outline

Mandatory Minimum Policy Reform and the Sentencing of Crack Cocaine Defendants: An Analysis of the Fair Sentencing Act David Bjerk*

The Fair Sentencing Act of 2010 (FSA) affected the U.S. federal mandatory minimum sentencing laws for crack cocaine offenders, and represented the first congressional reform of sentencing laws in over 20 years. A primary goal of this legislation was to lessen the harshness of sentences for crack cocaine offenders and decrease the sentencing gap between crack defendants and powder cocaine defendants. While the mean sentence length for crack offenders fell following the implementation of the FSA, these changes appear to primarily reflect the continuation of ongoing sentencing trends that were initiated by a variety of noncongressional reforms to federal sentencing policy that commenced around 2007. However, the FSA appears to have been helpful in allowing these trends to continue past 2010.

I. Introduction

The Anti-Drug Abuse Act of 1986 implemented mandatory minimum sentences upon

conviction for trafficking in quantities of drugs exceeding specific quantity thresholds

for each drug. Although crack cocaine is produced from powder cocaine, and therefore

is pharmacologically identical in its active ingredient, the mandatory minimum trigger-

ing quantities prescribed under the law were extremely different for the two drugs, with

the ratio of powder cocaine relative to crack cocaine needed to trigger eligibility for a

mandatory minimum being 100:1.

The reasons Congress initially cited for justifying the differential treatment of

crack cocaine relative to powder cocaine under the mandatory minimum legislation

were eventually revealed to have little merit. Moreover, the disparate treatment of crack

cocaine relative to powder cocaine also became viewed as racially biased, as over 80 per-

cent of those convicted for crack have generally been African American, while less than

33 percent of those convicted for other drugs such as powder cocaine, methamphet-

amine, and marijuana have been African American (Vagins & McCurdy 2006).

*Associate Professor, Department of Economics, Claremont McKenna College, 500 E. Ninth St., Claremont, CA 91711; email: [email protected]. Special thanks to J. J. Prescott, Eric Helland, Hani Mansour, and other members of the University of Colorado–Denver Department of Economics for their valuable input on this project.

370

Journal of Empirical Legal Studies

Volume 14, Issue 2, 370–396, June 2017

In response to this controversy, there developed strong bipartisan support to less-

en the disparate treatment of crack relative to powder cocaine under the mandatory

minimums, which eventually led to the passage of the Fair Sentencing Act of 2010

(FSA). The primary policy change inherent in the actual FSA legislation was that the

mandatory minimum triggering quantities for crack cocaine were increased roughly five-

fold while holding the mandatory minimum triggering quantities for other drugs

(including powder cocaine) constant. Thus, following the implementation of the FSA,

the ratio of powder cocaine relative to crack cocaine needed to trigger eligibility for a

mandatory minimum dropped from 100:1 to about 20:1.

Many U.S. legislators and ideologically disparate groups, including the Open Soci-

ety Policy Center, the American Humanist Association, and the National Association of

Evangelicals, heralded the FSA as a major policy change that will increase the fairness of

U.S. sentencing policy. Such praise might at first glance appear warranted, as the mean

sentence for crack defendants fell by 17 months, or almost 16 percent, in the two years

just after the passage of the FSA relative to the two years just before. However, the evi-

dence presented in this article suggests that this simple observation presents a somewhat

misleading picture the FSA’s impact on sentencing.

In particular, this article argues that the FSA should not be viewed as a mechanism

by which the U.S. Congress pushed the U.S. district courts into reducing sentences for

crack offenders (and thereby lessening the crack/powder cocaine sentencing gap), but as

a congressional assist to a variety of other reforms to the U.S. federal sentencing guide-

lines that have allowed federal judges and prosecutors to act on their preferences for

shorter sentences for crack offenders. More specifically, the U.S. Supreme Court’s deci-

sions in the Booker case and subsequent Rita, Gall, and Kimbrough cases, which made the

U.S. sentencing guidelines advisory rather than mandatory by 2007, and the U.S. Sentenc-

ing Commission’s implementation of the Crack Amendment of 2007, which altered the

U.S. sentencing guidelines for crack offenders, initiated a trend toward more leniency in

sentencing of crack offenders. The subsequent adjustments to mandatory minimum trig-

gering quantities as directed by the FSA, in conjunction with further reforms to the U.S.

Sentencing Guidelines for crack offenders that occurred concurrently, allowed this evolu-

tion toward more lenient sentencing norms for crack offenders to continue. Therefore, in

assessing whether the impact of the FSA should be considered a major policy reform, the

answer appears somewhat nuanced. While the reforms to mandatory minimums inherent

in the FSA appear to have done little in and of themselves to accelerate downward trends

in sentencing of crack offenders and the crack/powder cocaine sentencing gap, the

implementation of the FSA likely helped preclude the mandatory minimums from sub-

stantially impeding the continuance of the pre-FSA trends toward greater leniency.

II. Background on the Fair Sentencing Act of 2010

Crack cocaine first started to be manufactured and used in the early 1980s as an alterna-

tive to freebasing cocaine, or heating powder cocaine and inhaling the vapors. Crack

cocaine use expanded quickly, particularly in low-income, predominantly African-

371Crack Cocaine Defendants and the Fair Sentencing Act

American neighborhoods (Chitwood et al. 1996). With the crack trade expanding, vio-

lence and cocaine-related deaths increased (Kerr 1986; Klein et al. 1991; Grogger & Wil-

lis 2000; Fryer et al. 2013) to the point where it was being termed an epidemic.

The issue of drug addiction and the proliferation of crack became prominent

enough such that by the mid-1980s Congress passed the Anti-Drug Abuse Act of 1986.

The major policy aspect of the Anti-Drug Abuse Act was mandated minimum sentences

triggered by strict quantity thresholds that differed by drug. Table 1 shows the different

mandatory minimums and their triggering quantities associated with different drugs and

different criminal histories as dictated by the Anti-Drug Abuse Act. As can be seen, there

are two triggering quantity thresholds for each drug, where the higher one corresponds

to a higher mandatory minimum (unless the defendant has been convicted for two pre-

vious drug felonies and there was bodily injury associated with the current conviction).

Notably, these mandatory minimum triggering quantities were far lower for crack

cocaine than for all other major drugs. Possibly most controversially, the ratios between

the mandatory minimum triggering quantities for powder cocaine relative to crack

cocaine were 100:1.1

This disparate treatment of crack cocaine relative to powder cocaine was increas-

ingly criticized as being racially biased, as crack cocaine defendants were far more likely

to be black than white. Indeed, in the year 2000, the U.S. Sentencing Commission

reported that over 80 percent of federal crack cocaine offenders were black, a rate far

above the analogue for other drugs (USSC 2002).

At the urging of several civil rights groups, including the American Civil Liberties

Union and the Sentencing Project, by the early 2000s, both the U.S. Sentencing

Table 1: Federal Mandatory Minimum Drug Sentences, by Drug and Quantity, Pre-FSA

1st Drug Offense 2nd Drug Offense 3rd Drug Offense

Marijuana

100 kg or more 5 yrs 10 yrs 10 yrs 1,000 kg or more 10 yrs 20 yrs Life

Methamphetamine

0.05 kg or more 5 yrs 10 yrs 10 yrs 0.5 kg or more 10 yrs 20 yrs Life

Heroin

0.1 kg or more 5 yrs 10 yrs 10 yrs 1 kg or more 10 yrs 20 yrs Life

Powder Cocaine

0.5 kg or more 5 yrs 10 yrs 10 yrs 5 kg or more 10 yrs 20 yrs Life

Crack Cocaine

0.005 kg or more 5 yrs 10 yrs 10 yrs 0.05 kg or more 10 yrs 20 yrs Life

1For a good review of the background on reforming the Anti-Drug Abuse Act of 1986, see Beaver (2010), as well as Reinarman et al. (1997) and Vagins and McCurdy (2006).

372 Bjerk

Commission and a bipartisan coalition of congressmen voiced support for decreasing or

even eliminating the differential treatment of crack cocaine relative to powder cocaine.

Even with such support, any such legislation faced opposition and numerous hurdles.

The Fair Sentencing Act (FSA) was eventually introduced as a compromise plan, and

signed into law August 3, 2010. The primary policy reform associated with the Act was

to change the mandatory minimum triggering quantities for crack cocaine. In particu-

lar, while the FSA did not fully eradicate the disparate treatment of crack cocaine rela-

tive to powder cocaine, it decreased the 100:1 ratio of mandatory minimum triggering

quantities to a roughly 20:1 ratio. Table 2 shows the key details associated with the FSA

legislation. For crack cocaine offenders, the lower mandatory minimum triggering quan-

tity rose from 0.005 kilograms to 0.028 kilograms, and the upper mandatory minimum

triggering quantity rose from 0.05 kilograms to 0.28 kilograms. The FSA did not alter

the mandatory minimum triggering quantities for other drugs.

