# Reliability

Indexes are built because they are expected to provide a more accurate measurement of the theoretical dimension than any single variable can do alone. Hence, the variables included in the index should all measure aspects of the same dimension. For example, if you were asked to create an index for body size, it would be sensible to include height and weight, but it would make no sense to include eye colour. One way of evaluating the extent to which an index measures one dimension that underlies all of its items is to perform a reliability analysis. Cronbach’s alpha is a measure of the internal consistency of the items in a scale, and it ranges from 0 to 1.1 An alpha greater than 0.7 is desirable for indexes that are used as a scale.

1. Find the alpha for each of the indexes. SPSS

Perform ten reliability analyses, one for each value. The example below gives the syntax for power. You could use either the original variables or the recoded variables (as in the example). If you prefer to use the menu: Analyse – Scale – Reliability Analysis.

RELIABILITY
/VARIABLES=nimprich niprspot
/FORMAT=NOLABELS
/SCALE(ALPHA)=ALL/MODEL=ALPHA.

You should not be surprised if the internal reliabilities of several PVQ indexes are relatively low. This reflects two facts. First, the items in the indexes were constructed and selected to cover the different conceptual components of each value, not to be nearly redundant measures of a narrowly defined concept. For example, the power value items tap both wealth and authority, and the universalism items tap understanding, concern for nature, and social concern. If items with more similar meanings were chosen for each index, alpha would be higher, but at the cost of poorer coverage of the breadth of meaning of each of the types of values. Second, each index includes only two or three items. With so few items it is virtually impossible to obtain high alphas unless the items are very similar to one another. Considering the small number of items used to measure each of the ten values and their necessary heterogeneity, even reliabilities of 0.4 are reasonable.

Go to next page >>

#### Footnotes

• [1] The explanation of Cronbach's alpha is rather complex, but it could be simplified to the following: The alpha depends upon the number of indicators in the index and the average intercorrelation among the items comprising the index. The more items there are in the index and the higher the average intercorrelation among them, the higher the alpha.