How the First Step Act Shortened Federal Sentences

The guest list for the bill signing on December 21, 2018, would have seemed like a practical joke a few years earlier. In the Oval Office, President Donald Trump put his signature on a criminal justice reform bill that Democratic senators had spent years pushing, that conservative policy groups funded by some of the country's biggest libertarian donors had lobbied for, and that civil rights organizations had called modest but real. The Senate had passed it 87 to 12. The House followed with 358 votes in favor and 36 against. It was a moment when almost everyone in Washington agreed on something about prisons, and the name the authors gave the law was a deliberate signal of humility: it was the First Step, not the last.

In short: The First Step Act of 2018 shortened some federal mandatory minimum sentences, made the Fair Sentencing Act's crack cocaine reforms retroactive, fixed a calculation that had shortchanged prisoners on good-conduct time, and created earned time credits for people who complete rehabilitation programs. It applies only to federal prisoners, it excludes many offenses, and how much it helps has depended heavily on how the Bureau of Prisons carries it out.

Two bills that became one

The law grew out of two separate ideas that had trouble passing on their own.

The first was sentencing reform. For years, senators led by the Republican Chuck Grassley of Iowa and the Democrat Dick Durbin of Illinois had tried to soften federal mandatory minimums, especially for drug offenses, that had swelled the federal prison population since the 1980s. Their efforts kept stalling, partly over objections from law enforcement groups and some Republican senators who argued that lighter sentences would mean more crime.

The second was prison reform: changing what happens to people while they are locked up so they leave better prepared and less likely to return. That was an easier sell for conservatives, and the House passed a prison-focused bill in 2018, sponsored by the Georgia Republican Doug Collins and the New York Democrat Hakeem Jeffries.

Many Senate Democrats and reform groups argued the House version was too narrow without sentencing changes. The final Senate bill combined the prison-reform framework with a set of targeted sentencing changes. That compromise is what passed, and it explains the law's shape: part sentencing law, part prison management law.

The coalition behind it was unusually broad. The president's son-in-law, Jared Kushner, pushed for it inside the White House. Conservative organizations associated with the Koch network and a "Right on Crime" movement among Republicans argued that mass incarceration was a big-government failure that wasted taxpayer money. Progressive and civil rights groups, sentencing reform advocates and formerly incarcerated people pressed for broader changes but supported the final bill. A few senators, notably Arkansas's Tom Cotton, led opposition from the right, warning that early release would endanger the public.

The sentencing changes

These provisions apply to how federal judges sentence people, and some apply to people already in prison.

Making the 2010 crack reforms retroactive. In 2010 the Fair Sentencing Act reduced the infamous gap between penalties for crack and powder cocaine. But it applied only to people sentenced after it took effect. The First Step Act let people sentenced under the old crack penalties ask a judge to resentence them as if the 2010 law had been in effect. Thousands of people, overwhelmingly Black men, have had their sentences reduced through this provision.

Shorter enhanced drug minimums. Federal law imposes higher mandatory minimums when a person with certain prior convictions is convicted of a serious drug offense. The First Step Act lowered the enhanced 20-year minimum to 15 years and the mandatory life sentence for a third qualifying offense to 25 years. It also narrowed which prior convictions count, requiring a "serious drug felony" or "serious violent felony" rather than any prior drug felony.

A broader safety valve. The "safety valve" lets judges sentence some low-level drug defendants below the mandatory minimum. Before 2018 it was available only to people with almost no criminal history. The law widened eligibility somewhat. The Supreme Court later read that expansion narrowly: in Pulsifer v. United States (2024) the Court held that a defendant is disqualified if he has any one of the listed kinds of criminal history, rather than only if he has all of them.

An end to "stacking." Federal law imposes mandatory, consecutive prison terms for using or carrying a gun during a drug or violent crime. Prosecutors had been able to charge multiple counts in a single case, so that a first-time defendant could face decades of mandatory time, with the second count treated as a "second" offense carrying 25 years. The law clarified that the 25-year term applies only to people who already had a prior conviction for the offense. This change was not made retroactive, which left many people sentenced under the old practice serving terms that would be illegal today.

The prison changes: earned time and good time

The bigger daily effect for many prisoners comes from the prison-management side.

The good-conduct fix. Federal law had long promised up to 54 days of credit for each year of good behavior. The Bureau of Prisons calculated it in a way that yielded only about 47 days per year of the sentence imposed. The First Step Act clarified that the 54 days apply to each year of the sentence the judge imposed. When that change took effect in 2019, it shortened sentences for most federal prisoners at once, and thousands of people were released in a single wave.

Earned time credits. The law directed the Justice Department to create a risk and needs assessment system, which became a tool known as PATTERN, and to offer "evidence-based recidivism reduction" programs and productive activities: things like education, job training, drug treatment and cognitive behavioral programs. Eligible prisoners earn 10 days of credit for every 30 days of successful participation. Those assessed as minimum or low risk who maintain that rating earn an additional 5 days per 30.

Those credits can be used two ways. They can move a person into prerelease custody, such as a halfway house or home confinement, earlier. And up to 12 months of credits can be applied to start supervised release sooner, effectively shortening time in prison.

Who is left out. The law lists a long series of offenses that make a person ineligible to use earned time credits, including many violent crimes, terrorism offenses, many sex offenses and certain serious drug and firearm offenses. People in the highest risk categories also face limits on applying credits until their risk score drops.

The quieter provisions that matter to families

Some parts of the law got less attention but changed real lives.

