study were collected and analyzed. In broad strokes, these data are comprised of three related sets of criminal cases from five-year period (2013-2017) in the four largest counties in Maryland (Baltimore City, Prince George’s County, Montgomery County and Baltimore County).
Maryland has a two-tiered court system for criminal trials. All cases begin in District Court, but more serious cases tend to move up to Circuit Court - as do all cases tried before a jury.19 This study focused on cases that began and ended in District Court during the 2013-2017 time period.20 In addition, it only looked at cases where the criminal defendant appeared before a District Court Commissioner to determine whether that defendant would be released or detained pretrial.21 In sum, 167,022 cases fit this bill.22
Charging, outcome, and demographic information about these 167,022 cases can be explored via this portal as explained in Appendix B. The initial focus of this study, however, concerns nolle pross rates. Specifically, I found that all charges were dropped before trial in 101,558 cases - a total nolle pross rate of 60.81%.23 As shown in Figure 2 below, if we focus only on the disposition of the top charge, the nolle pross rate was 67% - compared to a top-charge guilty rate of just under 11%.
This nolle pross rate is significant. It means that in 60.81% of District Court cases, proof of guilt was not forthcoming, thus confirming the legal innocence of the accused.24 On its face, a 60% nolle pross rate would appear to empirically validate the POI for the jurisdictions in the study. In other words, even if the Constitution did not mandate a POI, it would make sense to implement one in Maryland’s four largest counties given that it is more likely than not that an accused person will have all the accusations against her dropped.25
After determining a nolle pross rate, the second stage of the study calculated unjustified incarceration time. This was obtained by identifying all cases in the study where the defendant was jailed pretrial only to have all charges dropped. As shown in Figure 3 below, there were 11,842 cases in this category.26 This represents 7.09% of all District Court cases (or 11.66% of cases that were nolle prossed in their entirety).27
While 7% may sound like a low percentage in the abstract, the context here bears repeating. These are people whose liberty was completely deprived despite their presumed legal innocence. Figure 3 further shows that each person spent an average of 46 days in a cage - or a median of 39 days - before her actual legal innocence was formally acknowledged. Consider the intense disruption and hardship caused by 39 to 46 days of incarceration on individuals and families - job loss, eviction, child custody and more. This should not be shrugged off lightly. Imagine if 7% of the Maryland Judiciary or 7% of Congress were subjected to 39 days of unjustified incarceration every year. Heads would roll!
Yet nearly 12,000 souls suffered this fate in just a single state in a just single five-year period without any widespread alarm. They spent a total of 1485 years and 121 days of pretrial incarceration before all charges against them were dropped. For these people, the POI was an empty formality; it provided no meaningful protection from unjustified incarceration.
This then is then central empirical claim of this Essay: for significant numbers of people, the POI does not protect against unjustified pretrial incarceration. This POI problem has real magnitude, affecting many thousands of Americans every year. Mass-pretrial-incarceration-despite-the-POI is a problem that should concern us far more that POI complaints surrounding Supreme Court nominations or other cases where jail time is not at issue. In a world where the POI had real teeth, the problem of unjustified pretrial incarceration would not be as common as the hard numbers just presented show it is.
In the next part of this Essay, I will analyze why the POI does not have real teeth. Before turning to that task, however, I want to raise possible objections to my central empirical claim. Skeptics will naturally question whether the data really show that unjustified pretrial incarceration is a significant problem and their objections deserve a response.
A first objection might question the asserted nexus between nolle pross and unjustified incarceration. Just because a prosecutor chooses to drop all charges against a defendant, the argument goes, does not mean that the defendant did not engage in dangerous criminal acts. Prosecutors drop charges for all kinds of reasons including lack of cooperation from complaining witnesses; such failures of proof do not mean that the charges were not warranted when filed. Therefore, pretrial incarceration on charges ultimately nolle prossed may have been justified.
