Impressions from the oral argument in Kansas v. Glover

On reasonable suspicion and how to calculate it.


The Supreme Court held argument Monday in Kansas v. Glover, the Fourth Amendment case I blogged about last week.  The transcript is up here, and I thought I would blog some reactions.

1. Does the officer need to testify?  Several Justices were particularly interested in whether the officer needs to take the stand and testify that, in his training and experience, a positive hit on an owner-license-suspension query correlates sufficiently with the premise that the owner with a suspended license is driving.  A few Justices suggested that this could be important to provide actual evidence in support of the inference that this was likely.

I gather this question was a focus in part because the parties in Glover stipulated the facts. It was agreed that the officer assumed that the positive hit meant the owner was driving, and as a result he did not  testify about his "training and experience" with that.

There was also a recognition that, as Justice Gorsuch noted, requiring an officer to take the stand might be pretty formalistic. After all, the officer will take the stand and say that he was trained to draw that distinction, or that his impression is that the inference is justified.  If that's all that is required, then officers will be trained accordingly and cross examination about that training and experience may not add much.

This is a really interesting issue.  The general question, I take, is this: In assessing cause, what is the zone of judicial assessment and what is the zone of officer experience?  When are conclusions properly a call for judges to make, and when are conclusions something for an individual witness?  Off the top of my head, I don't think there's a well-developed theory for what the line is between those two.

For what it's worth, my own intuitions tend to make the general question something for judges but the experience with special cases something up to witnesses.  If that's right, I would think that the officer's testimony shouldn't be required in a case like Glover, but that there will be Glover cases where it would be relevant.

Here's my thinking.  If we want to get a ballpark sense of whether an intuition is justified, the ballpark sense is probably something best generalized by courts.  I would think it would be a relatively uncommon case in which the officer's training and experience can shed substantial light on the generalized answer.   Relying on the officer's training seems particularly unhelpful because it's presumably circular.  In the future, the officer's training should depend on how the Court rules in Glover.  So that doesn't seem all that useful.

I would think that an officer's testimony about his experience will be pretty limited in most cases.  As the Chief Justice noted, some officers will be new.  Others will have no strong recollections of the correlation between a positive database hit and the owner driving.  And none will be well-equipped to offer a statistical analysis of the inference anyway, such as whether the factors that went into their past experience were relevant to this case.  Whether the inference is justified strikes me as something often or typically beyond the officer's own personal experience.  Given that, I'm skeptical that such testimony should be required.

I see this as a lot like Harris, the dog sniff case.  The officer didn't need to testify that, in his experience, the alert of a well-trained dog was ordinarily sufficient to create probable cause.  That was the Court's assessment, not the officer's.  Here was the key language from Harris:

If a bona fide organization has certified a dog after testing his reliability in a controlled setting, a court can presume (subject to any conflicting evidence offered) that the dog's alert provides probable cause to search. The same is true, even in the absence of formal certification, if the dog has recently and successfully completed a training program that evaluated his proficiency in locating drugs.

The ballpark intuition came from the Court.

With that said, I would also think that the officer's experience is certainly relevant in some Glover cases.  Officer experience can be important when an officer has a lot of experience with a recurring fact pattern.  That experience can impart significant knowledge that may not be obvious to outsiders.  For example, an officer who does undercover buys often may know how drug deals work, and that may give the officer knowledge of what lessons to draw from something he sees during a drug buy that someone without that experience would lack. The idea is that the officer can tell us about a world we don't know, serving as a sort of expert to guide us.

If I'm right about that, I would think there will be circumstances in which testimony about an officer's experience is relevant.  Imagine an officer has a lot of localized experience with a particular kind of stop in a particular place and time.  If the officer has a lot of experience with past stops similar to that one, the officer's experience could give useful localized information about whether the stop in those same circumstances justifies the usual inference or instead a departure from it.

Again, it seems a lot like Harris.  The Court gives a ballpark assessment, and then you can get testimony about whether a particular situation fits that ballpark assessment.

