Concept
Sentencing Changes in Practice
Ex Post Facto Laws
Sentencing Changes in Practice
Modern punishment is rarely fixed by the number in the offense statute. How long a person actually spends in custody depends on a sentencing range calculated under guidelines, on credits that shorten the term, on the timing and procedure of parole consideration, and on whether earlier convictions enlarge the penalty for a new crime. In each of these places a legislature or a sentencing body can change the consequences of an old offense without altering a word of the offense statute.
Article I, § 9 forbids Congress to pass an ex post facto law, and § 10 imposes the same bar on the states.1 Justice Chase’s third category supplies the test for sentencing changes: it reaches every law that changes the punishment and inflicts a greater punishment than the law annexed to the crime when it was committed.2 The test compares two scales, the one in force on the day of the offense and the one the prosecution now asks the court to apply.
Labels do not decide these cases. A legislature may call a rule administrative and may intend no harm, yet the label says nothing about whether the person will spend more time confined. Both functions of the Clause turn on what the person faces. Someone who weighed the cost of an offense by the consequences the law then announced was warned about those consequences, and a retroactive change that lengthens them withdraws the warning whatever the legislature’s reason. A body that already knows whom the change will reach also has the opportunity for targeted lawmaking that the Clause exists to check. The cases therefore look to how the new rule actually operates, and effect, not intent, drives the outcome.
One question ties the four settings together: does the change create a significant risk (the cases sometimes say substantial) of a longer or harsher punishment for the earlier crime? A change that makes a heavier penalty a realistic prospect for an offense already committed falls within the third category. A change whose bearing on time in custody is speculative, or that reaches only conduct occurring after it takes effect, does not. Each setting below is taught by contrast, a change that crossed the line set against one that did not, so that you can see which fact moved the answer. The guidelines come first, because there the comparison is between two numbered ranges.
Sentencing Guidelines
A guidelines range is the first number a judge sees, and the body that writes the manual revises it from time to time. A defendant whose offense was complete under one edition may therefore face sentencing under a later edition that sets a higher range for the same conduct. The federal Guidelines Manual anticipates the problem. It directs the sentencing court to use the manual in effect on the date of sentencing unless doing so would violate the Ex Post Facto Clause, in which case the court must use the manual in effect on the date the offense was committed.3
Miller v. Florida shows when the Clause is violated. When Miller committed his offense, Florida’s guidelines gave a presumptive sentence of 3½ to 4¼ years. By the time of sentencing, revisions had raised it to 5½ to 7 years, and the trial court imposed seven.4 The Court held the revised guidelines an ex post facto law as applied, because they substantially disadvantaged the defendant, and it gave no weight to the possibility that he might have received the same sentence under the old version.5 The trial court was not bound to give the presumptive sentence, but its discretion to depart was quite limited, so as a practical matter the change increased his sentence.6 Nor was the revision merely procedural. It increased the quantum of punishment, and the Florida guidelines had been legislatively enacted and carried the force and effect of law.7 The Clause looks to the standard of punishment the law prescribes, not to the sentence a judge happens to choose.
United States v. Turner carries the principle to the federal Guidelines and shows why the date of the offense matters so much. Turner was convicted of conspiracy to defraud the United States. The 2000 Guidelines set a base offense level of 10 for the crime, a level that later amendments raised to 14, so the 2006 Guidelines the district court applied produced a range of 33 to 41 months where the earlier manual would have produced 21 to 27.8 The court imposed 33 months.9 It reached the later manual by ruling that the conspiracy continued through January 2006, on the theory that Turner and a co-conspirator, Andrews, had agreed to conceal the scheme, an agreement that could be inferred only from Andrews’s encounter with federal agents in 2005.10 The D.C. Circuit disagreed. Questions about when a conspiracy ended usually arise in limitations and hearsay disputes, but the cases that decide them control at sentencing too, and under the Krulewitch-Lutwak-Grunewald pattern the 2005 encounter did not carry the conspiracy past 2001.11 The earlier manual therefore marked the date of the offense, and a later section of this chapter returns to how that date is fixed.
