Survey of Public Opinion on Sentencing Federal Crimes

SUBJECT: Public Opinion on Sentencing Federal Crimes

DATE: March 14, 1997

The Sentencing Commission is pleased to release this national sample survey of public opinions on sentences for federal crimes. Pursuant to our enabling statute, the Commission continues to assess the sentencing guidelines to reflect, to the extent practicable, advancement in the knowledge of human behavior as it relates to the criminal justice process.

The attached research study addresses the Commission's obligation to "develop means of measuring the degree to which the sentencing, penal, and correctional practices are effective in meeting the purposes of sentencing," including the purpose of providing "just punishment for the offense" (28 U.S.C. § 991(b)(2)). Two of the seven factors the statute directs the Commission to consider in developing guideline offense categories are the community view of the gravity of the offense and the public concern generated by the offense. This study will prove helpful in our efforts to improve the guideline system.

The Commission contracted with Dr. Peter H. Rossi of the University of Massachusetts, Amherst, and Dr. Richard A. Berk of the University of California at Los Angeles to prepare a general report summarizing the survey data. We thank them for their contributions in improving the federal sentencing system. While the Commission accepted their final report as fulfilling the terms of the contract, the report is the work product of these independent contractors and does not necessarily reflect the views of the U.S. Sentencing Commission.

Executive Summary: Public Opinion on Sentencing Federal Crimes

Completed under contracts with:
Peter H. Rossi, University of Massachusetts at Amherst
Richard A. Berk, University of California at Los Angeles
and Response Analysis Corporation

October 1995


Many persons contributed to the study reported in this monograph. Members of the staff of the U. S. Sentencing Commission participated in making many of the critical decisions on the design of the study, especially Ms. Phyllis Newton. Dr. Linda Drazga Maxfield served as contract manager. We are especially indebted to Dr. Linda Drazga Maxfield, Mrs. Christine Kitchens, and Ms. Ronnie M. Scotkin, for mapping guidelines sentences on to the variables used in this study.

The sample design and data collection for this study was carried out by the Response Analysis Corporation of Princeton, New Jersey: Staff members, Dr. Lynne Firester and Dr. Andrew Kulley, managed those tasks with great skill and devotion.

At the University of California at Los Angeles, Alec Campbell served as research assistant, producing many of the tables and graphs that are the foundation of Chapters 5 through 9.

Americans' willingness to answer questions posed by strangers make surveys like this one possible. More than 1,700 patiently answered our questions in sessions lasting almost an hour each. Our interviewers thanked each person interviewed, a gesture that is not enough. We hope that this report and the knowledge derived from their patient attention to our questions adds to their rewards for cooperation.

We are grateful for all the help we received and are certain that these efforts have measurably improved the quality of this monograph. Whatever faults this report may have arise out of our inability to take maximum advantage of this help.


This monograph reports the findings of a national sample survey commissioned by The U. S. Sentencing Commission. Conducted early in 1994 through face-to-face interviews with a probability sample of the American populace, the main purpose of the study was to find out how the American public would sentence persons convicted of crimes under the Federal Criminal Code. This report describes the sentences given, examines the latent principles used in giving sentences and compares the sentences to those recommended in the guidelines. The study is unique both in the crimes covered and in the methods of measurement used.

Methodology: Chapter 2

To find out how Americans would sentence persons convicted in the federal courts, each selected respondent was asked to sentence convicted persons whose crimes and personal characteristics were described in short vignettes. Using examples of crimes covered in the guidelines, short vignettes were written incorporating descriptions of the crimes, circumstances of crimes that might warrant sentence modification (either increases or decreases), and relevant characteristics of the offender. For each of 42 vignettes describing crimes and offenders respondents gave what each considered to be an appropriate sanction.

The vignettes were put together using the factorial survey method, a research design that makes it possible to use unusually rich descriptions of offenders. The vignettes were constructed by randomly varying their contents, constituting a number of factorial experiments. The vignettes randomly varied the crime described, some of the circumstances surrounding the crime, and offender characteristics.

The sentencing alternatives included probation, imprisonment for a fixed period, life imprisonment, and death. All available evidence suggests that the vast majority of respondents took the task seriously and were able to provide informed judgements.