Up to and after the implementation of the FSA, individuals and groups from

across the political spectrum hailed the Act as a substantial policy reform aimed at miti-

gating the crack cocaine/powder cocaine sentencing disparity, and in conjunction less-

en the racial disparity of the federal justice system. For example, on the political left,

Patrick Leahy, a Democratic U.S. Senator, said: “After more than 20 years, the Senate

has finally acted on legislation to correct the crack-powder disparity and the harm to

public confidence in our justice system it created.” Similarly, Roy Speckhardt, Executive

Director of the American Humanist Association, said “[t]his corrects a historical injus-

tice within our legal system,” and Nkechi Taifa, a senior policy analyst at the Open Soci-

ety Policy Center, stated: “This victory in drug sentencing reform is extraordinary;

advocates have been fighting for nearly two decades to eliminate the egregious disparity

between crack and powder cocaine. We have significantly ‘cracked the disparity’ with

unprecedented bipartisan support and will continue the critical work to achieve a fair

and just criminal justice system.” These sentiments are not only relegated to the political

left. For example, Galen Carey, Director of Government Affairs at the National Associa-

tion of Evangelicals, remarked: “The legislation makes significant progress toward parity

in criminal penalties for possession and use of crack and powder cocaine.”

As will be seen in more detail below, however, in evaluating the impact of the

FSA, it is important not to view the FSA in isolation, but in the broader context of a

variety of sentencing reforms not specifically related to mandatory minimums that were

Table 2: Changes in Mandatory Minimum Cut-Off Quantities for Crack Cocaine, Pre-

and Post-FSA

1st Drug Offense 2nd Drug Offense 3rd Drug Offense

Pre-FSA

0.005 kg or more 5 yrs 10 yrs 10 yrs 0.05 kg or more 10 yrs 20 yrs Life

Post-FSA

0.028 kg or more 5 yrs 10 yrs 10 yrs 0.28 kg or more 10 yrs 20 yrs Life

373Crack Cocaine Defendants and the Fair Sentencing Act

occurring concurrently and even well before the implementation of the FSA. Important-

ly, while the FSA itself only affected the quantity cut-off thresholds for mandatory mini-

mums for crack cocaine, the U.S. Sentencing Commission (USSC) responded to the

FSA in November 2010 by altering the entire sentencing guideline structure for crack

cocaine to make it consistent with the new mandatory minimum cutoffs. To put it

another way, while the FSA itself only affected sentencing for crack offenders convicted

for quantities between 0.005 kilograms and 0.028 kilograms (who now faced no manda-

tory minimum) and crack offenders convicted for quantities between 0.05 kilograms

and 0.28 kilograms (who now faced a lower mandatory minimum), the changes to the

sentencing guidelines ostensibly applied to all crack offenders.

Moreover, the whole of the USSC sentencing guidelines underwent some quite radi-

cal changes a few years prior to passage of the FSA. The guidelines are determined not by

Congress, but by the U.S. Sentencing Commission, and map each quantity of each drug

into an “offense severity” score, which in conjunction with a criminal history score, map

into a cell in a sentencing grid that is associated with a relatively narrow sentence range.

Initially, the USSC guidelines were mandatory, but a series of U.S. Supreme Court deci-

sions, starting with the Booker decision of 2005, and maybe most notably the Kimbrough deci-

sion in late 2007, substantially weakened the role for the guidelines. In particular, the

Kimbrough decision held that a judge could depart from the guidelines if he/she disagreed

with the policy choices underlying the guideline sentence—most notably the disparate

treatment of crack cocaine relative to powder cocaine. In other words, the guidelines

changed from being mandatory to advisory. Moreover, also in late 2007, the USSC lowered

the “offense severity” score for all crack cocaine offenders in the sentencing guidelines,

effectively lessening the recommended guideline sentence for all crack offenders (this

change to the guidelines is often referred to as the 2007 Crack Amendment). As I discuss

in more detail below, these changes in the structure and force of the USSC guidelines

appear to have had large implications on sentencing, particularly for crack offenders.

To my knowledge, there have been no formal analyses of the impact of the FSA

on sentence lengths for crack defendants or on the sentencing gap between crack

cocaine defendants and powder cocaine defendants. However, there is a diverse litera-

ture looking at mandatory minimum sentencing laws more broadly. For example,

LaCasse and Payne (1999), Bjerk (2005), Kurlycheck et al. (2007), Ulmer et al. (2007),

and Starr and Rehavi(2013) look at how mandatory minimums impact pretrial bargain-

ing. Helland and Tabarrok (2007), Shepherd (2002), Marvell and Moody (2001), and

Loftin et al. (1992) examine the extent to which mandatory minimums may affect the

behavior of potential criminals. Loftin et al. (1983) look at both of these issues with

respect to a state mandatory minimum law for gun crimes. Somewhat relatedly, Fisch-

man and Schanzenbach (2012) argue that mandatory minimums may have played a role

in increasing racial disparities in sentencing following the U.S. Supreme Court’s deci-

sions that declared the sentencing guidelines to be advisory rather than mandatory.2

2See Tonry (1992) for a detailed discussion of many of the issues that have been highlighted regarding mandato- ry minimums.

374 Bjerk

Studies that focus explicitly on mandatory minimums for drug crimes are fewer,

but still encompass a wide spectrum of analyses. For example, Caulkins et al. (1997)

conduct a cost/benefit analysis of the mandatory minimums for cocaine, Kautt and

Spohn (2002) look at whether defendants of different races or genders are more or less

subject to federal drug mandatory minimum sentences, and Hartley et al. (2007) look at

how prosecutors apply “substantial assistance” to circumvent guidelines and mandatory

minimums for crack and powder cocaine cases.

Some of the most comprehensive work looking at federal mandatory minimum

sentences for drug crimes has been done by the U.S. Sentencing Commission (USSC

2011). Using federal sentencing data from the 2010 fiscal year (a subset of the data

used in the analysis below), the USSC finds that mandatory minimum eligible federal

drug defendants appear to receive sentences consistent with the mandated minimums

less than half the time, primarily due to the federal “safety-valve” statute that allows

defendants meeting certain criteria (notably a minimal criminal history) to avoid a man-

datory minimum. The USSC report also finds that eligible black defendants are more

likely to receive a sentence consistent with the mandated sentence than are white or His-

panic defendants. While the USSC study suggests that part of the reason black defend-

ants were more likely than defendants of other races to receive a sentence consistent

with the mandatory minimum might be because they were more likely to have substan-

tial criminal records, making them ineligible for safety-valve relief, they did not explicitly

analyze this issue.

III. Methodology

The goal of the first part of this article is the estimate the direct impact of the FSA legis-

lation on the expected sentence length for crack offenders. The “na€ıve” approach would

be to just compare the expected sentence for crack offenders sentenced after the FSA

took effect to those sentenced prior to the FSA (i.e., 2010 and earlier), or if we let d represent the impact of the FSA, this would correspond to d 5 E[Scrack|post-FSA] – E[Scrack|pre-FSA], where Scrack refers to sentences for crack offenders and E[] is the

mathematical expectation operator.

However, there are a couple of key issues that will cause the na€ıve approach men-

tioned above to potentially dramatically overstate d, or overstate the direct impact of the FSA. First, there might be other trends impacting changes in expected sentence for

crack offenders beyond the FSA. For example, one might be concerned that the charac-

teristics of crack defendants are changing over time (most notably their criminal histo-

ries and/or their conviction quantities), and these compositional changes among crack

defendants are impacting average sentence lengths for crack. Similarly, there may be

other policies and reforms (such as those highlighted in the previous section) that may

have been affecting sentences for crack offenders over time. Given the possibility of

such other trends, E[Scrack| post-FSA] – E[Scrack| pre-FSA] will be a biased estimate of d, or, in other words, not capture just the impact of the FSA, as it will conflate the impact

of the FSA with these other trends. Therefore, to assess the impact of the FSA, not only

375Crack Cocaine Defendants and the Fair Sentencing Act

do we have to control for any changes in crack defendant characteristics over time, but

also, rather than evaluate whether there was a change in expected sentences for crack

offenders following the implementation of the FSA, we need to evaluate whether there

was a change in the trend in the expected sentencing for crack offenders following the

implementation of the FSA.

