Help 4

a t

Designing Research

78

TYPES OF RESEARCH DESIGN

If the authors of this book have learned anything in their varied career lives, it is that there is no right or wrong method or research design. The proper method and design, if one can actually use the term proper here, depends on what you are trying to. achieve in your research. In turn, what

L you are trying to achieve depends heavily on your research idea. Thus, the starting point of any

U. t~ cc

research study is the research idea. Your research idea then becomes the engine that pulls the rest G

of the research train. From one's research idea, one develops his or her design. The research re

design literally becomes the plan the researcher will use in undertaking the study. This will in- re

elude identification of any theoretical framework that may bring together the overall orientation w

of the study, as well as identification and selection of the methodological technologies to be used St

in collecting data. Furthermore, the design will indicate how data, once collected, will be organ- (p ized and analyzed. of

Regardless of these truisms—that there are no right or wrong methods or designs—many inexperienced researchers (along with a fair number of so-called experienced ones) regularly use do methodological designs and data collection strategies that are quite inappropriate for their partic- ular research idea. What we mean by this, of course, is that when you want to simply examine gr

what may be going on in some natural setting, undertaking some sort of experimental design will simply not work. Similarly, if you are interested in testing whether some action or variable seems

trc

to influence some other action or variable, undertaking a long involved descriptive ethnographic field—based study is not likely to yield appropriate data or results. Thus, it is useful for re-

ex

searchers to understand the distinctions between each of these types of general research design tre

orientations. The following sections will undertake to offer a description of each of the major re- ate

search designs, namely, experimental, inferential, descriptive, and exploratory designs. so. jec In

Experimental designs typically seek to address questions having to do with cause and effect pe. (Shadish, Cook, & Cambell, 2001). In scientific notation, this is often depicted in the following way:

no do

X > Y Fu alc cre

Experimental Designs

tF

This phrase is usually read as "X causes Y." In causal relationships, the X variable is referred to as the independent variable, while the Y variable is referred to as the dependent variable. One way to remember this is that the Y variable is dependent on the X variable to occur. In other words, in the absence of the X, there would be no Y or changes in Y.

In order for a causal relationship to be accepted, three conditions have to be met. These conditions include that (1) X precedes Y in time (a temporal element), (2) changes in X result in changes in Y (a covariant relationship between the two variables X and Y), and (3) there are no other alternative variables that may be influencing the relationship so that it appears X causes Y In other words, there is no alternative variable (usually depicted as Z) actually creating the

tha lev we tro: wh inc net

Chapter 4 • Designing Research

appearance that X causes Y (what is usually referred to as a spurious effect). This spurious rela- tionship also has a scientific notation as follows:

there is no tctually use i turn, what point of any tills the rest e research his will in- orientation to be used

11 be organ-

gns—many :gularly use :heir partic- ly examine design will iable seems hnographic eful for re- arch design to major re- fs.

and effect owing way:

referred to riable. One ur. In other

met. These X result in

:here are no X causes Y. :reating the

Z

X > Z'

In order to accomplish an experiment, the researcher uses cases or people, typically referred to as subjects, whose features or characteristics—what may be labeled variables—are examined in terms of certain hypothesized relationships; namely, that some variable has a causal relationship with another variable (Ryan, 2007).

Ideally, experiments should have at least two groups drawn randomly from a research pop- ulation: a control group and an experimental or treatment group (the distinctions between these two groups will be drawn out later). In what is sometimes referred to as a classical experiment, certain important conditions must be met in developing these two groups (Babble, 2007). First, the research must identify the full research population in question, including the parameters of this population; in other words, all of the elements that make up this full working population. Let's say, for example, we hypothesize that when college students at the State University—eat Granny Smith apples, they develop stomachaches. Thus, our causal statement becomes Granny Smith apples [X] cause stomachaches in State University students [Y]. In this example, the full research population would be all students attending State University at a given point in time when the study was being conducted.

A second condition for our experiment is that subjects taken from our full population (all students attending State University) must be drawn randomly. This means that every student (potential subject) must have an equal opportunity or chance of being selected into one or the other of the two groups to be used in the experiment.

A third condition is that as subjects are selected from the full population, they must be ran- domly assigned to one or the other of the two groups (the control group or the experimental group).

At this point, many readers are probably asking, "What is the difference between the con- trol and experimental groups?" Well, the control group will be used as a comparison to the exper- imental group. This means we will not do anything to the control group. On the other hand, the experimental group will have the variable in question, what may also be referred to as the treatment applied to it. In order for this to be an effective comparison, however, we need to cre- ate our two groups as equivalent to each other as we possibly can.

In an ideal world, we could clone one group to create the other, and we would have ab- solutely identical subjects in each group. Since this is not possible, we must settle on similar sub- ject groupings, at least in terms of various demographic characteristics of subjects in each group. In this way, when the experimental group is given some treatment—that is, subjected to the ex- perimental variable—any difference that appears among subjects in the experimental group, and not in the control group, may be attributed to the treatment (the independent variable). So, how do we create this equivalency between the two groups? There are essentially two major ways. First, we can create a matched sample in each group, where we will have absolute equivalence along certain limited demographic elements among subjects in each group, and second, we can create approximate equivalence in each group through randomization of selection.

In the first scenario, the matched sample, we would identify certain demographic factors that seem important or relevant to the study focus. This might include age, gender, educational level, family income, religious affiliation, and similar demographic factors (Rubin, 2006). Next, we seek to include subjects with specific aspects of each of these demographics in both the con- trol and the experimental groups. In other words, if we identify a subject for our control group who is a 22-year-old man, a sophomore at State University, comes from a family whose annual income last year was $65,000.00, and who declares membership as a Lutheran; then, we would need another subject with these identical demographic factors to be placed into the experimental

I

Chapter 4 • Designing Research

group. In a matched sample, the researcher can feel fairly confident that at least along the matched variables, the two groups are absolutely equivalent.