It banned the use of restraints on pregnant women in federal custody, with narrow exceptions. It directed the Bureau of Prisons to place people in facilities within 500 driving miles of their primary residence where possible, making family visits more realistic. It allowed prisoners to ask a court directly for compassionate release, for example because of terminal illness, after first asking the warden and either being denied or waiting 30 days. Before 2018 only the Bureau of Prisons could bring such a motion, and it rarely did. It required help getting identification documents before release, and training in de-escalation for corrections officers.

The compassionate release change proved especially important during the COVID-19 pandemic, when prisoners with serious medical conditions used it to ask courts for release and some judges granted those requests.

Two people the law reached differently

It helps to picture how these provisions land on actual sentences. Consider two hypothetical federal prisoners, both serving time for drug offenses.

The first was sentenced in 2005 for selling crack cocaine, in an amount that triggered a ten-year mandatory minimum under the old 100-to-1 rules. Under the First Step Act, he can ask the judge who sentenced him, or her successor, to resentence him as though the 2010 thresholds applied. If the judge agrees, his mandatory minimum may fall, and the judge can also look at what he has done in prison: classes completed, jobs held, discipline records. He might go home years earlier. Nothing about this is automatic; he needs a motion, a lawyer or the patience to file one himself, and a judge willing to act.

The second was sentenced in 2016 for a powder cocaine offense and is assessed as low risk. Nothing about her sentence changes. But if she enrolls in qualifying programs, such as a drug treatment course and a vocational class, and keeps up her participation, she earns credits month by month. Those credits can move her to a halfway house or home confinement sooner, and some can be applied to begin supervised release early. Her day-to-day choices inside matter more than before.

A third person, convicted of an offense on the exclusion list, would find that most of the law passes him by, apart from the good-conduct fix and protections such as placement closer to home. That is the law's design: a set of targeted doors rather than a wide gate.

Why implementation became the real story

A law is only as good as the agency that carries it out, and the First Step Act asked a great deal of the Bureau of Prisons.

The earned time system needed programs, staff to run them, a working risk tool and software to calculate credits. Reports from the Justice Department's inspector general and the Government Accountability Office pointed to delays and problems: program shortages at some facilities, staffing gaps, and slow progress in actually applying credits. The Bureau finalized its rule on how time credits would be calculated in early 2022, years after passage, and a significant number of people became eligible for transfer or release once credits were applied.

The risk tool itself drew criticism. Researchers and advocates raised concerns that PATTERN's reliance on criminal history and other factors could rate Black prisoners as higher risk than white prisoners with similar records, limiting their access to credits. The Justice Department made revisions over time, but the question of whether algorithmic risk tools can be fair is still debated.

There is also the problem of halfway house space. If credits move someone into prerelease custody but no bed is available, the credit is less valuable than it looks on paper.

The case critics made, and what happened

Opponents in 2018 made a straightforward argument: releasing people early, or sentencing them less harshly, would mean more crime, and federal prisoners are not mostly low-level offenders. That deserves an answer rather than a dismissal.

The Justice Department's own reporting on people released under the act has generally shown rearrest rates lower than those for federal prisoners overall, though comparisons are complicated because people who qualify for credits are by design those assessed as lower risk. The fair conclusion is that the law has not produced the public safety disaster its critics predicted, and that its early-release provisions were targeted at people the system already considered likely to succeed.

The critique from the left is different and also deserves a hearing. The law is small relative to the problem. Federal prisons hold only a minority of the country's prisoners, and most people incarcerated in the United States are in state prisons and local jails beyond the reach of a federal statute. Many sentencing changes were not made retroactive. The exclusion list is long. And mandatory minimums remain on the books, still shaping plea negotiations every day.

Here is our view: both critiques point to the same conclusion. The First Step Act worked well enough, on its own limited terms, to justify a second step. It also showed that a bill can pass with a Republican president, a Republican Senate and Democratic votes when it is framed around redemption, taxpayer cost and fairness rather than partisan blame.

What a second step could include

Reform advocates and some lawmakers in both parties have proposed next moves, none of which is law as of this writing:

• Making the 2018 sentencing changes, such as the end of stacking and the reduced enhanced minimums, retroactive.
• Ending the remaining crack and powder cocaine disparity entirely, a change the House passed with strong bipartisan support in 2021 that then stalled in the Senate.
• Creating a "second look" process letting judges review long sentences after a person has served many years.
• Funding programs and staff so that earned time is a real option at every federal facility.

The new Congress that takes office in January 2027 will decide whether any of these move. Prediction markets and pundits can guess; only votes count.

What you can do

• Write to your members of Congress about the next step. Ask your representative and senators whether they support making the 2018 sentencing changes retroactive or ending the crack disparity. Our representatives page gives you their phone numbers, contact forms and recent votes.
• Watch for bipartisan bills. Criminal justice measures have a record of passing with support from both parties. The common-ground tracker lists bills that drew votes on both sides, and our Congress page follows votes and bills as they move.
• Help someone come home. Re-entry organizations help people released from federal and state prison find housing, work and identification. Find groups near you in our nonprofit directory under justice reform.
• Comment when the rules change. Federal agencies, including the Bureau of Prisons, sometimes open proposed rules for public comment. When they do, public comments become part of the official record, and our comment page lists rules open now, closing soonest first, with guidance on writing a useful comment.

The First Step Act will not be remembered as the law that ended mass incarceration. It may be remembered as the law that proved Washington could still agree that people in prison are people, and that a sentence can be both just and shorter.