In addition to the crime vignettes, a questionnaire was used to collect information on respondents' backgrounds and responses to a number of attitudinal questions thought to be related to sentencing preferences. These items also seemed to have been taken seriously by the vast majority of respondents.

The sample design called for interviews with persons over 18 with sufficient literacy skills to perform the sentencing task. A total of 1,737 interviews were conducted with an overall response rate of 70 percent.

Overall Sentencing Patterns: Chapter 3

Twenty general crimes were included in the study, each represented by several concrete examples. For example, the general crime "fraud" was represented by 13 examples, ranging from insider trading to false statements on a mortgage application. All told, there were 73 concrete Crime Examples each representing one of the 20 general crimes. The general crimes included some ordinarily considered street crimes, such as robbery, drug trafficking, and drug possession, some "white collar crimes" such as bribery, money laundering, and antitrust violations, as well as some civil rights violations and environmental crimes.

In giving sentences, respondents made clear distinctions among general crimes and among Crime Examples indicating that the sentencing task was taken quite seriously. For example, the median sentence given to a kidnapping in which the victim was killed was death whereas the median sentence for marijuana possession was probation. In general, common sense knowledge about the seriousness of the offense would lead one to make correct predictions about either the median or mean sentence given by respondents. But there were some exceptions.

Median sentences were universally shorter than mean sentences indicating that sentence distributions had long tails. For each general crime and each Crime Example, a minority of respondents gave very lengthy sentences, severely influencing the means. Accordingly, the analyses tend to emphasize median sentences because such measures were closer to the central tendencies found among the respondents considered collectively. In most cases mean sentences are also reported.

Although only the most serious and the least serious Crime Examples each received close to the same sentences from the respondents, for most crimes the distributions of sentences given did not indicate a clear consensus. Nevertheless, there was some convergence among the American public.

Respondents saw many distinctions among specific Crime Examples that are formally considered identical. For example, different examples of fraud were given quite dissimilar sentences. It also appeared that respondents were more sensitive to threats to the physical safety of victims than they were to monetary losses. Accordingly, victimless crimes were treated much less harshly than ones in which the physical safety of victims was threatened.

Guideline and Respondent Sentences Compared: Chapter 4

By design, the crimes and their descriptions followed the guidelines as closely as possible. Hence it was feasible to calculate a guideline sentence for each vignette. The calculation of guideline sentences started with the base offense level for each crime and was then modified by those crime circumstances recognized in the guidelines as a sentence enhancement or mitigation, including the guidelines' adjustments in sentencing for previous felony convictions.

Comparisons were made at three levels: 1) At the level of individual vignettes, comparing the sentencing judgments made by individual respondents to the calculated guideline sentences for those vignettes; 2) At the level of Crime Examples, comparing the median and mean sentences for the 73 Crime Examples with the guidelines' medians or means; and, 3) At the level of general crime groups, comparing median or means sentences for the 20 such groups included in the study.

The comparisons do not lead to clear-cut conclusions concerning the degree of comparability between guideline and respondent sentences. On the individual vignette level, only a modest level of agreement was found, the coefficient of determination, R2, being 0.22 for the regression equation linking the guideline and respondent sentences. To add to the complexity, respondents tended to be both more severe and more lenient in sentencing than the guidelines, expressing preferences for longer sentences for minor offenses and shorter sentences for major crimes.

However, much greater comparability was found when mean and median sentences for either general crimes or Crime Examples were compared. When the mean guideline sentences for the 20 general crimes are compared to the median respondent sentences, the resulting R2 is .87 indicating quite strong agreement on that level. Agreement over the 73 Crime Examples is somewhat lower but still quite strong, the R2 being .78. In short, comparability is much higher when the central tendencies of the public's sentences are taken into account.

Major points of disagreement over sentencing centered around certain types of crimes. The strongest sentencing disagreements occur over drug trafficking crimes: The guidelines call for drug trafficking sentences that vary according to the type of drug sold, roles played in the crime and the amount of drugs involved. In contrast, respondents did not make such distinctions nor did they weigh these crime elements the same way as do the guidelines. The result is strong differences in sentencing drug trafficking crimes with the guideline sentences being much harsher. Weaker sentencing differences were found with respect to sentences for bribery, civil rights, and environmental crimes. And, generally, respondents gave greater enhancements to crimes in which personal safety was endangered.