While this objection has some appeal,28 it proves too much. Consider that the same logic applies to defendants acquitted after trial. After all, it’s no secret that juries sometimes acquit factually guilty defendants. Yet nobody would suggest that just because some juries get it wrong, we should randomly sentence some percentage of acquitted defendants to a median of 39 days prison time. That would surely constitute unjustified incarceration.
The point is that “unjustified incarceration” represents a legal concept of justification tethered to legal norms. Per these norms, nolle pross vindicates legal innocence as surely as acquittal (if not as finally).29 It would be unjustified, on this understanding, to ignore the legal conclusion of innocence entailed by a nolle pross based on metaphysical uncertainty about facts.30
Of course, it is fair to distinguish between ex ante and ex post justification. Dropping charges ex post does not necessarily mean that decisions to charge or to incarcerate pretrial were fatally flawed or violated due process ex ante. For this reason, I do not argue that prosecutors should either hesitate to bring charges or to nolle pross as they see fit. Instead, I argue that the numbers show that (ex post unjustified) pretrial incarceration happens too often and that the POI should do better at protecting against this outcome.
This brings us to a second and final objection. This objection accepts the validity of the nolle-pross-unjustified-pretrial-incarceration nexus, but questions whether the problem identified is truly widespread. Proof that Maryland suffers from this injustice is not the same as proof that the problem extends beyond Maryland’s boundaries. According to this argument, purely local conditions likely explain the observed data, meaning it is not likely that there is a national pretrial POI crisis.
The short response here is that any debate over Maryland’s likely representativeness currently cannot be resolved empirically. This is because, as mentioned above, there is a dearth of good data speaking to the specific questions raised herein.31 Unless and until others pursue parallel investigations in other jurisdictions, the representativeness of the nolle-pross-unjustified-pretrial-incarceration situation in Maryland’s four largest counties will necessarily involve guesswork.32
With this caveat in mind, I would offer a few observations. First, nothing in Maryland’s history suggests that it has taken a radically different path from its neighboring states or from the nation as a whole.33 Second, in the general criminal context, Maryland does not appear to be an outlier state. Although it has a relatively high violent-crime rate, Maryland sits below the national median for property crimes.34 In short, the high-level picture offers no strong reason to believe that Maryland counties have radically atypical prosecution patterns.35
In closing this Part, let us return to our opening statistics: at any given moment 462,000 legally innocent individuals are detained pretrial across the country; 10 million people cycle through jail over the course of a year.36 At minimum, the data presented herein make a prima facie case that some non-negligible percentage of these detainees spend considerable time behind bars without ever being convicted of any crime. In this light, the phenomenon of mass pretrial incarceration becomes even more alarming.
Bear in mind: this study’s bottom-line estimate of 1485 years of unjustified incarceration in Maryland alone over 5 years is already very conservative. Multiply that by whatever factor you please and it becomes undeniable that every year, the POI fails to stop many lifetimes of unjustified incarceration. By any fair measure, the POI is not alive and well.
The argument that cries Wolfish – and its relationship to the death of the pretrial POI – is best introduced though a recent, concrete case. As you consider the facts of Commonwealth v. Duse,37 note how the court uses Wolfish to defeat the proposition that the POI has relevance to a pretrial release decision for an accused-but-unconvicted individual.
