Optimizing Longitudinal Studies in Offending Research Paper

This sample Optimizing Longitudinal Studies in Offending Research Paper is published for educational and informational purposes only. If you need help writing your assignment, please use our research paper writing service and buy a paper on any topic at affordable price. Also check our tips on how to write a research paper, see the lists of criminal justice research paper topics, and browse research paper examples.

Three aspects of potential future longitudinal studies will be discussed: (1) good practices for the execution of a study and the changes that have taken place over time; (2) opportunities to enrich the field of enquiry, thereby enabling new questions to be addressed; and (3) optimizing the use of a data set, the collection of which most likely has been very expensive.

Introduction

Longitudinal studies, with their repeated measurements spanning decades, by their nature are very expensive and should be used only for those research questions that require the study of within-individual changes over time. Thus, they should only be used when a developmental approach to offending will yield information for questions that cannot be answered by cross-sectional studies. Two topics will be discussed: the execution of longitudinal studies and increasing the yield of these studies.

Execution Of Longitudinal Studies

One of the problems in executing longitudinal studies is that generally the management skills and practical experience needed to conduct such studies are rarely taught during graduate training in universities and need to be acquired by researchers by mentors or through books (Stouthamer-Loeber and van Kammen 1995). Moreover, information on how to run such studies is rarely found in articles published in scientific journals. Although we can now learn from a fair number of longitudinal studies dealing with antisocial and delinquent behavior (summarized in Loeber et al. 2008, p. 20; Killias et al. 2012), it is of great help to have a mentor who has handson experience and can help prevent errors.

Investigators need to be involved in all matters of data collection and management and not count on staff to run a study for them. Before researchers can withdraw behind their desks to write papers, they have to be entirely assured that the data are as good as they can possibly be. Involvement in the daily running of a study is not a distraction, but a necessity.

The optimization of the execution of a longitudinal study starts in the planning phase. Running a longitudinal study is not something one can start to think about when the grant has been awarded. The aims, budget, and plans for data collection need to be coherent. Questions that need to be answered ahead of time are the following: Can the aims be measured by the proposed data collection? Is the data collection designed to capture developmental changes? Is the sample large enough and the sample acquisition realistic? Is the timeline realistic? What efforts will be made to continue to follow up the sample? And does the budget fit all the tasks to be accomplished? Also, thought needs to be given to how the potential study results may have a social impact. Although it may be necessary to make adjustments after the project has been funded due to budget cuts, the basic framework should have been worked out in detail. This makes it also easier to discuss with a granting agency the effects of a cut and possibly argue for a lesser cut.

Longitudinal studies have changed in method of data collection and storage. In addition, issues of consent and confidentiality as well as research conditions have changed as well.

Method of Data Collection. One of the questions confronting the researcher early on is whether to contract out the sample selection and data collection or whether to keep these tasks “in-house.” There are pros and cons for either decision. If the research team has no specialized knowledge of sample selection and data collection, it may be easier to use a survey organization. Also, if data collection needs to take place across the country, it may make sense to use an existing network of interviewers affiliated with a survey organization. In addition, some researchers may not have the inclination or the time to be involved in data collection and prefer to receive the data “ready-made.” There are also several arguments that can be made against the use of a survey organization. First, the expense is generally larger than for in-house data collection. More important, however, is that the researchers do not have direct control over the number of contact attempts with potential participants and the quality and completeness of the data and how the participants have been treated.

Survey organizations generally do their own quality control which leaves the researchers to take the quality of the final product on faith. It would be better if researchers would take some of the quality control in their own hands to ensure that the product they receive is as good as it should be. It is not wise to be just a passive recipient of a data set.

If data collection is done in-house, the greatest care should be taken in the selection and training of interviewers. Apart from finding the right persons, it is very important to be up-to-date on the current hiring and firing regulations. These regulations change from time to time and a lot of unpleasantness can be encountered by not being informed. Once the staff is on board, training needs to take place. It is advisable to have an actual written training manual which includes job responsibilities, expectations, contact with participants, and training materials for data collection. A written manual is a help for the interviewers who can consult it to refresh some information. It is also important if interviewers are trained in batches to have the same training materials. In addition, a manual can serve the purpose of evidence that rules and regulations have been laid out.

