2006 Questionnaire Design and Experimental Research Survey:
Demographic Questions Analysis
Elizabeth Nichols, Jennifer Hunter Childs and Rolando Rodríguez
KEY WORDS: Date of birth question; Age confirmation; Relationship question; Real-time data editing; Mode consistency
Two small studies were embedded into the 2006 Questionnaire Design and Experimental Research Survey (QDERS). Both of these studies pertained to testing different ways to collect and verify demographic data using a computerized data collection mode. These results can be applied to any demographic data collection including surveys or censuses.
The first study is an experiment focusing on how to confirm age, given a valid date of birth. The second study is an investigation exploring a real-time edit to fix problematic relationship data by verifying with the respondent seemingly misreported relationships.
Traditionally, the decennial census has gathered the age of all household members as of a particular date (Census Day, which has typically been April 1st of the year in which the census is conducted). Carter and Brady (2002) found respondents had problems reporting age when a person’s current age differed from that person’s age on a specific date. Norris (2005) also found many respondents had difficulty reporting an age in the past. The QDERS experiment attempted to determine if respondents had problems confirming an age in the past which differed from the current age, and whether confirming the current age would solve any cognitive demands of the former method. The 2006 QDERS did not find the same level of misreported data associated with confirming an age in the past as Norris or Carter and Brady found with reporting an age in the past. However, QDERS did demonstrate that confirming an age in the past does take significantly longer than confirming current age, and thus, causes more interviewer and respondent burden.
The decennial census also collects data on the relationship of each person in the household to a resident owner or renter. Love and Byrne (2005) found that respondents sometimes report relationships between members of a household in the reverse direction from the actual relationship (e.g., reporting “father” rather than “son”). Love and Byrne found that many people who were reported as parents were actually younger than the reported children; and similarly many “children” were older than the reported parents. Although these relationships could be true with blended families, the high instances of this kind of reporting cited in the Love and Byrne paper indicate that it is actually mistaken reporting. The QDERS study attempted to determine if a real-time edit to seemingly inconsistent relationship data could clean up the data while the interviewer was still speaking to the respondent. However, the 2006 QDERS did not find the same level of inconsistent relationship to age data as found by Love and Byrne. The edit was only invoked three times, clearing up an apparent confusion one of those times.
In conclusion, the 2006 QDERS did not find the same level of misreported data that the motivating papers, Carter and Brady (2002), Norris (2005) or Love and Byrne (2005), found. Yet, the strategies the QDERS paper used, confirming current age and adding an edit to seemingly inconsistent relationship data, at least preserved, if not improved the original questions without creating additional interviewer or respondent burden. Thus, we recommend using these strategies in future demographic data collections.