On July 26, 2017, a man named Rex Olsen was shot and killed outside of his workplace in Warrenton, Virginia.38 Six days later, police arrested Bernard Duse, Jr., a 76-year-old African American with no prior criminal history.39 Duse sought pretrial release pending his trial on murder charges and a bail hearing was eventually held on December 28.40
At the hearing, Duse presented evidence showing his high level of education, strong community support, and consistent lifelong employment.41 The Commonwealth opposed bail, citing evidence of Duse’s contentious employment-dispute history and mental health struggles.42 Duse countered with his own mental health expert and with a bail bondsman’s testimony that GPS monitoring could track Duse’s location and ensure his appearance at trial.43
The circuit court concluded that the conflicting evidence “counter act[ed] and balance[d] each other off.”44 The court then set a $75,000 bail and ordered GPS monitoring. In justifying this decision, the court noted:
Furthermore, we’re not trying this case. We’re not… making an adjudication of Mr. Duse here as to his guilt beyond a reasonable doubt. There may have been some probable cause determinations, but those were done with less than 50 percent certainty. That’s what probable cause is. It’s a reasonable belief. He hasn’t been adjudicated guilty yet. He’s entitled to a presumption of innocence.45
After the Commonwealth unsuccessfully appealed to the Court of Appeals, the case proceeded to the Virginia Supreme Court.46
Before Virginia’s top court, the Commonwealth asserted that the “circuit court abused its discretion by applying the doctrine of presumed innocence to a pre-trial bail hearing.”47 The Virginia Supreme Court agreed: “by applying the presumption of innocence, the circuit court utilized an erroneous legal standard to guide its consideration… and its decision regarding bail, premised on that consideration, was an abuse of discretion.”48
For the key proposition that it was erroneous to apply the POI to a pretrial release decision, the Duse court cited Bell v. Wolfish.49 Indeed, the Duse court quoted the precise language that forms the crux of the argument that cries Wolfish: “The presumption of innocence is a doctrine that allocates the burden of proof in criminal trials… [b]ut it has no application to a determination of the rights of a pretrial detainee during confinement before his trial has even begun.”50
The Duse court exemplifies how the argument that cries Wolfish operates in practice. First, the Duse court uses the argument to deny any relevance of the POI to a pretrial release decision. Second, this sweeping conclusion is justified solely by dint of Wolfish’s unimpeachable authority; Wolfish’s reasoning receives no scrutiny, the justification is pure argument ex concessis.51 Third, the bottom-line result is pretrial detention of a legally innocent defendant.52
Perhaps unsurprisingly, I assert that the argument that cries Wolfish is both dangerous and wrong. It is dangerous because its widespread acceptance has helped facilitate mass pretrial incarceration. And it is wrong because the argument misunderstands the true import of Wolfish and also ignores post-Wolfish doctrinal developments. These positions are developed below.
Definitively proving that the Wolfish argument has widespread acceptance in the doctrine is tricky. This is because, as I have explained in prior work, the tangled hermeneutic nature of legal doctrine renders precise empirical analysis of its evolution extremely difficult.53 Doctrine develops through complex dialectics between legal texts and legal readers; judges may be influenced by texts that they do not cite, lawyers may not make not make promising arguments simply because they seem foreclosed.54 Such dynamics elude measurement. Thus, my claim that the argument that cries Wolfish has widespread acceptance is not fundamentally empirical.55
Instead, I offer different kinds of proof. First, key cases like Duse show that the argument that the POI has no relevance to pretrial release decisions is accepted without serious dissent or reference to a widely recognized contrary view.56 Alas, many Duse-like cases exist.57 Second, I see significance in the fact that the Wolfish argument has traction in academic circles, even among scholars working to decrease our pretrial incarcerated population.58 When even scholars who support more pretrial release over mass pretrial incarceration embrace the Wolfish position, the argument’s ascendancy seems clear.59
Turning to my claim that the argument that cries Wolfish is dangerous, once again definitive proof is tricky. The basic form of my claim is counterfactual - if Wolfish had not killed the pretrial POI, pretrial mass incarceration would not have reached the epidemic proportions it has today - and philosophers dispute what even “counts” as counterfactual proof.60 At the risk of offending such philosophers, I base my claim about Wolfish’s danger on intuitive appeal and this revealing graphic:
Figure 4 shows the number of citations to “presumption of innocence”, “pretrial detention”, and “bell v. wolfish” from American caselaw between 1939 and 2017.61 The graph clearly suggests a parallel between when cites to Wolfish began in the late 1970s and when courts started speaking about pretrial detention. Meanwhile, POI cites in caselaw started to spike before pretrial detention’s rise but then rose in parallel fashion. In the mid 2000s, cites to POI declined rapidly and now courts invoke pretrial detention about as frequently as POI.