Before interviewers are allowed to do “real” interviews, they need to be thoroughly familiar with the procedures and the interview. None of the interviews done with participants should ever be used as practice or training interviews. A study is only worthwhile if the data are as close to perfect as possible. There are enough circumstances (over which a researcher has little control) that influence the quality of the data, but not enough interviewer preparation should not be one of them. Although this description generally deals with data collection by interviewing participants, similar procedures can be developed for staffs who collect and code existing data.

Once interviewers have been trained, it is helpful to have a system in place that provides the supervisor with information about quality and quantity of work done by each interviewer. This should also include information about search and contact efforts as well as address changes which can be used in the next assessment phase. If necessary, retraining can take place where procedures need to be clarified. Also errors can often be fixed if noticed early enough but not at the stage when the researcher requires the data for analyses. The information can also be used to give staff members positive feedback and even problems can be turned into learning opportunities. It is useful for interviewers to know that all aspects of their work are important enough for the researchers to be informed about on an ongoing basis. Databases should be created in which the relevant information is deposited so that reports can be created on a regular basis rather than ad hoc when one feels something is not going right.

If interviewers are working in the field, it is a good idea to have a weekly session with their supervisor. During that session, completed materials can be handed in or downloaded and problems can be discussed, including errors that still can be fixed. The supervisor can brainstorm with the interviewer about search strategies for difficult to find participants and how to deal with reluctant participants. It may be useful to convene all interviewers to share solutions to particular problems and to learn from each other’s techniques.

Various changes have taken place over the past years in the form and content of data collection. Data collection has largely moved away from unstructured interviews which required extensive training and coding afterward to interviewing with structured answer formats. The advantage of the latter method is that the questions are more likely to be asked in a uniform manner of each participant and that there is less coding to be done after the interview is completed.

Other changes have to do with the recording of the interview. Most surveys now use electronic data collection, avoiding data entry (Groves et al. 2009). An additional change has been that most studies now use multiple rather than single informants to ensure a more complete picture of the participant and his/her environment. The frequency of repeated measures has increased, and studies are continuing to cover longer periods of time, allowing a larger portion of development of offending to be studied and to move from studying persistence to also taking into account desistance. In these studies desistance from offending can be better estimated if the time window for desistance is sufficiently large to measure stable desistance. These studies also allow for the detection of cases with a late onset.

Although face-to-face interviewing will always have the advantage of ensuring a person’s identity, the ease and the reduction in cost of web-based data collection will increase over time when more people have access to online electronic devices (e.g., Celio et al. 2011; Gosling et al. 2004). Such devices do not necessarily have to be a computer but may be a phone as long as it can connect to a survey program which can provide the questions and can store the data. Various devices may also collect motion or sleep data or can alert the participant to provide some specific information (Anokwa et al. 2009). Web-based data collection makes it easier for participants to continue to participate in a study from practically all over the world. However, web-assisted interviewing still has at least three problems: getting the participant to do it, completing the survey in a confidential manner, and ascertaining the identity of the participant. Thus, face-to-face interviewing, even though it is costly, may still be the most certain way of collecting data.

Data Management. The term “audit worthiness” stands for how well data are handled, cleaned, and analyzed (Freeland and Carney 1992). Errors that may be found in an audit may not be evidence of intentional deception, but more likely the result of carelessness and/or lack of documentation. Any error reduces the usefulness of the data set, and proper data management is of the utmost importance (Stouthamer-Loeber 1993). Magnusson and Bergman (1990) lay out a set of standards on how to manage data.

With regard to data storage, we have gone from mainframe storage with backup tapes that seem to go out of date continuously to in-office servers or just desktop computers. Further changes in data storage are likely to take place. Cloud storage and computing will make it possible to use one’s desktop computer just as a typing device with all the software and data stored somewhere else (Armbrust et al. 2010). Although security problems still have to be worked out, this may also be a way to make data available to certified users. Sharing data collected in large, expensive longitudinal studies is now often a requirement for obtaining funds in the first place. Cloud computing may be a way to deal with the technical, if not the confidential, end of data sharing.

Confidentiality. Robins (1976) remarked on the change that the informed consent procedures had brought to research in the United Sates. Databases that in the past could have been used to select potential participants became off bounds because prior consent was required. The more recent HIPAA law, enacted in 1996, has further restricted access to health and health-care information. Consent forms are now complicated documents. The consent and confidentiality requirements have become stricter over the years and researchers, at least in the United States, have to go to great length to explain procedures and reasons for data collection and how the data will be kept confidential. The proper execution of obtaining consent cannot be stressed strongly enough and should covered thoroughly during training. Consent forms are legal documents and no short cuts can be taken. In addition, all staff should sign a statement that assures the confidentiality of the data and identity of the participants.