To control for changes in crack defendant characteristics and test for a break in

the sentencing trend for crack offenders following the FSA, I estimate a regression of

the following form for all crack defendants:

Si 5 a 1 b1 � t1 d � postFSAi 1 UXi 1 gD1ei; (1)

where Si, corresponds to the sentence given to a defendant i, t is a linear monthly time

trend, and postFSAi is a dummy variable equal to 1 if the defendant was sentenced after

the FSA became effective. To account for any changes in defendant characteristics over

time, the vector Xi includes race and gender dummies, criminal history dummies, and

quantity category dummies (corresponding to the cells in USSC sentencing grid severity

scores), while D is a vector of district fixed effects to control for the possibility that

more lenient districts may have received a larger or smaller fraction of crack cases over

time. The final term, ei, is the residual error term. 3 The parameter of interest in Equa-

tion (1) is d, which measures the extent to which there was a discrete shift in the trend of expected sentences for crack offenders in the time following the implementation of

the FSA.

Given that first-time offenders can often avoid a mandatory minimum via the

“safety-valve” designation, which is available only to those with minimal criminal histo-

ries, one might think that the FSA had very heterogeneous impacts on non-first-time

offenders relative to first-time offenders. Therefore, I also estimate a version of Equation

(1) where I put in separate time trends for first-time offenders and repeat offenders,

and put in separate postFSA dummies for first-time offenders and repeat offenders.

There is a second reason why the na€ıve approach discussed at the outset of this

section might be a biased assessment of the FSA. Namely, as alluded to in the previous

section, the changes to crack mandatory minimums inherent in the FSA were not the

only changes to sentencing policy in the fall of 2010, as just after the implementation of

the FSA, the USSC also altered the guidelines for all crack offenders to make them con-

sistent with the revised mandatory minimum cutoffs specified by the FSA. This means

that to the extent expected sentences for crack offenders did fall following the imple-

mentation of the FSA (at least relative to pre-FSA trends), some of this might not be

due to the FSA changes to mandatory minimums, but due to the changes in the sen-

tencing guidelines.

To consider this issue, I examine the degree to which expected sentences for

crack offenders fell for those whose sentences should have been directly impacted by

the reforms to mandatory minimums inherent in the FSA relative to those whose

3I also use natural log of sentence length as the dependent variable as a robustness check.

376 Bjerk

sentences should not have been impacted by the reforms to the mandatory minimums

(but may well have been impacted by the changes to the sentencing guidelines). In par-

ticular, as discussed above, the FSA changed the mandatory minimum sentences only

for crack defendants convicted for quantities between 0.005 kilograms to 0.028 kilo-

grams and those convicted for quantities between 0.05 kilograms and 0.28 kilograms

(hereafter FSA-affected crack quantities). Sentences for crack defendants convicted for oth-

er quantities may have been affected by the changes to the sentencing guidelines, but

should not have been impacted by the changes to mandatory minimums due to the

FSA. Therefore, I look at how sentences changed for offenders convicted for FSA-affected

crack quantities surrounding the implementation of the FSA relative to the analogous

changes in sentencing for those crack offenders convicted for quantities ostensibly unaf-

fected by the FSA. The extent to which expected sentences changed differentially for

offenders convicted for FSA-affected crack quantities relative to offenders convicted for

quantities unaffected by the FSA in the time surrounding the implementation of the

FSA will also provide a measure of the direct impact of just the changes to the mandato-

ry minimums as directed by the FSA, relative to other policies affecting crack

sentencing.

IV. Description of the Data

The data used for this article come from the annual Monitoring of Federal Criminal

Sentences datasets. The primary analyses that follow are based on the 2009 through

2012 versions (though later I also look at data from versions going all the way back to

2002). These data are collected and organized by the U.S. Sentencing Commission, and

contain information on all cases sentenced under the federal court system. These data-

sets contain a wealth of defendant information, including conviction charge—and for

drug convictions, conviction drug type and quantity—as well as adjudication district and

demographic and criminal history category for each defendant. Moreover, sentencing

information is provided.

From these data, I extract only cases where the defendant was charged with only

one type of drug, where that drug is crack cocaine, powder cocaine, or methamphet-

amine. I also limit the sample to only those with valid data on drug amount and sen-

tence length, and exclude cases in non-U.S. districts (e.g., Puerto Rico, Guam, and the

Virgin Islands).4

The initial analyses in this article will use data covering the 2009 to 2012 fiscal

years. This initial time period of 2009–2012 contains 11,479 defendants convicted for

crack who are included in the analysis. However, data going back to 2002 will be used

subsequently to analyze earlier policy changes to crack cocaine sentencing.

The Fair Sentencing Act (FSA) was made effective August 3, 2010. Therefore, for

ease of reference throughout the article, a year will be counted as going from August 1

of the previous calendar year to July 31. So, for example, when I refer to the year 2010, I

4Including observations without valid drug amounts generally has no appreciable impact on the results.

377Crack Cocaine Defendants and the Fair Sentencing Act

am referring to the time period going from August 1 of 2009 to July 31 of 2010. This is

done so that all months in and prior to year 2010 are prior to the FSA being made effec-

tive, and all months in and after year 2011 are after the FSA was made effective.

Table 3 shows basic pre-FSA summary statistics by drug (i.e., for defendants con-

victed in the years 2009 and 2010). As can be seen, crack defendants on average received

substantially longer sentences when compared to those convicted for powder cocaine.

Table 3 also shows that while over 80 percent of crack defendants are black, less than

30 percent of defendants convicted for powder cocaine are black. Finally, as can be seen

in the last row of Table 3, a much higher fraction of crack defendants had previous felo-

ny convictions than those defendants convicted for other drugs. Hence, as has been

highlighted before, the sentencing of crack defendants is an issue that by and large is

something that primarily impacts black Americans and repeat offenders.

V. Evaluating the Impact of the Fair Sentencing Act

on Average Sentence Lengths for Crack Defendants

Figure 1 shows a na€ıve assessment of the impact of the FSA by simply looking at how

the mean sentence for crack defendants changed between the 24 months prior to the

Table 3: Summary Characteristics by Drug Type (Pre-FSA)

Crack Powder Cocaine Meth

Mean sentence 109 mo 80 mo 97 mo Fraction black 0.83 0.26 0.02 Fraction Hispanic 0.07 0.56 0.30 Fraction “white” 0.10 0.18 0.67 Previous felony conviction 0.80 0.39 0.55

Figure 1: Mean sentences for crack defendants pre- vs. post-FSA.

NOTES: “Pre-FSA” time period spans August 1, 2008--July 31, 2010. “Post-FSA” time period spans August 1, 2010-- July 31, 2012.

378 Bjerk

FSA and the 24 months following the implementation of the FSA. As can be seen, the

mean sentence fell by 17 months, or about 16 percent.

However, as discussed above, this simple pre- to post-FSA difference in mean sen-

tence length for crack defendants (with relatively large time blocks on either side of the

passage of the FSA) may be a problematic assessment of the direct effect of the FSA on

sentence lengths for crack defendants. The top line in Figure 2 shows the time trend in

mean sentence length for crack defendants, from the years 2009 to 2012. As stated earli-

er, each year refers to time periods going from August 1 to July 31 (e.g., year 2010 here

refers to August 1, 2009 to July 31, 2010), meaning the FSA effective date of August 3,

2010 means it became effective just after the year 2010 (vertical line on Figure 2). As

can be seen quite clearly, the mean sentence for crack defendants was already falling

well prior to the implementation of the FSA, and at least at this aggregated level, there

does not appear to be any obvious change or jump in the trend following the imple-

mentation of the FSA.

As a point of comparison, Figure 2 also shows the trends in the sentencing for

powder cocaine offenders, which should not have been directly affected by the FSA. As

can be seen, the mean sentence length for cocaine offenders fell a small amount follow-

ing the implementation of the FSA. This at least suggests that there were no countervail-

ing sentencing policy changes that were pushing up sentences for drug crimes as a

whole during the time the FSA was implemented. Interestingly, the trend in mean

cocaine sentences was actually quite flat prior to the implementation of the FSA, but

started a downward trend right around the implementation of the FSA. Therefore, one

thing this picture makes clear is that even though cocaine sentences should not have

been impacted by the FSA, it would not be appropriate to do a difference-in-difference

estimate using the trends in cocaine sentencing as the “missing counterfactual” trend,

since the parallel pretreatment trends requirement for such a procedure is clearly not

met (the same is true for methamphetamine cases).