The parallel lines in Figure 4 suggest that the judicial conversation around Wolfish, the POI, and pretrial detention moved together. And when these conversation lines are triangulated with Figure 1 (showing the steady rise in the number of pretrial detainees since the early 1980s), the overall picture supports the intuition that our current state of mass pretrial incarceration was facilitated by Wolfish’s blow to the pretrial POI.
Even absent definitive proof that Wolfish shares blame for pretrial mass incarceration, it remains the case that the argument should be rejected. This is because, like the boy who cried wolf, the argument that cries Wolfish is usually wrong and cannot be trusted.
The argument fails for four basic reasons. First, the argument overstates Wolfish’s limited holding. Second, the argument ignores pre-Wolfish POI Supreme Court precedent that undermines Wolfish’s aggressive dicta. Third, the argument fails to account for post-Wolfish doctrinal developments that redeem a stronger view of the POI. Finally, the argument’s insistence that the POI is only relevant at trial is neither logically necessary nor compatible with the in-the-trenches reality of today’s criminal justice system.
We begin with Wolfish’s limited holding. The litigation that led to the Supreme Court’s decision started as a class action challenge to the conditions of confinement for pretrial detainees in New York City’s federal Metropolitan Correctional Center (MCC).62 The District Court enjoined a number of MCC practices towards detainees including double-bunking and requiring body cavity searches after contact visits, and the Second Circuit affirmed.63
Ultimately, a 5-4 majority on the Supreme Court reversed, holding that the conditions in the MCC did not violate the detainees’ rights.64 At one point in his majority opinion, then Justice Rehnquist rebuffed the District and Circuit courts below for relying on the POI “as the source of the detainee’s substantive right to be free of [the contested] conditions of confinement.”65 Then Justice Rehnquist famously wrote:
The presumption of innocence is a doctrine that allocates the burden of proof in criminal trials… But it has no application to a determination of the rights of a pretrial detainee during confinement before his trial has even begun.66
On its face then, Wolfish only rejects application of the POI as relevant to pretrial detainees’ conditions-of-confinement challenges. The case says nothing about the POI’s relevance to pretrial detention decisions. Indeed, Justice Rehnquist specifically noted: “It is important to focus on what is at issue here. We are not concerned with the initial decision to detain an accused and the curtailment of liberty that such a decision necessarily entails.”67
By suggesting that Wolfish supports exiling the POI from pretrial detention decisions, the argument that cries Wolfish violently misreads the case’s consciously limited holding. As some perceptive courts have recognized, this misreading alone is enough to reject the argument.68
A second reason to reject the argument is that Justice Rehnquist’s aggressive dicta seeking to limit the POI’s scope was belied by existing Supreme Court precedent. As Justice Stevens pointed out in his Wolfish dissent, the Court had twice previously “relied on [the POI] as a justification for shielding a person awaiting trial from potentially oppressive governmental actions” in McGinnis v. Royster and Stack v. Boyle.69
For Justice Stevens, McGinnis and Stack demonstrated that the POI “colors all of the government’s actions toward persons not yet convicted”; Stevens quite correctly faulted his brother Rehnquist for ignoring these cases and “relying on nothing more than force of assertion” to prove his questionable point.70 Rehnquist’s ipse dixit argument was thus weak from the start.71
This brings us to the third reason to reject the argument that cries Wolfish - subsequent doctrinal development. Two key Supreme Court cases - United States v. Salerno72 and Nelson v. Colorado73 - make clear that Rehnquist’s dicta never really held sway over the Court.