Research Conditions. The increased use of answering telephone machines and the availability of the caller number has impeded the contacting of participants. Also the extensive use of cell phones makes contacting more difficult since phone numbers are not listed and cell phone numbers tend to change more rapidly than landlines. On the other hand ways of electronic searching for people have multiplied. For instance, many states have extensive databases online of incarcerated individuals. Different search and contact methods are described in Cotter et al. (2002); Haggerty et al. (2007), and Stouthamer-Loeber and van Kammen (1995).

The success of longitudinal studies depends very much on the sustained cooperation of the participants. In order for a participant to cooperate, the participant first needs to be found, contacted, and agree to another round. Although there are well-developed methods of imputation of missing data, they work best for analyzing group data (McCartney et al. 2006). However, longitudinal studies also allow the study of the developmental paths of individual participants over time, and actual data is far to be preferred over imputation. In many studies, the group of most interest (e.g., families of conduct disordered boys) requires the most attempts to find, contact, and interview (see Cotter et al. 2005; Haggerty et al. 2007). Curtailing search and contact attempts does in general yield a biased subject loss. The cost of searching for participants generally rises over time because of moves. Even though most longitudinal studies ask the participants to provide some names, addresses, and telephone numbers of people who would always know their whereabouts, this information is not always helpful because the contacts themselves may have moved or may have lost track of the participant. Therefore, it is important that the budget reflects the costs of extensive searching for participants.

Longitudinal surveys, by their nature, cannot provide extensive feedback to the participants or provide help or treatment. Thus, in general, a participant may not directly benefit from participating. The reason why people do participate in these studies is because they are curious, they are made to feel special, they are civic-minded, and/or they are paid for their time. Further factors that may influence participation are the time it takes to participate, the ease of participation, and their earlier experience in being interviewed. Since longitudinal studies are expensive and the topic of antisocial behavior and crime may be in or out of favor with funding agencies over time, it is not possible to anticipate if funds available for existing or planned longitudinal studies may fluctuate over time.

Optimizing The Content Of Longitudinal Studies

Whereas the older longitudinal studies relied solely on official records, later studies also used self-reports. Considering the low proportion of crimes that are caught, the addition of self-reports is crucial. More recent studies also try to include community and socioeconomic data as well as reports of routine activities and situational information surrounding the commission of crime. Some biological information has been collected in the past such as heart rate and skin conductance; however, lately the range of biological information has increased by the collection of data on hormones, DNA, and brain activity (fMRI). The combination of individual, environmental, and biological data will make it possible to endeavor to answer complicated questions about interactions within and between these three data domains.

Content. Expansion in the array of factors included in longitudinal studies has been going on for a number of years. Measures have been developed for routine activities (Wikstrom 2006; Wortley and Mazzerole (2008)) and for situational circumstances surrounding the commission of antisocial behavior and crime (Wikstro¨ m 2006). Also, with the enormous amount of demographic information available on the web, community information can be more fine-tuned than it was in the past, and with less effort.

Another development that will take place is that genetic databases will be formed that collect data from different studies to allow for the study of rare factors. This requires careful documentation of design and variables in the individual studies contributing to such a database and a willingness to collaborate.

The study of girls’ antisocial behavior and crime is still underdeveloped. The few studies that included girls often have too small a sample to yield enough girls with problems for analyses. Very few studies have a large enough sample to focus on low base-rate behaviors (review Hipwell and Loeber, 2006). The study of girls is particularly important because they will bring up the next generation of children, often without a male in the picture. Thus, their influence on the next generation is considerable. Antisocial girls often have babies early, which may mean that they are not ready for the responsibility of motherhood. In addition, their sexual partners are likely to be antisocial themselves. This leads to the question of transmission of problem behavior across generations.

An area where progress can be made in the future is the measurement of positive outcomes. Currently in many studies the overwhelming focus is on negative outcomes and on negative factors influencing an outcome. Positive factors are often implied from the lack of negative factors. Well-measured positive factors, such as skill development and competence, and outcomes, such as jobs and stable relationships, will be of enormous help in devising treatment plans.

Most current studies have not included information on the procedures involved in moving through the justice system. So far longitudinal studies have mainly relied on official records which generally do not tell how individual cases are dealt with and how long an individual was in what kind of institution and whether treatment was applied. Such information may be very relevant to the rate of re-offending.

