The Comparison That Misleads

In political argument, the Violent Crime Control and Law Enforcement Act of 1994 and the First Step Act of 2018 are treated as opposite answers to a single question. The earlier statute stands in for the punitive turn in American criminal justice policy, and the later statute stands in for its reversal. Commentators, advocates, and officeholders reach for the pairing to narrate an arc: the country built mass incarceration with one law and began dismantling it with the other. That telling has moral force but little structural content. The two statutes did not address the same population through the same instruments, and the comparison that matters is not between an era of severity and an era of mercy. It is between two different theories of what federal legislation can do about crime and punishment in a country where criminal law is enforced almost entirely by the states.

Each statute arrived at the end of a long negotiation about federal power. The 1994 act was the product of years of bargaining over how Washington could respond to violent crime that was, by every measure, rising through the late 1980s and early 1990s. Its answer was money: a spending program on a scale no federal crime bill had attempted, aimed at state and local institutions that the federal government could fund but not command. The 2018 act arrived after years of debate about the federal prison population and the cost of incarcerating it. Its answer was administrative: changes to federal sentencing rules and to the operation of federal prisons, applied directly to the system Congress controls outright. Understood this way, the pairing stops being a morality play and starts being a map of the federal government’s two levers: the spending power, which can influence the states, and direct legislation, which reaches only the federal system.

The Earlier Statute

The Violent Crime Control and Law Enforcement Act of 1994 became Public Law 103-322 on September 13, 1994, when President Bill Clinton signed it. The House agreed to the conference report by 235 to 195 on August 21, and the Senate agreed by 61 to 38 on August 25, after the Senate had passed its version 95 to 4 the previous November. The statute authorized 30.2 billion dollars in funding authority over six fiscal years, 1995 through 2000, spread across more than sixty law enforcement, prison construction, and crime prevention programs. No federal crime bill had ever authorized sums on that order.

The core contents were a mix of grants and new federal criminal provisions. The largest single program was the Community Oriented Policing Services hiring program, which provided 8.8 billion dollars to put 100,000 additional police officers on patrol in state and local departments. The Violent Offender Incarceration and Truth-in-Sentencing incentive grants offered states 9.7 billion dollars for prison construction and expansion, conditioned on states requiring persons convicted of designated violent offenses to serve at least 85 percent of the imposed sentence. Title VI, the Federal Death Penalty Act, created approximately sixty new death penalty offenses under forty-one federal capital statutes, covering terrorism, murder of federal law enforcement officers, large-scale drug trafficking, drive-by shootings resulting in death, and carjackings resulting in death. A federal three-strikes provision mandated life imprisonment for a third serious violent felony conviction under federal law, or for a violent felony conviction following a serious violent felony and a serious drug conviction. Title XI, Subtitle A, known as the Federal Assault Weapons Ban, barred the manufacture of nineteen specified semi-automatic firearms and newly manufactured magazines holding more than ten rounds, with a ten-year sunset provision that took the ban off the books on September 13, 2004. Title IV created the Violence Against Women Act with 1.6 billion dollars for law enforcement, prosecution, and victim services. Title XVII, the Jacob Wetterling Crimes Against Children and Sexually Violent Offender Registration Act, set federal guidelines for state sex offender registries. A further 1 billion dollars went to drug courts and substance abuse treatment for nonviolent offenders.

The bill began as H.R. 3355, introduced in the House in October 1993 by Representative Jack Brooks of Texas, and the Senate version was drafted by Senator Joe Biden of Delaware. A conference committee reconciled the two chambers’ versions in August 1994. Beyond policing and prisons, the act set aside 6.9 billion dollars for crime prevention: after-school and weekend safe-haven programs for youth, in-school anti-crime programs, anti-gang initiatives that funded midnight sports leagues and boys and girls clubs, and 1 billion dollars for drug courts and substance abuse treatment for nonviolent offenders. That mix, grants for punishment on one side and grants for prevention on the other, made the statute’s theory of action unmistakable: Washington would pay for the local policies it wanted, because it could not order them.

The mechanism mattered more than any single provision. The great majority of the money flowed outward, to state and local police departments, state prison systems, and community programs. Congress in 1994 understood that it could not rewrite state sentencing codes by fiat; what it could do was attach conditions to money and wait for states to accept the bargain. The statute was, at bottom, a spending instrument with criminal provisions attached, and its reach into American punishment came through the states that took its grants.

The Later Statute

The First Step Act of 2018 became Public Law 115-391 on December 21, 2018, when President Donald Trump signed it. The Senate had passed it 87 to 12, and the House 358 to 36. The Congressional Research Service described it as the culmination of several years of congressional debate about reducing the federal prison population while maintaining public safety, with some members motivated by the growth of the Bureau of Prisons budget and others by the social costs of incarceration and the policy choices of the 1980s and early 1990s.

The statute had three major components: correctional reform through a risk and needs assessment system at the Bureau of Prisons, sentencing reform for certain federal offenses, and reauthorization of the Second Chance Act of 2007. It authorized 75 million dollars per fiscal year for fiscal years 2019 through 2023, with eighty percent directed to the Bureau of Prisons for implementation, a total authorization of 375 million dollars. The Congressional Budget Office, scoring the House version of the bill, estimated a net reduction in discretionary costs of 342 million dollars over the 2019 through 2028 period alongside increases in direct spending tied to earlier releases, figures that placed the statute’s fiscal footprint in the hundreds of millions rather than the tens of billions.

Its contents operated directly on federal machinery. Title I directed the Department of Justice to develop a risk and needs assessment system, later implemented as the Prisoner Assessment Tool Targeting Estimated Risk and Needs, to guide housing, program assignments, and eligibility for earned time credits. Federal prisoners could earn ten days of prerelease custody for each thirty days of successful participation in recidivism reduction programming, fifteen days for those assessed at minimum or low risk. The statute changed the computation of good time credit so that a prisoner could earn up to fifty-four days per year of the sentence imposed, rather than per year of the sentence served. On sentencing, Section 401 reduced the twenty-year mandatory minimum for drug trafficking with one prior qualifying conviction to fifteen years and the mandatory life term for two or more priors to twenty-five years, while narrowing the predicate convictions to serious drug felonies and serious violent felonies. Section 402 expanded the safety valve, previously available only to offenders with no more than one criminal history point, by rewriting the criminal-history criterion into a three-part test, extending coverage to Maritime Drug Law Enforcement Act offenses, and adding use-immunity for the defendant’s truthful disclosure. Section 403 eliminated the stacking of the twenty-five-year mandatory minimum under 18 U.S.C. 924(c) for second or subsequent firearm convictions charged in the same proceeding, limiting the enhancement to cases where the prior conviction was already final. Section 404 made the Fair Sentencing Act of 2010, which reduced the crack-to-powder cocaine sentencing disparity, retroactive, allowing qualifying prisoners to petition for resentencing. Section 105 stated that the act did not authorize prerelease custody or supervised release for an individual serving a prison term for a state offense.

Title V, the Second Chance Reauthorization Act of 2018, reauthorized many of the grant programs first created by the Second Chance Act of 2007 and extended a Bureau of Prisons pilot program for early release of elderly prisoners. It was the one part of the statute that faced the states rather than the federal system, and it did so the 1994 way: through grants, reauthorizing money for state and local reentry programs rather than commanding state policy. The statute as a whole passed by margins that signaled a broad coalition, 87 to 12 in the Senate and 358 to 36 in the House, and it drew public support from the American Bar Association, whose president described it as an important step toward improvements in the criminal justice system.

Thirty Billion Dollars and Three Hundred Million

The headline contrast is fiscal, and it is real. The earlier statute authorized 30.2 billion dollars. The later statute authorized 375 million dollars over five years. The Congressional Budget Office’s estimates for the later statute’s budget effects ran in the low hundreds of millions. The gap between the two figures is roughly a factor of eighty. But the fiscal contrast points to a deeper structural one: the two statutes bought different kinds of power.

The 1994 act purchased influence over systems Congress could not command. At the end of 2022, the Bureau of Justice Statistics counted 1.23 million people in state and federal prisons, of whom 1,070,834 were in state custody and 159,309 in federal custody, roughly eighty-seven percent and thirteen percent. The federal share of all incarceration, including local jails, was smaller still. Congress cannot set the sentences served by the seven-eighths of prisoners held by the states, so the 1994 act’s drafters converted federal dollars into state policy: prison construction money for states that adopted truth-in-sentencing rules, hiring money for local police departments, prevention money for community programs. The mechanism was indirect by constitutional necessity.

Not every provision of the 1994 act ran through the states. The expanded federal death penalty, the federal three-strikes provision, the assault weapons ban, and the federal sex offender registry guidelines were exercises of direct federal authority over federal crimes and federally regulated conduct. The assault weapons ban operated as a direct federal prohibition on manufacture and transfer, not as a condition on state money. But these direct provisions touched only the federal system’s small share of criminal cases. The statute’s transformative reach, the 100,000 officers, the prison construction incentives, the prevention grants, moved through state and local institutions that chose to participate. The 1994 act therefore demonstrates both levers in a single statute, with the spending lever carrying nearly all the weight.

The First Step Act took the opposite route into the opposite scale. Its sentencing and release provisions applied to federal prisoners alone, a population of roughly 160,000 against more than a million in state prisons. Section 105 made the boundary explicit. What the statute lost in reach it gained in directness: where it applied, it changed rules rather than offering grants, recalculating good time, limiting charging practices under 924(c), and opening crack-cocaine sentences imposed before 2010 to judicial review. No state legislature had to act, no governor had to accept a bargain, and no state prisoner saw a sentence change because of it.

The contrast, then, is not severity against leniency. It is scale of spending against directness of control. One statute spent enormously to reach, at one remove, the large population it could not touch directly. The other spent modestly to change, at first hand, the small population it could. Each chose the lever that fit its object, and each accepted the limits of that lever.

Two Errors Worth Retiring

Two mistakes recur in discussions of the two statutes, and both flatten the comparison.

The first is the claim that the First Step Act repealed the 1994 crime bill. It did not. No title of the later statute repeals, amends, or references Public Law 103-322. The provisions of the 1994 act that are gone are gone for reasons internal to that act: the assault weapons ban expired on September 13, 2004, under its own ten-year sunset provision, after Congress declined to renew it. The three-strikes provision, the expanded federal death penalty, the truth-in-sentencing grant programs, and the Violence Against Women Act provisions all remained on the books after December 21, 2018. The First Step Act adjusted federal sentencing rules that had been shaped by several statutes across three decades; it undid none of them wholesale.

The second error is the assumption that either statute reached state sentencing directly. Neither did. The 1994 act’s truth-in-sentencing grants could only offer states money for adopting the eighty-five-percent rule; a state that declined the money kept its own sentencing law untouched. The First Step Act’s earned time credits, good time recalculation, and retroactivity provisions applied to federal sentences and Bureau of Prisons administration alone, and Section 105 drew the line at state offenses in so many words. The two statutes thus share a constitutional boundary even as they sit on opposite sides of the policy debate: the federal government could pay the states to change, or it could change itself, but it could not legislate a state sentence.