Generally speaking, Salerno is a hugely important case since it established the constitutionality of modern pretrial detention based on future dangerousness.74 For our purposes, however, Salerno’s specific import lies in what the case did not say. The vital context comes from Justice Marshall’s Salerno dissent, in which he vociferously objected to what they saw as a drastic expansion of the “police state”:
for the first time a statute  declares that a person innocent of any crime may be jailed indefinitely, pending the trial of allegations which are legally presumed to be untrue, if the Government shows to the satisfaction of a judge that the accused is likely to commit crimes, unrelated to the pending charges, at any time in the future.75
Over and over again in his dissent, Justice Marshall invoked the POI and protested that pretrial detention contradicted that axiomatic principle.76
Despite Marshall’s repeated sharp references, now Chief Justice Rehnquist remained entirely mute regarding the POI in Salerno. Rehnquist quoted Wolfish as support for the claim that pretrial detention does not inherently constitute punishment before trial, but he utterly failed to suggest that the POI was irrelevant pretrial or cite Wolfish for that proposition.77 Such silence from the author of Wolfish speaks volumes against the argument crying its name.
Nelson provides even more pointed proof of the argument’s apparent demise at the Supreme Court. Decided in 2017, Nelson concerned a challenge by persons who had criminal convictions invalidated by reviewing courts but were denied refund of restitution amounts, fines, and other fees previously assessed upon conviction.78 A 7-1 majority found that a scheme requiring such persons to prove their innocence by clear and convincing evidence violated Due Process.79
In her majority opinion, Justice Ginsburg explained that once a person’s conviction is reversed, the person’s POI is “restored.”80 After stressing the POI’s foundational nature, Justice Ginsburg dropped a footnote directly responding to a variation on the argument that cries Wolfish.81
Relying on Wolfish, Colorado had asserted that the POI “applies only at criminal trials” and therefore had no application in the non-trial situation before the Court.82 Ginsburg flatly rejected this interpretation, writing “Wolfish held only that the presumption does not prevent the government from ‘detain[ing a defendant] to ensure his presence at trial … so long as [the] conditions and restrictions [of his detention] do not amount to punishment, or otherwise violate the Constitution.’”83
In this crisp response, Justice Ginsburg kills any zombie notion that Wolfish means the POI applies only at trial. It means no such thing. Indeed, Nelson confirms that the POI remains a vital general principal and a source for rights beyond trial.84 Let us shout Nelson from the rooftops! The argument that cries Wolfish is dead in doctrine!
The fourth and final reason to reject the argument that cries Wolfish looks beyond the doctrinal debate that Nelson appears to have ended. Despite scholarly claims to the contrary, the notion that the POI only has trial relevance is logically unnecessary and also incompatible with daily reality of today’s criminal justice system.
Here, I begin by accepting that the POI’s reference to a presumption can be read as invoking formal procedural mechanisms of proof. On this reading, the POI establishes that a burden of proof falls on the State to provide evidence overcoming the presumption in order to establish the accused’s culpability.85 Strictly speaking, the POI does not constitute evidence of innocence to be weighed on the scales of justice; rather the POI calibrates the scales themselves.86
This procedural conceptualization of the POI means the first step in the Wolfish argument is technically correct - the POI does require the prosecution to introduce proof beyond a reasonable doubt to secure a conviction at trial.87 However, it does not then follow (as per the argument) that POI only has relevance at trial. The criminal process has contexts other than trial and these contexts also implicate burdens of proof.
Think of it this way: the POI means that police officers cannot stop and frisk presumed innocent pedestrians without reasonable suspicion;88 it means that law enforcement cannot arrest presumed innocent members of the public absent probable cause;89 and it means that presumed innocent arrestees cannot be detained without clear and convincing evidence of their dangerousness or risk of flight.90 At each stage, the POI requires recalibration of the scales of justice.