Longitudinal data may, in the future, be used in creative ways to look at issues that have not been extensively studied so far. With the large amount of data collected at regular intervals, modeling exercises can be undertaken to examine what the effect certain interventions could have if applied to a population. Assuming a certain level of success of an intervention, the effect on rates of later problems can be estimated. These exercises can be very useful to convince funders and politicians that certain interventions may pay off over time. Better still would be the integration of an intervention study within a survey study, something that David Farrington has advocated (2003). Such a study would have to be planned, taking into account the number of subjects that will have to be “set aside” for intervention. Such a design would make a longitudinal study even more expensive by increasing the N and instituting an intervention for a part of the sample. After the fact, not many longitudinal researchers are willing, or in a position, to reduce their N by allowing an intervention to take place with a random subsample. This would reduce the power of the survey analyses.

One puzzling question that has not yet had much attention is secular changes in the agecrime curve. Even though in general the agecrime curve has a recognizably similar shape, for different age cohorts, there is also variation is the width and/or height of the curve. Some age cohorts may start delinquency earlier and stop later or may have a higher frequency (height of curve) of offending. The variables that influence the width and the height of the curve have not been studied extensively. Information about which factors influence the age-crime curve is of great importance for public policy and for reducing future crime waves. In order to study influences on the age-crime curve, one would need several age cohorts studied over a period of time so that the ages overlap in the course of the study. If substantial differences are found in the age-crime curves of the different cohorts, then the exercise is to see if one can find variables influencing the curves. It is difficult to predict the optimal age interval between the cohorts because potential changes affecting the agecrime curve lie in the future.

Optimizing The Use Of Longitudinal Studies

The expense of longitudinal studies, generally funded with tax payers’ money, requires serious thought about how to optimize the use of the study.

Collaboration Within Studies. With most studies aspiring to cover many disparate areas such as genetics and environmental influences, future studies will require a team of collaborators with a lead investigator willing and able to commit a large portion of his/her professional life to keeping such a study going and attract collaborators who are experts in specific fields to work with over many years.

Since the academic system values principal investigator ship rather than co-investigator ship, a structure needs to be found that makes it worthwhile to be part of a larger team. One way to do this is to devise substudies each with their own principal investigator and funding. However, such a solution presupposes a stable line of funding for the main study.

Collaboration Across Studies or Study Sites. It is expected that in the future, there will be a better integration of biological and environmental and individual factors. This will probably be accomplished by pooling data from different studies in order to make sure the sample is large enough for small effects to show. Data pooling is often more complicated than it sounds. Presumably “comparable” data may have a slightly different format, or different collection sites may have left some questions out, leading to a cumbersome process of making a combined data set.

The funding agencies are at present the driving force for data pooling and data transfer to agencies set up to deal with data storage and data requests from researchers outside of the original group of investigators. The advantage of wider collaboration, apart from a larger sample size, is that more extensive use is made of the large investment needed for longitudinal studies. The original investigators cannot be experts on all questions that could potentially be examined with the collected data. However, the danger is that regardless of how well the data are documented, “new” researchers may misinterpret the meaning of some variables. It is, therefore, always useful to be in contact with the original investigators. Data sharing provides a challenge for the promise of confidentiality. Even if the most obvious identifiable information is removed from a file, it may still be possible to identify a participant through a combination of variables (called deductive discovery). This puts a participant at risk of being identified and information being misused. In addition, it also puts the original investigators at risk for having to pay a fine for a breach in confidentiality.

In our experience, the safest way of dealing with these confidentiality problems is to have each person wanting to use data write a small proposal that is approved by his/her university and has the appropriate IRB approval, shifting the responsibility of confidentiality to the “new” researcher who was not involved in the original data collection. Since data sharing is now more or less the norm, I expect that procedures for de-identifying data and for certifying that “new” investigators will protect the confidentiality of the data will become standard soon. De-identifying a data set, however, will reduce the number of questions that can be answered after sensitive data has been removed.

Agency Involvement. The size of future studies and the range of variables that should be included may require funding from several agencies. Collaboration between agencies and the acceptance of new thoughts in the field will require that the investigators play a very active role in preparing agencies for the scope of the planned research and for the idea of funding from different agencies. Researchers cannot just wait for a request for proposals to come out that just fits their plans or expect agencies to take the lead role in searching for collaborations with other agencies. Instead, prospective investigators need to work together with agency staff to develop a blueprint for future longitudinal studies and for finding the necessary funding.

