Table 1.

Sample Construction, by wave

	W1	W2	W3	W4	W5	W6	W7	W8
Usable sample, previous wave		5004	5,070	5,125	6,005	6,037	6,065	7,071
- Lost since previous wave		−492	−428	−446	−513	−522	−441	−562
Carry over into current wave		4,512	4,642	4,679	5,492	5,515	5,624	6,509
% retained from previous wave		90.2%	91.6%	91.3%	91.5%	91.4%	92.7%	92.1%
Total individuals current wave	12,652	19,642	17,991	21,384	19,579	18,167	20,129	18,469
- Not female	−5,868	−8,227	−7,479	−8,959	−8,112	−7,458	−8,350	−7,584
- Not between ages 51-61*	−1,710	−6,189	−5,179	−6,228	−5,262	−4,458	−4,409	−3,537
- Missing family/children information	−5	−73	−120	−62	−79	−128	−220	−215
- Missing income information	−53	−78	−84	−127	−83	−53	−64	−58
- Missing other explanatory variables	−12	−5	−4	−3	−6	−5	−15	−8
Final admissible sample this wave:	5,004	5,070	5,125	6,005	6,037	6,065	7,071	7,067
(1) Carry over from last wave		4,512	4,642	4,679	5,492	5,515	5,624	6,509
(2) Newly added this wave	5,004	558	483	1,326	545	550	1,447	558
Cumulative total sample:	5,004	10,074	15,199	21,204	27,241	33,306	40,377	47,444

^*Abbreviated definition. The exact definition of this criterion is that an individual “either was >61 at first entry into the survey or whose age in the current wave is <51”. Therefore, if an individual is deemed “admissible” in an earlier wave (i.e., in addition to meeting the other criteria, they have met the age 51-61 criterion in an earlier wave), they continue to be included in subsequent waves, even if they age beyond the 51-61 age range. As noted in the paper, a large proportion of the observations deleted due to this criterion are respondents that were part of the original Assets and Health Dynamics of the Oldest Old (AHEAD) survey, that sampled people born prior to 1923 and their spouses. Consistent with the HRS sampling design regarding the addition of cohorts, if a respondent is initially deemed inadmissible due to being younger than 51, she joins the analysis sample upon reaching 51.