In the final analysis, the POI is a grundnorm - a first principle that stands at the back of a legal system’s chain of validity.91 The POI stands at the back of criminal law where it is logically prior to even Due Process.92 The POI thus defines the essential context for all criminal processes, dictating that the State must always have some level of compelling evidence before denying liberty. This is why the POI is “axiomatic and elementary, and its enforcement lies at the foundation of the administration of our criminal law.”93
In a system dominated by plea bargaining, it’s no secret that criminal trials are increasingly rare events.94 Relegating the POI to the cramped trial corner of the criminal stage (as the argument that cries Wolfish would do) is simply wrong. The POI deserves starring role and should be center stage.
The idea that liberty is our default and that denying liberty requires context-appropriate proof should inform our criminal justice process at every stage. The POI reminds us of this core value, this grundnorm. A lack of proper respect for the POI, egged on by the argument that cries Wolfish and its dismissive attitude, has permitted mass pretrial incarceration to flourish. It’s past time to restore to the POI the respect it deserves.
Prison and jail are seriously bad places. In an age of mass incarceration, we’ve become desensitized to this basic reality. Nonetheless, the fact remains that nobody should not suffer incarceration lightly. We are presumed innocent and should live free.
This Essay has presented original empirical evidence showing that the POI is not properly honored. In just one five-year period in Maryland’s four largest counties, legally innocent folks spent a combined 1486 years improperly incarcerated before all charges against them were dropped. While their POI was ultimately vindicated, these nearly 12,000 individuals languished behind bars for an average of 46 days, needless disrupting lives and exacerbating structural inequality.
There is no good reason to believe that Maryland is an outlier within the United States. The problem of unjust incarceration before charges are all dropped most likely occurs across the country. Our nation’s pretrial incarcerated population is nearly half a million at any given moment with over 10 million people cycling through jails annually. The scope and scale of pretrial mass incarceration’s injustice is staggering.
The argument that cries Wolfish shares the blame for this situation. Widely accepted for too long, this argument improperly degrades and limits the POI. It has successfully defended our grim status quo by deeming the POI irrelevant to pretrial detention decisions.
Yet the argument that cries Wolfish is wrong. Like the boy who cried wolf, it fooled people with false claims and has ultimately made us less safe. Luckily, the Supreme Court in Nelson95 finally sealed the argument’s demise. Lower courts everywhere should recognize this new doctrinal reality and the common sense upon which it stands. The POI should be enforced more rigorously pretrial.
Although the law now theoretically requires “clear and convincing evidence” to justify detention before conviction, the evidence presented shows that, at an absolute minimum, this high standard is not properly enforced in Maryland.96 The Court’s original blessing of the clear-and-convincing standard in Salerno was too tepid; the POI and the Constitution demand a more full-throated articulation.97
The Salerno opinion closed with soaring language: “In our society liberty is the norm, and detention prior to trial or without trial is the carefully limited exception.”98 This now rings hollow. Since Salerno, the exception of detention has swallowed the norm of liberty. This must change. Only when the POI is respected and the clear-and-convincing-evidence standard has real teeth will we be in a position confront mass pretrial incarceration.
The District Court data underlying Part II's empirical claims regarding nolle pross rates and unjustified pretrial incarceration derive from three concentric datasets: an "outer ring" (the "Baseline" set), a "middle ring" subset of the baseline (first subset - the "All Nolle Pross" set), and then an "inner ring" subset of middle ring (second subset - the "All Nolle Pross - Detained" set).
The outermost "Baseline" set consists of all closed criminal cases from Maryland District Court in Baltimore City, Baltimore County, Montgomery County, and Prince George's County which issued between January 1, 2013 and December 31, 2017 where (a) the defendant appeared before a Commissioner in order to determine whether he or she would be detained or released pretrial; and (b) the case was resolved at the district court level. There are 167,022 cases in the baseline set.
The middle ring consists of those cases in the baseline set where every single criminal charge lodged against a defendant was dropped before trial. In Maryland, dropping charges in this manner is referred to by a Latin phrase: Noelle Prosequi. There are 101,558 cases in the "All Nolle Pross" subset. In other words, in 60.81% of criminal District Court cases during this period, every charge was dropped before trial.
