Word Length in Personal Advertisements: Controlling a Common Confound

Objectives:

To demonstrate the confounding influence of word length on content analyses of personal advertisements and to present conditional logistic regression as a technique for testing hypotheses drawn from evolutionary, assortative and exchange theories of partner selection while controlling for word length.

Design:

A Quantitative Content Analysis.

Methods:

A sample of 2934 advertisements was drawn from newspapers over a period of eight months. Individual advertisements were coded for all deBive phrases with computer assistance. Data was reduced to categories indicating sex of advertiser, whether the advertiser claimed to be attractive and/or to have socio-economic status, and whether the advertiser sought a partner who was attractive and/or had socio-economic status. Total number of phrases of self and other deBion were also recorded.

Results:

A Significant correlation was found between number of phrases of self and other deBion used (r=0.039, p=0.033). The probability of attractiveness or status being described was significantly related to the number of deBive phrases used (e.g. r=0.352, p<0.001). Sex differences were found in the number of self and other deBive phrases used (e.g. Z=6.55, p<0.001). This demonstrated the presence of a network of inter-correlated influences of word length that confound tests of sex differences in, as well as assortative matching and exchange of, attractiveness and status.

Conditional Logistic Regression (CLR) is presented as a technique that provides full control of the confounding influence of word length. Results of CLR to examine hypotheses drawn from evolutionary, assortative and exchange theories of partner selection are presented, which provide partial support for all three theories.

Conclusions:

Advertisement length is a confounding influence that needs to be taken account of when analysing this type of data. CLR provides an analytical technique that controls for this and provides results that are largely in accord with established theory.