2.  Quantifying probable cause and reasonable suspicion.  Chief Justice Roberts pushed a line of questioning in which he discussed ballpark numbers for probable cause and reasonable suspicion.  He quantified probable cause, which he said had been held to be less than fifty percent:

CHIEF JUSTICE ROBERTS: Okay. Well, we know that probable cause is not 50 percent.


CHIEF JUSTICE ROBERTS: It's somewhat less than 50 percent.


He then asked about a quantification of reasonable suspicion:

CHIEF JUSTICE ROBERTS: Well, what—what reasonable suspicion cutoff do you think? Do you think it's one out of five?

MS. HARRINGTON: I can't say because this Court has said repeatedly that —


MS. HARRINGTON:—none of us can say, right?

As President of the  Why Probable Cause Shouldn't Be Quantified Club, I thought I would push back on this.  First, contrary to the Chief's statement, the Court has not held that probable cause is less than 50%.  Granted, there are cases saying that the government doesn't need to show more than a 50% likelihood to establish probable cause.  But that's because probable cause is not quantifiable, not because it is quantifiable and the target number is below 50%.

The Court has put he point this way: "The test for probable cause is not reducible to precise definition or quantification. Finely tuned standards such as proof beyond a reasonable doubt or by a preponderance of the evidence  have no place in the probable cause decision." Florida v. Harris (2013), (cleaned up)   I think the same goes for reasonable suspicion. Although some Justices have hinted at wanting to quantify it, such as in Justice Scalia's dissent in Navarette, I don't think it is a quantifiable idea.

3.  Can cause be based on purely statistical information?  At the bottom of page 30, Justice Kagan asked whether reasonable suspicion could be based on no more than a statistical study:

Suppose that a municipality has a law that says everybody has to carry their driver's license with them at all times. And suppose that a particular police department actually did a kind of survey or, you know, a study of their practices and found that actually 50 percent of teenagers do not carry their driver's license with them at all times. All right? So now it's like common sense that if you see a teenager, she won't be carrying her driver's license with her. Does that give the police officer the ability to stop every teenager that he sees?

The scholarship on cause thresholds has pondered similar hypotheticals.  For example, in my quantifying probable cause article, I asked whether the government could get a warrant to search a Harvard dorm for marijuana based solely on a reliable study showing that 60% of Harvard dorms have marijuana.  It's the same basic question, I gather: When can the government rely on statistics alone?

In the context of a warrant, I argued, the problem with relying on statistical evidence and no more is that the actual probability of finding evidence was likely no where near like the percent found in the study.  The reason is the missing information.  Someone had selected that particular dorm room to search, and investigators apparently had found no other evidence justifying the search beyond the study to include in the warrant affidavit.  To know the actual likelihood drugs would be found, I argued, we really need to know the backstory of why that room was selected and why no other evidence had been offered.  When you filled in that missing information, the actual probability quickly became wildly different from 60%.

A question raised by Justice Kagan's hypothetical is whether the same is true for a reasonable suspicion determination in the field in a Glover-like case.   If there are statistics showing how often a license-suspension hit leads to a driver behind the wheel, can a court rely on them in a particular case?  I'm not sure.  Off the top of my head, I think the basic dynamic is at least partly valid.  There could be lots of reasons to think that the particular case where a stop is made is not representative of the whole.

At the same time, I can also think of reasons why the dynamic might be less pronounced in that setting.  I gather that the decision to run a license check is normally not as involved as that for obtaining a warrant, and therefore involves less question of why a particular search or seizure was made and what investigation was tried.  Not sure where I come out, but it's an interesting question.

4. A note on the criminal/civil divide.  One additional small thought I had is that I gather all parties are assuming that driving without a license is a crime, not just a civil offense.  Under Terry and its progeny, evidence of an ongoing crime permits the stop.  So the court is treating Glover as a Terry case, and asking if a Terry stop is justified. That makes sense, as driving on a revoked or suspended license is a misdemeanor offense in Kansas.

I flag that because driving on a suspended license is a civil offense instead of a criminal offense in at least some jurisdictions, and the standard for a stop in the case of a civil infraction is less settled.  Whren held that probable cause to believe a civil traffic violation was committed justifies a stop. My recollection is that some lower courts have held that mere reasonable suspicion of a civil traffic violation also permits a stop, and that lower courts are divided on the question.