The innermost ring consists of those cases in the "All Nolle Pross" subset where the defendant was incarcerated pretrial and not released from jail until all charges were dropped. There are 11,842 cases in this second "All Nolle Pross - Detained" subset. This represents 7% of all cases in the outermost ring and nearly 12% of the middle ring.
Obtaining this original data was a multi-step process. The data used here ultimately originates from public records available through the Maryland Judiciary's Casesearch site. I specifically used a database called CLUE, which scrapes cases from Casesearch and places them in a fully searchable relational database. (Thank you to Matthew Stubenberg and Maryland Volunteer Legal Services for providing me with access to CLUE.)
1. SQL Queries + Create CSVs
The first step in my process was to design a series of Structured Query Language (SQL) queries to pull relevant cases from CLUE. All the queries used are available on GitHub. The basic strategy was to pull the following information for every case in each of the three datasets: casenumber, county, pretrial commissioner outcome, pretrial commissioner hearing date, defendant race, sex, zipcode, number of charges, top charge, top charge disposition, and disposition date.
The main query to get the outermost data set limited the search to closed criminal cases in the target jurisdictions where the case issued during the relevant 2013-17 time period and where the defendant appeared before a commissioner to determine pretrial release. This main query pulled all the information except number of charges per case; a second query pulled that information.
The "middle ring" query paralleled the main query but with the added condition that the outcome for every charge in a case must equal "nolle prosequi". As with outermost ring, a second middle ring SQL query was necessary to calculate the number of charges each defendant in this set faced.
Finally, the query to get the innermost ring was quite complex. The basic strategy was to find those cases where a defendant was released from custody (event code = "RELS") on the same day that his/her charges were dropped. In addition, this query deliberately excluded from the dataset those cases where an FTA (failure to appear) warrant had issued during the case and where there had been more than one RELS event. As with the outer and middle rings, a second query calculated the number of charges.
Once the SQL queries were designed, they were executed via this Python code. In addition to executing the queries and fetching the output, this code also merges the "main query" and "number of charges" results, and then outputs master Comma Separated Value (CSV) files for use in analysis. The CSV files can be directly examined at this GitHub repo.
2. Create Python App to Analyze Data
With the data in place, the next step was to create an application to permit user analysis of the data. My basic strategy was to load the data into a series of Pandas dataframes and then create a Flask application with Bootstrap design templates.
As explained more in Appendix B, the Flask app enables users to examine data a variety of ways. This includes breaking down each of the three datasets (outer-, middle-, and inner-ring) by race, gender, top charge, and so on. All of these looks employ the Pandas GroupBy function to perform aggregations.
The calculation for "time of unjustified incarceration" is relatively straightforward once all the data were obtained. It proceeds by reference to the innermost "All Nolle Pross - Detained" subset. For each defendant in this subset, the number of days was calculated from the appearance before the commissioner when the defendant was incarcerated (event code = CMIT) to the date when all charges were dropped and the defendant was released from custody (event code = RELS). Across the 5-year period in the four jurisdictions, the total amount of unjustified pretrial incarceration is 1485 years and 121 days.
It should be noted that the 1485 years of unjustified incarceration is very conservative and in all likelihood undercounts the actual amount of time unnecessarily spent in jail by pretrial defendants across the jurisdictions in the relevant period.
The first reason why this number is very conservative is that it entirely excludes defendants whose cases were transferred to Circuit Court. Even though there were 124,606 Circuit Court cases that were resolved during this time period across the four jurisdictions. and 26,268 of these cases were nolle prossed in their entirety (21.08%), none of the cases were included in 1485 years calculation.