Figure 2: Mean sentence length by drug (August 2008--August 2012).

NOTES: The FSA was made effective August 3, 2010. So for ease of reference, each “year” is counted as running from August to July (e.g., “2010” runs from August 2009 to July 2010), meaning all of “year 2010” was before FSA took effect.

379Crack Cocaine Defendants and the Fair Sentencing Act

Table 4 shows the results of several OLS regression analyses corresponding to

Equation (1) for crack cases from years 2009 to 2011. Specification (1) controls for only

a linear monthly time trend and a post-FSA dummy. Specification (2) further controls

for defendant characteristics including race, gender, drug quantity, and criminal history

via dummy variables. Recognizing that the FSA is essentially being imposed on judicial

districts (as it is the combination of the judges and prosecutors in a district who essen-

tially apply their interpretation of federal sentencing policy), and districts may differ not

only in their general sentencing severity but also in their variance in sentencing policy,

specification (3) further controls for district fixed effects and uses two-way clustered

standard errors by district and time (monthly).5 As can be seen, in all of the specifica-

tions (1)–(3) the coefficient on the linear time trend is always significantly negative,

Table 4: OLS Analysis for Trend Break Following Fair Sentencing Act of 2010

(Dependent Variable: Sentencing Length in Months)

Specification

(1) (2) (3) (4) (5)

Time trend (monthly) 20.77*** 20.83*** 20.80*** (0.123) (0.093) (0.112)

Coefficient on post-FSA indicator 2.56 1.51 1.54 (3.367) (2.559) (2.740)

[–4.0, 9.2] [–3.5, 6.5] [–3.8, 6.9] Post-year 2009 indicator 29.82***

(2.350) Post-year 2010 indicator 27.39***

(2.220) Post-year 2011 indicator 210.67***

(2.024) Coefficient on post-FSA

indicator (crim hist cat 5 1 only) 1.32

(3.706) [–5.9, 8.6]

Ttime trend (crim hist cat 5 1 only) 20.53*** (0.162)

Coefficient on post-FSA indicator (crim hist cat�2 only)

1.63 (3.073)

[–4.4, 7.6] Time trend (crim hist cat�2 only) 20.86***

(0.118) Race and gender indicators No Yes Yes Yes Yes Criminal history category indicators No Yes Yes Yes Yes Quantity (by guideline cell) No Yes Yes Yes Yes Two-way clustered standard errors

by district and month No No Yes Yes Yes

NOTES: Crack cases only. Data run from August 1, 2008 to July 31, 2012. Standard errors in parentheses, brackets show 95 percent confidence interval. ***indicates significance at 1 percent level.

5Thanks to Colin Cameron for making his two-way clustered standard errors code available.

380 Bjerk

suggesting expected sentences have been falling by about 0.8 months per month, or a

little over 9 months per year. However, the coefficient on the post-FSA indicator variable

is never significantly different than zero (and positive in sign). Indeed, even if one looks

at the 95 percent confidence intervals surrounding the point estimates on the post-FSA

coefficients (in brackets below standard errors), there is very little evidence of a notable

change in trend following the implementation of the FSA. For example, if we take the

lower bound on the 95 percent confidence interval in specification (3) (what I would

argue is the preferred specification), the strongest interpretation of the results suggests

that the FSA could have pushed sentences for crack offenders down by a little under 4

months relative to the pre-FSA trend. Given the mean sentence for crack offenders pre-

FSA was 109 months, even this would be a pretty small shift.

As a robustness check, rather than regressing sentence length on a linear time

trend and post-FSA indicator, specification (4) of Table 4 regresses sentence length on

indicators for post-2009, post-2010, and post-2011 (along with race and gender indica-

tors, drug quantity, criminal history indicators, and district fixed effects). If there is a

notable acceleration in the time trend following the implementation of the FSA, then

the coefficients on the post-2010 and/or post-2011 indicators should be significantly

more negative than the coefficient on the post-2009 indicator. As can be seen, the coef-

ficients on all three of these indicators are quite similar (indeed a joint chi-squared test

cannot reject the hypothesis that they are all equal at any standard level of significance),

all suggesting that expected sentences for crack offenders fall roughly 8–10 months per

year between the year 2009 and the year 2012 (which is consistent with the coefficient on

the time trend in specification (3)).

As discussed previously, one might think that the FSA had a larger impact on

crack defendants with previous criminal histories, as prior to the FSA a large fraction of

mandatory minimum eligible first-time crack offenders were able to avoid the mandatory

minimums due to their eligibility for the federal “safety-valve” relief (USSC 2011). Given

this, specification (5) of Table 4 puts in separate time trends and post-FSA indicator var-

iables for first-time offenders (criminal history category 1) and repeat offenders (crimi-

nal history category 21). As can be seen, while again the coefficients on both of the

time trend variables are negative and statistically significant, the coefficients on the two

post-FSA dummies provide very little evidence that sentencing trends changed for either

of these two groups following the implementation of the FSA.

Given that distribution of sentence lengths is actually quite right skewed, simply

using raw sentence length as the outcome variable in the regression specifications might

not be the best fitting specification. So, Table 5 shows the results of specifications identi-

cal to those in Table 4, but using natural log of sentence length as the outcome vari-

able. As can again be seen, the results are essentially the same as in Table 4, again

suggesting that sentencing trends for crack offenders did not change following the

implementation of the FSA.

As discussed in Section III, another way to consider the direct impact of the

changes to mandatory minimums dictated by the FSA is to examine whether sentences

changed more in quantity categories directly impacted by the FSA relative to quantity

categories not directly impacted by the FSA. To the extent to which raising the

381Crack Cocaine Defendants and the Fair Sentencing Act

mandatory minimum quantity thresholds as dictated by the FSA was the primary driver

of any changes in sentence lengths around the passage of the FSA, then we should see

larger declines in sentence lengths among defendants convicted for crack quantities

that were directly impacted by the FSA than among defendants convicted for quantities

not directly impacted by the FSA.

To look at this, Table 6 shows simple changes in mean sentence length for crack

offenders in the time following the passage of the FSA relative to the time just prior to

its passage across seven different quantity categories (to ensure ample number of cases

in each cell, the pre-FSA period consists of the two years prior to the implementation of

the FSA and the post-FSA consists of the two years following the implementation of the

FSA). As alluded to above, the FSA itself only impacted defendants convicted for crack

quantities between 0.005 kilograms to 0.028 kilograms (subject to a mandatory mini-

mum before the FSA but not after) and those convicted for crack quantities between

0.05 kilograms and 0.28 kilograms (subject to a lower mandatory minimum after the

Table 5: OLS Analysis for Trend Break Following Fair Sentencing Act of 2010

(Dependent Variable: Natural Log of Sentencing Length)

Specification

(1) (2) (3) (4) (5)

Time trend (monthly) 20.01*** 20.01*** 20.01*** (0.002) (0.001) (0.002)

Coefficient on post-FSA indicator 0.02 0.00 0.01 (0.046) (0.035) (0.048)

[–0.07, 0.11] [–0.07, 0.07] [–0.08, 0.11] Post-year 2009 indicator 20.17***

(0.034) Post-year 2010 indicator 20.12***

(0.035) Post-year 2011 indicator 20.12***

(0.031) Coefficient on post-FSA

indicator (crim hist cat 5 1 only) 0.07

(0.126) [–0.18, 0.31]

Time trend (crim hist cat 5 1 only) 20.02*** (0.005)

Coefficient on post-FSA indicator (crim hist cat�2 only)

0.00 (0.035)

[–0.07, 0.07] Time trend (crim hist cat�2 only) 20.01***

(0.001) Race and gender indicators No Yes Yes Yes Yes Criminal history category indicators No Yes Yes Yes Yes Quantity (by guideline cell) No Yes Yes Yes Yes Two-way clustered standard errors

by district and month No No Yes Yes Yes

NOTES: Crack cases only. Data runs from August 1, 2008 to July 31, 2012. Dependent variable equals 0 if no prison sentence. Standard errors in parentheses, brackets show 95 percent confidence interval.

382 Bjerk

passage of the FSA than before). Any mandatory minimums associated with the other

quantity categories were unaffected by the FSA. However, as also discussed above, the

sentencing guidelines for all the crack categories also changed just after the implemen-

tation of the FSA. Notably, though, by the time the FSA was implemented in 2010, the

guidelines were advisory rather than mandatory.