Given the apparent disagreement, it seems important to note that Glover is only about when the facts of a case satisfy the reasonable suspicion standard of Terry.  It is not about the distinct question of whether a stop can always be made when reasonable suspicion exists that a person is driving on a suspended license.


NEXT: ByteDance, bitten. By CFIUS.

Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Report abuses.

  1. Seems like the statistical argument for search thresholds must be off base. The 4A has an explicit requirement for particularity. Whatever you can say about statistical methods, they work on the basis of generalization, not particularity.

    1. The particular basis in this case was “The car being driven has an owner whose license is suspended”

      1. Exactly. The stop was based on information about that particular car.

        Probable cause is a combination of statistics and specific information. You have to have a relevant reason to think a specific vehicle is more likely than the average vehicle to be involved in the crime. Whatever the threshold for “probable” is, it has to be higher than the background probability, and met due to information about the particular instance you’re looking at increasing that probability.

        1. Except that there’s nothing particular about the owner having a suspended license that makes the car more likely to be driven by the owner. If anything, that particular information makes it less likely that the owner drives.

          1. Except that there’s nothing particular about the owner having a suspended license that makes the car more likely to be driven by the owner. If anything, that particular information makes it less likely that the owner drives.

            You’re misunderstanding the argument the state is using. It’s not whether that car is more likely to be driven by the owner than another car; it’s whether that car is more likely to be driven by someone with a suspended license than another car.

            If you pick a random car off the street, that driver is not likely to be suspended.
            If you pick a car owned by someone with a suspended license, that driver is a lot more likely to be suspended.

            The issue is whether that latter, true statement is sufficient to justify a stop.

        2. “You have to have a relevant reason to think a specific vehicle is more likely than the average vehicle to be involved in the crime. Whatever the threshold for “probable” is, it has to be higher than the background probability, and met due to information about the particular instance you’re looking at increasing that probability.”

          But in your argument, what is the background probability and what is the particularized probability is arbitrary.

          To use the Harvard dorm room analogy, let’s say I can show that a a Harvard dorm room is more likely than the average residence to contain MJ. So by your argument, I get to search the Harvard dorm room precisely because that particular residence is a Harvard dorm room.

          Your background probability argument would also apply to a vehicle in a high-crime area. Let’s say I can show that a vehicle in a high-crime area is more likely to contain drugs than a vehicle that is not. By your argument, the cops could search a particular vehicle because it is located in a high-crime area. But we know that that is not the case.

          1. Wasn’t there also a ruling that discrimination cases couldn’t rely on statistical evidence but rather had to show direct evidence in that particular case? That may be for a lawsuit, I don’t recall.

            It seems to me that is the important difference. The whole point is to stop government from hassling people at will. 40 or 50% seems to leave “at will” wayyyyyy too open.

    2. The 4A has an explicit requirement for particularity.

      It does, but only in terms of describing what a valid warrant must contain. All it says about the kind of warrantless search at issue here is that it must not be “unreasonable”.

  2. “The general question is this: In assessing cause, what is the zone of judicial assessment and what is the zone of officer experience? When are conclusions properly a call for judges to make, and when are conclusions something for an individual witness?”

    Given this framing of the question, I thought the trial judge’s initial reasoning (as quoted by the Court of Appeals) might be of some interest. The Court of Appeals writes:

    “The district court ultimately granted Glover’s motion to suppress, holding that the officer did not have reasonable suspicion to initiate the stop. The [trial] court reasoned: “I mean, just as a personal observation, I have three registered in my name. My husband drives on every day; my daughter’s in [Washington, D.C.] with one every day, and I drive the other. And I think that’s true for a lot of families that if there are multiple family members and multiple vehicles, that somebody other than the registered owner often is driving the vehicle.””

    The Court of Appeals ultimately upheld the stop and I think it says something that the trial judge’s remark is quoted as the “reason” for trial court’s decision to suppress. As I read the COA opinion (which was reversed by the Kansas Supreme Court), the trial judge drew from her personal experience to refute the argument that it is reasonable to infer that the registered owner of a vehicle is the person driving the vehicle.