In the second scenario, approximately equivalent groups, the researcher relies on chance obtained through randomization to create approximate equivalence in the two groups. In other words, if each subject is selected from the full population and is then randomly assigned to one or the other of the two groups, the laws of probability suggest that approximately the same pro- portion of each demographic element among subjects in each group should be obtained. The ex- periment relies on this idea of group equivalence as a basis of making a comparison between groups after the experimental group has been subjected to whatever the treatment (the experi- mental variable) may be. Again, the notion is that since the groups are approximately equivalent, when the experimental group is exposed to the treatment variable and the control group is not, and a change occurs among the experimental group subjects, it can be attributed to their exposure to the experimental variable.

In our current example, that would mean that the experimental group would be given some Granny Smith apples to eat to see if subjects in this experimental group develop stomachaches, while the subjects in the control group would be given nothing (no apples). The initial basic logic to this experiment, then, would be to see if the State University students given Granny Smith ap- ples to eat did develop stomachaches (our experimental group) while the other group (the control group) of State University students did not (since they did not eat the Granny Smith apples). In other words,

Bating Granny Smith apples [X] Stomachaches {Y]

The problem with this simple version of the experiment, as some readers may have noticed, is that we do not, from the current information offered, know if there may have been other alterna- tive variables at play that are causing the stomachaches (a spurious relationship). So, what we need to do is to control a number of other conditions to ensure the only variable that may be caus- ing the stomachaches is, in fact, the ingestion of the Granny Smith apples. This is usually re- ferred to as manipulating the independent variable or controlling external conditions. How might we accomplish this in our current experiment? First, we would need to ensure that all subjects in the experiment (both in the control and in the experimental groups) are healthy and free from any flu or stomach illnesses. We might accomplish this by having each student examined by a physi- cian. We might also want to ensure that none of the students are allergic to apples. Furthermore, we might want to ensure that, as much as possible, these students have not ingested anything else that might cause stomachaches. So, we might house both groups of students in a similar dormi- tory for 24 hours prior to the actual experiment (when we give the experimental group the Granny Smith apples to eat), thereby feeding them and controlling what all the students eat, thus guaranteeing that everyone eats exactly the same diet of foods for these 24 hours. By controlling these environmental elements, we effectively control the possibility of a spurious relationship creating the appearance of cause (Gravetter & WalInau, 2004; Rubin, 2006).

When a researcher takes all of the precautions outlined above to create the two groups and undertake the experiment, the process is usually described as a classical experiment. However, there are many occasions when a researcher is unable to determine the parameters of a full population and thus cannot legitimately undertake a genuine random selection from a bounded full population when creating the two comparison groups (the control and experimental groups). In these cases, the research may be said to be undertaking a quasi- experimental design.

The necessary aspects of undertaking a classical experiment may be listed as follows:

1. Identify a research population, and determine its parameters. 2. Randomly select subjects to be placed into an experimental or control group. 3. Arbitrarily (randomly) assign subjects, as selected from the population, to one or the other

of the two groups (the experimental or control group). 4. Manipulate the independent variable (to control for spuriousness). S. Apply the treatment variable (the experimental variable). 6. Observe if a change occurs in the experimental group as compared with the control group.

Under many circumstances, it may be impossible to include all of the elements that go into the requirements of a classical experiment, but some sort of experimental design, often referred to as quasi-experimental, may still seem desirable.

QUASI-EX experimen assigned si interested. sible to ide a family n taming the churches v such a pol placed intc tors among appear in been rando groups to t lions, the re This is usu;

Single Grou school read was effectii graders in t half to a cot tablish a ba program fot the end of tl in the exper elude that ti serious ethic ing is a verb els, in fact, advanced pt experiment over the stu types of cla ing skill de` dents may I possible anc measure, an gram, all the rendering of perimental c group has be

Inferentw~

Inferential r• and to readl use inferent about a give: ity that anX as the term is conditions;; ing on in ow

Most c known as the of covariance analysis, mul the importan become fami

Chapter 4 • Designing Research 81

g the

lance other

one : pro- se ex- ween :peri- alent,

not, sure

some ches, logic h ap- 'ntrol ~). In

;d, is aa- ►t we :aus- y re- iight ;ts in iany hysi- iore, else

the thus lung iship

oups tent. titers :lion and

casi-

)ther

•oup.

the to as

c,.

QUASI-EXPERIMENTAL DESIGN A quasi-experimental design is one that looks a bit like an experimental design but lacks the key ingredient, namely, a completely randomly selected and assigned set of subjects into the two groups (Shadish et al., 2001). For example, let's say you are interested in studying asthma among Latino families in Madison, Wisconsin. It would be impos- sible to identify all of the Latino families residing in Madison, Wisconsin, let alone all those with a family member with asthma. On the other hand, you could create a research population con- taining these factors by advertising in local Latino-oriented newspapers and recruiting in local churches where the congregations are composed largely of Latino families. Once you created such a population, you could now employ randomization in your selection of subjects to be placed into the two study groups (control and experimental). Thus, the various demographic fac- tors among those individuals who found themselves in your created research population should appear in approximately the same proportions in each of the two groups for which they have been randomly selected. In some situations, you may be unable to create two completely separate groups to use in comparison with one another during the course of an experiment. In such situa- tions, the researcher may use the same group as both the control group and the experimental group. This is usually referred to as a single group before and after comparison experimental design.

