react to the context of the study itself. Hence, the results obtained are more likely to hold outside the study in other settings, thereby increasing external validity.
Again, exceptions must be appended to these conclusions. Because in panel studies the same respondents are interviewed repeatedly over time, they may grow sensitized to the fact that they are under study and thus become less typical of the population they were originally chosen to represent. Th is problem may be allevi- ated by limiting participation in the panel to a short period of time. Case studies are less tractable with respect to external validity. Th e case is usually selected pre- cisely because there is something distinctive, atypical, or particularly interesting about it (for example, a failure of a nuclear power plant, or a “whistleblower”
whose courageous persistence saves a state government millions of dollars). As a consequence, it is diffi cult to judge the extent to which the results of a case study have relevance for other cases. Case study researchers should devote serious attention to considering the population of cases to which they may legitimately generalize their results. Unfortunately, researchers and readers alike often are so captivated by the details of an arresting case that they fail to ask the important question: What can be learned from this case to apply to other cases?
Chapter Summary
Research designs involve setting up a research project so that the research questions can be answered as unambiguously as possible. Th e objective of a good research design is to establish causal relationships and to assess their generalizability.
Th e basic building block in constructing causal explanations is the concept, which pinpoints an idea or element thought to be essential in accounting for the class of events under study. Concepts are defi ned in two ways: with a nominal defi nition, which is the standard dictionary defi nition, and with an operational definition, which translates the nominal definition into a form in which the concept can be measured empirically. Once concepts have been operationalized and measured in a sample of data, they are called variables. Th e two major types of variables are independent (anticipated causes) and dependent (the variables thought to be aff ected by them). A hypothesis formally proposes an expected relationship between an independent and a dependent variable.
Social scientists have identifi ed four criteria as necessary for establishing a rela- tionship as causal: time order, covariation, nonspuriousness, and theory. A research design is a program for evaluating empirically proposed causal relationships. Th e evaluation is based on two criteria: internal validity (Did the independent variable lead to changes in the dependent variable?) and external validity (Can the results obtained in the study be generalized to other populations, times, and settings?).
Two major families of research design were outlined: experimental designs and quasi-experimental designs. Both types were evaluated with regard to the four cri- teria for causal relationships that defi ne internal validity and with regard to exter- nal validity. Briefl y, the experimental design has its primary strengths in internal validity, and the quasi-experimental design has its strengths in external validity.
Problems
3.1 A researcher asserts that the relationship between attitude toward the fi eld of public administration and taking courses in a public administration degree program is causal.
(a) What evidence must the researcher provide about this relationship to prove that it is causal?
(b) Given your answer, what aspects of the researcher’s argument are likely to be strongest, and what aspects of her argument are likely to be weakest?
Your discussion should include clear defi nitions of each element of a causal relationship.
3.2 Develop a model that includes at least four concepts. Elaborate any theoretical or literature support underlying it. Present the model in an arrow diagram that shows schematically the relationships the model proposes. Provide operational defi nitions for all concepts, and state hypotheses derived from the model.
(a) Which types of research designs would be best suited to testing the model?
(b) Which types of research designs would be least suited to testing the model?
3.3 Professor George A. Bulldogski has taught social graces to athletic teams at a major southeastern university for the past 15 years. Based on this experience, he insists that table manners are causally related to leadership. Professor Bulldogski has data showing that athletes who have better table manners also demonstrate greater leadership in athletic competition. Th e university gymnastics coach, who wants to build leadership on her team, is considering asking Professor Bulldogski to meet regularly with her team. She hopes that after he teaches table manners to team members, they will become better leaders. Should she invite Professor Bulldogski to meet with the gymnastics team? If she does so, can she expect his involvement to develop greater leadership on the team? Explain your answers.
3.4 For the entire month of January 2011, the local domestic violence shelter ran a series of ads on the government access cable television channel. Th e director of the shelter hypothesized that the ads would bring more attention to the problem of domestic violence and make victims of domestic violence more aware of the shelter’s programs. Th e director has monthly data on the number of new clients.