As can be seen in Table 6, mean sentences fell significantly for all crack defend-

ants other than those convicted for quantities less than 0.005 kilograms. Most notably,

both in raw terms and percentage terms, the fall in average sentence length for those in

FSA-affected quantity categories were roughly in line with the fall in average sentence

lengths of those in quantity categories that were not directly affected by the FSA. Again,

these findings suggest that much of the changes in sentencing of crack offenders

around the passage of the FSA was not necessarily due to the changes in the mandatory

minimums as dictated by the FSA.

Overall, the analyses in this section suggest that the FSA per se does not appear to

be wholly responsible for the recent declines in the mean sentence for crack defend-

ants, as sentences for crack offenders convicted for quantities that were ostensibly not

directly affected by the FSA changes to the mandatory minimums saw similar declines as

those that were, and maybe even more notably, there was a downward trend in sentenc-

ing for crack offenders starting well before the implementation of the FSA. This leaves a

couple of questions. First, if not the FSA, what precipitated the downward trend in sen-

tencing for crack offenders? Second, while the FSA may not have been the proximal

cause of the downward trend in crack sentencing, what would have happened in its

absence? In particular, to what extent would the downward trends in sentencing for

Table 6: Pre- vs. Post-FSA Changes in Mean Sentence Length, by Quantity, Crack Cases

Only (2009–2012)

Mean Sentence (Months)

Quantity (kg) Affected by FSA? Pre-FSA Post-FSA Raw Difference Percent Difference

< 0.005 No 54.98 50.62 24.36 27.9 (2.558) (3.163) (4.068) (7.173)

0.005–0.028 Yes 81.57 67.78 213.78*** 217*** (1.478) (1.991) (2.480) (2.868)

0.028–0.05 No 94.49 79.98 214.51*** 215*** (2.986) (2.780) (4.080) (3.976)

0.05–0.28 Yes 120.69 97.93 222.76*** 219*** (1.782) (1.775) (2.515) (1.896)

0.28–1 No 144.81 126.84 217.97*** 212*** (4.488) (4.171) (6.127) (3.958)

1–3 No 176.00 140.49 235.50*** 220*** (6.913) (6.538) (9.515) (4.861)

> 3 No 209.14 168.75 240.39*** 219*** (7.838) (7.171) (10.623) (4.572)

NOTES: “Pre-FSA” spans August 1, 2008–July 31, 2010. “Post-FSA” spans August 1, 2010–July 31, 2012. *** indicates significance at 1 percent level.

383Crack Cocaine Defendants and the Fair Sentencing Act

crack offenders have been impeded were it not for the reforms to the crack mandatory

minimums inherent in the FSA? The following sections consider these questions in

more detail.

VI. So why have Crack Sentences been Falling

so Much?

As discussed above, while the FSA is the only congressional sentencing reform to occur

for over 20 years, there have been a variety of other reforms to the U.S. sentencing guide-

lines starting in the mid-2000s. With the Booker decision in 2005, the Supreme Court held

that the U.S. sentencing guidelines could no longer be deemed mandatory, but only advi-

sory, with sentences being reviewed for “reasonableness.” This still left considerable ambi-

guity regarding the role of the guidelines. The further decisions coming out of the Rita,

Gall, and Kimbrough cases in 2007 clarified the role of the guidelines and indeed substan-

tially weakened the role for the guidelines. In particular, the Kimbrough decision held that

a judge could depart from the guidelines if he/she disagreed with the policy choices

underlying the guideline sentence—most notably the disparate treatment of crack cocaine

relative to powder cocaine. It is also important to note that the Booker, Rita, Gall, and Kim-

brough decisions had no impact on the mandatory minimum sentencing statutes.

In addition to the Kimbrough decision in late 2007, the U.S. Sentencing Commis-

sion also altered the sentencing guidelines. Specifically, the USSC sentencing guidelines

assign a sentencing range to a convicted offender based on a criminal history score and

what is termed a “final offense level,” which is a numeric score meant to capture the

seriousness of the crime. Any quantity of a given drug corresponds to a particular

offense level. In November 2007, the USSC amended the guidelines to reduce by two

offense levels the base offense level assigned to each crack quantity threshold, a policy

change sometimes referred to as the 2007 Crack Amendment. This effectively lowered

the guideline sentence range associated with any crack cocaine conviction. As stated

above, while these U.S. sentencing guidelines were no longer mandatory at the time of

this change, they were still advisory.

To get a sense of the impact of these reforms to the U.S. sentencing guidelines

on sentencing, I use the Monitoring of Federal Criminal Sentences datasets spanning all

the way back to 2002 to extend Figure 2 backward. Figure 3 shows this extended trend

in mean sentence lengths for crack and powder cocaine from year 2002 to year 2012.

There are four vertical lines shown in Figure 3. The dotted line simply shows the begin-

ning date used in Figure 2 (August 2008 or 24 months prior to the implementation of

the FSA). The solid lines show when significant changes to sentencing policy occurred

over this time period. The vertical line furthest to the right indicates the implementa-

tion of the FSA. The vertical line all the way to the left indicates the Booker decision by

the U.S. Supreme Court, and the middle solid vertical line indicates the Supreme Court

decision regarding the Kimbrough case as well as the Crack Amendment to the USSC sen-

tencing guidelines. The vertical line for the FSA is placed between years 2010 and 2011

for the reasons discussed above (i.e., the year 2010 is defined to go from August 2009 to

384 Bjerk

July 2010, with the FSA being implemented August 3, 2010). By contrast, the vertical

line is placed mid-year 2005 for the Booker case because it was decided in January 2005,

halfway between August 2004 and July 2005 (i.e., the year 2005), meaning some of the

year 2005 cases were decided before Booker while others after. The vertical line for Kim-

brough cases and the Crack Amendment is again placed between year 2007 and year 2008,

as these events occurred during the fall of 2007, meaning they would primarily only

apply to cases after year 2007 as defined here (i.e., August 2006 to July 2007).

As can be seen in Figure 3, the darker line shows that, other than a slight uptick

in 2009, the mean sentence length for crack has been falling from 2007—the year of

the Kimbrough decision and the Crack Amendment. By contrast, the lighter gray line

shows that the mean sentence length for powder cocaine was roughly constant from

2005 until a slight downtick between 2010 and 2011 (the time when the FSA was imple-

mented). One interpretation of this graph is that the changes to the force and structure

of the USSC sentencing guidelines that occurred in late 2007 allowed judges and prose-

cutors to start acting on their desires to lessen sentence lengths for crack offenders.

To look at this more formally, analogous to the regression specification (4) in

Table 4, I regress sentence length on postyear indicators for each year and other defen-

dant characteristics, or

Si 5a1 X2012

j 52003

dj �post-Yearj 1UXi 1gD1ei; (2)

where again Si, corresponds to the sentence given to a defendant i, and postYearj is a

dummy variable equal to 1 if the defendant was sentenced in Yearj or after (for years

Figure 3: Mean sentence length by drug (August 2001--August 2012).

385Crack Cocaine Defendants and the Fair Sentencing Act

2003–2012). To again account for any changes in defendant characteristics over time,

the vector Xi includes race and gender dummies, criminal history dummies, and quanti-

ty category dummies, while D is a vector of district fixed effects to control for the possi-

bility that more lenient districts may have received a larger or smaller fraction of crack

cases over time. The final term, ei, is the residual error term. Like before, to test for a change in trend happening between year j and j – 1, we can examine whether the coeffi-

cient on the postYearj (i.e., dj) is significantly different from the coefficient on postYearj-1 (i.e., dj-1). Intuitively, if mean sentences are constant over a given time period, then the djs should equal zero over that time period. By contrast, if there is a constant downward trend over a given time period, meaning mean sentences are falling by n months each

year over that time period, then the djs should roughly equal –n over that time period. If there is a change in trend between year j and j – 1, we should see a change in dj rela- tive to dj-1.

Table 7 shows the results from estimating Equation (2) over a few different

specifications for crack offenders. Specification (1) just regresses sentence length on

all of the postYearj indicators, specification (2) further controls for defendant race,

gender, and criminal history (i.e., Xi in Equation (2)), and specification (3) further

controls for district fixed effects (i.e., D in Equation (2)). Specification (3) also uses

two-way clustered standard errors by district and time (monthly). The stars indicate

the level of significance with respect to a chi-squared test of the null hypothesis that

the coefficient on the postYearj indicator equals the coefficient on the postYearj-1 indicator.