Single Group Before and After Comparison Experimental Design Let's say you are an elementary school reading teacher interested in testing to see if a new experimental modular reading program was effective on sixth graders. One way to undertake this experiment is to create a list of all the six graders in the school, and then randomly assign half of the students to an experimental group and half to a control group. Next, you would administer a reading test to all subjects (both groups) to es- tablish a baseline reading level of students. Following this, you would begin the modular reading program for only the experimental group. Let's say this program takes four months to complete. At the end of the program, all of the students will be administered the reading test again. If the students in the experimental group have higher reading scores than those in the control group, we might con- clude that the modular reading program was effective. Success! Or is it? Actually, there are some serious ethical problems with the experimental design outlined above. Here is the problem: Read- ing is a very important skill in education, especially during elementary school years. Reading lev- els, in fact, are used to determine, for example, whether a student may be qualified for special advanced programs in science, math, and social studies. If the experiment allows the students in the experimental group to improve their reading skills appreciably, then they may have an advantage over the students in the control group when it comes to assignment into these various specialized types of classes. Conversely, if the program does not work well, and perhaps even slows the read- ing skill development of the students in the experimental group, then the experimental group stu- dents may be permanently disadvantaged as compared with the control group students. One possible and more ethical alternative would be to give all the students the reading test as a baseline measure, and then all of the students the reading module program. At the completion of the pro- gram, all the students would again be administered a reading test. By using the scores of the first rendering of the reading test as the control, and the second rendering of the examination as the ex- perimental offering, one can measure whether the modular reading program was effective the same group has been used as both the control group and the experimental group (Creswell, 2008).

Inferential Research Designs

Inferential research designs use inferential statistics to draw out associations between variables and to reach conclusions that extend beyond the immediate data alone. For instance, we night use inferential statistics to try to infer from the sample data what the population might think about a given issue. Or, we might use inferential statistics to make judgments about the probabil- ity that an X variable will causeY variable to change in a hypothesized causal relationship. Thus, as the term implies, we use inferential statistics to make inferences from our data to more general conditions; as will be indicated later, we use descriptive statistics simply to describe what's go- ing on in our data.

Most of the major inferential statistics derive from a general family of statistical models known as the general linear model. This includes the t test, analysis of variance (ANOVA), analysis of covariance (ANCOVA), regression analysis, and many of the multivariate methods such as factor analysis, multidimensional soling, cluster analysis, discriminant function analysis, and so on. Given the importance of the general linear model, it's a good idea for any fledgling social researcher to become familiar with its workings: however, there is neither time nor mace in this text to fully

82 Chapter 4 • Designing Research

elaborate on this model or the various inferential statistics associated with it. Suffice it to say that the

logic of the general linear model would imply only consideration of a straight-line model. However, upon some investigation, the reader will learn that linear models may include curved lines as well, sometimes referred to as curvilinear analysis (Creswell & Piano-Clark, 2007).

Descriptive Research Designs

Descriptive research designs, as implied in the terminology, are intended to describe what is go- ing on during some event or among some groups of people (Babbie, 2007). The general process involves the researcher carefully observing what is occurring in some setting, or during some event, and then detailing what he or she has witnessed. Because scientific observation is careful, systematic, and deliberate, descriptive research studies are typically more accurate than mere reminiscence from a casual observer who may also have been witness to some event. Further- more, descriptions may derive from lengthy field explorations in which the researcher collects reams of detailed field notes or systematically collects form surveys during a single or series of administrations. The U.S. Census is often pointed to in textbooks as an excellent example of a study that uses a descriptive research design. The major goal of the U.S. Census is to describe a variety of characteristics of the U.S. population at the time of the data collections (the U.S. Cen- sus is administered every 10 years). Other examples of descriptive studies might include an examination of individuals who have been successful at long-term maintenance of a large weight loss, or a product-marketing survey that describes the kinds of people who wait in lines—sometimes for days at a time—in order to purchase new cutting-edge electronic devices, such as the latest telephone or video-game player. Or, a political researcher who assesses the voting patterns of state legislators may also be creating a descriptive study.

While many qualitative studies may be aimed at describing some condition or phenome- non, it should be obvious from the preceding examples that descriptive studies do not necessar- ily need to be qualitative in their orientation and that many quantitative studies are also constructed from descriptive research designs (Rubin & Babbie, 2007).

Exploratory Research Designs

Sometimes research is conducted in an effort to better understand exactly what is going on dur- ing an event or phenomenon, or to simply explore some topic where little empirical research may have previously been conducted (Marczyk, Dematteo, & Festinger, 2005). For example, when gas prices soared during the summer of 2008, rising to nearly $5.00 a gallon across the United States, what would have happened if many U.S. citizens had decided to boycott the gas pumps for one day a week for a month? Or perhaps a better question might be, what would need to oc- cur to motivate people to undertake such a boycott? Either of these questions might have re- quired an exploratory research design. One might be interested, for example, in considering how such a movement might be organized? Which types of people would be most likely to take part in such a boycott? An exploratory study would allow one to learn, at least approximately, what the answers are to these sorts of questions.

Exploratory designs are also useful for more pedestrian types of issues and phenomena. For example, let's say you are unhappy with having been refused admission into some profes- sional school, such as a law school or medical school. You might study the history of admissions to such programs and those of the particular schools to which you have been unsuccessful in gaining admission. You may, as a consequence of this research, learn that your characteristics did not meet those of the schools you applied to, but may be better suited to some other school's pro- gram and admissions standards.

Like descriptive studies, sometimes the research design calls for various qualitative strategies; but in other situations, quantitative strategies such as surveys may be used. In the example above, one might research admissions requirements using a more qualitative archival strategy (locating ad- missions requirements in college bulletins either in libraries or on the Internet). Or, one might survey students who were successful at gaining admission in order to determine what characteristics they seem to share in common (e.g., GPA levels, collecting butterflies, extracurricular undergraduate col- lege activities, certain scores on professional examinations such as the MCATS and LSATS).

