Mon, March 20, 2017, noon:
Dean Yang, Taken by Storm
Conrad, Frederick G., Lance J. Rips, and Scott S. Fricker. 2009. "Seam Effects in Quantitative Responses." Journal of Official Statistics, 25(3): 339-361.
The seam effect is an artifact of panel surveys in which respondents are interviewed every few months (for example) and asked for data about each of the intervening months. The effect is defined by small changes in responses between adjacent months within a reference period (i.e., within an interview) relative to large changes between adjacent months across reference periods (i.e., across interviews). In the present studies, we used an experimental method to examine the seam effect for quantitative questions about dollar amounts. The studies produced robust seam effects, accompanied by forgetting of correct amounts and constant responding (no change between months) within the reference periods. The results also showed decreases in the seam effect when questions about the same topic appeared in different parts of the interview. In some conditions, dependent interviewing (providing respondents with their earlier answers) also reduced the size of the seam effect. Neither manipulation, however, improved the overall accuracy of responses. Dependent feedback may encourage respondents to make less drastic changes at the seam. But since the earlier answers may themselves contain error, dependent interviewing can also perpetuate the same incorrect response from one reference period to the next.
Country of focus: United States of America.