Two experiments were implemented with the round 20 (2002) survey. Each is described in more detail below.
Early Bird Experiment
The first experiment, called the Early Bird Experiment, sought to encourage early participation in the survey and to save resources on later locating and conversion efforts. A group of over 2800 respondents who have been friendly and cooperative in round 19 (2000) were assigned to an "Early Bird" sample of respondents. These respondents were further assigned randomly to one of 12 replicates which were fielded sequentially at regular intervals. A toll-free telephone line was established, staffed with interviewers working remotely around the country.
An initial advance letter was sent to the respondents in each replicate as their replicate was to be activated for interviewing. They were offered a higher participation fee (randomly assigned $60/$80) if they called into the toll-free line to schedule an interview within a four week window. Respondents who had not contacted the toll-free line in the first two weeks of their initial letter were sent a reminder postcard. Respondents who did not contact the toll-free line to complete their interview by the time their Early Bird window expired reverted to a $40 standard fee.
Income Recall and Estimation Experiment
The second experiment was aimed at determining the most effective way for respondents to estimate income and asset amounts. This experiment was implemented in two stages.
First, a select group of historically cooperative respondents were asked to recall their own income and, if applicable, their spouse/partner's income for the calendar year 1999. These respondents were assigned randomly to one of three types of questions and asked to estimate his/her and/or his/her spouse/partner's 1999 income. The three types of questions are described below.
- Unfolding brackets: This involves asking the respondent first if the amount is above a certain amount, and based on that response whether it would be above a second amount. This procedure establishes a loose range for the amount.
- Estimate to the nearest $10,000: The respondent is asked to estimate the amount to the nearest $10,000.
- Self-reported ranges: The respondent is asked to report an upper and lower range for the amount. The intent of this type of question is to have the respondent to narrow the possible range in his or her own mind.
In the second stage of this experiment, these three types of questions were included in the instrument following selected income questions. Respondents who answered either "don't know" or "refuse" to the exact income amount were asked to estimate the amount using one of the three types of questions, depending upon the type of questions to which they were pre-assigned.
The results from this experiment were analyzed to determine which yielded 1) estimates of 1999 income that were closest to the actual amounts given in the 2000 interview; 2) estimates of income that most closely coincided with income reports in previous interviews and; 3) the method most successful in gaining income estimates from respondents who otherwise refuse to give such information. Users interested in obtaining the results of this analysis can request the report entitled "NLSY79 Income Experiment" (Aughinbaugh, Gardecki, 2003) by e-mailing NLS User Services.