There are three major reasons one might seek to undertake an exploratory study. These rea- sons are as follows: (1) to satisfy a researcher's curiosity about some situation, issue, or phenom- enon, where little or no previous empirical research may exist; (2) to test the feasibility of a

larger; hypotl

Expl~

When occur why o come; tory r identi: planat disco} crime to det crime one tb sider explai

the fo

1. 2.

3.

4.

DEV

As Et startir ing ar daunti taking comrr: report report been these

thing searcl the in those wear some the of for at ten is: about anythi Ideas witho the re: refine suit h resear

f: Chapter 4 • Designing Research

say that the 1. However, ies as well,

what is go- ral process iring some is careful, than mere tt. Further- ier collects )r series of ample of a describe a U.S. Cen-

include an irge weight -sometimes s the latest patterns of

phenome- t necessar- :s are also

ng on dur- ;earch may iple, when the United gas pumps teed to oc- it have re- lering how o take part ately, what

henomena. me profes- idmissions xessful in xistics did pool's pro-

strategies; iple above, ccating ad- ight survey ristics they aduate col-

rs).

These rea- r phenom- Lbihty of a

larger, more extensive study (perhaps a larger inferential study); and (3) to develop methods or hypotheses to be tested in some subsequent study.

i

Whereas descriptive and exploratory research designs look at what, where, when, and how things occur during some situation, event, or phenomenon, explanatory research designs examine the

' why of things. Thus, the desire to explain the purpose of something, or some phenomenon, be- comes the main objective in explanatory research designs (Creswell, 2008). Sometimes, explana- tory research will build on descriptive and/or exploratory research studies by going further in identifying the reasons for something that occurred, or that occurred in a particular manner. Ex- planatory research looks for causes and reasons. For example, a descriptive research study may discover that 60% of juveniles officially identified as delinquents after adjudication from some

f crime were physically abused by their parents. An exploratory orientation, however, might seek to determine why having been physically abused by one's parents led the youth to commit a crime; or even why the parents of delinquents tended to abuse their children. In other words, it is one thing to determine crime rates for some juvenile category of crime; it is quite another to con- sider why these juveniles committed the crime. This latter effort becomes the principal focus of

I explanatory research. There are a number of generally held goals in explanatory research designs. These include

the following:

1. Explaining phenomena in an effort to enrich previous theories. 2. To ascertain which among a series of possible theories or explanations seems to be the best

theory or explanation. 3. To further, enrich, or refine, current levels of understanding about some phenomenon,

event, or situation. i 4. To provide evidence to support or refute an explanation or prediction about something.

7 I

Explanatory Research Designs

DEVELOPING RESEARCH IDEAS

As Berg (2007, p. 21) suggests, "Every research project has to start somewhere; typically, the starting point is an idea." The obvious next question, of course, is "How does one go about find- ing an idea that will serve as a good research topic?" For many new researchers, this can be a daunting task in itself. Ironically, many researchers arrive at their initial research idea simply by taking stock of their personal interests or by looking at what may be happening in their local community, the nation, or society. For example, listening to a new broadcast, you might hear a report indicating that there is an increase in juvenile crime, that several young girls had falsely reported that their gym teacher had fondled them in the locker room, or even that an orderly had been arrested for sexually assaulting several women in a maternity ward of some hospital. Any of these stories may spawn ideas for conducting research.

Similarly, many people have personal experiences that cause them to take interest in some- thing that they would like to know more about, and this too may lead to their undertaking re- search. For example, while sitting in your Introduction to Criminal Justice class, you notice that the instructor seems to repeatedly call on students who are dressed in nice clothes and to ignore those wearing sloppy, torn, or very casual clothing. You might begin to wonder if what people wear has anything to do with teacher affinity, or even, perhaps grading. This too might lead to some sort of a research study. A person taking a ride along with a police officer might notice that the officer's interactions seem different depending on the age of the person whom he or she stops for a traffic citation—perhaps being more lenient about not issuing a ticket to adults and more of- ten issuing tickets to young people. Again, this might lead to the ride-along passenger thinking about how a study of this deference might be undertaken. The long and short of it is that almost anything you see or do, learn about, hear about, or experience may be fodder for a research study. Ideas are all around us; the real trick is to select one of these ideas for which to undertake a study, without being continuously sidetracked by all the other possible research ideas. The next step in the research process is to take this potentially broad and even unstructured idea and turn it into a refined and sleek research topic or question to actually study. Toward this end, you need to con- sult the literature to uncover any theories that may inform you about this idea as it morphs into a research topic or question.

84 Chapter 4 • Designing Research

THEORY AND RESEARCH

Once you have identified the general idea for the study, the next thing you need to do is examine what other researchers have previously thought about and studied with regard to this topic. Say, as an example, you have an idea about preventing juveniles caught up in minor acts of delin- quency from becoming further entangled in the juvenile or criminal justice system. You might begin by examining literature on previous research-related topics such as delinquency preven- tion, primary delinquency interventions, secondary delinquency interventions, alternative to ad- judication programs, net widening, and similar related topics. The logical way to accomplish this is through a library search and a Web-based search on the Internet. Once you have identified lit- erature on your topic and begin reading through it, you should notice that your research topic is becoming refined and starting to sound more like a traditional research problem or question.

So, using our current example, you may read various studies and theories offered as expla- nations in these studies, and become interested with the notion of net widening. In a nutshell, net widening involves placing juveniles who have committed minor infractions of the law under some form of observation or control of the justice system, or an agency working for the goals of the justice system (see Bechard, 2006; Macallair & Roche, 2001). Now you might reformulate your idea into a more viable research question, such as "Does net widening target appropriate juvenile populations?" Your direction and orientation, then, has been informed and shaped by the current theories and research on the topic found during your search and reading of the existing literature. In other words, your research question didn't just spring forth spontaneously but was influenced by the literature on delinquency prevention and net widening that you read about.