  3. So, the “percentage argument” is a real one. Now the courts don’t want to closely define what % reasonable percentage is, because then you get into a % argument every time a cop stops someone on reasonable suspicion. I.E. “Well, it was only a 19% chance I was going to rob the store, not 20%, so you didn’t have reasonable suspicion to stop me”

    And that’s not an argument the courts want coming up repeatedly. They’d prefer ballpark numbers that are more flexible.

    2. Now in Kagan’s “60% marijuana chance” Harvard dorms example, there are a few things.
    a. You likely need probable cause (IE, your 60%).
    b. You also likely need a search warrant
    c. It should ideally be enforcing the law evenly in such a situation. IE, not just searching one dorm, but ALL dorms. (Otherwise you’re just selectively enforcing the law, which is problematic).

    And in situations like this, where there is a desire to remove marijuana from dorm rooms, a clear search of ALL dorms rooms has been done, and applied. Usually by the University, not the police, but it’s reasonable.

    1. Are we talking about dorms, or dorm rooms?

      In any case, I’m not sure the analogy is a good one. Searching a dorm room for marijuana – legal in MA, by the way, so pick a different school – is way more intrusive than stopping a car, all the more so if the room is shared.

      1. It’s not a great analogy (apologies, it appears it was Kerr’s analogy, not Kagan’s), but I’m responding to what was provided in good faith, without quibbling over the details.

        Kagan’s analogy about mandating teenagers have driver’s licenses on them at all times isn’t good either (If for no other reason that many teenagers “can’t” get driver’s licenses due to their age).

        Perhaps a better example would be a study that comes out saying that “60% of the excess opioids in a community come from a local pharmacy selling off the books”. And then police (who are aware of the opioid problem in the community) use that study to get a search warrant on the local pharmacies and search their books.

    2. I would add,

      d. You need a reason why you’re searching that particular room.

      The general probability applicable to all rooms in the dorm doesn’t get you there. You need a reason to think that particular room contains evidence of a crime.

      If you use a general probability to justify a specific search, all you’ve done is recreated the general warrant that the 4th amendment was supposed to prohibit.

      1. d) I would say, if you’re not searching all rooms, or a high percentage of rooms, you need a reason why you’re search that particular room.

        If you’re searching all rooms, then you don’t need a particular reason for “that” room.

      2. Brett, I look at this as more of a general search than a specific, criminal search.

        Sort of like a police road block for drunk stops.

        Just like people authorize the drunk stops (when they sign for their drivers license), I imagine dorm dwellers agree to safety/welfare checks on their dorms rooms when they sign for them.

        1. There’s no such thing as implied consent under the Fourth Amendment

          Checkpoints are justified by a special need, not consent.

          1. Well, there does seem to be one area in which implied consent is being recognized: statutes that mandate consent post-arrest blood alcohol testing as a condition of obtaining a driver’s license. See, e.g. Cal. Pen. Code section 23612. It seems to me such statutes are on shaky constitutional ground, but they’re still out there, and they are enforced.

            1. Nope. Those statutes SAY implied consent, because states want to mislead their citizenry into not challenging them.

              But they have never been upheld as a consent search. They have been upheld based on reasonableness.

              1. I meant Vehicle Code section 23612, not Penal Code.

                I agree with you (I think we agree, anyway) that the so-called implied consent laws are crap. You’re correct that in the cases considering these statutes, the courts are not finding implied consent, but are affirming convictions based upon other exceptions to the warrant requirement, such express consent (the defendant actually agreed to the test), or (in the case of breath tests at least) search incident to arrest. I think there have also been some cases that found exigent circumstances. I have a case on appeal right now where I am arguing that defendant’s consent was not knowing and voluntary. It’s an uphill battle.

                1. Yes, the laws are crap.

                  Under current doctrines, it is of course very difficult to get a DUI test overturned. There’s no doubt of that. But there’s definitely no effective 4th Amendment consent to a search that is implied by conduct that happens months in advance.

      3. Yeah.

        And there’s a historical argument against house to house searches. A search based on nothing but a statistical inference without any evidence of actual criminal activity going on in the place may be per se unreasonable given it is precisely what the 4th Amendment was supposed to prohibit.