As can be seen in Table 7, the coefficient on the postYear2008 indicator is negative

and large in magnitude across all specifications, and a chi-squared test rejects the equali-

ty of the coefficients on the postYear2008 and postYear2007 indicators. This suggests that

there was a substantial downward trend break happening in the time between year 2007

and year 2008, or right after the Kimbrough decision and the 2007 Crack Amendment to

the USSC sentencing guidelines. Although the chi-squared tests also indicate that there

may be trend breaks between the years 2004 and 2005, between the years 2008 and 2009,

and between the years 2009 and 2010, there is again no indication of a trend break

between the year 2011 and the year 2010 (i.e., when the FSA was implemented). This is

obviously consistent with the graphical evidence from Figure 3 showing a big decline in

mean sentence length for crack offenders between year 2007 and year 2008, followed by

a slight uptick between year 2008 and year 2009, followed by a steady downturn from year

2009 and onward.

As a comparison, Table 8 shows regressions analogous to those in Table 7, but

using only the sample of powder cocaine defendants. As can be seen, there is little evi-

dence of anything but a flat trend in sentencing of powder cocaine offenders through-

out the whole time period from 2002–2012.

Overall, the results shown in Tables 7 and 8 are consistent with the notion that

the Kimbrough decision and the 2007 Crack Amendment to the USSC sentencing guide-

lines precipitated the beginning of a downward trend in mean sentence lengths for

crack offenders, but not powder cocaine offenders, that was unaltered by the implemen-

tation of the FSA.

386 Bjerk

VII. What Would Have Happened to Crack Sentences

in the Absence of the FSA?

The results above show that the trend toward more leniency for crack offenders started

well before the implementation of the FSA and, moreover, that the FSA did not appear

to accelerate this trend. However, this does not mean that the FSA had no impact on

sentencing. Indeed, in the absence of the FSA, the mandatory minimums may have

impeded the downward trend in sentencing for crack offenders from continuing. This

section attempts look at this issue in more detail.

In thinking about what sentences for crack offenders would have been in the

absence of the FSA we first have to consider whether the FSA impacted other aspects of

the judicial process beyond just sentencing. Maybe most notably, did the changes inher-

ent in the FSA cause defendants to alter how much crack they carried and/or prosecu-

tors’ willingness and ability to bargain over the conviction quantity? The dataset used

Table 7: OLS Analysis for Trend Breaks in Sentencing for Crack Since 2002

(Dependent Variable: Sentencing Length in Months)

Specification

(1) (2) (3)

Post-Year 2003 indicator 1.92 1.19 1.41 (2.308) (1.713) (1.586)

Post-Year 2004 indicator 10.77** 5.91 5.73 (2.228) (1.654) (2.157)

Post-Year 2005 indicator 24.21*** 23.35*** 22.63** (2.269) (1.684) (2.230)

Post-Year 2006 indicator 25.16 20.80 20.42 (2.210) (1.641) (2.270)

Post-Year 2007 indicator 9.70*** 4.08* 2.53 (2.181) (1.618) (1.671)

Post-Year 2008 indicator 218.23*** 213.09*** 212.10*** (2.160) (1.603) (2.180)

Post-Year 2009 indicator 5.27*** 22.71*** 22.36** (2.196) (1.630) (2.164)

Post-Year 2010 indicator 29.13*** 28.94*** 29.48* (2.359) (1.750) (2.365)

Post-Year 2011 indicator 27.40 29.08 27.96 (2.511) (1.863) (2.291)

Post-Year 2012 indicator 28.13 210.48 210.13 (2.742) (2.035) (2.095)

Race and gender indicators No Yes Yes Criminal history category indicators No Yes Yes Quantity (by guideline cell) No Yes Yes Two-way clustered standard errors by district and month No No Yes

NOTES: Crack cases only. Post-Year X refers to a dummy equaling 1 if the case was sentenced in or after year X. Standard errors in parentheses. Asterisks indicate statistical significance of chi-squared test of null hypothesis of equality of coefficients between Post-Year j and Post-Year j – 1 (* indicating 10 percent level, **indicating 5 per- cent level, ***indicating 1 percent level).

387Crack Cocaine Defendants and the Fair Sentencing Act

here does not provide information on the initial quantity any given defendant was

charged with, so these issues cannot be looked at directly. We can, however, look at how

the distribution of conviction quantities for crack defendants changed before and after

the FSA was implemented.

Figure 4 looks at this issue in detail. This figure shows side-by-side histograms of

the quantity distributions for each year from 2009–2012 (where the FSA was only in

affect for the years 2011 and 2012). The two boxes in Figure 5 indicate the quantity bins

that would have been directly impacted by the FSA. Again, defendants convicted for

quantities in these two bins would be subject to a shorter mandatory minimum post-FSA

than pre-FSA, while defendants in all other bins would be subject to the same mandato-

ry minimum pre- and post- FSA. Curiously, both in the years before and the years after

the FSA, the two FSA-impacted quantity ranges contain more than half the cases (this

will be something to return to later). Moreover, the fraction of cases in each of these

two bins changes somewhat before and after the FSA was implemented. Notably,

Table 8: OLS Analysis for Trend Breaks in Sentencing for Powder Cocaine Since 2002

(Dependent Variable: Sentencing Length in Months)

Specification

(1) (2) (3)

Post-Year 2003 indicator 3.53 0.09 21.03 (1.712) (1.273) (1.406)

Post-Year 2004 indicator 1.05 1.06 0.04 (1.656) (1.231) (1.478)

Post-Year 2005 indicator 3.63 21.50 21.29 (1.675) (1.245) (1.090)

Post-Year 2006 indicator 4.33 1.37 0.55 (1.624) (1.207) (0.926)

Post-Year 2007 indicator 22.30** 21.36 20.51 (1.569) (1.166) (1.340)

Post-Year 2008 indicator 0.02 21.44 20.95 (1.589) (1.181) (0.682)

Post-Year 2009 indicator 1.99 1.31 0.63 (1.612) (1.199) (.804)

Post-Year 2010 indicator 22.61* 0.97 1.56 (1.601) (1.191) (1.446)

Post-Year 2011 indicator 26.67 22.11 21.64 (1.633) (1.214) (1.608)

Post-Year 2012 indicator 20.28** 23.00 21.58 (1.656) (1.231) (0.899)

Race and gender indicators No Yes Yes Criminal history category indicators No Yes Yes Quantity (by guideline cell) No Yes Yes Two-way clustered standard errors by district and month No No Yes

NOTES: Powder cocaine cases only. Post-Year X refers to a dummy equaling 1 if the case was sentenced in or after year X. Standard errors in parentheses. Asterisks indicate statistical significance of chi-squared test of null hypoth- esis of equality of coefficients between Post-Year j and Post-Year j – 1 (* indicating 10 percent level, **indicating 5 percent level, ***indicating 1 percent level).

388 Bjerk

however, these changes go in different directions across these two quantity bins. The

fraction of crack defendants convicted for quantities between 0.005 kilograms and 0.028

kilograms falls in the two years following the implementation of the FSA, while the frac-

tion convicted for quantities between 0.05 kilograms and 0.28 kilograms rises somewhat

in the two years following the implementation of the FSA.6

Figure 4: Distribution of crack conviction quantities by year.

NOTES: “Pre-FSA” spans Year 2009--Year 2010 (i.e., August 1, 2008--July 31, 2010), “Post-FSA” spans Year 2009--Year 2010 (i.e., August 1, 2010--July 31, 2012). Boxes highlight FSA-affected quantities.

Figure 5: Fraction of crack convictions with safety valve invoked.

6Note, the dates we use here are the dates at which the defendants were sentenced. Clearly, however, a plea bar- gain stipulating the conviction quantity could have been struck well before the FSA was implemented even if the actual sentence was given post-FSA. This means that if there were substantial changes in quantity manipulation following the implementation of the FSA, the above issue means that such changes may be somewhat obscured when comparing the year 2010 to the year 2011. Therefore, it might be cleaner to focus mostly on the year 2012 results relative to the pre-FSA years of 2009 and 2010.

389Crack Cocaine Defendants and the Fair Sentencing Act

These findings are a little ambiguous regarding whether defendants or prosecu-

tors significantly changed their behaviors in response to the FSA. While the distribution

of conviction quantities does seem to change slightly in the quantities that would osten-

sibly be affected by the FSA, it does not do so in a uniform manner. From a defendant’s

perspective, the consequences from being convicted for a quantity between 0.005 kilo-

grams and 0.028 kilograms, or for a quantity between 0.05 kilograms and 0.28 kilograms,

should be lower post-FSA. Similarly, post-FSA, judges and prosecutors should ostensibly

have more discretion over the sentences given to defendants convicted for quantities

between 0.005 kilograms and 0.028 kilograms, and between 0.05 kilograms and 0.28

kilograms, than they did before. Both forces would seem to suggest that we would see a

significantly higher fraction of defendants in these two categories post-FSA than pre-FSA

but, as stated above, we find that is only true in one of these categories, with exactly the

opposite occurring in the other. In general, Figure 4 does not suggest large changes in

criminal or prosecutor behavior with respect to real or charged crack quantities.