References

Babble, E. (2007). The practice of social research (11th ed.). Belmont, CA: Thomson/Wadsworth. Berg, B. L. (2007). Qualitative research methods for the social sciences. Boston, MA: Allyn & Bacon. Bechard, S. (2006). The effects of net widening in a youth diversion program. Dissertation Abstracts Inter-

national, 45(05). (U1VII No. 1442646) Creswell, J. W. (2008). Research design: Qualitative, quantitative, and mixed methods approaches (3rd

ed.). Belmont, CA: Sage. Creswell, J. W., & Plano-Clark, V. L. (2007). Designing and conducting mixed method research. Belmont,

CA: Sage. Gravetter, R I., & Wallnau, L. B. (2004). Statistics for the behavioral sciences (6th ed.). Belmont, CA:

Wadsworth. Macallair, D., & Roche, T. (2001). Widening the net in juvenile justice and the dangers of prevention and in-

tervention. National Criminal Justice Reference Service Abstracts. Retrieved June 8, 2007, from http://www.ncjrs.gov/

Marczyk, G. R., Dematteo, D., & Festinger, D. (2005). Essentials of research design and methodology. Malden, MA: John Wiley & Sons.

Rubin, A., & Babble, E. (2007). Research methods for social work (6th ed.). Belmont, CA: Brooks/Cole. Rubin, D. B. (2006). Matched samples for causal effects. New York: Cambridge University Press. Ryan, T. (2007). Modern experimental design. San Francisco, CA: Wiley Interscience. Shadish, W. R., Cook, T. D., & Cambell, D. T. (2001). Experimental and quasi-experimental designs for

generalized causal inference (2nd ed.). Boston: Houghton Mifflin.

EXE

AD1

TtJE

wwv

NUJ

WA decry Bure of fe

and porte

Coal

bight AID;

203 i cent:

1,59: AID; in 20 100,(

lems, STD: rents

half r either

portef

prior inmal a nee. feden

eral it state i

of fec perce:

were http:/, mpp/t gram:

examine::; )pic. Say;'.: of delin- ou might preven-

Ve to ad- plish this .tified lit- h topic is ;tion. as expla- shell, net tw under goals of

ormulate ?ropriate ~d by the existing but was

bout.

acon. cts Inter-

rhes (3rd

Belmont,

ont, CA:

n and in- 37, from

odology.

)ks/Cole.

signs for

Exercise 41 Descriptive Research

ADVANCE FOR IMMEDIATE RELEASE

TUESDAY, APRIL. 22, 2008

www.oip,usdoj.gov/bjs

Chapter 4 • Designing Research 85

Bureau of Justice Statistics

Contact: Stu Smith: 202-307-0784

After hours: 301-983-9354

NUMBER OF HIV-POSITIVE STATE AND FEDERAL INMATES CONTINUES TO DECLINE

About 44 percent of state inmates and 39 percent of federal inmates have medical problems

WASHINGTON—Between 2005 and 2006 the number of state and federal prisoners who were HIV-positive decreased 3.1 percent—from 22,676 to 21,980 inmates, according to a report by the Justice Department's Bureau of Justice Statistics (BJS). Another BJS report estimated that 44 percent of state inmates and 39 percent of federal inmates reported a current medical problem other than a cold or a virus.

Sixteen states and the federal system reported a decrease in the number of HIV-infected prisoners and 25 states reported an increase from 2005 through 2006. Texas, with 293 more HIV-positive inmates, re- ported the largest increase. New York with 440 fewer HIV positive prisoners reported the largest drop.

On December 31, 2006, an estimated 5,977 inmates had confirmed AIDS, up from 5,620 in 2005. Confirmed AIDS cases accounted for more than a quarter of inmates known to be HIV positive.

At yearend 2006, the rate of confirmed AIDS in state and federal prisoners was more than 2i; times higher than in the U.S. population. About 46 in 10,000 prison inmates were estimated to have confirmed AIDS, compared to 17 per 10,000 persons in the general population.

During 2006, the number of AIDS-related deaths in state and federal prisons totaled 167, down from 203 in 2005. In 2006, nearly 5 percent of state inmate deaths were attributable to AIDS, down from 34 per- centin 1995.

At yearend 2006, 0.9 percent of federal inmates (1,530) were known to be HIV positive, down from 1,592 in 2005. Of the federal prisoners known to be HIV positive at year-end 2006, 656 had confirmed AIDS, up from 594 in 2005. During 2006, 12 federal inmates died from AIDS-related causes, down from 27 in 2005. During 2006, the rate of death due to AIDS-related causes among federal prisoners was 6 per 100,000 inmates.

Current medical problems among prisoners include arthritis, asthma, cancer, diabetes, heart prob- lems, hypertension, kidney problems, liver problems, paralysis, problems due to a stroke, hepatitis, fly, J l'Us, or tuberculosis. More than a half of female inmates and over a thud of male inmates reported a cur- rent medical problem in a national inmate survey.

Among inmates admitted to state and federal prisons, about an eighth reported having surgery, and half reported having a dental problem. A third of state inmates and 28 percent of federal inmates reported either an accidental injury or a fight-related injury since admission to prison.

More than a thud (36 percent) of state inmates and nearly a quarter (24 percent) of federal inmates re- ported having an impairment, including a learning, speech, hearing, vision, mobility, or mental impairment.

More than half of state (51 percent) and federal inmates (56 percent) who were homeless in the year prior to arrest reported a current medical problem compared to 43 percent of state and 38 percent of federal inmates who were not homeless. Sixty percent of state inmates and 58 percent of federal inmates who used a needle to inject drugs reported a current medical problem compared to 40 percent state and 36 percent of federal [inmates] who did not.