The Choice Every Federal Reform Faces

Drawn structurally rather than morally, the comparison yields a single durable lesson, and it is the one the remaining sections of this article will test against evidence. Any federal criminal justice reform must choose its lever. It can operate directly on the federal system, where Congress writes the rules and the Bureau of Prisons executes them, accepting that the federal system holds roughly one prisoner in eight. Or it can spend money to influence the states, where the great majority of prisoners are held, accepting that influence is slower, conditional, and hostage to fifty separate decisions.

The 1994 act chose the second lever at a scale no federal crime bill had attempted, and its results track the logic of the choice: where states took the money, policy followed, and where they did not, the statute’s reach stopped. The First Step Act chose the first lever at a fraction of the cost, and its results track that logic too: federal sentences changed by operation of law, while state sentencing remained outside its field of action. The question for any future reform is not which of these two statutes to imitate in spirit, but which lever to pull given what is known about what each one can and cannot do. That is the decision framework the rest of this article builds.

The framework points first to the population each statute could touch. Before direction, coalition, or instrument can be judged, the comparison needs the constitutional map: the thirteen percent of prisoners Congress governs directly against the eighty-seven percent it can only influence with money. The earlier statute spent thirty billion dollars to cross that boundary; the later statute stayed inside it and spent a fraction. Reach is therefore the axis on which the other three turn, and the evidence for it is arithmetic.

Reach: Who Each Statute Could Actually Touch

Reach, for a criminal justice statute, is an operational question before it is a political one. It asks which population the law changes the legal position of, which system of courts and prisons bears the resulting movement of people, and which constitutional doorway the statute must walk through to get there. The 1994 statute and the 2018 statute answer that question in mirror image. One law reached deep into the much larger state systems without ever legislating for them directly, because Congress cannot command state sentencing policy. It instead purchased influence, conditioning federal dollars on state choices. The other law governed only the small federal slice of American incarceration, but it governed that slice directly, rewriting the sentences and prison conditions of people in federal custody by its own terms. The difference in their cost, roughly thirty billion dollars against a few hundred million, follows from that same division of authority. The size gap between the two statutes is structural, fixed by the Constitution’s assignment of criminal justice power, rather than the product of any single drafting decision.

The Earlier Statute and the Purchase of Influence

The Violent Crime Control and Law Enforcement Act of 1994 authorized approximately 30.2 billion dollars in federal funding over six years, divided among more than sixty law enforcement, prison construction, and crime prevention programs, according to the California Legislative Analyst’s Office analysis of the bill as enacted. State and local law enforcement received roughly 10.8 billion dollars of that total. Prison programs received roughly 9.9 billion dollars. Crime prevention programs received roughly 6.9 billion dollars. President Clinton signed the bill on September 13, 1994, after it passed the 103rd Congress, making it Public Law 103-322. The law also expanded federal criminal penalties directly, establishing the federal three strikes provision mandating life without parole for a third serious violent federal felony and extending mandatory minimums for federal drug offenses. Those direct federal penalty provisions reached only defendants charged under federal law, a population then and since dwarfed by the state systems.

The larger, and analytically more important, half of the statute’s reach ran through the spending power. Criminal sentencing policy sits primarily with the states under the Constitution’s federal structure. Congress cannot order a state legislature to rewrite its sentencing code or a state parole board to alter its release standards. What Congress can do, under the doctrine the Supreme Court set out in South Dakota v. Dole in 1987, is condition federal money on state adoption of particular policies. The Dole decision held that the Spending Clause permits Congress to attach conditions to funds disbursed to the states even when the regulation would exceed Congress’s authority to mandate directly, provided the conditions serve the general welfare and remain short of coercion. A state always retains what the Court called the simple expedient of declining the money. The 1994 statute used that mechanism as its principal instrument for reshaping state criminal justice systems.

The Community Oriented Policing Services program, known as the COPS program, was the most visible grant instrument: an 8.8 billion dollar program aimed at placing 100,000 additional community policing officers on patrol nationwide, with the money flowing to state and local agencies that accepted the grants and their terms. The program’s federal matching arrangement, the diminishing federal share, and the awards it recorded are set out in full where the instrument analysis below takes up the grant design. The Violent Offender Incarceration and Truth in Sentencing incentive grants, created in Title II of the act, operated more directly on sentencing outcomes. To qualify for these grants, a state had to demonstrate that persons convicted of designated violent crimes would serve a minimum of eighty-five percent of the sentence imposed, a requirement known as truth in sentencing. The incentive funding authorized for prison construction and expansion under this program totaled roughly 9.7 billion dollars. The Bureau of Justice Assistance described one purpose of the grants as enabling states to manage prison capacity by funding additional beds for violent offenders, and the practical effect of the eighty-five percent condition was that states seeking the money had to tighten release rules across their own prison systems. By 1998, twenty-seven states and the District of Columbia had qualified for the incentive grants, and former National Institute of Justice director Jeremy Travis later described twenty-eight states and the District of Columbia as having followed the grant money toward stricter sentencing laws for violent offenses.

This is the precise sense in which the 1994 statute bought influence rather than exercised command. No provision of the law dictated that a state must require eighty-five percent time served. No provision ordered state legislatures to build prisons. The conditions attached to the grant money did that work, and states that declined the funds could keep their existing policies untouched. The mechanics of the transaction were administrative as well as political: a state seeking VOI/TIS money had to apply to the Bureau of Justice Assistance and demonstrate that its laws required violent offenders to serve the qualifying portion of their sentences, which meant that acceptance of the money and conformity of state law arrived as a single packaged decision. The Supreme Court in Dole had reasoned that this structure survives constitutional scrutiny precisely because the state’s acceptance remains voluntary, however unattractive refusal might be as a practical matter. The coercion line that limits the spending power, later sharpened when the Court struck down the Medicaid expansion condition in NFIB v. Sebelius in 2012, requires that the condition remain an inducement rather than a compulsion, but the incentive grants of the 1994 statute were designed well within that boundary: construction money for prisons, offered in exchange for tougher release rules. Operationally, reach therefore meant the populations of every state prison system whose legislature accepted federal construction and policing dollars and conformed state law to federal conditions in exchange. Because the state systems collectively hold the overwhelming majority of prisoners in the United States, and because the grant programs offered sums large enough to reshape state correctional policy, the 1994 statute’s effective reach extended far beyond the federal defendants its penalty provisions covered. It reached the state systems by spending on them.

The Later Statute and the Federal Thirteenth

The First Step Act, signed by President Trump on December 21, 2018, took the opposite approach. Every substantive provision of the law applied to people prosecuted under federal law and held in the custody of the Federal Bureau of Prisons. No provision altered a state sentence, changed a state prisoner’s release date, or conditioned state funding on anything at all. The statute reached no state prisoner, in either direction, expanding or contracting. Its entire operational reach was the federal prison population, the population over which Congress exercises direct and plenary legislative authority.

The law’s sentencing reform title addressed how federal sentences are calculated and imposed. Section 401 narrowed the enhancement provisions for prior drug and violent felonies, reducing the mandatory minimum for a second violation from twenty years to fifteen years and reducing the mandatory minimum for a third violation from life imprisonment to twenty-five years, while restricting the qualifying predicates to serious drug felonies and serious violent felonies. Those changes applied to offenses committed before enactment where sentence had not yet been imposed, but they were not fully retroactive. Section 402 broadened the safety valve that allows federal judges to sentence low-level nonviolent drug offenders below mandatory minimums, extending eligibility beyond defendants with no more than one criminal history point. The rewrite of the criminal-history criterion into a three-part test, the extension to Maritime Drug Law Enforcement Act offenses, and the new use-immunity for the defendant’s disclosure are examined in the direction analysis that follows, which holds the authoritative treatment. Section 403 eliminated the stacking of mandatory consecutive penalties under 18 U.S.C. section 924(c) for firearms offenses charged in a single prosecution, so that enhanced penalties attach only to offenders with a prior final conviction for a qualifying gun offense. The reach of these sentencing provisions was further bounded by how Congress structured their timing. The section 401 narrowing of enhancements applied to offenses committed before enactment only where sentence had not yet been imposed, and Bloomberg Law’s analysis estimated that full retroactivity would have reduced the sentences of just over 3,000 inmates, while the enacted prospective version was estimated to affect only dozens of defendants per year. Section 404, by contrast, was fully retroactive, and the two provisions together show how a directly legislating Congress can calibrate reach with precision: the statute opened the prison doors for people sentenced under the old crack ratios while declining to reopen sentences that had already become final under the old enhancement rules. Section 404 gave retroactive effect to the Fair Sentencing Act of 2010, which had reduced the disparity between crack and powder cocaine sentences from a one hundred to one quantity ratio to eighteen to one, allowing people sentenced under the old ratio to petition federal courts for resentencing. Floor debate in the House estimated that more than 2,600 people would have the opportunity to seek resentencing under that provision.

The prison administration title addressed how federal prisoners serve the time that remains. It corrected the Bureau of Prisons’ calculation of good conduct time, restoring the full fifty-four days per year that the statute as originally written had contemplated, in place of the forty-seven days the Bureau had been awarding through its own formula, and applied the correction retroactively. That correction alone led to the release of 3,100 people from federal prisons on July 19, 2019, the implementation date the law set. The title also directed the Attorney General to develop a risk and needs assessment system within one hundred eighty days of enactment, the system later known as PATTERN, to classify prisoners by recidivism risk, match them to evidence-based recidivism reduction programming, and allow eligible prisoners to earn time credits under 18 U.S.C. section 3632(d)(4) toward prerelease custody or supervised release upon successful program completion. Congress built boundaries into the earned time credit mechanism itself: the statute lists nearly seventy conviction categories that render a prisoner ineligible to accrue these credits, and prisoners subject to a final order of removal are likewise excluded. The Bureau was required to have programming available to all eligible prisoners within three years, a deadline the law set as a statutory obligation rather than an aspiration. Beyond the headline sentence reforms, the law placed prisoners closer to home by requiring designation within five hundred driving miles of the primary residence where possible, expanded opioid and heroin treatment programming, prohibited the shackling of pregnant prisoners, and ended juvenile solitary confinement in federal facilities. Each of these provisions presupposed the same jurisdictional foundation: they applied to the Bureau of Prisons because the Bureau is a federal agency holding federal prisoners, and no state facility entered the picture at any point. It expanded programming obligations for the Bureau, expanded home confinement options for elderly and terminally ill prisoners, broadened compassionate release so prisoners could petition federal courts directly, and reduced the age threshold for the elderly offender program from sixty-five to sixty while lowering the required portion of the sentence to be served from seventy-five percent to two-thirds.

The operational reach of all of these provisions was therefore bounded by the walls of the federal system. A person serving a state drug sentence in a state prison experienced no change in sentence calculation, no access to earned time credits, and no revised good conduct formula under the First Step Act. The statute’s authors could have attached conditions to federal grants to encourage states to adopt similar reforms, as the 1994 law did in reverse, but the First Step Act contained no such grant instrument. Its reach was direct and narrow: it rewrote the law governing federal prisoners, and only federal prisoners.

