As the first Barack Obama administration began, experts assembled by the National Academy of Sciences’ National Research Council issued a report calling for improvements at the U.S. Justice Department’s Bureau of Justice Statistics (BJS)—one of the smallest federal statistical agencies, but one which, as the panel said, “shoulders one of the most expansive and detailed legal mandates.”
In October 2009, President Obama nominated a member of the study panel, James P. Lynch, then a distinguished criminologist at John Jay College of Criminal Justice in New York, to run the agency. Lynch was confirmed by the Senate in June 2010.
After two and a half years, Lynch is leaving to become chairman of the Department of Criminology and Criminal Justice at the University of Maryland. As he packed up his office last month, Lynch sat down with The Crime Report’s Washington bureau chief, Ted Gest. In a wide-ranging conversation, Lynch discussed how methodologies for tracking key indices like crime victimization, sex crimes and recidivism are being revamped; and why there remain significant “holes” in data about the U.S. justice system.
The Crime Report: In June 2010, you came into an agency that had not received any significant budget increases in many years. How are things now that you are departing?
Lynch: We're in reasonably good shape, but it's fragile. I'd call our budget situation “precarious.” The good news is that for the first time in many years, we have more money because of a “set aside” that the Department of Justice (DOJ) requested and Congress provided for us in the budget of our parent agency, the Office of Justice Programs, for statistics and research. This has meant an increase from about $34 million in 2007 to $60 million overall in 2010.
TCR: What was your biggest challenge at the outset?
Lynch: As outlined by the National Academies panel on which I served before my appointment here, the National Crime Victimization Survey (NCVS) needed to be stabilized. The survey is based on interviews with U.S. residents on whether they have been crime victims in the previous six months. Over the previous 20 years the sample of persons interviewed had been cut substantially, by 25 percent or more. In the last two years, we were able to increase the sample size by 24 percent, back to its 1997 level.
Training of interviewers had been suspended during the late 1990s. It has now been resumed, which should make the survey estimates more accurate. The performance metrics by which interviewers are assessed were changed to reflect the content of the training. All this restoration was done in a manner that would allow us to detect its effects on the crime rates.
TCR: When you came into this job, the NCVS produced only national data. In an earlier discussion, you mentioned a plan to study whether the survey could be configured to yield “subnational” data so that crime trends in specific states and cities could be produced. Can you report any progress?
Lynch: Yes. We conducted various simulations, and determined that, for reasonable increases in sample and cost, we could obtain reliable data from the 22 most populous states, and we also could develop statistical models to produce estimates from the other, smaller states. Now we are conducting a pilot project in the seven largest states. Within about two years, BJS should have results from this test which, if successful, would allow us to move to the full 22 states in 2016.
TCR: Does NCVS measure anything beyond basic victimization rates?
Lynch: The major purpose of the NCVS has been and remains annual estimates of the level and change in crime, but it can also be used to ask the public about things related to crime. For example, NCVS includes a supplement done every three years on contacts between the police and public. The public has contact with the police for many reasons other than to report a victimization, and [such] encounters can substantially affect citizen perceptions of the police.
It is useful to get descriptions of these encounters directly from citizens in a systematic way. The most recent report on police-public contacts, issued for 2008, found that the number of people with face-to-fact contact with the police in a year dropped five million to 40 million between 2002 and 2008. The survey also found that nine out of ten residents believed the police were respectful or acted properly. NCVS could be used to learn other “social intelligence” like this.
TCR: Haven't questions been raised about the accuracy of the survey's results on sex crimes?
Lynch: There are a number of agencies and private groups that have done surveys of sexual violence, including crimes such as rape and sexual assault; and the results of these surveys are often different from those in the NCVS. Comparisons of these various surveys have raised questions about the most appropriate methodology to use in self report surveys of sexual violence.
In an effort to resolve these disputes over methodology, BJS has asked the National Academies of Science to begin a 21-month study to “assess the quality and relevance of statistics on rape and sexual assault,” both from the National Crime Victimization Survey and similar surveys by federal agencies and private organizations. We hope that the results will identify the optimum approach to surveying for sexual violence, and suggest whether this methodology can be accommodated in the NCVS or a completely different survey would be required.
TCR: Questions often are raised about the difference between NCVS and the Uniform Crime Report (UCR) issued by the FBI, which is a compilation of reports submitted voluntarily by law enforcement agencies. Are both reports useful?
Lynch: It is essential that we have multiple measures of the crime problem that do not suffer from the same limitations, so that we can have a more nuanced and reliable data on this issue. We have two measures of unemployment: for example, one from a self-report survey (the Current Population Survey), and one from state administrative records of unemployment claims.
Each of these data sources has its own limitations, but together they provide a more reliable picture of unemployment. Policy changes at the state level can affect administrative record data but they will not affect the survey data and vice versa The same is the case for crime. We have a survey based indicator and one computed with administrative records data.