Another aspect of behavior that may have changed in response to the FSA is the will-

ingness of prosecutors to invoke the federal “safety-valve” provision. As alluded to in Sec-

tion II, defendants ostensibly eligible for a federal mandatory minimum can be given a

sentence less than the mandatory minimum if they are deemed eligible for the federal

“safety-valve” program, which can occur if the prosecutor determines that the defendant

has a minimal criminal history, there was no gun or violence used in the commission of

the crime, the defendant was not a leader or organizer, and the defendant was determined

to be cooperative. While some of these criteria are quite objective (e.g., minimal criminal

history), most leave substantial discretion up to the prosecutor (e.g., the defendant was

cooperative, the defendant was not an organizer or leader). Hence, one might wonder

whether prosecutors became less likely to grant safety-valve eligibly after the FSA raised the

quantity thresholds determining eligibility for the mandatory minimums for crack.

Figure 5 shows that the rate at which the safety valve was invoked for crack defend-

ants was essentially unchanged surrounding the implementation of the FSA, hovering right

around 12 percent. To look at this issue somewhat more rigorously, Table 9 shows the

results of OLS linear probability models, where a binary variable equaling 1 if the safety

valve was reported to be invoked (and 0 otherwise) was regressed on a post-FSA indicator

variable, a monthly time trend, and other defendant characteristics. The results in specifi-

cation (1) mirror the result shown in Figure 6, showing that without conditioning on any-

thing else, there is no significant time trend or change in the likelihood of the safety valve

being invoked post-FSA. Specification (2) further controls for race, gender, criminal histo-

ry, and quantity, while specification (3) further adds district fixed effects (and uses two-way

clustered standard errors on district and monthly time). As can be seen by the extremely

small in magnitude coefficients on the post-FSA indicator (along with relatively tight confi-

dence intervals), even controlling for defendant characteristics, there is no evidence that

use of the safety valve changed following the implementation of the FSA.7

7I used a linear probability OLS model for simplicity given the number of fixed effects and clustered standard errors. However, little changes if I use a probit model.

390 Bjerk

Given the results above, let us consider again the extent to which the FSA pre-

vented mandatory minimums from impeding the pre-FSA downward trends in sentenc-

ing for crack offenders from continuing. While determining a precise answer to this

question may not be possible, we can potentially determine a reasonable bound. Specifi-

cally, let us suppose that pre-FSA, the mandatory minimums were already generally bind-

ing for defendants convicted for quantities between 0.005 and 0.028 kilograms and for

defendants convicted for quantities between 0.05 and 0.28 kilograms (the two quantity

categories directly impacted by the FSA). In other words, pre-FSA, prosecutors and judg-

es generally wanted to give sentences to such defendants at or less than required by the

mandatory minimum legislation but could not. Given this, in the absence of the FSA,

even if prosecutors and judges wanted to continue the trend of becoming more lenient

with respect to sentencing crack defendants, they would have been constrained to give

sentences to defendants in these quantity categories similar to what they were giving to

like defendants pre-FSA. Therefore, we can compute “counterfactual” sentences for

Table 9: OLS Analysis for Trend Break Following Fair Sentencing Act of 2010

(Dependent Variable: Binary Variable for Invocation of Safety Valve)

Specification

(1) (2) (3)

Coefficient on post-FSA indicator 0.00 0.00 0.00 (0.014) (0.009) (0.008)

[–0.02, 0.03] [–0.02, 0.02] [–0.02, 0.01] Time trend (monthly) 0.00 0.00 0.00

(0.001) (0.0) (0.0) Race and gender No Yes Yes Criminal history category No Yes Yes Quantity (by guideline cell) No Yes Yes Two-way clustered standard errors by district and month No No Yes

NOTES: Crack cases only. Data run from August 1, 2008 to July 31, 2012. Standard errors in parentheses, brackets show 95 percent confidence interval.

Figure 6: Actual and “counterfactual” mean sentence lengths for crack.

391Crack Cocaine Defendants and the Fair Sentencing Act

defendants convicted post-FSA for quantities in the 0.005–0.028 kilogram range and the

0.05–0.28 kilogram range by estimating what they would have received given their char-

acteristics had they been convicted just prior to the implementation of the FSA. Given

the FSA did not affect the mandatory minimums relevant to those convicted for quanti-

ties of crack outside the two above quantity categories, the “counterfactual” sentences

for defendants convicted for such quantities post-FSA simply equal their actual

sentences.

To estimate these counterfactual sentences for crack defendants convicted for

quantities in the 0.005–0.028 kilogram range and the 0.05–0.28 kilogram range post-

FSA, I first regress sentence length on quantity guidelines category dummies, criminal

history category dummies, district dummies, and indicators for race and gender, using

just the sample of crack defendants convicted for quantities in the 0.005–0.028 kilogram

range in the 12 months prior to the implementation of the FSA. I then used the coeffi-

cients from this regression to predict “counterfactual” sentences for each post-FSA crack

defendant convicted for a quantity in the 0.005–0.028 kilogram range based on their

characteristics. I then do an analogous procedure for each post-FSA defendant con-

victed for a quantity in the 0.05–0.28 kilogram range. In other words, somewhat analo-

gous to a Oaxaca-Blinder decomposition, I estimate “counterfactual” sentences for

defendants convicted for quantities affected by the FSA by using their characteristics

along with regression coefficients calculated with a pre-FSA sample to map the charac-

teristics of the post-FSA sample into counterfactual expected sentences. As stated above,

though, for post-FSA crack defendants whose conviction quantities were not affected by

the FSA, their “counterfactual” sentence is the same as their actual sentence.

Figure 6 shows the results from this “counterfactual” exercise. The darker gray

line (with diamonds) shows the actual trend in mean sentence lengths for crack

offenders from the year 2009 to year 2012 as shown previously in Figure 2. The lighter

gray line (with squares) shows the trend in mean sentence lengths for crack offenders

from 2009–2012 under the “counterfactual” sentences as described above. As can be

seen, this counterfactual suggests that mean sentences for crack offenders would have

stagnated at pre-FSA levels, meaning that in the absence of the FSA the overall mean

sentence for crack offenders would be almost 8 percent higher in year 2011 (104 months

vs. 97 months in actuality) and almost 20 percent longer in year 2012 (105 months vs.

88 months).

The magnitude of the difference between what actually happened and this coun-

terfactual arguably may seem somewhat implausible given much of the discussion above,

particularly given these counterfactual sentences equal the actual sentences for all those

convicted for crack quantities outside of the 0.005–0.028 kilogram and 0.05–0.28 kilo-

gram ranges post-FSA. However, recall from Figure 5 that almost 65 percent of crack

defendants are convicted for quantities within one of these two quantity categories

(both before and after the FSA). Moreover, the above results likely overstate how much

the downward trend in sentencing for crack offenders would have been impeded by

mandatory minimums in the absence of the FSA, as it assumes that the mandatory mini-

mums were already binding with respect to the sentencing of crack defendants in the

affected quantity categories pre-FSA. This is likely a strong assumption. Regardless, these

392 Bjerk

results are at least suggestive that even though the FSA did not accelerate the downward

trend in sentencing of crack offenders, without the FSA, this downward trend would

have notably decelerated.

VIII. The FSA and the Crack/Powder Cocaine

Sentencing Gap

As alluded to previously, another of the main motivations for the FSA was to reduce the

sentencing disparity between crack defendants and powder cocaine defendants, particu-

larly for those convicted for similar amounts. This section examines this issue more

directly.

The dark gray line with diamonds in Figure 7 shows the evolution of the crack/

powder cocaine gap for a sample of defendants convicted for quantities under 3 kilo-

grams (which constitutes over 95 percent of crack cases).8 Once again, vertical lines are

put into to show the Booker decision, the Kimbrough decision and the 2007 Crack Amend-

ment to the sentencing guidelines, and the FSA. As can be seen, the crack/powder

cocaine sentencing gap experienced a precipitous decline following the Kimbrough deci-

sion and the 2007 Crack Amendment to the sentencing guidelines, with the ratio of

mean crack sentences to mean powder cocaine sentences falling from around 2.5 to 1

Figure 7: Ratio of mean sentence for crack vs. powder cocaine (< 3kg).