The Arithmetic of Authority

The cost figures attached to the two statutes make the constitutional structure visible in dollars. The 1994 statute authorized roughly 30.2 billion dollars over six years. The Congressional Budget Office’s August 2018 cost estimate for the House-passed version of the First Step Act projected a net reduction in discretionary costs of 342 million dollars over the 2019 to 2028 period, driven by the projected decline in the federal prison population, alongside an increase in direct spending of 214 million dollars and a four million dollar revenue reduction as earlier releases expanded eligibility for federal benefit programs. The good conduct time correction alone was estimated to save 40 million dollars in its first year, because a smaller federal prison population costs less to house. The arithmetic underneath these estimates ran entirely through the federal headcount: fewer federal prisoners, fewer federal dollars spent on federal confinement, and earlier releases shifting some costs to benefit programs administered by the federal government. No state correctional budget appeared in the estimate, because no state prisoner was touched. A later Senate version drew a comparable estimate of several hundred million dollars in net costs. The two orders of magnitude are not an accident of drafting priorities. They record what each statute had to do to reach the population it targeted.

The Bureau of Justice Statistics reported that at year end 2022, the United States held 1,230,143 persons in state or federal prisons, divided between federal and state custody in the thirteen-to-eighty-seven proportion established in the fiscal overview above. Local jails held several hundred thousand additional people, almost all of whom fall outside both the federal system and the state prison systems covered by the prison construction grants. For the purposes of comparing the two statutes’ reach, the decisive proportion is the federal share: roughly one prisoner in eight is a federal prisoner, and roughly seven in eight are state prisoners. Congress legislates directly for the thirteen percent it controls and can only spend money to influence the eighty-seven percent it does not.

That arithmetic explains why the First Step Act could cost so little while changing law directly, and why the 1994 statute had to cost so much to change anything at all. Direct legislation over the federal population requires no grant apparatus, no construction subsidy, and no state compliance machinery. Congress writes the sentencing statute, the federal courts apply it, and the Bureau of Prisons implements it. The First Step Act’s hundreds of millions of dollars in costs reflect programming obligations, administrative implementation, and the benefit-program effects of earlier releases, not the purchase of state cooperation. The 1994 statute, by contrast, had to buy its way into state systems one grant at a time. Thirty billion dollars purchased 100,000 police officers for local agencies, prison construction funds conditioned on eighty-five percent time served, and dozens of smaller programs, because that was the price of influencing systems Congress could not govern. The grant structure also explains why the 1994 figure is measured in authorizations spread over six years rather than a single appropriation: conditional spending reaches states through successive annual grant cycles, each of which must be funded, administered, and accepted anew, so the thirty billion dollars represents a sustained purchasing commitment rather than a one-time expenditure.

The reach of each statute therefore followed the constitutional division of criminal justice authority with near mechanical fidelity. The 1994 statute was large because it was indirect. The 2018 statute was small because it was direct. A law that must reach the eighty-seven percent through conditional spending will always be measured in billions, since it has to compensate fifty separate sovereigns for doing what Congress wants. A law that reaches only the thirteen percent through direct federal authority can be measured in the hundreds of millions, since it changes law rather than purchasing compliance.

Why the Size Gap Is Constitutional

The enforcement machinery behind each statute’s reach differed as sharply as the reach itself. The 1994 statute’s state-facing ambitions depended entirely on uptake: grants had to be applied for, conditions had to be met, and legislatures had to pass new laws to qualify. Twenty-eight states moved to truth in sentencing in the wake of the incentive program, which means the statute’s reach into the eighty-seven percent was real but uneven, tracking state willingness to trade policy conformity for construction dollars. The First Step Act’s reach, by contrast, was automatic. Its sentencing provisions took effect as federal law upon enactment, the good conduct correction released 3,100 people on a date the statute itself specified, and no state action could expand or contract the population affected. One statute reached the majority of American prisoners by persuading fifty sovereigns; the other reached the minority by commanding no one, because it needed no one’s permission. The full record of that state-side uptake, and of what the earlier statute’s prison construction money purchased, is set out in the earlier statute’s incarceration impact.

If Congress had written a direct-legislation version of the 1994 crime bill that applied only to federal defendants, it would have reached roughly thirteen percent of the incarcerated population and cost a fraction of thirty billion dollars. If Congress had attempted to export the First Step Act’s reforms to the states by conditioning grant money on their adoption, the bill would have needed a grant architecture and appropriations measured in the billions. Neither statute’s reach was a free choice among alternatives. Each followed from the constitutional instrument available for the population it addressed.

The Constitution assigns the general police power, including the definition and punishment of most crimes, to the states. Congress holds enumerated powers, including the authority to define and punish federal offenses, but it cannot commandeer state legislatures or state officers into enforcing federal criminal policy, a principle the Supreme Court confirmed in the anti-commandeering cases of the nineteen nineties. The two instruments left open are the direct regulation of federal offenses and the conditional spending recognized in Dole. The 1994 statute used both, reserving direct regulation for the federal penalty provisions and deploying conditional spending as the vehicle for its state-facing ambitions. The First Step Act used only the first. The size gap between the two laws is therefore not a policy choice that either Congress could have reversed at will. It is the shape of the Constitution’s division of criminal justice authority, expressed in budgets.

With reach established, the comparison turns to direction: what each statute did to the machinery it could touch. The earlier law used both its instruments to turn that machinery outward, toward longer and more certain confinement. The later law turned the same gears inward, toward narrower triggers, lower floors, and more doors out. The gears themselves did not change between the two enactments. Only the direction of the turn did.

Two Statutes Pulling Opposite Ways on the Same System

The Violent Crime Control and Law Enforcement Act of 1994, signed on September 13, 1994 as Public Law 103-322, and the First Step Act of 2018, signed on December 21, 2018 as Public Law 115-391, operated on the same federal sentencing and corrections machinery in opposite directions. The earlier statute widened the funnel: it made more conduct eligible for the harshest federal penalties, lengthened the mandatory terms for repeat offenders, and paid for the prison beds to hold them. The later statute narrowed the funnel: it restricted which prior convictions could trigger enhanced penalties, lowered the resulting floors, and built new paths out of federal custody. Direction, in this comparison, is the sign of each amendment. The two laws share the same gears: the mandatory minimum as a sentence floor, the sentencing guidelines grid, good time credit, and the release mechanisms at the back end. The 1994 Act turned those gears toward longer, more certain confinement; the 2018 Act turned them back.

The Outward Movement: More Conduct, Longer Terms, More Beds

Title VII of the 1994 Act created the federal three-strikes provision, codified at 18 U.S.C. 3559(c). A defendant convicted in federal court of a serious violent felony faced mandatory life imprisonment upon a third qualifying strike: two prior serious violent felonies, or one serious violent felony plus one serious drug offense, each conviction final, each committed on a separate prior occasion, and each offense after the first committed after conviction for the preceding one. The statute enumerated qualifying violent felonies such as murder, rape, kidnapping, and robbery, and it treated any other ten-year felony involving force or the threat of force as qualifying; it also gave defendants an affirmative defense for robbery or arson convictions that involved neither the fact nor the threat of injury to a person. The mechanism was a sentence floor raised to its absolute maximum and made immune to judicial adjustment: notwithstanding any other provision of law, the court imposed life.

Title VI, the Federal Death Penalty Act, created sixty new death penalty offenses under forty-one federal capital statutes. The new capital offenses fell into three categories: homicide offenses, espionage and treason, and non-homicidal narcotics offenses. The expansion reached drive-by shootings resulting in death, carjackings resulting in death, murders of federal law enforcement officers, and large-scale drug trafficking enterprises even where no killing occurred. Congress thus moved the ceiling of federal punishment outward at the same time it raised the floor.

The 1994 Act also added mandatory-minimum provisions of its own. It imposed a one-year mandatory minimum for distributing narcotics within one thousand feet of a public housing project, and it created new penalties for distributing or possessing with intent to distribute a controlled substance in or on, or within one thousand feet of, a truck stop or safety rest area. At the same time, and often overlooked, the Act’s Title VIII, styled “Applicability of Mandatory Minimum Penalties in Certain Cases,” created the federal safety valve, codified at 18 U.S.C. 3553(f). Under that provision a court could sentence below an otherwise applicable drug mandatory minimum, pursuant to the guidelines, when the defendant had a minimal criminal history, had not used violence or a weapon, had not caused death or serious bodily injury, had not led the offense, and had truthfully disclosed everything known about it. The 1994 statute therefore did two things at once on mandatory minimums: it added new ones and cut a narrow door through the existing ones. The 2018 statute later widened that same door.

Title XI, subtitle A, the Public Safety and Recreational Firearms Use Protection Act, banned the manufacture for civilian use of defined semiautomatic assault weapons and large-capacity ammunition magazines. The ban lasted ten years and expired under its own sunset provision on September 13, 2004.

The 1994 Act also expanded the physical plant of punishment. It authorized 9.7 billion dollars in prison-related assistance over six years, including 1.8 billion dollars to reimburse states for the costs of incarcerating criminal aliens and 7.9 billion dollars in Violent Offender Incarceration and Truth in Sentencing incentive grants to help states expand correctional capacity for violent offenders. The Truth in Sentencing grants conditioned federal aid on states requiring violent offenders to serve at least eighty-five percent of their sentences. Alongside the prison money, the Act funded the hiring of one hundred thousand additional police officers through community policing grants, and it ended federal Pell grant eligibility for incarcerated students, closing off an education channel inside the prisons it was filling. The Truth in Sentencing condition carried a further mechanical significance: the eighty-five percent service requirement it imposed on the states mirrored the federal design under the Sentencing Reform Act, where the intended fifteen percent good time reduction left a prisoner serving eighty-five percent of the imposed term. The 1994 Act thus exported the federal service ratio to state systems as a condition of funding. The outward movement was therefore complete: more conduct reached the harshest penalties, mandatory terms grew longer, the safety valve stayed narrow, the beds were funded, and the doors at the back narrowed.

The Inward Movement: Narrower Triggers, Lower Floors, More Doors

The First Step Act reversed direction on the same gears. Section 401 rewrote the recidivist enhancements in 21 U.S.C. 841(b)(1)(A) for high-quantity drug offenses. Where one prior felony drug offense had triggered a mandatory minimum of twenty years, the Act required one prior “serious drug felony or serious violent felony” and set the minimum at fifteen years. Where two or more prior felony drug offenses had triggered mandatory life imprisonment without release, the Act required two or more prior serious drug or serious violent felonies and set the minimum at twenty-five years. The new terms of art did the narrowing work. A “serious drug felony” meant an offense described in 18 U.S.C. 924(e)(2) for which the offender served more than twelve months and was released within fifteen years of the start of the current offense. A “serious violent felony” meant an offense described in 18 U.S.C. 3559(c)(2) for which the offender served more than twelve months, or the equivalent of an 18 U.S.C. 113 assault offense in the special maritime and territorial jurisdiction. Priors that had once triggered the enhancements automatically now had to clear a service-time and recency screen, and the resulting floors dropped by a quarter or more. The narrowing was not purely one-directional in its details: while the Act tightened which drug priors could count, it simultaneously widened the predicate class to include serious violent felonies, which the old felony-drug-offense standard had not covered. A defendant with a qualifying violent history could now trigger the enhancement where previously only drug priors counted. The dominant effect was still contraction, because the service-time and fifteen-year recency screens disqualified large numbers of stale drug priors that the old standard had swept in, but the provision’s architecture shows direction operating through precision rather than blanket erasure.