For the most part, the NCVS and UCR have reported consistent results over time. One exception was for 2011, when NCVS found an estimated national increase in violent crime of 18 percent, although that was due in large part to an increase in “simple assaults” and crimes not reported to the police. The FBI reported a decrease in violent crime reports of 3.8 percent, but that includes what the FBI defines as “aggravated assaults” and not “simple assaults,” and in crimes reported to the police.
TCR: Given budget limitations, can anything be done to improve the system of compiling reports to law enforcement agencies?
Lynch: One of the major limitations of the UCR is the fact that it provides only jurisdiction-level counts of eight types of crime, which tells us very little about the crime problem. It would tell us, for example, that 100 rapes occurred in Washington, D.C., but it would not tell us anything about the victims or the offenders, or anything about the social context of the event.
For many years, the FBI has maintained a National Incident-Based Reporting System (NIBRS) that collects much more detailed information about 46 kinds of crimes than are captured in the basic Uniform Crime Report. This has grown slowly over the years, yet it still covers less than one-third of the U.S. population.
BJS has proposed a plan for expanding NIBRS to a nationally representative sample of jurisdictions so that these data can be used to guide national crime control policies. The plan is called the National Crime Statistics Exchange (NCS-X), and we hope it will produce more detailed data from a nationally representative sample of 400 police agencies, including those in the largest cities. Our agency plans to make funding available each year from 2014 to 2016 to encourage more local police departments to provide incident-level data.
In anticipation of the availability of much more detailed information on crime through NCS-X, BJS recently convened a meeting of police chiefs and sheriffs from all of the major police membership organizations, including the International Association of Chiefs of Police, the Major City Chiefs Association, the Police Executive Research Forum, and the Major County Sheriffs Association (MCSA).
We asked the chiefs how they would use this new information to better characterize the crime problem. The chiefs had a number of suggestions, including the production of gun crime rates and domestic violence rates that are not currently available in the UCR.
TCR: What are BJS' main priorities in the years ahead, taking into account the agency's modestly increased resources?
Lynch: In general, BJS, like all statistical agencies, must capitalize on the massive increases in the volume and quality of operational data in the criminal justice system to provide statistical data on the functional of the system. In the last two decades, police, courts and correctional agencies have awakened to the importance of information for working more efficiently and they have begun to collect and analyze more information.
BJS must find ways to use this information for statistical purposes.
I would cite four specific areas in which this is being done. One is recidivism. BJS has invested half a billion dollars to improve criminal history records over two decades. More recently, we have invested in technology that would standardize these RAP sheets across states and automatically code the information in the RAP sheets making them suitable for statistical use. This will allow larger and more accurate recidivism studies to be done faster, and it raises the possibility of routinely producing state level estimates of recidivism, which cannot be done easily now.
We also are working to improve the National Judicial Reporting Program. For 20 years, this data collection has extracted information on sentencing in felony cases from a sample of state courts in 300 counties.
If state court management information systems have grown sufficiently in coverage and accuracy, it may now be possible to simply export data from these management information systems. This would reduce the costs of collection so that we could broaden the coverage of the collection to include all states and not simply a national sample. This would permit state level estimates of sentences that are not possible currently.
It might also be possible to obtain data on misdemeanor cases. To these ends, we are assessing the capabilities of management information systems in state courts.
In the corrections area, we have been redesigning the National Corrections Reporting Program (NCRP). Since the early 1980s, this data collection has been used to monitor admissions and releases to state prison using annual cohorts of persons admitted and released. We used these data in a model to estimate time served by age, race, gender and offense of the prisoner. With the redesign, we will be able to link a person's admission record to his specific release record to obtain a much more accurate and more data-based estimate of time served in prison
Finally, the NCS-X program mentioned earlier will only succeed if the incident-level police data can be extracted from management information system of local police departments. The NCS-X project is assessing this as we speak.
TCR: Do any BJS initiatives go beyond collection of basic crime and punishment data?
Lynch: Yes. We are trying to fill in the holes in other parts of the justice system. A good example is a pilot study now going on regarding services provided by agencies that serve crime victims. As we said in the announcement of this project, “Victim service agencies are an important and untapped source of information on victims of crime and the services they receive. These organizations know which victims have been referred to them from the police and other sources. They know the source of referral, the crime for which they were referred, the services they received, the cost of those services, the source that funded service delivery, and possibly the outcome of any service provided.”
We need data like this to determine what is and is not being done for crime victims, as well as what is effective and what is not effective.
TCR: Are you concerned that there is less interest in supporting basic federal work in gathering and analyzing crime statistics, now that crime rates clearly have dropped in recent years?
Lynch: No. Basic statistical data of the type collected by BJS are what tell us definitively that crime has dropped. You don't hear that question being asked about areas like unemployment and health as indicators get better or worse. No one would suggest that we only collect data on unemployment when lots of people are out of work. It shouldn't be asked about crime either.
We should not wait until crime problems get worse to gather basic data about what is happening.
Ted Gest is president of Criminal Justice Journalists and Washington DC bureau chief of The Crime Report. He welcomes comments from readers.