Disagreement between guideline and respondent sentences also was found over the weights to be given to some features of crimes. In particular, respondent sentences did not increase as sharply as guideline sentences when the economic gain increased. In contrast, the weights given by respondents to offenders' criminal records were quite similar to the guideline recommendations.

Factors Influencing Sentencing: Chapter 5-9

Because certain features of the crime as well as characteristics of the offender were varied from vignette to vignette, it was possible to calculate the effects of these factors on sentencing. Most of the characteristics varied were specific to groups of similar crimes requiring separate analyses for the major crime groups.

As a group, drug trafficking crimes were given relatively long sentences. Persons convicted of trafficking in very large amounts of drugs were punished more severely. However, respondents did not treat trafficking in heroin, powder cocaine or crack cocaine very differently from each other. In contrast, marijuana trafficking was sanctioned much less severely. Median sentences for trafficking in crack cocaine, powder cocaine, and heroin all topped out at about 12 years, even for defendants with four prior prison terms.

Drug possession, as contrasted with drug trafficking, was treated as one of the least serious crimes by respondents. For possession of crack cocaine, powder cocaine, and heroin, average sentences were about a year. For marijuana, the average sentence was essentially probation.

Street crimes elicited somewhat shorter sentences than drug trafficking except when victims or bystanders were put at risk, injured or killed. Variation in sentences for street crimes was driven most dramatically by concerns about personal safety. The number of prior prison terms on the defendant's record affected sentences, but far less than actual or potential harm to victims.

White collar crimes, environmental crimes, and hate crimes were given far shorter sentences than drug trafficking and street crimes except when safety of persons was endangered. Sentences were more punitive as the economic gains from the crime increased, but the increments were modest.

From these findings four general conclusions follow that apply to the majority of the crimes we studied. First, most of the variation in sentences given was a function of the crimes committed, not the background of the defendant. The defendants' gender, employment situation, or family circumstances rarely had any significant impact on sentencing. What mattered was what the defendant did, not who he or she was.

Second, even though about a third of the defendants in the crime vignettes had a record of four prior prison terms, there was little support for sentences consistent with most habitual offender legislation. To be sure, longer previous criminal records led to longer sentences, but at substantially smaller increments than under such initiatives as "three-strikes-and-you're-out."

Third, sentence length did not increase at the same rate as the economic gains from crime. In some sense, respondents were allowing crime to pay. Whether this resulted from a lack of support for deterrence in economic crimes or from a failure to appreciate the large monetary inducements for certain kinds of felonies, cannot be determined. In this respect, respondent sentencing patterns are similar to the guidelines' treatment of economic gain in which sentences increase logarithmically with increases in economic gain.

Fourth, the impact of variation in characteristics of the crime and the defendant on sentences seems to have been multiplicative. Given some baseline sentences, sanctions were more or less punitive by some multiplier. Thus, the impacts of crime characteristics and defendant characteristics was generally larger the more serious the kind of crime. This pattern is also found in the guidelines.

Regional and Community Size Differences: Chapter 10

Regional and community size differences in sentencing given to crime groups were analyzed by the nine Census regions and by residence in two sizes of Metropolitan Statistical Areas and counties outside of metropolitan areas.

There are two main findings:

First, there are strong and consistent regional differences. Residents of the New England states were the most lenient in sentencing. In contrast, residents of the two South Central regions, from Texas and Oklahoma on the west and Alabama, Kentucky and Tennessee on the east, gave almost consistently longer sentences to offenders of all kinds.

Second, community size differences were somewhat weaker. In general, longer sentences were given by those residing in counties that were outside of major metropolitan areas. Whether respondents lived in large or smaller Metropolitan Statistical Areas seemed not to make any difference.

Social and Individual Factors in Sentencing: Chapter 11

In order to analyze differences in sentencing among individual respondents, several summary sentencing indexes were calculated. These nine indexes were computed for each respondent and expressed the ways each respondent differed from the sample as a whole in giving sentences of various sorts to grouped crimes. The analysis looked at demographic differences in sentencing as well as describing correlations with certain political and social attitudes.