Among inmates who reported a medical problem, 70 percent of state inmates and 76 percent of fed- eral inmates reported seeing a medical professional because of the problem. More than 8 in 10 inmates in state and federal prisons reported receiving a medical exam or a blood test since admission.

Among females in state prisons, 4 percent said they were pregnant at the time of admission; 3 percent of federal inmates were pregnant. Of those in state prisons who said they were pregnant at admission, 94 percent received an obstetric exam. More than half (54 percent) received some type of pregnancy care.

The reports, HIV in Prisons, 2006 (NCJ-222179), and Medical Problems of Prisoners (NCJ-221740) were written by BJS Statistician Laura M. Maruschak. Following publication, the reports can be found at http://www.ojp.usdoj.gov/bjs/pub/html/hivp/2006/hivp06.htm and http://www.ojp.usdoj.gov/bjs/pub/htmll mpp/mpp.htm. For additional information about the Bureau of Justice Statistics's statistical reports and pro- grams, please visit the BJS Web site at http://www.ojp.usdoj.gov/bjs.

86 Chapter 4 • Designing Research

Application Exercise 4.1 Exerc

Name of Student:

Student ID No.:

Course/Section No.:

Date:

1. The Bureau of Justice Statistics press release on inmate HIV status discusses several key findings. These are primarily descriptive, meaning that it describes a sample but makes no argument about causes or implications. Name three primary findings detailed in this report?

2. What empirical research questions might be sparked from this information?

3. What are the implications of these findings for prison administrators?

4. What implications do the findings have for public health?

5. What fiscal implications do the findings have?

Sexual

SUMP

Since t] ous typ ing up relative leave,a ited we hornich high ris this prc main of explora

PROJE

This sti

1., I

2.] 7

3. ] 4. ]

9. '

attentic equally social does n~ womer.

legally womer here, I iting re

19805,

Chapter 4 • Designing Research

Exercise 4.2 Exploratory Research

Sexual Assault During and After Separation/Divorce: An Exploratory Study

Walter S. DeKeseredy

MARCH 2007 U.S. DEPARTMENT OF JUSTICE

:y findings. bout causes

SUMMARY

Since the 1970s, social scientists have greatly enhanced an empirical and theoretical understanding of vari-

ous types of woman abuse in ongoing heterosexual relationships. However, although we know that break- ing up with a violent man greatly increases a woman's risk of experiencing lethal and non-lethal violence,

relatively little empirical and theoretical attention has been paid to the victimization of women who want to leave, are in the process of leaving, or who have left their marital/cohabiting partners. Furthermore, the lim-

ited work that has been done on this topic focuses primarily on physical violence, such as beatings and

homicide. Abuse, of course, is multidimensional in nature, and a few studies show that women are also at

high risk of being sexually assaulted during and after separation/divorce. Still, almost all of the research on

this problem, regardless of whether it is qualitative or quantitative, was conducted in urban areas. Thus, the main objective of this report is to help fill two major research gaps by presenting the results of a qualitative, exploratory study of separation/divorce sexual assault in rural Ohio.

PROJECT GOALS AND OBJECTIVES

This study was specifically designed to provide answers to the following questions:

1. Are survivors of separation/divorce sexual abuse also victims of physical and psychological abuse, or

is sexual assault the only type of abuse they experience? 2. Is there evidence indicating that separation/divorce sexual assault is a major problem in rural com-

munities? 3. Is sexual assault more frequent and severe during or after separation/divorce? 4. Is separation/divorce sexual assault multidimensional in nature? For example, do survivors of this

abuse, like survivors of sexual assault in ongoing relationships, experience different types of forced sexual activity?

5. Based on survivors' point of view, what are the major characteristics of men who sexually assault their ex-partners?

5. What are the psychological, physical, and economic effects of separation/divorce sexual assault? 7. Based on survivors' perspectives, what types of social support and intervention strategies are most

effective? 8. What new directions should be taken to develop and test explanatory models of separation/divorce

sexual assault? 9. What are the implications for further qualitative and quantitative research on separation/divorce sex-

ual assault?

In addition to trying to enhance a social scientific understanding of a problem that has garnered limited attention from the media, the scientific community, and the criminal justice system, the research team was equally concerned with generating policy-relevant data that can be used to tailor more effective prevention and social support services for a group of women who continue to suffer in silence. Too often, separation/divorce does not end abuse and thus it is necessary to develop policies and practices that meet the unique needs of women victimized by sexual violence during and after the process of leaving marital/cohabiting relationships.

A broad definition of separation/divorce guides this study. For example, a woman does not need to be legally tied to a man to experience sexual assault during or after exiting a relationship. Further, many women cannot leave a relationship for a host o£ reasons but emotionally separate from their partners. Thus, here, I use the term separation/divorce to mean physically, legally, or emotionally exiting a marital/cohab- king relationship. Further, guided by empirical work done by Dr. Mary Koss and her colleagues in the late 1980s, the types of sexual assault described by 43 rural Ohio women was classified as follows;

• Sexual contact includes sex play (fondling, kissing, or petting) arising from menacing verbal pres- sure, misuse of authority, threats of harm, or actual physical force.

• Sexual coercion includes unwanted sexual intercourse arising from the use of menacing verbal pres- sure or the misuse of authority.

• Attempted rape includes attempted unwanted sexual intercourse arising from the use of or threats of force, or the use of drugs or alcohol.

• Rape includes unwanted sexual intercourse arising from the use of or threats of force and other un- wanted sex acts (anal or oral intercourse or penetration by objects other than the penis) arising from the use of or threat of force, or the use of drugs or alcohol.