At the end of April 2011, the director asks you to come up with a research design to help assess whether the ads had an eff ect on the number of clients served.
(a) What type of research design would you use to test the hypothesis?
(b) Th e director has monthly fi gures for the number of clients served in 2010.
Are these data relevant for testing the above hypothesis? If so, why?
3.5 Th e head of the Teen Intervention Center believes that troubled youth are not getting the message if they complete the center’s 5-week education program but still have subsequent encounters with law enforcement. When they fi rst come to the center, teens are broken up into groups of 15 to meet with a counselor and answer questions about what society defi nes as acceptable versus unacceptable behaviors.
(a) If you were the head of counseling at the center, what knowledge might you gain by asking each group of teens similar questions at the end of the 5-week program?
(b) Is there any reason to expect diff erent responses to the initial questions and those asked at the end of the program? Explain.
(c) What type of research design would you be using if you administered a ques- tionnaire to the same group of individuals at the outset of treatment, at the end of treatment, and 6 months after treatment?
Descriptive Statistics
2
4
D
escriptive statistics is nothing more than a fancy term for numbers that summarize a group of data. Th ese data may be the number of arrests each police offi cer makes, the amount of garbage collected by city work crews, the number of fund-raising events held by a nonprofi t organization in a year, the number of volunteers assisting a government agency, the number of high school students participating in community service projects, or the size of various gov- ernment agencies (measured by personnel or budget). In their unsummarized or nontabulated form, data (aff ectionately known as “raw data”) are diffi cult to com- prehend. For example, the list below gives the number of tons of trash collected by the Normal, Oklahoma, sanitary engineer teams for the week of June 8, 2011.Each entry is the number of tons of trash collected by a team during the week.
57 70 62 66 68 62 76 71 79 87
82 63 71 51 65 78 61 78 55 64
83 75 50 70 61 69 80 51 52 94
89 63 82 75 58 68 84 83 71 79
77 89 59 88 97 86 75 95 64 65
53 74 75 61 86 65 95 77 73 86
81 66 73 51 75 64 67 54 54 78
57 81 65 72 59 72 84 85 79 67
62 76 52 92 66 74 72 83 56 93
96 64 95 94 86 75 73 72 85 94
Clearly, presenting these data in their raw form would tell the administrator lit- tle or nothing about trash collection in Normal. For example, how many tons of trash do most teams collect? Do the teams seem to collect about the same amount, or does their performance vary?
Th e most basic restructuring of raw data to facilitate understanding is the frequency distribution. A frequency distribution is a table that pairs data values—or ranges of data values—with their frequency of occurrence. For exam- ple, Table 4.1 is a frequency distribution of the number of arrests each Morgan City police offi cer made in March 2011. Note that the entire table is labeled, as is each column. Here, the data values are the number of arrests, and the frequencies
Frequency
Distributions
are the number of police offi cers. Th is procedure makes it easy to see that most Morgan City police offi cers made between 16 and 20 arrests in March 2011.
Some defi nitions are in order. A variable is the trait or characteristic on which the classifi cation is based; in the preceding example, the variable is the number of arrests per police offi cer. A class is one of the grouped categories of the variable. Th e fi rst class, for example, is from 1 to 5 arrests. Classes have class boundaries (the lowest and highest values that fall within the class) and class midpoints (the point halfway between the upper and lower class boundaries).
Th e class midpoint of the third class, for example, is 13—which is 11, the lower class boundary, plus 15, the upper class boundary, divided in half (by 2), or (11 1 15) 4 2. Th e class interval is the distance between the upper limit of one class and the upper limit of the next higher class. In our example, the class inter- val is 5. Th e class frequency is the number of observations or occurrences of the variable within a given class; for example, the class frequency of the fourth class (16–20) is 132. Th e total frequency is the total number of observations or cases in the table—in this case, 244. In the remainder of this chapter, we will discuss some important characteristics of frequency distributions and the procedures for constructing them.