Validity, often called construct validity, refers to the extent to which a measure adequately represents the underlying construct that it is supposed to measure. For instance, is a measure of compassion really measuring compassion, and not measuring a different construct such as empathy? Validity can be assessed using theoretical or empirical approaches, and should ideally be measured using both approaches. Theoretical assessment of validity focuses on how well the idea of a theoretical construct is translated into or represented in an operational measure. This type of validity is called translational validity (or representational validity), and consists of two subtypes: face and content validity. Translational validity is typically assessed using a panel of expert judges, who rate each item (indicator) on how well they fit the conceptual definition of that construct, and a qualitative technique called Q-sort.
Empirical assessment of validity examines how well a given measure relates to one or more external criterion, based on empirical observations. This type of validity is called criterion-related validity, which includes four sub-types: convergent, discriminant, concurrent, and predictive validity. While translation validity examines whether a measure is a good reflection of its underlying construct, criterion-related validity examines whether a given measure behaves the way it should, given the theory of that construct. This assessment is based on quantitative analysis of observed data using statistical techniques such as correlational analysis, factor analysis, and so forth. The distinction between theoretical and empirical assessment of validity is illustrated in Figure 7.2. However, both approaches are needed to adequately ensure the validity of measures in social science research.
Note that the different types of validity discussed here refer to the validity of the measurement procedures, which is distinct from the validity of hypotheses testing procedures, such as internal validity (causality), external validity (generalizability), or statistical conclusion validity. The latter types of validity are discussed in a later chapter.
Face validity. Face validity refers to whether an indicator seems to be a reasonable measure of its underlying construct “on its face”. For instance, the frequency of one’s attendance at religious services seems to make sense as an indication of a person’s religiosity without a lot of explanation. Hence this indicator has face validity. However, if we were to suggest how many books were checked out of an office library as a measure of employee morale, then such a measure would probably lack face validity because it does not seem to make much sense. Interestingly, some of the popular measures used in organizational research appears to lack face validity. For instance, absorptive capacity of an organization (how much new knowledge can it assimilate for improving organizational processes) has often been measured as research and development intensity (i.e., R&D expenses divided by gross revenues)! If your research includes constructs that are highly abstract or constructs that are hard to conceptually separate from each other (e.g., compassion and empathy), it may be worthwhile to consider using a panel of experts to evaluate the face validity of your construct measures.
Content validity. Content validity is an assessment of how well a set of scale items matches with the relevant content domain of the construct that it is trying to measure. For instance, if you want to measure the construct “satisfaction with restaurant service,” and you define the content domain of restaurant service as including the quality of food, courtesy of wait staff, duration of wait, and the overall ambience of the restaurant (i.e., whether it is noisy, smoky, etc.), then for adequate content validity, this construct should be measured using indicators that examine the extent to which a restaurant patron is satisfied with the quality of food, courtesy of wait staff, the length of wait, and the restaurant’s ambience. Of course, this approach requires a detailed description of the entire content domain of a construct, which may be difficult for complex constructs such as self-esteem or intelligence. Hence, it may not be always possible to adequately assess content validity. As with face validity, an expert panel of judges may be employed to examine content validity of constructs.
Convergent validity refers to the closeness with which a measure relates to (or converges on) the construct that it is purported to measure, and discriminant validity refers to the degree to which a measure does not measure (or discriminates from) other constructs that it is not supposed to measure. Usually, convergent validity and discriminant validity are assessed jointly for a set of related constructs. For instance, if you expect that an organization’s knowledge is related to its performance, how can you assure that your measure of organizational knowledge is indeed measuring organizational knowledge (for convergent validity) and not organizational performance (for discriminant validity)? Convergent validity can be established by comparing the observed values of one indicator of one construct with that of other indicators of the same construct and demonstrating similarity (or high correlation) between values of these indicators. Discriminant validity is established by demonstrating that indicators of one construct are dissimilar from (i.e., have low correlation with) other constructs. In the above example, if we have a three-item measure of organizational knowledge and three more items for organizational performance, based on observed sample data, we can compute bivariate correlations between each pair of knowledge and performance items. If this correlation matrix shows high correlations within items of the organizational knowledge and organizational performance constructs, but low correlations between items of these constructs, then we have simultaneously demonstrated convergent and discriminant validity (see Table 7.1).
Table 7.1. Bivariate correlational analysis for convergent and discriminant validity
An alternative and more common statistical method used to demonstrate convergent and discriminant validity is exploratory factor analysis. This is a data reduction technique which aggregates a given set of items to a smaller set of factors based on the bivariate correlation structure discussed above using a statistical technique called principal components analysis. These factors should ideally correspond to the underling theoretical constructs that we are trying to measure. The general norm for factor extraction is that each extracted factor should have an eigenvalue greater than 1.0. The extracted factors can then be rotated using orthogonal or oblique rotation techniques, depending on whether the underlying constructs are expected to be relatively uncorrelated or correlated, to generate factor weights that can be used to aggregate the individual items of each construct into a composite measure. For adequate convergent validity, it is expected that items belonging to a common construct should exhibit factor loadings of 0.60 or higher on a single factor (called same-factor loadings), while for discriminant validity, these items should have factor loadings of 0.30 or less on all other factors (cross-factor loadings), as shown in rotated factor matrix example in Table 7.2. A more sophisticated technique for evaluating convergent and discriminant validity is the multi-trait multi-method (MTMM) approach. This technique requires measuring each construct (trait) using two or more different methods (e.g., survey and personal observation, or perhaps survey of two different respondent groups such as teachers and parents for evaluating academic quality). This is an onerous and relatively less popular approach, and is therefore not discussed here.
Criterion-related validity can also be assessed based on whether a given measure relate well with a current or future criterion, which are respectively called concurrent and predictive validity. Predictive validity is the degree to which a measure successfully predicts a future outcome that it is theoretically expected to predict. For instance, can standardized test scores (e.g., Scholastic Aptitude Test scores) correctly predict the academic success in college (e.g., as measured by college grade point average)? Assessing such validity requires creation of a “nomological network” showing how constructs are theoretically related to each other. Concurrent validity examines how well one measure relates to other concrete criterion that is presumed to occur simultaneously. For instance, do students’ scores in a calculus class correlate well with their scores in a linear algebra class? These scores should be related concurrently because they are both tests of mathematics. Unlike convergent and discriminant validity, concurrent and predictive validity is frequently ignored in empirical social science research.
Table 7.2. Exploratory factor analysis for convergent and discriminant validity