Section 403 addressed 18 U.S.C. 924(c), the firearm enhancement for using or carrying a gun during a violent or drug trafficking crime. Before the amendment, prosecutors could charge multiple 924(c) counts in a single indictment and the twenty-five-year penalty for a “second or subsequent conviction” attached to each count after the first, producing consecutive twenty-five-year terms for a first-time offender in a single case. The Act replaced the phrase “second or subsequent conviction under this subsection” with “violation of this subsection that occurs after a prior conviction under this subsection has become final.” The twenty-five-year term survived, but it applied only to true repeat offenders whose earlier 924(c) conviction was already final when the new violation occurred.

Section 404 made the Fair Sentencing Act of 2010 retroactive. That 2010 law had reduced the crack-to-powder cocaine quantity ratio that triggered mandatory minimums from one hundred to one to roughly eighteen to one. Section 404 allowed courts, on motion, to resentence defendants for covered offenses, meaning violations of federal criminal statutes whose penalties the 2010 Act had modified and which were committed before August 3, 2010, as if the 2010 law had been in effect at the time of the offense. Relief was petition-driven rather than automatic.

The release-side reforms were the other half of the inward movement. Section 101 amended 18 U.S.C. 3624(b) so that federal prisoners could earn up to fifty-four days of good time credit for each year of the sentence imposed by the court, rather than for each year actually served. The Bureau of Prisons had long read the statute the second way, granting roughly forty-seven days per year of the imposed sentence under a declining-balance method that the Supreme Court upheld in Barber v. Thomas in 2010. The amendment restored the fifteen-percent design Congress had originally written into the Sentencing Reform Act of 1984, and it applied retroactively, recalculating time for prisoners already in custody.

Title I directed the Attorney General to develop a risk and needs assessment system and created earned time credits: eligible prisoners could earn ten to fifteen days of credit for every thirty days of successful participation in evidence-based recidivism reduction programming or productive activities. The credits could be applied toward prerelease custody, such as a halfway house or home confinement, without regard to the old statutory caps, or toward transfer to supervised release up to twelve months early for prisoners assessed as minimum or low risk. Inmates with certain disqualifying convictions, nearly seventy categories listed at 18 U.S.C. 3632(d)(4)(D), and prisoners subject to deportation could not earn the credits. The Act also directed the Bureau, to the extent practicable, to place lower-risk and lower-needs prisoners on home confinement for the maximum time permitted, expanded the pilot program placing elderly and terminally ill offenders in home detention under 34 U.S.C. 60541 by lowering the age threshold from sixty-five to sixty and the required portion of the sentence served from seventy-five percent to two-thirds, and expanded access to compassionate release.

These sentencing-reform provisions were prospective. They applied where a sentence had not yet been imposed when the Act took effect, or to convictions entered afterward, and Congress did not make them fully retroactive. Only the good time fix and the section 404 crack resentencing reached back into sentences already being served.

Why Contraction Was Not Repeal

A common error treats the First Step Act as the undoing of the 1994 crime bill. The statute did nothing of the kind. It repealed no title of the earlier law. The three-strikes mandatory life term at 18 U.S.C. 3559(c) survived intact. The sixty new death penalty offenses created by Title VI survived intact. The mandatory-minimum framework survived: the ten-year and five-year floors in 21 U.S.C. 841(b) for first offenses were untouched, and only the recidivist enhancements above them were recalibrated. The assault weapons ban needed no repeal because it had already expired by its own sunset on September 13, 2004, fourteen years before the First Step Act’s enactment. The prison capacity funded with the 1994 appropriations stayed built. The scale of what the 2018 statute did not touch becomes clearer when the 1994 Act is seen whole: a three-hundred-fifty-six-page omnibus whose titles covered community policing, prevention programs, violence-against-women provisions, immigration offenses, firearms licensing, and state sex-offender registries, among much else. The First Step Act amended a handful of sentencing and custody provisions. The overwhelming bulk of the 1994 law, the policing grants, the prevention money, the registry requirements, the immigration crimes, lay entirely outside the 2018 statute’s scope. Describing the later law as a repeal of the earlier one mistakes a targeted recalibration of penalty and release gears for a dismantling that never happened.

What the 2018 statute changed, it changed by amendment rather than erasure, and the safety valve illustrates the point most sharply. The provision the First Step Act broadened in section 402 was itself a creation of the 1994 Act’s Title VIII. The 1994 Congress had recognized that mandatory minimums were sweeping in low-level, nonviolent first offenders and had cut the valve to let judges sentence those defendants under the guidelines instead. Section 402 did not invent the relief; it rewrote the valve’s criminal-history criterion from “not more than one criminal history point” to a three-part test, extended the valve’s coverage to Maritime Drug Law Enforcement Act offenses, and barred the government from using a defendant’s truthful disclosure to enhance the sentence unless the information related to a violent offense. In 2024, the Supreme Court in Pulsifer v. United States construed the amended criterion, holding that a defendant must satisfy each of the three conditions to qualify. The safety valve’s effect on the guidelines calculation added a further inward push: a defendant who met the valve’s criteria also received a two-level reduction in the drug offense level under the guidelines, so the relief operated both below the statutory floor and within the range computed above it. The mechanism at work was continuity across twenty-four years: one Congress cut a narrow relief valve into its own mandatory-minimum regime, and a later Congress widened it.

The honest accounting of direction is therefore this. On selected gears the 2018 Act turned the 1994 settings back: enhancement triggers narrowed, penalty floors lowered, credit calculations restored, release paths multiplied. On every other gear the 1994 settings held: life for the third violent strike, death eligibility for the sixty offenses, the mandatory-minimum floors themselves, and the beds. Contraction on one axis and stasis on the others is not repeal; it is a partial reversal, and naming it precisely matters because the difference determines what a future Congress would still need to change.

The Machinery Both Statutes Worked

The two statutes make sense as opposites only once the machinery is visible. Federal sentencing starts with the statute of conviction. Where Congress has set a mandatory minimum, that minimum is the floor: whatever the guidelines calculation produces, the defendant serves at least the statutory term. The safety valve is the principal door through that floor for qualifying drug defendants that does not require a government motion, letting the court sentence within the guideline range instead. Above the floor, the United States Sentencing Guidelines map an offense level, from one to forty-three, against a criminal history category, from Roman numeral I to VI, to produce a range of months. The guidelines were mandatory from their effective date in 1987 until the Supreme Court’s 2005 decision in United States v. Booker made them advisory, but even as advisory they remained the starting point of every federal sentencing calculation.

The back end of the machinery determines how much of the imposed term is actually served. The Sentencing Reform Act of 1984 abolished federal parole, effective for offenses committed after November 1, 1987, and replaced it with supervised release, a period of supervision served after the prison term rather than in place of part of it. Good conduct time under 18 U.S.C. 3624(b) shortens the term itself: the 1984 design aimed at a fifteen-percent reduction, which the Bureau’s reading had compressed to roughly forty-seven days per imposed year until the 2018 fix restored fifty-four. The final months of a term can be served in a halfway house or on home confinement under 18 U.S.C. 3624(c), and courts can reduce a term for extraordinary and compelling reasons under the compassionate release provision at 18 U.S.C. 3582(c)(1)(A).

The 1994 Act operated this machinery outward at every accessible point. It raised the floors, adding mandatory life for the third strike and new mandatory minimums for drug distribution near public housing, truck stops, and rest areas. It expanded the conduct reaching the harshest penalties through the sixty new capital offenses. It funded the beds that longer terms required and conditioned state aid on eighty-five-percent service. It also narrowed the education channel available for rehabilitation inside the prisons it was filling, by ending Pell grant eligibility for prisoners.

The 2018 Act operated the same machinery inward at every accessible point. It lowered the floors and narrowed their triggers in section 401, confined the twenty-five-year firearm enhancement to true repeat offenders in section 403, and widened the door through the floor in section 402. It restored the good time calculation to the design Congress had written in 1984. It created a new currency of earned time credits convertible into prerelease custody and early supervised release, expanded home confinement for low-risk and elderly prisoners, and reopened the crack sentences imposed under the old one-hundred-to-one ratio. Two distinctions sharpen the picture of the machine. The three-strikes life term bypassed the guidelines grid entirely, imposing life by statutory command no matter what range the offense level and criminal history category produced. The 924(c) enhancement, by contrast, operated through consecutive stacking: its terms had to run back to back with the underlying sentence, which is why limiting the twenty-five-year term to final prior convictions mattered so much to total time served. And earned time credits differed in kind from good time: good time shortened the sentence itself, while earned time credits left the sentence intact and changed where its final portion was served. Nothing in the machine was replaced. Each statute simply chose a direction and turned the gears that way, and the comparison between them is a record of how far each turn reached.

The machinery analysis settles what each statute did; the vote analysis settles how each statute passed. Direction can be read from statutory text, but coalition must be read from roll calls, and the roll calls complicate the story the statutes alone would tell. The narrow, contested majority of 1994 and the near consensus of 2018 belong to different conditions rather than different eras, and the members who voted for both prove it.

Axis Three: The Coalition Behind Each Vote

The two statutes reached the president’s desk through markedly different roll calls. One survived a procedural defeat and passed on narrow, contested margins after weeks of open warfare inside the majority party. The other moved through both chambers with near consensus in a lame-duck session, opposed on the floor by a small bloc of one party and by no member of the other. The vote records define each coalition more precisely than any retrospective framing, and they complicate the tempting story that the two laws belong to opposite political eras.

The Rule That Failed

The 1994 statute nearly died on a procedural motion. The conference report on H.R. 3355, filed on August 10, 1994, required a rule for floor consideration. The Rules Committee reported H. Res. 517, which would have waived all points of order against the conference report and its consideration. On August 11, 1994, the House defeated the rule 225 to 210. The Congressional Record’s action history for the bill records the outcome flatly: Rule H. Res. 517 failed passage of the House. Commentary published within days of the vote called the defeat extraordinary, because votes on rules are ordinarily tests of party loyalty that the majority leadership wins.

The coalition that sank the rule was itself bipartisan. An account of the fight published by the History News Network described House Democrats wary of new gun restrictions joining Republicans to defeat the procedural vote. The National Rifle Association had objected to the bill’s ban on certain semiautomatic firearms, and Cato Institute commentary published in the days after the vote argued that two-thirds of the Republicans who had supported the assault weapons ban as a stand-alone measure in May had voted against the crime bill on the rule. The same commentary attacked the bill’s prevention provisions as pork, contrasting funding sufficient for roughly 20,000 additional police officers with prevention money that, in the commentators’ telling, would have paid for some 40,000 social workers to run programs such as midnight sports leagues.

President Bill Clinton’s handwritten notes in the days after the defeat show the White House counting its own coalition piece by piece. The notes recorded that 65 Republican congressmen had voted for the crime bill in April, that 38 Republicans had voted for the assault weapons ban in May, and that 11 Republicans had withstood pressure from gun-rights groups to support the package. When Speaker Thomas Foley and other House leaders came to the White House after the rule’s defeat and urged Clinton to drop the assault weapons ban to save the bill, the president refused. Rahm Emanuel, then a senior White House policy adviser who attended the meeting, later recalled that the administration decided instead to seek Republican votes for the conference report, an approach he described as novel at the time.