    3. While a “search of ALL dorms rooms has been done”, it was not reasonable – at least not in any sense that connects the analogy to police. The University could justify the search of ALL dorm rooms because they are acting in loco parentis. The military has a similar right to search all barracks from their “health and safety” obligations.

      Kagan’s analogy made sense as a test of whether and when percentages should matter to a police reasonableness test. Extending that to a parent’s right to search for health and safety breaks the analogy.

      1. Most college students are adults. In loco parentis has no application.

        1. You’d think so. Courts have so far mostly disagreed. It (in combination with the landlord doctrine) is the justification that allows Universities to search dorm rooms without a warrant where police can not.

          Footnote: In loco parentis was a strongly held doctrine until the 1960s. A series of decisions weakened the doctrine as it applied to high schools and colleges but then in the 1990s the doctrine began making a resurgence.

          1. The landlord doctrine may allow some level of searches. But any court who says “in loco parentis” is making a category error. No parent has any general right to search the belongings of an 18 year old or over offspring without the offspring’s consent, outside of the context of the offspring living at the parent’s residence.

            1. Again, that makes perfect sense to me and I agree that’s what the rule should be. Courts have not agreed with us. In loco parentis is still good law in at least some cases involving college students.

      2. Sounds like the Latin fallacy: believing if it’s in Latin it must be true.

        In loco parentis ended a long time ago at colleges and universities.

  4. Worrisome.

    In the context of driving – the use of data that comes up on a driver when running a tag check as a routine matter or say pass time when the officer has nothing else going on, for purposes of substantiating a Terry stop, where there is no actual driving violation, approximates a check point. Call it a scan check point if you will.

    Saturday. 2 a.m. No moving violation identified. Run tag. Officer screen displays, “owner, 5 previous dui”. Stop initiated.

    Saturday. 2 a.m. No moving violation. Run tag, screen shows “owner, probation, driving restricted to daytime/job”. Stop initiated.

    Personally, I’d like to see the unauthorized checkpoint angle explored. Soon a negative social rating or credit rating which equates highly with no license on person or some other offense, will be enough to initiate a stop.

  5. ” I think the same goes for reasonable suspicion. Although some Justices have hinted at wanting to quantify it, such as in Justice Scalia’s dissent in Navarette, I don’t think it is a quantifiable idea”

    If you simply give up and declare it’s not quantifiable, aren’t you effectively reading “reasonable” out of “reasonable suspicion”?

    If it’s not quantifiable, how/on what basis do you draw the line between reasonable and unreasonable suspicion?

  6. Until we attract a better class of people to law enforcement (with more accountability, better training, better equipment, and better management), and eliminate or severely curb qualified immunity, I am inclined to hope the government loses just about every search-and-seizure decision.

    I blame my libertarianism.

    (I always welcome Prof. Kerr’s observations, though.)

    1. If you want someone to blame, blame the man in the mirror.

    2. I don’t think it’s the class of people attracted to law enforcement per se (though there are of course people who merely want to bullies there), but – that there are systematic incentives that encourage impropriety. Remove the incentives – like unions, immunity, unequal application of the law, etc – and we won’t see the abuses. Both because abuser will be in prison, and because there won’t be a continual pull towards worse and worse behavior. Though again, the people who sought out to join the SS for power will still seek it out here.

      We could even enable some of these with simple legislation: any state could add a private cause of action for violation of state laws, and even a private criminal prosecution along the lines of a qui tam case. The problem with the rest is that States generally aren’t accepting liability for their own unlawful acts (which is total crap, but that ship sailed centuries ago), so the only option for violations in most cases is under Federal law, and the Judiciary decided to grant impunity regardless of what the statute actually says.

  7. Degree of cause / suspicion can sometimes be quantified. Unknown thief escapes into one of two units on floor of am apartment building. Each unit would have 50% (this was ky v king, roughly). Or unknown thief escapes into one of 3 units. Or 4 units.

    This quantification thing will become more important as police look to “artificial intelligence” to find and support RS and PC.