The harder question is whether Miller survives United States v. Booker, which made the federal Guidelines advisory.12 The Seventh Circuit, in United States v. Demaree, concluded that the Clause reaches only laws and regulations that bind rather than advise. After Booker, it reasoned, judges may sentence outside the range, and judges barred from using a new guideline could simply say they were taking into account the information behind it.13 The D.C. Circuit answered on practical grounds. It declined to assume that judges would misstate the basis for their sentences, and it observed that appellate courts may presume a within-Guidelines sentence reasonable, which gives judges a reason to stay within the range.14 Most federal sentences still fall within the range after Booker, because the applicable range gives judges a starting point, an “anchor,” that shapes the sentences they impose.15 Discretion likewise does not foreclose a claim. The controlling inquiry asks how discretion operates in practice and whether it actually creates a significant risk of prolonging the punishment, which calls for an as-applied analysis and not a facial one.16 The Supreme Court later confirmed the point in Peugh v. United States: applying amended guidelines that raise the range creates a sufficient risk of a higher sentence to violate the Clause even though the guidelines are advisory.17
The as-applied standard asks for less than certainty. The district court did not pull 33 months out of thin air, and Turner did not have to show definitively that he would have received a lesser sentence under the 2000 Guidelines. It was enough that using the 2006 Guidelines created a substantial risk that his sentence was more severe, and that risk violated the Clause.18 The reason for the lower showing follows from how ranges work. Because the range is the starting point and most sentences land inside it, a higher range raises the likely sentence even though the judge keeps discretion. A rule that made the defendant prove what one judge would have done under a different range would turn the protection into speculation about that judge’s mind, and the defendant would almost never be able to supply the proof.
The first step does the real work, because the date of the offense picks the baseline manual. Had the conspiracy continued into 2006, the manual in effect when it ended would have been the one used at sentencing, and the later manual would have raised nothing. A later manual that lowers the range creates no risk of a harsher sentence either, so the default instruction to use the manual in effect at sentencing governs.19
Good-Time Credits
A sentence announced in years seldom fixes the day of release. Statutes commonly let a prisoner earn good-time credits through conduct in custody, and each credit subtracts from the time the prisoner must serve before early release. A later law that rescinds credits that would have applied to a prisoner’s sentence therefore increases the time actually served, and that increase may violate the Clauses.21 The puzzle is why a change in what looks like a prison housekeeping benefit counts as greater punishment under Justice Chase’s third category.
Weaver v. Graham supplies the answer. A Florida statute enacted after the prisoner’s crime had slowed the rate at which good-time credits could accumulate, and it would have extended his confinement beyond what the law in effect at the time of the crime provided. The Court struck it down.22 The state’s natural reply is that credits are a gift. No provision of the Constitution requires a legislature to offer them, and a benefit that may be withheld altogether seems to be one that may be trimmed. The Court rejected the reply: even a statute that only alters penal provisions accorded by the grace of the legislature violates the Clause if it is both retrospective and more onerous than the law in effect on the date of the offense.
The answer rests on what a credit schedule does. A prisoner who knows the schedule can compute a release date, so the schedule in force when the crime was committed formed part of what the legislature announced the crime would cost. Cutting the credits afterward leaves the term in years untouched but lengthens the time behind bars. It substantially alters the consequences attached to a crime already completed, and that means it changes the quantum of punishment even if good time is technically not part of the sentence.23 The Clause looks to the standard of punishment the law prescribes, and the credit schedule is part of that standard because it decides how long the prescribed term lasts in practice.
Lynce v. Mathis shows the same principle at work on credits already earned. Lynce was convicted in 1986 of attempted murder and sentenced to 22 years. The statutes then in force gave prisoners credits for conduct and also provisional credits, awarded under a formula tied to prison overcrowding, which shortened the time of most prisoners, attempted murderers included, whenever the prisons were crowded.24 Lynce earned both kinds and was released. A 1992 statute, as the state attorney general read it, cancelled provisional credits for anyone convicted of murder or attempted murder, and the cancellation sent Lynce back to prison to serve more of his original sentence.25 The Court held that the cancellation violated Article I, § 10.26
The Court asked two objective questions: whether the law was truly retrospective, and whether it made the sentence longer.27 Both answers were plain. The 1992 change was aimed at people convicted of murder-related offenses who had accumulated overcrowding credits, so it operated on the very offense Lynce had committed, and it put him back in custody for time the earlier law had already forgiven. A court asks about the effects of a credit rescission and gives no weight to legislative intent, so a sincere description of the cancellation as prison administration would not have saved it.28 The case also illustrates the second function of the Clause, because a legislature that writes after the fact knows which prisoners its change will keep in custody.