The Conference Report Votes

Ten days after the rule’s defeat, the House took up the conference report again. The vote came on August 21, 1994, during a rare Sunday session held in a summer in which Congress had foregone its customary district work period, according to the House’s own historical account of the day. The chamber agreed to the conference report by a vote of 235 to 195, recorded as Roll No. 416. The Clerk’s breakdown shows the partisan shape of the majority: Republicans voted 46 in favor and 131 against, with one member not voting; Democrats voted 188 in favor and 64 against, with four not voting; the chamber’s one Independent, Bernie Sanders of Vermont, voted in favor. The 46 Republicans who voted yes were the bloc the White House had courted after the rule fight.

Four days later, on August 25, 1994, the Senate agreed to the conference report by a vote of 61 to 38, Record Vote No. 295. Earlier the same day, the Senate had invoked cloture on the conference report by the same 61 to 38 margin, Record Vote No. 294, and had waived a Budget Act point of order against the report by 61 to 39, Record Vote No. 293. A tally published the following week put the 61 yeas at 55 Democrats and 6 Republicans, with 38 senators voting no and one not voting. Clinton issued a statement that afternoon accusing the Republican Senate leadership of putting political interests ahead of public safety, and noting that a number of Republicans who had voted for the crime bill the previous November had used the bill’s own funding provisions to block it.

The shape of the 1994 coalition is worth stating plainly. The statute was signed by a Democratic president and passed by Democratic majorities in both chambers, but it needed Republican votes in each: 46 in the House and 6 in the Senate, plus the Independent Sanders in the House. The opposition was also bipartisan in composition: 131 House Republicans and 64 House Democrats voted no, and the Senate no votes included members of both parties. The earlier stages of the same bill had been less partisan still. The Senate’s own version of the legislation had passed 95 to 4 in November 1993, and the House had passed its version in April 1994 with 65 Republican votes, according to Clinton’s notes. The 1994 law was not a party-line product; it was a Democratic-led bill that picked up a minority of Republicans at each stage while losing a minority of Democrats.

The Democratic losses included a notable split inside the Congressional Black Caucus. Of the caucus’s 38 members in the 103rd Congress, 26 voted for the conference report and 12 voted against, according to a 2015 retrospective account. The Clerk’s roll call confirms that Representatives John Lewis of Georgia and John Conyers of Michigan were among the no votes, while caucus chairman Kweisi Mfume voted yes. Individual Democrats who would remain prominent in later decades recorded their positions in the same roll call: Chuck Schumer of New York, who had been described in contemporary profiles as one of the 1994 bill’s major architects and the driving force behind the Brady gun legislation and the House assault weapons ban, voted yes. Dick Durbin of Illinois voted yes. Steny Hoyer of Maryland voted yes.

The First Step Act Majorities

The 2018 statute moved on a different trajectory. The Senate acted first on the final package. On December 17, 2018, the chamber invoked cloture on the motion to concur in the House amendment, as amended, by a vote of 82 to 12, Record Vote No. 267. The next evening, December 18, 2018, the Senate concurred by a vote of 87 to 12, Record Vote No. 271. The Senate’s official roll call shows a chamber in which the only organized opposition came from one side of the aisle: Democrats voted 47 to 0 in favor, Independents 2 to 0 in favor, and Republicans voted 38 in favor to 12 against, with Lindsey Graham of South Carolina not voting. Every one of the 12 no votes was cast by a Republican: John Barrasso of Wyoming, Tom Cotton of Arkansas, Mike Enzi of Wyoming, John Kennedy of Louisiana, Jon Kyl of Arizona, Lisa Murkowski of Alaska, James Risch of Idaho, Mike Rounds of South Dakota, Marco Rubio of Florida, Ben Sasse of Nebraska, Richard Shelby of Alabama, and Dan Sullivan of Alaska.

The opposition had made its stand the same day through amendments. Senators Cotton and Kennedy offered three amendments that would have expanded the list of offenses barring prisoners from earning time credits, imposed victim-notification requirements, and required recidivism tracking; supporters of the bill argued that any amendment would sink months of negotiation. The Cotton-Kennedy amendments were rejected 37 to 62. A separate amendment offered for Senator Ted Cruz of Texas, expanding the crime exclusion list and adjusting a faith-based provision, was adopted by voice vote, and Cruz, who had originally opposed the bill, then voted for it.

Two days later, on December 20, 2018, the House suspended its rules and agreed to the Senate amendment by a vote of 358 to 36, Roll No. 448, a motion that required a two-thirds majority. The breakdown showed Democrats voting 176 to 0 in favor and Republicans voting 182 to 36 against. No Democratic member of either chamber voted against the bill. President Donald Trump signed it on December 21, 2018, and it became Public Law 115-391.

The bill’s path to those margins had been uncertain for months. In the Senate, Chuck Grassley of Iowa, Dick Durbin of Illinois, Cory Booker of New Jersey, and Mike Lee of Utah built the bipartisan coalition, while in the House, Doug Collins of Georgia, Hakeem Jeffries of New York, and John Lewis of Georgia promoted the legislation. Grassley, who chaired the Judiciary Committee, insisted that the Senate package include sentencing provisions that had been left out of the House-passed version of May 2018, which the House had approved 360 to 59. The White House’s posture shifted during the fall: Attorney General Jeff Sessions, a staunch opponent of sentencing reductions, departed in November 2018, after which Trump endorsed the more expansive Senate package following an intensive lobbying effort by his son-in-law and senior adviser Jared Kushner. Contemporary reporting described the bill as a significant victory for advocates on the left and the right who had pressed Congress to reduce the prison population, and Trump called the result a great bipartisan achievement.

The Advocates Who Agreed on Little Else

The organizational coalition behind the 2018 law was the feature that drew the most comment at the time. Right on Crime, a criminal justice campaign of the Texas Public Policy Foundation, cosigned an endorsement letter to Congress alongside 41 other organizations. Reporting on the bill’s passage described the alliance in terms of its improbability: the American Civil Liberties Union and Americans for Prosperity, a conservative advocacy group, had pushed the bill through together. A University of Minnesota scholar writing in The Conversation listed the bill’s backers as ranging from progressive nonprofits such as the Brennan Center for Justice and the ACLU to the conservative Koch brothers. A scholarly account published by Stanford University Press described the pairing of the American Conservative Union on the right with the ACLU on the left as the “strange bedfellows” element of the campaign, and noted that the Business Roundtable and the U.S. Chamber of Commerce had endorsed the act as well. A Koch-network organization’s own retrospective described Van Jones and Mark Holden setting aside their differences to unite a coalition stretching from Representative Jeffries and the ACLU to Senator Lee and the White House. Senators who backed the bill cited support from law enforcement organizations alongside criminal justice reform advocates, and Susan Collins of Maine described the legislation as modeled on the efforts of dozens of states.

The opposition, by contrast, was narrow and almost entirely intraparty. Published accounts of the Senate debate named Cotton, Kennedy, Sasse, and Murkowski among the Republican opponents, with Cotton arguing that the bill would free dangerous criminals and voicing concern about reduced sentences for drug traffickers and violent felons. No Democrat in either chamber voted no. The 2018 coalition, in other words, was a consensus measure with a one-party dissenting bloc; the 1994 coalition had been a contested majority with dissenters in both parties.

The Members Who Voted for Both

The vote records undercut any account that treats the two statutes as products of opposite political eras. Several members of Congress voted for both. Chuck Schumer voted yes on the 1994 conference report in the House, where he had been one of the bill’s principal architects, and voted yes on the First Step Act in the Senate, where the roll call lists him among the 87 yeas. Dick Durbin voted yes in the House in 1994 and yes in the Senate in 2018. Bernie Sanders, the Independent who voted yes on the 1994 conference report, voted yes on the First Step Act in the Senate. Their votes are recorded in the Clerk’s 1994 roll call and the Senate’s 2018 roll call alike.

Other members traveled the distance between the two votes in the opposite direction. John Lewis voted against the 1994 crime bill as one of the 12 Congressional Black Caucus members in opposition, and a quarter century later he was among the House promoters of the First Step Act. The point is not that individual members reversed themselves on a single question; the two bills asked different questions. It is that the personnel of the two coalitions overlapped, and that both coalitions were bipartisan in composition. The 1994 law drew 46 House Republicans and 6 Senate Republicans; the 2018 law drew 182 House Republicans and 38 Senate Republicans. The partisan labels of the signing presidents differed, but neither vote was a party-line reversal of the other.

What Moved the Votes

The change between the two votes is better explained by changed conditions than by partisan realignment. The first condition was the crime rate itself. The FBI’s Uniform Crime Reporting data show the national violent crime rate peaking in 1991 at 758.2 violent crimes per 100,000 people, the highest level in the series. The 1994 bill was debated and passed while violent crime was near that peak; the House’s historical account notes that the legislation was debated rigorously in the run-up to the midterm elections, and the politics of the moment were shaped by fear of the crack cocaine epidemic, which Durbin later recalled in a Senate hearing by saying that the arrival of crack had frightened lawmakers into imposing severe new sentencing rules. By the time the First Step Act came to the floor, the violent crime rate had fallen by more than half: the FBI reported a 2018 rate of 368.9 violent crimes per 100,000 inhabitants, down nearly 4 percent from 2017 alone, and a Congressional Research Service analysis dated December 2022 confirmed that the rate had generally declined since the 1991 peak, remaining well below the early-1990s levels even after a mid-2010s uptick.

The second condition was fiscal pressure on state corrections budgets. The Pew Center on the States reported in February 2008 that one in every 99.1 American adults was behind bars at the start of that year, and that states had spent more than 49 billion dollars on corrections in the prior year, up from 11 billion dollars twenty years earlier, even as the national recidivism rate held steady with about half of released inmates returning within three years. A later Pew analysis found that state general fund spending on corrections had risen from 19 billion dollars in 1987 to 49 billion dollars in 2007 after adjusting for inflation, that one in 14 state dollars went to corrections by 2007 compared with one in 20 in 1987, and that corrections was the second-fastest-growing category of state spending over that period, trailing only Medicaid. The cost argument gave reform a constituency that the 1994 debate had lacked: lawmakers who favored punishment but balked at the bill.

The third condition was accumulated evidence from the states. Beginning in 2007, Texas confronted projections that its prison population would require massive new construction; instead of spending the nearly 2 billion dollars that new prison construction and operations would have cost, the legislature reinvested 241 million dollars in treatment, diversion, and community supervision programs, according to a Pew-cited account. The Council of State Governments Justice Center, which worked with Texas on the strategy, reported that the state’s justice reinvestment effort produced 1.5 billion dollars in construction savings and 340 million dollars in annual averted operating costs. Senator John Cornyn of Texas told the Senate in 2015 that between 2007 and 2012 the state’s incarceration rate had fallen 9.4 percent, its total crime rate had dropped 16 percent, and taxpayers had saved more than 2 billion dollars. Since 2007, 17 states worked with the Justice Center to develop similar justice reinvestment strategies. When the First Step Act’s Senate backers described the bill as taking lessons from the states, they were invoking a record that did not exist in 1994: more than a decade of state-level experiments in which prison populations fell and crime did not rise to replace them.

The two coalitions, read together, describe a legislature responding to different emergencies. In 1994, with violent crime near its historic peak and an election approaching, a Democratic president and Democratic majorities assembled a narrow majority that included dozens of Republicans and lost dozens of Democrats, and the bill survived a procedural defeat that exposed how fragile that majority was. In 2018, with violent crime at less than half its peak, state budgets strained by two decades of corrections growth, and state experiments offering an alternative model, a Republican president signed a bill that drew nearly nine-tenths of the Senate and more than nine-tenths of the voting House, with dissent confined to a minority of one party. The bipartisan character of both votes is the constant; the conditions that made each coalition possible are what changed.

