A Note on Correlated Measurement Error in Wife-Husband Data

The potential effects of correlated measurement errors among responses of wives and their husbands are examined using data from 349 couples on the utility of and expectations for another child. No significant between-partners error correlations were found, which may be due to the absence of common nonrandom method errors in partners' responses to the same survey item. On the other hand, positive within-partner error correlations were found; when these correlations were estimated and partialed out, the structural effect of child utility on expectations for another child was somewhat reduced. However, there was no change in the previous finding that wife's child utility and husband's child utility had equal effects on the couple's expectations for another child. The collection of survey data from both wife and husband is usually designed to improve measures of family life. However, once these data are in hand, measurement error and its effects on estimates of structural relations often are forgotten. In a previous paper (Thomson and Williams, 1982), we discussed the possible effects of differential errors in wives' and husbands' responses in a model of couple's expectations for another child. Contrary to our initial hypothesis, we found that wives and husbands were equally reliable reporters of the utility of another child to themselves. Therefore, estimates of the effect of each partner's child utility on expectations for another child were reduced by controlling for measurement error, but the ratio of the two effects was not changed. In fact, we were able to accept the hypothesis that each partner's utility has equal effects on the couple's expectations for another child.