How Do We Measure Concepts? - SAGE Publications

1 5 How Do WeMeasure Concepts? Characteristics of a Good measurement ProcedureLevels of MeasurementLevels of measurement : NominalLevels of measurement : OrdinalLevels of measurement : IntervalLevels of measurement : RatioSummary of Levels of MeasurementWhat Are Reliability and Validity?ReliabilityStabilityConsistency ValidityFace ValidityCriterion ValidityConstruct ValidityContent ValiditySome Thoughts on Reliability and ValidityStudy QuestionsFor Further Reading5105-Greenstein 1/12/2006 11:58 AM Page 51 Social researchers are faced with something of a dilemma when theyattempt to Measure concepts.

2 Take, for example, a concept such as mari-tal satisfaction. This concept is abstract; you cannot touch, feel, or see maritalsatisfaction. Yet, we somehow have to translate this abstraction into somekind of concrete have to figure out a way to get people totell us, as accurately as possible, how satisfied they are with their of our thinking about families and children begins with what Babbie(2004) calls conceptions. We each have a conception of what defines maritalsatisfaction. But it s clear that marital satisfaction is not a physical object orquantity that can be readily observed and measured.

3 How do we translatethe relatively abstract and unobservable concept of marital satisfaction intosomething that we can Measure and study?Kaplan (1964) suggested three types of things that can be measured. Adirect observableis a physical characteristic that can be observed directly,such as the number of people present at a meeting. A person s response to aquestionnaire item about the number of people in a household is an indirectobservable an indirect representation of a characteristic or object.

4 A con-structis a creation based on observation, but it cannot be observed directly orindirectly. Marital satisfaction is an example of a construct. We may be ableto observe consequences of marital satisfaction ( , marital stability), and wemay be able to observe causes or antecedents of marital satisfaction ( , anequitable division of household labor between husband and wife); but wecannot observe marital satisfaction the name given to this process of translating abstrac-tions into concrete measurement processes.

5 In this process, we need to makea number of decisions about how to translate an abstract construct, such asmarital satisfaction, so we can Measure the marital satisfaction of our of a Good measurement ProcedureIn chapter 2, I distinguished between variables and their states, levels, orattributes. Gender is a variable; female is an attribute of gender. Income isa variable; $45,000 is a state or level of income. One important decisionto make in the operationalization process is what and how many attributes,states, or levels a variable will have in the measurement good measurement procedure is characterized by three key , the set of attributes representing a variable must be exhaustive and inclu-sive.

6 That is, all possible responses or states must be represented by a state,level, or attribute of the variable. If we were operationalizing the variable52 Methods of Family Research05-Greenstein 1/12/2006 11:58 AM Page 52marital status, we would find that using only the categories marriedand singlewould produce many blank stares from our respondents. We would probablyneed to include never married, divorced but not remarried, married but sepa-rated,andwidowedas second concern is that the attributes, states, or levels chosen to repre-sent a variable must be mutually exclusive.

7 No respondent should be able tobe placed in more than one category. In the five-category scheme to measuremarital status (never married, currently married, married but separated,divorced but not remarried, and widowed), no person could meaningfullyclaim to fall in more than one , we need to be concerned with our attribute scheme s , our category scheme should use more, rather than fewer, distinctionswhen feasible. For example, rather than coding certain respondents as Protes-tant, it is probably more useful to include the choices Baptist, Episcopalian,Methodist, and so on.

8 In this way you can determine how many respondentsare of each denomination. But if you fail to allow respondents to describewhether they re Baptist, Episcopalian, or Methodist, you will never of MeasurementThese three qualities of measurement processes inclusive and exhaustive,mutually exclusive, and precise are important to consider for all variables themselves, however, may be measured in vastly different of measurementrefers to the properties that define the measurementprocess itself.

9 Most social and behavioral data are measured at one of fourlevels of measurement : nominal, ordinal, interval, or of measurement : NominalSome variables have attributes that meet only the minimal requirementsof being exhaustive and mutually exclusive. Some examples of nominal vari-ables are religion, race, gender, city of residence, and marital status. Each ofthese variables, when properly defined, has a set of logically exhaustive andmutually exclusive categories or attributes. In other words, nominal variablesidentify the attributes associated with a variable without making any state-ments about whether one state or level is more or less than any other state orlevel.

10 With variables measured at the nominal level, it s not possible to rankcategories along any particular dimension, nor is it possible to state with greatprecision how much of the attribute is s also important to note that for nominally scaled variables, the differ-ences between categories are not mathematically meaningful. For example,How Do We Measure Concepts? 5305-Greenstein 1/12/2006 11:58 AM Page 53let s say that we gave the following response choices to the question Whatis your current marital status?