Coalition explains who voted; instrument explains what their votes could build. The earlier Congress faced violent crime committed almost entirely in state jurisdiction and answered with money, because money was the only constitutional channel into state systems. The later Congress faced a federal prison population it could reorder at will and answered with amendment and administration. The instrument is therefore not a stylistic choice layered over the policy. It is the statute: the tool the target system admitted, and nothing else.

The Instrument Is the Statute

Grants Because the States Ran the System

The Violent Crime Control and Law Enforcement Act moved its largest sums of money through conditional grants to states and local governments. The statute authorized about $30.2 billion over six years, ending in federal fiscal year 2000, and the dominant design choice inside that number was the grant. The Public Safety Partnership and Community Policing Act, Title I of the statute, authorized $8.8 billion for the COPS on the Beat program, whose stated purpose was to put 100,000 community policing officers on the street; the Attorney General created the Office of Community Oriented Policing Services in October 1994 to administer the awards, which flowed overwhelmingly to municipal, county, and state agencies with general arrest powers. A separate grant stream aimed at incarceration capacity: the statute authorized $9.7 billion for states to build and operate prisons, divided between the Violent Offender Incarceration grant program and the Truth in Sentencing incentive grant program. The second of those two programs made its logic explicit in the name. A state could draw the construction money only if it enacted truth-in-sentencing rules requiring persons convicted of certain violent crimes to serve at least 85 percent of the imposed sentence, and the programs carried a 25 percent matching requirement so that the federal dollar would pull state money behind it. The grant was therefore never a gift. It was a contract in which Congress purchased a particular shape of state criminal justice policy with federal money, and the condition attached to the money was the policy. The COPS program embodied the same logic in personnel form. Its federal contribution was structured as a five-year matching arrangement in which the federal share diminished over time and eventually disappeared, leaving local agencies with the full cost of retaining the officers hired with Washington’s help. By the summer of 2000 the COPS Office had awarded more than $7.6 billion to more than 12,000 law enforcement agencies, and the Government Accountability Office later judged the program a modest contributor to the crime declines of the decade. The scale of the spending, however, was not the same as control. A locality that declined the match kept its own staffing decisions, a state that declined the prison money kept its own sentencing rules, and Congress possessed no instrument for reaching either one except the offer itself. The conditional grant made federal criminal justice policy national in footprint and voluntary in operation, which is why its effects were uneven across states in exact proportion to how badly each state wanted the money.

The reason for this design was structural rather than stylistic. Ordinary street crime, the violence the statute was written against, is prosecuted under state law in state courts and punished in state prisons. Congress could not simply order the states to hire more officers or to hold violent offenders for 85 percent of their terms. The Constitution leaves the police power to the states, so the only lever available at the federal level for changing what happened inside state criminal systems was the spending power, exercised through grants whose conditions states were free to refuse by leaving the money on the table. The COPS money could not compel a single locality to hire an officer. The truth-in-sentencing money could not compel a single legislature to change its sentencing code. Both could only offer a bargain, and the bargain worked: states rewrote their sentencing laws to chase the incentive funds, because turning down nearly ten billion dollars in construction money while neighboring states accepted it was a posture few governors found sustainable. The grant condition thereby functioned as a policy mandate without ever bearing the name, reaching millions of state prisoners through rules Congress itself lacked the authority to write.

Direct Federal Provisions Where Congress Had the Authority

The same statute did not rely on grants where it did not need them. For conduct within federal jurisdiction, Congress legislated directly, because direct legislation was available. Title VII established the federal three-strikes provision, Title VI created the sixty new federal death penalty offenses, and Title VIII established the federal safety valve; the definitions, qualifying offenses, affirmative defenses, and eligibility criteria of those provisions are examined in full in the direction analysis above. These provisions amended federal law directly. They applied in federal court on the day they took effect, required no state assent, carried no matching requirement, and depended on no appropriation being spent. They touched only the federal criminal docket, a fraction of national criminal justice, but within that fraction they operated with complete authority.

The two kinds of provision therefore map exactly onto the two kinds of reach. Where Congress controlled the system, it wrote rules. Where it did not control the system, it wrote checks with conditions attached. The statute’s reputation as a law of punishment rests on the direct provisions, but its mass rested on the grants: the federal three-strikes law could only reach defendants prosecuted in federal court, while the truth-in-sentencing grants reached every state that wanted prison construction money. The instrument followed the reach, and the reach was constitutional before it was political.

Amendment and Administration: The Later Statute’s Instrument

The First Step Act of the 115th Congress used the opposite instrument because it operated on the opposite system. Signed on December 21, 2018, as Public Law 115-391, it was a federal-system statute from its first line to its last. Its sentencing reforms were direct amendments of federal sentencing statutes. The act reduced the mandatory life sentence for a third qualifying federal drug offense to a mandatory minimum of 25 years, widened the federal safety valve so that more offenders could qualify for a sentence below the mandatory minimum, and made retroactive the sentencing reductions of the Fair Sentencing Act of 2010, allowing federal prisoners to petition for the reduced crack-cocaine disparity years after that statute’s enactment. Each of these provisions worked the way federal three-strikes had worked in the earlier law: by rewriting the rules of federal court directly, with immediate effect and no state intermediary. The difference was direction rather than method. The earlier statute had used direct federal authority to lengthen federal sentences; the later statute used the same authority to shorten them. The direction analysis above examines each change in full.

The act’s more original instrument, however, was administrative rather than legislative. Title I required the Attorney General to develop a risk and needs assessment system for the federal prison population, and the Bureau of Prisons implemented that requirement as PATTERN, the Prisoner Assessment Tool Targeting Estimated Risk and Needs. The statute then tied a new earned time credit system to participation in the programming the assessment identified. Under 18 U.S.C. section 3632(d)(4), an eligible prisoner earns 10 days of credit for every 30 days of successful participation in evidence-based recidivism reduction programming or productive activities, and prisoners assessed as minimum or low risk of recidivism earn 15 days per 30-day period. The credits do not shorten the sentence itself. They apply toward earlier transfer to prerelease custody, meaning a residential reentry center or home confinement, or toward earlier supervised release, with eligible prisoners able to transition to supervised release up to 12 months earlier than before. Eligibility excludes large categories: offenses involving terrorism, espionage, human trafficking, sex offenses, and other statutorily listed crimes, and prisoners assessed at medium or high risk may earn privileges but cannot apply credits toward release.

This design is administration as legislation. Congress did not attempt to reorder state sentencing systems, because it had no need to reach them and no authority to do so directly; the federal prison population was its own domain, and within that domain Congress could build an entire administrative apparatus, a risk assessment instrument, a catalog of approved programs, a credit formula, and a transfer mechanism, all run by a federal agency under federal law. The price of that completeness was scope. The apparatus reached only the federal prison system, and the federal prison system holds only a minority of the nation’s incarcerated population. As of the year of enactment’s surrounding BJS reporting, federal facilities held roughly 12 to 13 percent of all incarcerated people, with state prisons holding about 59 percent and local jails the remainder. A statute that operated with perfect authority inside federal walls could therefore move only a small share of American incarceration, however sophisticated its machinery.

Why Each Instrument Follows Its Reach

The contrast clarifies itself when the two statutes are read as answers to the same federalism problem. The earlier Congress faced a wave of violent crime committed almost entirely in state jurisdiction and answered with money, because money was the only constitutional channel into state systems; it conditioned the money on truth-in-sentencing because an unconditional grant would have left state policy untouched; and it accepted that the condition would be enforceable only so long as states preferred the money to their own sentencing autonomy. The later Congress faced a federal prison population it could reorder at will and answered with direct amendment and administrative machinery, because no bargain with any state was required; it built a risk assessment system and a credit ledger because, once Congress committed to operating inside its own walls, the interesting questions were all administrative, about which programs counted, how risk was scored, and how credits converted to release. Neither Congress chose its instrument for style. Each took the only instrument that its target system admitted.

This is why the two statutes are mirror images rather than opposites. The earlier statute was broad in reach and shallow in authority: it could touch the 85 percent rule in dozens of state capitals, but only by persuasion with money, and a state that declined the grant kept its own rules. The later statute was narrow in reach and deep in authority: it could rewrite federal sentencing and build a credit system from nothing, but the population it could reach was the federal one alone. The size of each statute’s ambition therefore inverted the size of its footprint. The earlier statute’s grants moved state prison policy across the country while its direct federal provisions touched a sliver of the docket. The later statute’s direct provisions touched every federal courtroom while its grants did not exist at all.

The four axes have their evidence. Reach, direction, coalition, and instrument each carry a structural consequence, and the table below gathers them into a single view before the verdict draws them together.

The Four-Axis Comparison Table

Axis	The 1994 crime bill	The First Step Act	Structural consequence
Reach	Conditional grant money aimed at state prisons holding roughly eighty-seven percent of prisoners	Direct rule changes confined to the federal system holding roughly thirteen percent	A large bill can buy influence only where states accept the bargain
Direction	Turned the sentencing machinery outward: more conduct, longer floors, funded beds	Turned the same machinery inward: narrower triggers, lower floors, new release paths	Contraction on selected gears is recalibration, not repeal
Coalition	Narrow contested majority, 235 to 195 in the House and 61 to 38 in the Senate, with dissent in both parties	Near consensus, 358 to 36 in the House and 87 to 12 in the Senate, with dissent confined to one party	Broad margins attach most easily to statutes with contained stakes
Instrument	Conditional grants plus direct federal provisions where authority existed	Amendment plus administration inside the federal system	Congress has two instruments, and the target system dictates the choice

The Verdict

The deciding factor is named. Any federal criminal justice reform must choose between operating directly on the small federal system, where Congress has full authority and limited reach, and offering money to states, where the reach is large and the authority is only conditional. The two statutes are the clearest illustrations of each choice and its price, and no federal criminal justice statute has ever escaped that trade. The largest crime bill in American history, authorizing more than $30 billion, worked through grants, because its target was the state system and grants were the only instrument that could reach it. The most bipartisan criminal justice reform in a generation, passing the Senate 87 to 12 and the House 358 to 36 in December 2018, was necessarily small, because its target was the federal system and direct authority could reach no further.

The namable claim is therefore this: authority or reach, pick one. Congress can legislate directly for the roughly thirteen percent of prisoners it controls, or it can spend money to influence the eighty-seven percent it does not. The earlier statute picked reach. It accepted conditional authority, matching requirements, and the political reality that a state could walk away from the bargain, in exchange for shaping sentencing policy in capitals where no federal rule could ever apply. The later statute picked authority. It accepted a small footprint, a long list of ineligible offenses, and the administrative burden of scoring the risk of every federal prisoner, in exchange for machinery it could actually command. Neither choice is a failure of the other. Each is the honest price of the federal structure, and every argument that pretends one statute could have done the other’s work ignores the constitutional fact that made both instruments what they were.