88 Chapter • Designing Research

THEORETICAL FRAMEWORK

Developed by Walter DeKeseredy, McKenzie Rogness, and Martin Schwartz, the integrated theoretical model that informs this study includes the following variables: societal patriarchy; male proprietariness; exiting; threats to masculinity and patriarchal control; patriarchal male peer support; and separation/divorce sexual assault. Referred to by these sociologists as a feminist/male peer support model, some empirical sup- port for it is presented in Chapter 4. Still, after carefully analyzing the results and reading more rural socio- logical and criminological literature, it is evident that further theoretical work needs to take in account factors such as rural social and economic transformations that have occurred since the end of the last cen- tury. Thus, in Chapter 5, a new theoretical model is offered, one that addresses this variable, as well as male peer support and rural challenges to masculine identity.

METHODS

As described in this report, researchers encounter many obstacles while conducting rural studies of woman abuse. However, some of these problems were overcome or minimized using a variety of methods. The first step was preparatory research. This involved several meetings, electronic mail exchanges, and in-depth telephone conversations with leading researchers in the field, local shelter staff, sexual assault advocates, police officers, mental health workers, and others with a vested interest in curbing separation/divorce sex- ual assault and other types of woman abuse. Then, techniques like those used by Dr. Lee Bowker approxi- mately 24 years ago in Milwaukee were used to generate a sample. For example, an advertisement was placed twice in a free newspaper available throughout Athens County, Ohio. Also, posters about the study were pinned up in public places, such as courthouses, and were given to social service providers who came into contact with abused women.

In addition:

• Two local newspapers gave considerable coverage to the project. • Ohio University sent out a press release to newspapers and other Ohio-based media. • Three local radio stations and Ohio University's television station carried public service announce-

ments about the study. • The director of the local shelter and I appeared on a local television news show to discuss this proj-

ect and broader issues related to it. • The Ohio Domestic Violence Network and other agencies told interested parties (e.g., rural shelter

workers) about the study and helped to recruit participants. • Local shelter staff, a police department social worker, employees of the county sheriff's department,

Planned Parenthood, Women's Center staff at a local two-year college, and employees of the local Sexual Assault Survivor Advocate Program informed possible respondents about the study.

• Ohio University sociologist Judith Grant told women who participated in her addiction study about this research.

• Index-like cards with the information provided in the recruiting poster were routinely placed on top of newspaper boxes inside stores and on sidewalks in Athens, Ohio.

From early March 2003 until early April 2004, two female research assistants carried cellular phones 24 hours a day to receive calls from women interested in participating in the study. Callers were told the purpose of the project and were then asked a series of screening questions to determine their eligibility to be interviewed. The main criteria were being 18 years of age or older and having ever had any type of un- wanted sexual experience when they wanted to end, were trying to end, or after they had ended a relation- ship with a husband or live-in male partner. If they met the selection criteria, the women were invited to a semi-structured face-to-face interview at a time and place of their choosing, and they were paid $25.00 for their time. They were also given $7.75 for travel expenses and an index card listing the locations and phone numbers of local support services for survivors.

Female research assistants tape-recorded and transcribed all of the interviews. Most of them took about 90 minutes and a total of 43 women participated in this study. Posters placed in public places attracted most of our respondents (n = 27). Eight women called after exposure to ads or media stories about the study, and the same number were referred to us by individuals or organizations. Most respondents (n = 30) lived in Athens County, Ohio, three lived in Hocking County, Ohio, one lived in Vinton County, Ohio, and nine lived in other rural parts of the state. The mean age of the sample was 35 and the mean income for 2002 was $13,588. Sixty-five percent (n = 28) had some type of post-secondary education and close to half of the participants were unemployed. Of the 25 who had been married, all got divorced or legally separated, but only five remarried. Most of the respondents also lad children.

FINDINGS

The key findings are categorized under these headings: types and timing of abuse; characteristics of men who sexually assaulted their ex-partners; consequences of separation/divorce sexual assault; and social support.

Chapter 4 • Designing Research 89

Types and Timing of Abuse

)retical r Only a few respondents experienced just one of the above forms of separation/divorce sexual assault, uiness; and virtually all experienced rape or attempted rape. : ivorce Most (80%) of the women were victimized by two or more variants of other types of abuse, such as :al sup- physical violence and the destruction of prized possessions. I socio- • Nineteen percent o£ the respondents stated that their partners abused their children and one woman iccount believes that her ex-partner raped her as a means of killing her unborn child. sst cen- f • Seventy-four percent (n = 32) of the respondents said that they were sexually assaulted when they ex-

as male j pressed a desire to leave their relationships. Forty-rune percent (n = 21) were sexually abused while they were trying to leave or while they were leaving and 33% (n = 14) were victimized after they left.

• Compared to cohabiting women (33%, n = 6), married women (47%, n = 12) were more likely to report being abused while still in the relationship, before expressing a desire to exit, trying to exit, or exiting their relationships. At the next stage, when the women reported that they wanted to leave their

woman abusive relationship, 20 of the 25 married women (80%) stated that they were sexually assaulted, he first while 12 of the 18 cohabiting women (67%) stated that their assaults occurred at this point in time. i-depth ocates, ce sex- j Characteristics of Men Who Sexually Assaulted Their Ex-Partners

pprox'- • Sixty-seven percent (n = 29) of the interviewees reported on a variety of ways in which their part- :nt was ners' male peers perpetuated and legitimated separation/divorce sexual assault. Three methods in e study particular stand out: frequently drinking with male friends, informational support, and attachment to o came abusive peers.

• Seventy-nine percent of the sample said that their partners strongly believed that men should be in charge and control of domestic household settings.