Secondary Data Analyses. Secondary data analyses, whether it is by pooling data from different studies or analyzing data from one study for a new purpose, are a great way to increase the yield of the research investment. It is useful, however, to keep a record of secondary research projects so that researchers do not compete with each other on the same topic. That is not to say that it would not be useful to have two different views on the same research question, but researchers should be aware of other projects using the same data bordering on their areas of interest.

In summary, the complex art and science of longitudinal studies continues to be exciting and challenging. New questions can be pursued and new data collection tools spring up regularly.

Bibliography:

Anokwa Y, Hartung C, Brunette W, Borriello G, Lerer A (2009) Open source data collection in the developing world. Computer 42(10):97–99, http://dx.doi.org/10.1109/MC.2009.328
Armbrust M, Fox A, Griffith R, Joseph AD, Katz R, Konwinski A, Lee G, Patterson D, Rabkin A, Stoica I, Zahara M (2010) A view of cloud computing. Commun ACM 53(4):50–58
Celio MA, Vetter-O’Hagan CS, Lisman SA, Johanson GE, Spear LP (2011) Integrating field methodology and web-based data collection to assess the reliability of the Alcohol Use Disorders Identification Test (AUDIT). Drug Alcohol Depend. doi:10.10.1/ j.drugalcdep.2011.05.035
Cotter RB, Burke JD, Loeber R, Navratil JL (2002) Innovative retention methods in longitudinal research: a case study of the Development Trends Study. J Child Fam Stud 11:485–498
Cotter RB, Burke JD, Stouthamer-Loeber M, Loeber R (2005) Contacting participants for follow-up: how much effort is required to retain participants in longitudinal studies? Eval Program Plann 28:15–21
Farrington DP (2003) Developmental and lifecourse criminology: key theoretical and empirical Issues- The 2002 Sutherland Award address. Criminology 41:201–235
Freeland KE, Carney RM (1992) Data management and accountability in behavioral and biomedical research. Am Psychol 47:640–645
Gosling SD, Vazire S, Srivastava S, John OP (2004) Should we trust web-based studies? Am Psychol 59(2):93–104
Groves RM, Mosher WD, Lepkowski JM, Kirgis NG (2009) Planning and development of the continuous National Survey of Family Growth. Vital Health Stat 1:1–64
Haggerty KP, Fleming CB, Catalano RF, Petrie RS, Rubin RJ, Grassley MH (2007) Ten years later: locating and interviewing children of drug abusers. Eval Program Plann 31:1–9
Hipwell A, Loeber R (2006) Do we know which interventions are effective for disruptive and delinquent girls? Clinical Child and Family Psychology Review 9, 221–255
Killias M, Redondo S, Sarnecki J (2012) European perspectives. In: Loeber R, Farrington DP (eds) Transitions from juvenile delinquency to adult crime: criminal careers, justice policy and prevention. Oxford University Press, New York
Loeber R, Farrington DP, Stouthamer-Loeber M, White HR (2008) Violence and serious theft. Routledge, New York
Magnusson D, Bergman L (1990) Data quality in longitudinal research. Cambridge University Press, Cambridge
McCartney K, Burchinal MR, Bub KL (2006) Best practices in quantitative methods for developmentalists. Blackwell Publishing, Boston
Piquero AR, Farrington DP, Nagin DS, Moffitt TE (2010) Trajectories of offending and their relation to life failure in late middle age: findings from the Cambridge Study in Delinquent Development. J Res Crime Delinq 47(2):151–173
Robins LN (1976) Problems in follow-up studies. Am J Psychiatry 134:904–907
Schafer JL, Graham JW (2002) Missing data: our view of the state of the art. Psychol Methods 7:147–177
Stouthamer-Loeber M (1993) Optimizing data quality of individual and community sources in longitudinal research. In: Farrington DP, Sampson R, Wikstrom P-OH (eds) Integrating individual and ecological aspects of crime. BRA˚ -report, Stockholm
Stouthamer-Loeber M, Van Kammen WB (1995) Data collection and management: a practical guide. Sage, Newbury Park
Wikstrom P-O (2006) Individuals, settings and acts of crime. Situational mechanisms and the explanation of crime. In: Wikstrom P-O, Sampson RJ (eds) The explanation of crime: Context, mechanisms and development. Cambridge University Press, Cambridge
Wortley R, Mazzerole L (eds) (2008) Environmental criminology and crime analysis. Willan, Cullompton