The trade also explains each statute’s characteristic blind spot. The grant instrument let the earlier Congress reach the entire state system, but it could reach it only indirectly, so the national outcome depended on fifty separate state decisions, and the states that most resisted the bargain kept their own rules. The direct instrument let the later Congress command its own system completely, but it could command no other, so the national outcome left state sentencing entirely untouched, and the eighty-seven percent of prisoners outside federal walls saw no change from the statute’s most sophisticated machinery. Each statute therefore produced exactly the pattern its instrument predicted: broad but uneven change from the earlier law, deep but confined change from the later one. The verdict is not that one instrument is superior. The verdict is that Congress has only two, that they are constitutional before they are strategic, and that the choice between them is the single decision that determines everything downstream.

This comparison is drawn constantly in political argument and almost never structurally. The earlier statute is invoked as proof that federal action can remake criminal justice, and the later statute is invoked as proof that reform can pass with bipartisan support, but both invocations skip the mechanism. The earlier statute remade criminal justice where it could reach it, through state systems it could only bribe. The later statute passed with overwhelming margins in part because its reach was small enough to cost few members anything. The structural version of the comparison, the one this cluster of articles can offer that political argument cannot, is the instrument analysis: the earlier law’s tool was the conditional grant because its object was the states, and the later law’s tool was amendment plus administration because its object was the federal system. A reader who understands that sentence understands what Congress can and cannot do, which is the entire question.

What Each Reader Should Study First

The student seeking the shape of federal power should start with the comparison itself, before either statute in isolation. The two laws read side by side make visible what a single statute cannot: the boundary between direct federal authority and conditional federal influence, and the way that boundary dictates every design choice downstream. Once that boundary is clear, the earlier statute’s grant conditions and the later statute’s credit machinery read as two solutions to one constitutional problem rather than as two unrelated laws. The comparison is the fastest route to the structural lesson, and the structural lesson is the one that transfers to every other area of federal policy.

The congressional staffer drafting reform should study the grant-condition mechanics of the earlier statute and the administrative-credit machinery of the later one, in that order. The earlier statute is the case study in what conditional money can and cannot buy: the 85 percent rule as a funding condition, the matching requirement as a commitment device, the formula grant as the delivery mechanism, and the state’s freedom to decline as the ceiling on the whole design. The staffer’s practical question is always how tight the condition can be drawn before states walk away, and the earlier statute supplies the answer in the form of a five-year diminishing match on the personnel side and a hard sentencing-policy threshold on the construction side. The later statute is the case study in what direct administration requires: a risk and needs assessment built by statute but scored by agency, a credit formula set by law but tracked by the Bureau of Prisons, eligibility exclusions written into the text, and the conversion of credits into prerelease custody and supervised release managed entirely inside the executive branch. A staffer who has internalized both designs knows which problems call for which tool before drafting a single section, and, just as importantly, knows that no drafting technique exists for combining the two instruments into one, since the authority for each comes from a different constitutional source.

The researcher should study the incarceration-impact evidence, which is where the two statutes’ different instruments produced their most instructive asymmetries. The earlier statute’s direct federal provisions touched a small docket, while its grant conditions moved state sentencing policy for a generation, which means the evidence of its effect lies largely in state prison admissions and time-served data rather than in federal statistics. The later statute’s effects run in the opposite direction: its sentencing amendments and earned time credits operate entirely inside the federal system, so its measurable impact appears in Bureau of Prisons release dates, supervised release transitions, and federal recidivism cohorts. A researcher who looks for the earlier statute’s footprint in federal data, or the later statute’s footprint in state data, will find nothing and conclude nothing; the instrument determines where the evidence lives.

The journalist should study the roll calls and coalition records, because the vote margins tell the story that the policy language conceals. The earlier statute passed through a crime politics in which both parties competed to appear tougher, and the later statute passed the Senate 87 to 12 on December 18, 2018, and the House 358 to 36 on December 20, 2018, with a coalition that spanned the ideological spectrum, and the difference in those coalitions is itself evidence about the authority-reach trade. Bipartisan margins attach most easily to statutes whose reach is small enough that the stakes are contained; the later statute’s lopsided votes were possible in part because it touched only the federal system, where no member had to answer for a change in state sentencing policy back home. A journalist who reports the votes alongside the instruments, rather than the votes alone, gives readers the structural explanation for why Washington reforms criminal justice most confidently where the fewest prisoners live, and why the rare large-footprint federal action almost always arrives dressed as a grant.

The instrument analysis closes the four-axis comparison. What remains is to gather the argument into its final form: what the pairing proves about the limits of federal criminal justice reform, the verdict restated, and this article’s place in the series that contains it.

What the Comparison Proves

The comparison proves that federal criminal justice reform has a structural ceiling, and that the ceiling is constitutional rather than political. Congress commands the federal system outright and can only influence the states, so every reform proposal must accept one of two fates: direct change for the thirteen percent of prisoners in federal custody, or conditional money for the eighty-seven percent in state custody. No coalition, however bipartisan, can vote that boundary away. The First Step Act’s 87 to 12 Senate margin did not widen its reach by a single prisoner; the 1994 statute’s thirty billion dollars did not buy a single state sentence. The boundary held for both, and it will hold for the next statute too.

That limit reframes what counts as success. A reform aimed at state prisons must be judged by uptake, by how many states accepted the bargain and conformed their laws, which is why twenty-eight states qualifying for truth-in-sentencing grants is the meaningful measure of the earlier law. A reform aimed at the federal system must be judged by implementation, by whether the agency built what the statute ordered, which is why the risk assessment system, the credit formula, and the recalculated good time are the meaningful measures of the later law. Judging either statute by the other’s metric produces only confusion: state admissions data cannot measure the First Step Act, and federal sentencing data cannot measure the 1994 act’s grants.

The verdict restated is the one the article has carried from its opening: authority or reach, pick one. The earlier statute picked reach and paid in conditionality, matching requirements, and uneven results across fifty sovereigns. The later statute picked authority and paid in smallness, a long list of ineligible offenses, and the administrative burden of scoring every federal prisoner’s risk. Neither choice was a failure of the other, and neither Congress could have taken the other’s path without changing its target. The trade is the honest price of the federal structure.

Within this series, the comparison is the capstone the two single-statute guides point toward. The guide to the earlier law and the guide to the later law each explain what one statute did; this article explains what the pairing means, and the incarceration-impact evidence and the mandatory-minimum mechanics give the reader the two supporting studies the argument requires. Readers who want a fuller path through the subject can continue with the series study guide, and the VaultBook legislation study notebook and the ReportMedic civics study tool each carry a companion exercise built on this comparison.

Q: What is the difference between the 1994 crime bill and the First Step Act?

The difference is instrument and reach, not mercy against severity. The 1994 statute was a conditional spending program of thirty billion dollars aimed at state and local systems, while the 2018 statute was a direct rewrite of federal sentencing and prison rules confined to the federal system. One purchased influence over states; the other commanded machinery it owned outright.

The earlier law attached money to state choices: prison construction funds for truth-in-sentencing rules, hiring grants for local police departments, prevention grants for community programs. Its direct federal provisions, including the three-strikes rule and new capital offenses, touched only the small federal docket. The later law changed no state sentence. It narrowed recidivist enhancements, widened the safety valve, restored the good time formula, and built an earned time credit system, all inside federal walls. Reading the pair as opposites on punishment misses the point. They are two solutions to one constitutional problem: Congress can command the federal system and can only pay the states.

Q: Which was bigger, the 1994 crime bill or the First Step Act?

The 1994 statute was larger by roughly a factor of eighty. It authorized 30.2 billion dollars over six fiscal years, while the First Step Act authorized 375 million dollars over five fiscal years, with the Congressional Budget Office scoring its budget effects in the low hundreds of millions. The size gap is structural, not rhetorical.

The 1994 law had to buy its way into state systems, and purchasing compliance from fifty sovereigns costs billions: police hiring grants, prison construction conditioned on eighty-five percent time served, prevention programs. The 2018 law rewrote rules for a federal system Congress controls directly, which requires no grant apparatus and no state machinery. The budget figures therefore record the price of the instrument, not the importance of the policy. A small statute that changes law directly can matter more to its target population than a giant one that only offers money.

Q: Why was the First Step Act so much smaller than the 1994 crime bill?

The First Step Act was smaller because it reached only the federal system, which holds roughly one prisoner in eight, and Congress can legislate for that system directly. Direct legislation needs no grants, no construction subsidies, and no state compliance machinery. The 1994 statute spent billions because it tried to move state systems it could not command.

Under the Constitution’s division of criminal justice authority, Congress writes the rules for federal offenses but cannot commandeer state legislatures or rewrite state sentencing codes. Influence over the states must be purchased through conditional spending, and purchasing influence at national scale is expensive. The 1994 act’s thirty billion dollars was the price of a six-year purchasing commitment across more than sixty programs. The 2018 act’s hundreds of millions covered programming, administration, and benefit effects of earlier releases. Each figure was the cost of the lever its target population required, and neither Congress could have traded instruments at will.

Q: Did the First Step Act reverse the 1994 crime bill?

No. The First Step Act repealed no title of the 1994 law and referenced none of it. The three-strikes mandatory life term, the sixty new death penalty offenses, and the mandatory-minimum floors all remained on the books after December 21, 2018. The later statute recalibrated selected federal sentencing and release gears; it dismantled nothing.

The confusion is understandable because the later law turned some of the same gears the earlier law had turned outward: it narrowed recidivist enhancements, limited the twenty-five-year firearm enhancement to true repeat offenders, widened the safety valve, and reopened crack sentences imposed under the old ratio. But contraction on a few axes alongside stasis on all the others is a partial reversal, not a repeal. The assault weapons ban needed no repeal because it had expired by its own sunset in 2004. The prison capacity built with 1994 appropriations stayed built. Naming the change precisely matters because it determines what a future Congress would still need to amend.

Q: Which reached state prisons, the 1994 crime bill or the First Step Act?

The 1994 crime bill reached state prisons; the First Step Act reached none of them. The earlier statute’s prison construction and truth-in-sentencing grants moved state sentencing policy in dozens of state capitals, with twenty-eight states and the District of Columbia qualifying for the incentive grants. The later statute applied to federal prisoners alone.

The difference follows from the constitutional boundary neither statute crossed. Congress cannot legislate a state sentence, so the 1994 act conditioned money on state choices: adopt the eighty-five percent rule, draw construction dollars. States that declined kept their own rules untouched. The First Step Act used the other lever, direct legislation for the federal system, and its text drew the line explicitly: Section 105 stated that the act did not authorize prerelease custody or supervised release for a person serving a state term. One statute reached state prisoners through money; the other reached no state prisoner at all.

Q: Was the First Step Act more bipartisan than the 1994 crime bill?

By the roll calls, yes. The First Step Act passed the Senate 87 to 12 and the House 358 to 36, with no Democrat voting no in either chamber. The 1994 statute passed 235 to 195 in the House and 61 to 38 in the Senate, with dissenters in both parties.

Both coalitions were bipartisan in composition, which is the more useful point. The 1994 law drew 46 House Republicans and 6 Senate Republicans; the 2018 law drew 182 House Republicans and 38 Senate Republicans, with all twelve Senate no votes cast by Republicans. Members overlapped across the quarter century: Chuck Schumer, Dick Durbin, and Bernie Sanders voted yes on both, while John Lewis voted no in 1994 and helped promote the 2018 bill. The margin difference reflects conditions rather than eras: violent crime was near its historic peak in 1994, and the 2018 bill’s small federal reach contained its stakes.