    1. But that’s precisely the issue. If there’s a 50/50 chance between two apartments with nothing pointing specifically to one, I’m not sure courts would let the police blindly guess even between two choices. But, if let’s say one door slammed closed or they hear movement behind the door, that would let them point to one over the other even if it’s a complete coincidence.

  8. How does “probable” cause not imply statistics and the degree of probability?

    If a police agency tracked stops of cars owned by drivers with a suspended license and found in only 1 case out a 1000 the driver was the owner how is that not relevant to whether there is probable cause? Since much of this information could be tracked by police using terminals in police cars, why should this source of information not be tracked? It should be possible fairly soon (if it not already) to track this same information by voice recognition. Suppose it was found that most license checks were for cars owned by blacks?

    1. It’s almost like there’s a reason police stations resist any and all accountability metrics.

  9. There could be lots of reasons to think that the particular case where a stop is made is not representative of the whole.

    And are these reasons un-enumerable?

    For the dorm room/marijuana example, you refute that general probability is sufficient, and that particular supporting evidence must be provided. Why is that same bar not applicable to cars, requiring something beyond “some cars registered to a driver with an expired license are driven by that person”?

    If probability on general criteria is not sufficient, then it is not sufficient even for cars.

    1. But extending this case, if the cop got a picture of the owners drivers license on his screen in the patrol car, and then pulled up next to the driver and thought it was the same guy, that should itself be sufficient – the cop had a reasonable identification that the driver was the owner based on a poor sighting, and that person was prohibited from driving.

      Even if he turned out to be wrong.

      But if he sees an apparent blonde female driving, and the prohibited owner is a brunette male then a detention is unlawful, even if the actual driver was the prohibited owner wearing a wig.

      It should be that the process is a reliable one, executed competently, even if it’s not always correct. But if an unreliable process is done it’s always thrown out.

  10. Professor Kerr,

    I think the founders were trying to get to the philosophy of science rather than any specific quantum itself, but that wasn’t something that had been worked out in philosophy at the time.

    Science at its foundation is about predicting the future, with the scientific method a process to reliably predict the future.

    So when the constitution says “probable cause to believe….” they were really getting at “we’ve followed a method that will reliably get what we’re looking for because (reasons)” where those reasons are evaluated by a judge to determine if they do indeed reliably get the right answer.

    This is why science isn’t itself concerned with the right answer, and why nothing is ever proven – a scientific theory merely has not yet been disproven. That’s also why Newtonian physics and Keplerian mechanics are still taught, even though we know with certainty that they’re wrong – because they’re useful approximations in almost all circumstances (you need to operate across a strong gravity gradient for them to be materially wrong).

    Putting this to the case at hand, “training” should have literally no value in determining whether probable cause exists – after all the police could be taught that (insert stereotype you find morally objectionable) is a strong indicator of (bad behavior).

    Likewise “experience” should have minimal value in detecting probable cause unless the individual officer making the claim has tracked every instance of the event they claim, quantified the type 1 and type 2 errors, and the result is consistent with their claim. Otherwise we get (and I hope you remember one of the case name on this, because I don’t) officers claiming that their drug detection dogs have a 100% accuracy rate – every time they don’t alert the police don’t find drugs (because they don’t search), and every time they do alert the police either find drugs (true positive) or claim the dog detected “residual scent” (and entirely meaningless claim as it has no truth value at all).

    Under that rule at least a no-drug-dog-alert is dispositive proof that no drugs are present, but an alert is only potentially useful in determining probable cause if the officer tracked how often an alert occurred and no drugs were found – then a judge could determine if that rate was sufficiently high given the rest of the evidence, even if not explicitly quantified.

  11. When are conclusions properly a call for judges to make, and when are conclusions something for an individual witness?

    I would think conclusions of fact are for juries to make.

    1. Not for motions to suppress, which have always been considered a pretrial matter for a Judge to make a determination. Plenty of fact determinations are made by Judges (for example, cases of equity, certain foundational rules for evidentiary matters, etc.).

  12. It sounds less like you object to a probabilistic analysis and more feel that, properly done, that analysis will give a very different result than what we might get by looking at any average of hit rate. I mean the way you suggest the decision should be made sounds like how I would describe using what philosophers call epistemic probability.

Please to post comments