8The reason I truncated quantity at 3 kilograms for the analysis shown in Figure 7 is because I am most interest- ed in the crack cocaine/powder cocaine sentencing gap for defendants convicted for roughly similar amounts. Although over 95 percent of crack defendants were convicted for quantities under 3 kilograms, only about one- third of powder cocaine defendants were convicted for quantities under 3 kilograms. So without truncating, we would be comparing crack defendants to cocaine defendants who were on average convicted for much greater quantities.

393Crack Cocaine Defendants and the Fair Sentencing Act

to about 2 to 1. However, this ratio has stayed relatively constant every year since, includ-

ing following the implementation of the FSA.

The above result is not surprising given the results from Figure 2 showing that the

mean sentences for both crack and powder cocaine had been declining over the 2009–

2012 period, and that this downward trend in sentencing for crack defendants showed

no evidence of accelerating following the implementation of the FSA. However, it again

might be useful to consider a “counterfactual” of what would have happened in the

absence of the FSA. As discussed above, while the FSA may not have accelerated the

downward trend in sentencing for crack offenders, this downward trend may have been

stymied in its absence.

Looking again at Figure 7, the light gray line with squares shows the “counterfactual”

estimate of what would have happened to this ratio in the absence of the FSA. Again, this

counterfactual assumes that prior to the FSA, prosecutors/judges were constrained by the

mandatory minimums in 2010, so that in the absence of the FSA, those for whom the FSA

lessened the relevant mandatory minimum sentence (i.e., those convicted for quantities

between 0.005–0.028 kilograms and 0.05–0.28 kilograms post-FSA) would have received a

sentence consistent with the year 2010 practices in the absence of the FSA. As can be seen,

to the extent to which sentencing practices for defendants convicted for crack quantities

between 0.005–0.028 kilograms and 0.05–0.28 kilograms post-FSA would have been con-

strained to pre-FSA levels if the mandatory minimums for crack had not been adjusted by

the FSA, the ratio of crack to powder cocaine mean sentences may have risen again up to

2.3 to 1 rather than its current level of 2 to 1. In other words, by 2012, the crack/powder

cocaine sentencing gap may have been about 15 percent higher in the absence of the FSA.

IX. Conclusion

In contrast to some of the lofty rhetoric, the findings of this study suggest that the

changes to the mandatory minimum sentence quantity thresholds for crack offenders as

dictated by the Fair Sentencing Act of 2010 (FSA) were actually quite modest in their

impact. The FSA does not appear to be the primary force responsible for the fall in

average crack cocaine sentences following the passage of the FSA, nor does it appear

that the FSA in and of itself substantially decreased the sentencing disparity between

crack cocaine defendants and powder cocaine defendants. Rather, it appears that

changes in both the structure and implementation of the USSC sentencing guidelines

in the mid-2000s allowed prosecutors and/or judges to act on their preferences for

more leniency with respect to sentencing drug defendants, particularly crack defend-

ants. Although the FSA itself does not appear to have substantially accelerated these

trends toward greater leniency, the evidence presented in this article suggests that it

helped prevent the mandatory minimums from impeding the continuation of these

trends. Going forward, it is quite clear that to further reduce the crack/powder cocaine

sentencing gap, the U.S. Congress must work with the U.S. Sentencing Commission to

alter both the eligibility standards for mandatory minimums as well as the sentencing

guidelines for crack.

394 Bjerk

References

Beaver, Alyssa L. (2010) “Getting a Fix on Cocaine Sentencing Policy: Reforming the Sentencing Scheme of the Anti-Drug Abuse Act of 1986,” 78 Fordham Law Rev. 2531.

Bjerk, David (2005) “Making the Crime Fit the Penalty: The Role of Prosecutorial Discretion Under Mandatory Minimum Sentencing,” 48(2) J. of Law & Economics 591.

Caulkins, Jonathan P., C. Peter Rydell, William L. Schwabe, & James Chiesa (1997) Mandatory Mini- mum Drug Sentences: Throwing Away the Key or the Taxpayers’ Money? RAND Monograph Santa Monica, CA: RAND.

Chitwood, Dale, James Rivers, & James Inciardi (1996) The American Pipe Dream: Crack Cocaine and the Inner City. Austin, TX: Harcourt Brace College Publishers.

Fischman, Joshua B., & Max M. Schanzenbach (2012) “Racial Disparities Under the Federal Sen- tencing Guidelines: The Role of Judicial Discretion and Mandatory Minimums,” 9(4) J. of Empirical Legal Studies 729.

Fryer, Roland, Paul S. Heaton, Steven D. Levitt, & Kevin M. Murphy (2013) “Measuring Crack Cocaine and its Impact,” 51(3) Economic Inquiry 1651.

Grogger, Jeffrey, & Michael Willis (2000) “The Emergence of Crack Cocaine and the Rise in Urban Crime Rates,” 82(4) Rev. of Economics & Statistics 519.

Hartley, Richard, Sean Maddan, & Cassia Spohn (2007) “Prosecutorial Discretion: An Examination of Substantial Assistance Departures in Federal Crack-Cocaine and Powder Cocaine Cases,” 24(3) Justice Q. 382.

Helland, Eric, & Alex Tabarrok (2007) “Does Three Strikes Deter? A Non-Parametric Estimation,” 42(2) J. of Human Resources 309.

Kautt, Paula, & Cassia Spohn (2002) “Crack-ing Down on Black Drug Offenders? Testing for Inter- actions Among Offenders’ Race, Drug Type, and Sentencing Strategy in Federal Drug Sentences,” 19(1) Justice Q. 1.

Kerr, Peter (1986) “Anatomy of the Drug Issue: How, After Years, it Erupted” November 17 New York Times.

Klein, Malcolm W, Cheryl L. Maxon, & Lea C. Cunningham (1991) “‘Crack’, Street Gangs, and Vio- lence,” 29(4) Criminology 623.

Kurlychek, Megan C., Jeffery T. Ulmer, & John H. Kramer (2007) “Prosecutorial Discretion and the Imposition of Mandatory Minimum Sentences,” 44(4) J. of Research in Crime & Delinquency 427.

LaCasse, Chantale, & A. Abigail Payne (1999) “Federal Sentencing Guidelines and Mandatory Mini- mum Sentences: Do Defendants Bargain in the Shadow of the Judge?” 42(S1) J. of Law & Economics 245.

Loftin, Colin, Milton Heumann, & David McDowall (1983) “Mandatory Sentencing and Firearms Violence: Evaluating an Alternative to Gun Control,” 17(2) Law & Society Rev. 287.

Loftin, Colin, David McDowall, & Brian Wiersema (1992) “A Comparative Study of the Preventive Effects of Mandatory Sentencing Laws for Gun Crimes,” 83(2) J. of Criminal Law & Criminolo- gy 378.

Marvell, Thomas B., & Carlisle E. Moody (2001) “The Lethal Effects of Three-Strikes Laws,” 30(1) J. of Legal Studies 89.

Reinarman, Craig, Dan Waldorf, Sheigla B. Murphy, & Harry G. Levine (1997) “The Contingent Call of the Pipe: Bingeing and Addiction Among Heavy Cocaine Smokers,” In C. Reinarman & H. G. Levine, eds., Crack in America: Demon Drugs and Social Justice. Berkeley, CA: Univ. of California Press.

Shepherd, Joanna M. (2002) “Fear of the First Strike: The Full Deterrent Effect of California’s Two- and Three-Strikes Legislation,” 31(1) J. of Legal Studies 159.

Starr, Sonja, & M. Marit Rehavi (2013) “Mandatory Sentencing and Racial Disparity: Assessing the Role of Prosecutors and the Effects of Booker,” 123(2) Yale Law J. 2.

395Crack Cocaine Defendants and the Fair Sentencing Act

Tonry, Michael (1992) “Mandatory Penalties,” 16 Crime & Justice 243. USSC (2002) “Cocaine and Federal Sentencing Policy,” U.S. Sentencing Commission Special Report to

Congress, May. ———— (2011) “Mandatory Minimum Penalties in the Federal Criminal Justice System,” U.S. Sen-

tencing Commission Report to the Congress, October. Vagins, Deborah J., & Jesselyn McCurdy (2006) “Cracks in the System: Twenty Years of the Unjust

Federal Crack Cocaine Law,” American Civil Liberties Union Monograph New York: ACLU.

396 Bjerk

Copyright of Journal of Empirical Legal Studies is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.