We conduct a separate test for each category within each attribute, implying three separate
tests for attributes that have three categories. For example, as shown in Table 1, we test the
null hypotheses that the proportion of high-income targeted consumers is less than or equal
to the proportion of high-income neglected consumers, the proportion of medium-income
targeted consumers is equal to the proportion of medium-income neglected consumers, and
the proportion of low-income targeted consumers is greater than or equal to the proportion
of low-income neglected consumers.
For a particular attribute, rejection of the null hypothesis that the proportion of targeted
consumers is less than or equal to the proportion of neglected consumers indicates that credit
card companies use that attribute to target consumers. Similarly, rejection of the null
hypothesis that the proportion of targeted consumers is greater than or equal to the proportion
of neglected consumers for an attribute suggests that card companies neglect consumers with
that attribute.