Q: Did the First Step Act use grants like the 1994 crime bill?

No. The First Step Act contained no grant instrument for the states at all. It had no conditional spending, no construction subsidies, and no matching arrangements. Its sentencing reforms amended federal statutes directly, and its earned time credit system was administered by the Bureau of Prisons under federal law.

The omission was deliberate, not an oversight. The 1994 statute needed grants because its target was the state system, which Congress cannot command; the Second Chance Reauthorization title of the 2018 act was its only state-facing element, and even that merely reauthorized reentry grants rather than conditioning money on sentencing policy. The 2018 statute’s authors could have attached conditions to federal grants to export their reforms to the states, as the 1994 law did in reverse, but they chose direct authority instead. That choice confined the law to the federal system and kept its cost in the hundreds of millions.

Q: Which should a student study first, the 1994 crime bill or the First Step Act?

The comparison should come first, before either statute in isolation. The two laws read side by side reveal the boundary between direct federal authority and conditional federal influence, and that boundary dictates every design choice downstream. A student who grasps the boundary first reads both statutes as solutions to one constitutional problem.

After the comparison, the recommended order follows the reader’s question. A student seeking the shape of federal power studies the 1994 grant conditions next, then the 2018 credit machinery. A drafter studies the grant mechanics before the administrative design. A researcher starts with the incarceration evidence, because the instrument determines where the evidence lives: state data for the earlier law, federal data for the later one. A journalist studies the roll calls alongside the instruments, since the votes conceal what the mechanism explains.

Q: How did the two statutes differ in their use of federal money?

The earlier statute spent money to move systems it could not command; the later statute spent money only to run machinery it owned. The 1994 law authorized 30.2 billion dollars in conditional grants to states and localities, while the 2018 law authorized 375 million dollars, mostly for Bureau of Prisons implementation. One bought influence; the other paid for administration.

The 1994 grants carried conditions that did the policy work: prison construction money required the eighty-five percent rule, and COPS hiring money arrived as a five-year match whose federal share diminished over time. Eighty percent of the 2018 authorization went to the Bureau of Prisons, and the Congressional Budget Office found the law’s good time correction alone would save 40 million dollars in its first year by shrinking the federal headcount. The earlier money was a purchase; the later money was an operating budget. That distinction explains why the two figures differ by a factor of eighty.

Q: What constitutional limit shaped both statutes the same way?

Neither statute could legislate a state sentence. The Constitution leaves the general police power, including the definition and punishment of most crimes, to the states, and the anti-commandeering doctrine bars Congress from ordering state legislatures or officers to enforce federal policy. Both laws were designed inside that boundary, choosing different instruments for different targets.

The Supreme Court’s 1987 decision in South Dakota v. Dole supplies the grant-side rule: Congress may condition federal funds on state policy choices so long as the condition serves the general welfare and stops short of coercion, because a state retains the option of declining the money. The direct side needs no such doctrine: Congress writes federal criminal law under its enumerated powers. The 1994 statute used both channels, and the 2018 statute used only the second. The boundary is therefore not a policy preference either Congress could have waived. It is the architecture inside which every federal criminal justice bill is drafted.

Q: Which parts of the 1994 crime bill survived the First Step Act?

Nearly all of it. The First Step Act repealed nothing from the 1994 law. The three-strikes mandatory life term, the sixty new death penalty offenses, and the mandatory-minimum floors all survived, and the prison capacity built with 1994 appropriations stayed built. The assault weapons ban had already expired by its own sunset in 2004.

What the 2018 statute did was amend selected gears. Section 401 narrowed the recidivist enhancement triggers and lowered their floors, Section 402 widened the safety valve the 1994 Congress had created, Section 403 confined the twenty-five-year firearm enhancement to true repeat offenders, and Section 404 reopened crack sentences imposed under the old one-hundred-to-one ratio. Everything else in the three-hundred-fifty-six-page 1994 omnibus, the policing grants, the prevention money, the registry requirements, the immigration crimes, lay outside the 2018 statute’s scope. Describing the later law as a repeal mistakes recalibration for dismantling.

Q: How did the First Step Act change the federal safety valve?

Section 402 rewrote the valve’s criminal-history criterion instead of adding a new fifth criterion. The old rule required not more than one criminal history point; the amendment replaced it with a three-part test. It also extended coverage to Maritime Drug Law Enforcement Act offenses and added use-immunity for the defendant’s truthful disclosure.

The provision matters because the valve is the principal door through a drug mandatory minimum that does not require a government motion, letting a judge sentence within the guideline range instead. The 1994 Congress had cut that door narrowly into its own regime; the 2018 Congress widened it, and a qualifying defendant also received a two-level reduction in the drug offense level. In 2024 the Supreme Court construed the amended criterion in Pulsifer v. United States, holding that a defendant must satisfy each of the three conditions to qualify. The episode shows reform operating through continuity: one Congress cut the valve, a later Congress widened it.

Q: Why did the 1994 crime bill pass on narrow margins while the First Step Act passed overwhelmingly?

The conditions differed, not the parties. In 1994 violent crime was near its peak and a midterm election approached, so the bill moved through partisan warfare and nearly died when the House defeated the rule for its conference report. By 2018 violent crime had fallen by more than half, and the federal-only bill contained the stakes for every member.

Three forces moved the later votes. The FBI’s violent crime rate had fallen from 758.2 per 100,000 in 1991 to 368.9 in 2018. State corrections spending had grown from 19 billion dollars in 1987 to 49 billion in 2007 after inflation adjustment, creating a fiscal constituency for reform. And state experiments, led by Texas justice reinvestment beginning in 2007, showed prison populations falling without crime rising to replace them. The earlier bill faced none of those tailwinds; the later bill faced all three, and its small federal footprint meant few members had to answer for changed state policy back home.

Q: Which statute changed federal death penalty law?

The 1994 statute did. Its Title VI, the Federal Death Penalty Act, created sixty new death penalty offenses under forty-one federal capital statutes, covering terrorism, murders of federal law enforcement officers, large-scale drug trafficking, fatal drive-by shootings, and fatal carjackings. The First Step Act left those offenses entirely untouched.

The survival of the capital provisions illustrates the article’s central distinction between contraction and repeal. The later statute turned several sentencing gears inward while leaving the 1994 settings on every other gear exactly as Congress had left them. Death eligibility is one of the gears that never moved: no title of the 2018 law amends the capital statutes, and the three-strikes mandatory life term survived alongside them. A reader who treats the two statutes as moral opposites will expect the later law to have erased the earlier law’s harshest provisions; the record shows it erased none of them.

Q: Did either statute let Congress command state sentencing policy?

No. Neither statute could, and neither tried. Congress lacks the authority to rewrite state sentencing codes or order state parole boards to change release standards. The 1994 act purchased state cooperation with conditional grants; the 2018 act legislated only for the federal system. The constitutional boundary held for both.

The 1994 statute is sometimes misread as a federal takeover of state punishment, but its truth-in-sentencing grants could only offer money; a state that declined the funds kept its own law untouched, and the Supreme Court’s Dole doctrine permits such conditions precisely because acceptance remains voluntary. The First Step Act made the same boundary explicit in the opposite direction: Section 105 barred the law from authorizing prerelease custody or supervised release for anyone serving a state term. Two statutes, opposite directions, one shared limit: the federal government can pay the states to change or change itself, but it cannot legislate a state sentence.

Q: How did the good time credit fix work under the First Step Act?

Section 101 amended 18 U.S.C. 3624(b) so that federal prisoners earn up to fifty-four days of good time credit for each year of the sentence imposed, not each year served. The Bureau of Prisons had been awarding roughly forty-seven days per year under a method the Supreme Court upheld in Barber v. Thomas in 2010.

The amendment restored the fifteen-percent design Congress had written into the Sentencing Reform Act of 1984, and it applied retroactively, recalculating time for prisoners already in custody. The correction alone led to the release of 3,100 people on July 19, 2019, the implementation date the law set, and the Congressional Budget Office estimated it would save 40 million dollars in its first year. Good time differs from the act’s earned time credits: good time shortens the sentence itself, while earned time credits leave the sentence intact and change where its final portion is served.

Q: What made the First Step Act’s coalition unusual?

The coalition paired organizations that agreed on nothing else. The American Civil Liberties Union and Americans for Prosperity pushed the bill together, joined by groups ranging from the Brennan Center for Justice to the American Conservative Union. Law enforcement organizations backed it alongside reform advocates.

Senate sponsors Chuck Grassley, Dick Durbin, Cory Booker, and Mike Lee built the chamber coalition, while Doug Collins, Hakeem Jeffries, and John Lewis promoted the bill in the House. The dissent was narrow and almost entirely intraparty: all twelve Senate no votes came from Republicans, and no Democrat in either chamber voted against the bill. The Business Roundtable and the U.S. Chamber of Commerce endorsed the bill as well. The breadth was possible in part because the bill’s federal-only reach contained the political stakes. Members could support sentence reductions for federal prisoners without answering for changed sentencing policy in their home states, a luxury no state-facing bill can offer.

Q: Which statute cost taxpayers more, and why does the gap exist?

The 1994 statute cost far more: 30.2 billion dollars over six years, against the First Step Act’s 375 million dollars over five years. The gap exists because the two statutes bought different things: influence over states against direct changes inside the federal system.

Conditional spending is expensive because it must compensate fifty separate sovereigns for adopting policies Congress cannot order. The 1994 figure represents six years of grant cycles, each funded, administered, and accepted anew: hiring money for local police, construction money conditioned on the eighty-five percent rule, prevention grants. The 2018 figure covers programming, risk assessment implementation, and the benefit effects of earlier releases. Neither number measures the policy’s importance. The small statute changed law directly for its population; the large one offered money and waited for states to take it.

Q: How should a drafter choose between grants and direct federal rules?

The drafter should start with the target population. If the goal is to change outcomes in state prisons, only conditional grants can reach them, and the condition must be drawn tightly enough to bite but loosely enough that states accept the money. If the goal is to change federal sentences or prison operations, direct amendment and administration are the instruments.

The 1994 statute is the case study for the grant side: the eighty-five percent rule as a funding condition, the matching requirement as a commitment device, the formula grant as the delivery mechanism, and the state’s freedom to decline as the ceiling on the whole design. The 2018 statute is the case study for the direct side: a risk and needs assessment built by statute but scored by agency, a credit formula set by law but tracked by the Bureau of Prisons, and eligibility exclusions written into the text. No drafting technique combines the two, since the authority for each comes from a different constitutional source.

Q: Where does the evidence of each statute’s effect actually live?

In different places, because the instruments differ. The 1994 statute’s grant conditions moved state sentencing policy, so its measurable effect lies in state prison admissions and time-served data rather than federal statistics. The 2018 statute operated entirely inside the federal system, so its effect appears in Bureau of Prisons release dates, supervised release transitions, and federal recidivism cohorts.

A researcher who looks for the earlier statute’s footprint in federal data, or the later statute’s footprint in state data, will find nothing and conclude nothing; the instrument determines where the evidence lives. The Government Accountability Office judged the COPS program a modest contributor to the crime declines of the 1990s, and state truth-in-sentencing adoption is tracked through Bureau of Justice Assistance qualification records. For the later law, the 3,100 releases on the good time implementation date and the crack resentencing petitions under Section 404 are federal records by definition.