The two decisions sit at different points on the same line. In Weaver the legislature slowed the rate at which future credits would build, so the prisoner lost time he would have earned under the old schedule. In Lynce the legislature undid credits already awarded and reached a person who had already been released. In both the effect on confinement is a matter of arithmetic, and neither case required speculation about whether the change would prolong punishment. That certainty is what separates these cases from the parole-timing disputes that follow.
The rule runs in one direction only, because the Clause is aimed at disadvantage. A legislature remains free to lessen punishment, and a law that makes the penalty lighter inflicts none of the harms the Clause exists to prevent.29
Whether an earlier offender actually receives a more generous schedule is a matter of legislative intent and any saving clause. The constitutional question arises only when the change makes the schedule harsher: for an offense already committed, does the change lengthen the time the prisoner will actually serve?
Parole Timing
Parole is harder than credits because the law usually promises consideration for release and not release itself. A rule that changes when or how the board considers a prisoner leaves open whether anyone will actually serve longer. For such changes the question is whether the new system creates a significant risk of prolonging incarceration.31
That standard sits between two positions that would each defeat the purposes of the Clause. If any conceivable chance of a longer stay were enough, every adjustment to a hearing calendar would be barred, even adjustments that change no prisoner’s release date. If only a certain increase counted, a legislature could lengthen confinement by rewriting procedure, because the effect of a procedural change on any one prisoner is almost always uncertain. The standard demands less than certainty and more than speculation, which matches the guidelines cases, where a substantial risk of a more severe sentence sufficed and no one had to show that a judge would certainly have chosen a lower term.32
California Department of Corrections v. Morales drew the line. The statute in force when the prisoner committed his murder had provided for annual parole-suitability hearings. A later statute let the board defer the next hearing for up to three years for a prisoner with two or more homicide convictions, but only if the board found it “not reasonable to expect” that parole would be granted in the intervening years and stated its reasons.33 The prisoner argued that the statute eliminated any chance of release during the deferral and so increased the punishment for his crime. The Court held that the risk of enhanced punishment was too remote and speculative to violate the Clause.34
The Court emphasized two features of the statute, and each does separate work. The first was the class of prisoners it reached. A prisoner with multiple homicide convictions was extremely unlikely to be released at any hearing within the next three years, so skipping those hearings would rarely cost anyone an actual release. The second was procedure. Deferral required a finding and a statement of the bases for it, so the board could not use the power as a blanket delay, and the requirement confined deferrals to cases where release was not reasonably expected.35 Together the two features made it unlikely that the change would lengthen anyone’s time in custody, which is why the risk counted as remote.
Garner v. Jones reached the same result by a similar route. A parole board lengthened the interval between hearings for prisoners serving life sentences from three years to eight, and its rules provided for expedited reconsideration when there had been a “change in circumstances.”36 The Court concluded that the rules, on their face, did not create a significant risk that life prisoners would spend more time in prison than under the earlier rules. Expedited reconsideration plays the role that the findings requirement played in Morales: a prisoner whose prospects improve has a path to an earlier hearing, so a longer default interval does not lock in a longer stay. The Court also left open the possibility that the new rules, as applied to a specific individual, might enhance that person’s punishment in violation of the Clause.37
That open door matters because the first instinct on a parole problem is to say that the board has discretion, so no one can claim harm. The D.C. Circuit has rejected that move, holding that the existence of discretion does not foreclose an ex post facto claim. The controlling inquiry is how the authority exercises its discretion in practice and whether the exercise creates a significant risk of prolonging incarceration, which calls for an as-applied analysis and rules out a purely facial one.38 A facial challenge to a hearing-interval rule is therefore likely to fail when safeguards keep the risk remote across the class, while an as-applied claim asks whether those safeguards actually protect this particular prisoner.
Which parole changes fall on the other side of the line? A statute that reduces the minimum sentence but increases the time that must be served before parole may be granted is better viewed as increased punishment.39 The reason is a difference of kind. The changes in Morales and Garner altered how often a board looked at a prisoner whose release stayed possible at some later look, while a higher eligibility floor forecloses the possibility of release for a stretch in which the old law allowed release to be considered. A prisoner who could have been paroled after serving a third of the term and must now serve half has lost that possibility outright, and no forecast of the board’s judgment is needed to see it. Some older cases reached the opposite result by treating release on parole as an act of grace and not a right, but the same argument fails for good-time credits, since a statute altering provisions accorded by the grace of the legislature still violates the Clause if it is both retrospective and harsher than the law in effect at the time of the offense.40
Not every adjustment to parole eligibility adds punishment. In Holguin v. Raines, a correction in the method of calculating parole eligibility, made to conform to the state’s highest court’s reading of the state statute, did not violate the Clause.41 A correction that conforms the computation to what the statute already required leaves the price of the offense where it was, while a change that raises the eligibility floor increases it.