Given that Circuit Court cases are inherently more serious than District Court cases, it is highly likely that a large percentage of defendants were incarcerated pretrial before all their charges were dropped. However, it is impossible to precisely track which defendants in this group were/were not released pretrial because Circuit Courts in Maryland do not record commit (CMIT) or release (RELS) codes in Casesearch. Rather than inaccurately estimate, I omitted these cases all together.
The second reason why the total unjustified incarceration time is conservative is that it excludes all cases where defendants had any outcome other than nolle pross for any of their charges. So, for example, no time would be included if a defendant was incarcerated pretrial and then was acquitted of all charges at trial. Similarly, no time would be included if a defendant had all serious charges nolle prossed and then pled guilty to a single charge that carried no jail time. In short, various "unjustified incarceration" scenarios are possibly excluded. This was done to keep the calculation simple and unobjectionable.
Finally, the calculation also excludes any case where the defendant had a Failure to Appear (FTA) during the pendency of his or her case or was otherwise ever released and re-incarcerated pretrial. Although one can imagine arguments for including defendants who FTA'd if all their charges were ultimately dropped, excluding them both simplified the calculation and puts the final conclusion beyond cavil.
As noted above, the data herein comes from the CLUE database, which is in turn scraped from Maryland Casesearch. Two points about these sources bear mention.
First, entries into Casesearch are made by hand by clerks and other administrative staff in the courts. Occasionally, there are human-introduced errors. This kind of problem affects many real-world data sets and cannot be avoided.
Second, it is a consequence of Maryland's expungement laws that cases originally listed sometimes "disappear" from Casesearch because they have formally been expunged. However, these cases are not expunged from CLUE as they contain vital historical data. Indeed, cases that have been nolle prossed are eligible for expungement and more likely to not appear on Casesearch.
In the event that certain casenumbers used in this study are cross-referenced with Casesearch and an "invalid casenumber" error is thrown, the explanation lies with expungement.
The data collected for this project offer perspectives on Maryland's criminal justice system that go well beyond a single claim put forth in this Essay regarding unjustified incarceration. All of the datasets can be examined via this top-level portal maintained directly by the author.
The project actually offers five different data set endpoints for exploration: District Baseline, District All Nolle Pross, District All Nolle Pross - Detained, Circuit Baseline, and Circuit All Nolle Pross. In addition, there is a Compare Sets endpoint.
Via these endpoints, various attributes of the datasets can be examined broken up by any combination of the study’s years (2013-17) and jurisdictions (Baltimore City, Baltimore County, Montgomery County, Prince George’s Count). The available attributes include: Number of Cases, Race of Defendants, Sex of Defendants, Commissioner Hearing Outcome, Top Charge Disposition, Defendant Age, Length of Cases, Top 20 Charges, and Top 20 Zip codes.
Interested scholars, policymakers, and activists are encouraged to explore the various endpoints for themselves. Here is a small sample of queries that are possible:
The process described in Appendix A provides an overview of analysis undertaken by the author. After this article was accepted for publication, there was a further discussion between the author and the editorial team of the MIT Computational Law Report about how to best stage the content in order to provide a great user experience for reading the article, reviewing the data, and accessing the endpoints. This included developing strategies to port the article, originally written as a series of pages in HTML directly into the PubPub platform, as well as replicating the findings from the endpoints into a Google Sheets page to produce the native embeds for the document, and a link to a GitHub repository where this data is staged for the MIT Computational Law Report.
Colin Starger is a Professor of Law, University of Baltimore.
Author’s Note: I would like to thank Rachel Barkow, Shima Baradaran Baughman, Jeffrey Bellin, Leigh Goodmark, Corinna Lain, Zina Makar, Sandra Mayson, Tim Schnacke, Megan Stevenson and all participants in the 2020 Neighborhood Criminal Justice Roundtable (Lee Kovarsky, virtual host) for helpful feedback and comments on this project. All errors are, of course, my own. I would also like to thank Bryan Wilson for his hard work on editing and converting this essay from its original and unusual format.