• Regardless of whether they consumed it in groups, 65% of the sample's estranged partners viewed pornography, and it was reported to be involved in sexually abusive events experienced by 30% of

jounce- the interviewees. • More than half (58.14%) of the women said that male offenders had guns and some perpetrators even

is proj- threatened to use them. • Over 65% (n = 28) of the women interviewed said that their partners used illegal drugs and that their

shelter ; consumption of these substances contributed to abusive behaviors.

rtment, Consequences of Separation!Divorce Sexual Assault to local • Women experienced a wide range of negative outcomes, including low self-esteem, fear, nightmares,

y about I and a myriad of physical health problems. • All of the survivors interviewed developed a host of adverse post-assault psychological outcomes,

on top such as depression, sexual aversion, and fear. • Many respondents mentioned physical scars. • For many interviewees, exiting a relationship was financially devastating.

phones old the Social Support tytobe of un- • Data uncovered by this study strongly suggest that if there are high levels of collective efficacy in the

;lation- respondents' communities, they do not function to prevent and deter separation/divorce sexual assault.

:ed to a For example, most of the interviewees (84%) stated that women experiencing unwanted sex in their

..00 for community is a major problem and 81% reported that rape or sexual assault is also a serious problem.

I phone • That 81% of the respondents stated that they personally know other women who were sexually as- saulted provides further evidence that such victimization is a major problem in some rural Ohio com-

m took munities and that little is being done to prevent it.

xracted • Over half (58%) of the interviewees do not feel safe when they are at home.

out the • Eighty-four percent of the respondents stated that they could not count on their neighbors to help z = 30) solve their personal problems.

io, and • Sixty-seven percent of the sample did not get together with their neighbors in a typical week.

Dr 2002 • Fifty-eight percent of the women turned to at least one friend for help, but most of their friends did half of not live near them. Further, 44% sought assistance from the police, and 40% received help from a lo- ~arated, cal shelter.

• The interviewees' voices reveal that formal and better intervention by state authorities is more impor- tant for them than focusing on collective efficacy at this point in time.

• Only one of the respondents who turned to at least one element of the criminal justice system for help stated that it was the best assistance she received.

en who • Most interviewees turned to several different sources of social support. upport. • Most interview[ee]s found their friends to be the best source of social support.

i

90 Chapter 4 • Designing Research

SUGGESTIONS FOR FURTHER EMPIRICAL, THEORETICAL,AND POLICY WORK

This study shows that exiting or trying to exit a marital/cohabiting relationship increases women's chances of being sexually assaulted, especially if they are connected to patriarchal or abusive men. However, there

is still much that we do not know about separation/divorce sexual assault in rural and urban communities. Certainly, much more empirical and theoretical work is needed. Of course, it is also necessary to develop policies and practices that meet the unique needs of women who are terrorized by men who will not let them leave and men who they have left.

New Directions in Empirical Work

Regardless of whether separation/divorce sexual assault studies are conducted in rural or urban settings, data gathered from men are needed to more precisely determine the factors that motivate them to be abu- sive. Moreover, representative sample surveys of rural and urban populations would help determine the in- cidence and prevalence of separation/divorce sexual assault. Such rural research is undoubtedly difficult to do, given the methodological obstacles discussed in Chapter 3. Further, there are many other groups of men and women who need to be included in future research, such as those who are immigrants, living in public housing, have physical disabilities, and so on.

This is one of the first studies to apply collective efficacy theory to woman abuse in intimate, hetero- sexual relationships. Obviously, more research is needed, including studying the perceptions and experi- ences of rural women who are not abused. Another point to consider is that almost all studies of collective efficacy/social disorganization and crime use quantitative'techniques, such [as] analyses of census data. Nevertheless, many rural social problems are not easy to study using such methods, which is perhaps one of the key reasons why so few researchers focus on woman abuse in rural areas. Further, quantitative methods alone cannot adequately describe the complexities of rural woman abuse and community responses to it. Thus, it is essential to continue using other methods to examine community characteristics that affect sepa- ration/divorce sexual assault and other forms of woman abuse. One suggestion is to specifically design a qualitative project that focuses exclusively on the topics of central concern to this report and that uses in- depth interviews and participant observations of community relations.

POLICY RECOMMENDATIONS

The most common policy recommendations made by the 43 respondents are education, creating awareness, listening to the voices of survivors, criminal justice reforms, and subsidized housing. Based on data derived from these women, my previous empirical work, and a review of the extant literature on woman abuse in rural communities, I suggest that these and the following other policies be implemented as soon as possible and throughout all rural U.S. communities:

• Travel subsidies. • Job training and education. • Increased funding for rural service providers. • The development of and support for small, community-based businesses and small industrial dis-

tricts. • Community capacity building.

The policies proposed by 43 rural Ohio women and me are not the only effective solutions to prob- lems experienced by survivors of separation/divorce sexual assault. Rather, they are key elements of much need[edj community-based, collaborative efforts. Policy development must also be highly sensitive to the ways in which broader social forces contribute to the harms identified in this report.

Chapter 4 • Designing Research 91

Application Exercise 4.2 yes ere es. lop let

igs, bu- in- t to nen )lic

:ro- eri- tive ata.

of ods

it. ;pa- ;n a • in-

ess, ived e in .ible

dis-

rob- iuch i the

Name of Student:

Student ID No.:

Course/Section No.:

Date:

1. Name three primary findings in "Sexual Assault During and After Separation/Divorce: An Exploratory Study."

2. What makes this an exploratory study, rather than a descriptive study?

3. The report gives several policy recommendations. Do you think these are appropriate given the meth- ods described? Why or why not?

4. What future research directions are warranted given the results of these data?

5. If you were to design a follow-up study based on the research presented here, what kind of methods would you use? Be specific. a. How would you recruit your sample?

b. How would you ensure confidentiality?

92 Chapter 4 • Designing Research

c. What specific risks are present with this kind of population?

d. What research design would you employ?

One (vani bein sears if the both the is

(the: only accua

same ment;

from will t pect t the re

PRO

In a b that v from are se fied p proba rando popul posse; larger

pants: used p with selecti the hoi