The third and fourth steps carry the weight in a hearing-frequency problem, because the class and the safeguards are the facts that separated the upheld rules in Morales and Garner from a rule that simply defers release for everyone.
Recidivist Enhancements
A habitual-offender statute looks at first like a retroactive punishment, because it lengthens a repeat offender’s sentence on account of convictions that may long predate the statute. The question is which offense the added punishment attaches to. A statute that enhances the penalty for a later offense by reason of earlier convictions does not punish the earlier offenses, so the Clause is not triggered even when those convictions came before the statute was enacted.43
Three dates settle the question. A defendant commits crime A when no habitual-offender statute exists. The legislature then enacts a statute increasing the punishment for a second offense, and afterward the defendant commits crime B. Applying the statute to crime B falls outside the prohibition, because the statute prescribes no additional punishment for crime A; it prescribes it only for crime B, which was committed after the statute was passed.44 The conviction for crime A works as a fact about the defendant that the statute counts at the moment crime B occurs, and that is why the age of the earlier conviction does not decide the question.
Both functions of the Clause explain the result. A person deciding whether to commit crime B after the statute exists can read it, learn that earlier convictions will lengthen the sentence, and decline to offend. Nothing was withdrawn from anyone: the law in force when crime A was committed said nothing about crime B, and the law in force when crime B was committed announced the penalty in advance. The earlier conviction is a fixed fact the defendant cannot change, yet whether the enhancement applies still depends on the one thing the defendant controls, which is whether to commit the new offense. The check on targeted legislation is met for a similar reason. The legislature may know who already has convictions, but it cannot know which of those people will offend again, and the added penalty reaches only those who do.
State v. Hall applies the reasoning to a drunk-driving enhancement. New Mexico had amended its DWI statute so that a person convicted of a fourth offense is guilty of a fourth-degree felony, punishable by a jail term of not less than six months that may not be suspended or deferred.45 Hall’s sentence was enhanced because of predicate convictions that came before the amendment, and he argued that the enhancement was an ex post facto application. The court disagreed: a sentence statutorily enhanced because of a predicate conviction does not violate the prohibition even though the conviction occurred before the penalty provisions were enhanced.46 Such statutes do not punish a person again for the earlier offenses; they increase the penalty for repeating criminal conduct. Because Hall committed his offense after the amendment took effect, he had fair warning that earlier convictions could enhance his sentence. The court took from Dobbert v. Florida the idea that a statute’s existence on the books warns a defendant of the penalty to expect.47 Dobbert itself upheld applying a new death-penalty statute, with proper procedures, to a crime committed under an earlier death-penalty statute that had been declared unconstitutional.48 Hall also invoked Bouie v. City of Columbia, a due process decision about an unforeseeable construction of a criminal statute. The court found no changed construction, because Hall’s conduct was criminal under the earlier DWI statute and remained criminal under the new one.49
When a problem stresses that the prior convictions came years before the statute, the better question is when the offense being punished was committed.
Fixing that date can itself be contested when an offense continues across the effective date, a problem a later section takes up.
The same reasoning reaches beyond longer sentences. A legislature may define a crime as limited to conduct by persons who have already been convicted of some other offense, and may apply the statute to a person whose earlier offense and conviction predated it.50 In Salazar v. State, a statute making possession of a pistol by a person convicted of robbery a crime was held not ex post facto as to possession after the statute, even though the robbery conviction preceded it.51 The punished act is the possession, which occurred after the warning, and the robbery conviction is a status the statute counts. In In re Miller, a Michigan court reached a parallel result on credits: a statute denying a convict serving a sentence for a second offense the reductions allowed to other convicts was valid as applied to an offense committed after the statute, although the first conviction preceded it.52 That holding fits with the credit cases above, which struck down cuts in credits for offenses already committed, because the date of the offense for which the penalty is imposed separates them. A Kansas appellate court likewise held in State v. Beard that classifying a defendant’s three prior misdemeanors as a felony for criminal history purposes under the sentencing guidelines did not violate the prohibition, which shows that a criminal history score is another way prior convictions bear on the penalty for a new offense.53