Although social and individual characteristics of respondents were not strongly related to sentencing behavior, several salient findings emerged.

Among demographic groups, only those defined by educational attainment differed strongly and consistently, with the better educated being more lenient in sentencing than the less well educated. Men and women essentially did not differ in sentencing, although women gave fewer death sentences and wanted longer sentences for drug possession. Ethnic and racial differences also were small although Hispanics and blacks favored shorter sentences for some groups of crimes and did not give as many death sentences.

Personal experiences as a crime victim made little difference; generally, victims asked for more lenient sentences. Living in a community with high crimes rates also made little difference and then also in the direction of greater leniency. In contrast, persons who had some contact with our legal system as a juror or as a party to legal proceedings or experience in the criminal justice system through arrest or conviction were decidedly harsher in sentencing.

Respondent political and social attitudes affected sentencing if the attitudes were closely related to criminal justice issues. Respondents who favored more legal safeguards for persons accused of crimes were inclined to give shorter sentences whereas those who favored strengthening the investigative powers of the police gave harsher sentences. Although respondents who identified themselves as liberals were more lenient than conservatives, those contrasts were diminished greatly when other personal factors were taken into account. Attitudes towards minority rights, public welfare and environmental issues had slight effects on sentencing but those effects were virtually wiped out when other individual characteristics were taken into account.

Taking the individual characteristics into account in a multivariate analysis, regional differences along the lines discussed earlier tended to persist. The strong contrast between New Englanders and residents of the South persisted when such individual factors as educational attainment and the influence of experience with legal institutions were taken into account.

View or Download full text of individual chapters listed below in Adobe .PDF format



Methodology: Chapter 2
Overall Sentencing Patterns: Chapter 3
Guidelines and Respondent Sentences Compared: Chapter 4
Factors Influencing Sentencing: Chapter 5-9 Regional and Community Size Differences: Chapter 10
Social and Individual Factors in Sentencing: Chapter 11


Why Study the American Public's Views on Sentencing?
Existing Empirical Studies of Public Opinion on Sentencing and Closely Related Topics
Global Assessments of Sentencing Severity in the Courts
Studies of Crime Seriousness
Studies of the Public's Sentencing Preferences
Some Conclusions Concerning Previous Research
Objectives of the Research Project
Limitations of the Research
Plan of This Report


Instrument Design
Sampling and Data Collection
Data Analysis


Overall Sentencing Patterns
Crime Types
Holding Crime Constant
Some Sentencing Variations by Criminal Characteristics
Joint Effects of Gender, Family Status and Previous Record
Crime Examples


Calculating Guideline Sentences
Analysis Strategy
Distributions of Guidelines and Respondent Sentences Given to Vignettes
Vignette Level Regression Analysis
Taking Response Sets Into Account
Sample-Guidelines Rank Order Correspondences for Individual Respondents
Sentences for Crime Types
Crime Examples
Combining Prior Record and Crime Examples
Agreement Over Crime Dimensions
Summary and Conclusion


Major Substantive Issues
Data Analysis Techniques
Overview of Findings of Chapter 6 through 9


Trafficking in Illegal Drugs
Robbery of a Convenience Store
Extortion and Blackma
Drug Possession


Money Laundering
Promoting Illegal Tax Shelters
Tax Cheating
Forgery and Counterfeiting


Package Tampering
Pharmaceutical Drug Regulations
Illegal Sale of Firearms
Illegal Firearm Possession


Environmental Crimes - Air Pollution
Environmental Crime - Water Pollution Violations
Environmental Crime - Illegal Logging on Federal Land
Environmental Crimes - Endangered Species Violations
Civil Rights Crimes - Police Use of Unnecessary Force
Civil Rights Crimes - "Hate" Crimes


Analysis Strategy
Regional Differences in Sentencing


Analysis Strategy
Gender Differences in Sentencing
Age Differences in Sentencing
Race and Ethnic Differences In Sentencing
Sentencing Differences by Educational Attainment
Household Income, Occupation, and Sentencing
Net Effects of Respondent Demographic Characteristics
Net Effects of Respondent Demographic Characteristics
The Effects of Political and Social Attitudes on Sentencing
Independent Effects of Demographic, Experiential, and Attitudinal Factors





United States Sentencing Commission