Attachment 7 - Report on NonResponse

Attachment 7—Report on Nonresponse Through Round 26

1. Introduction:

The National Longitudinal Survey of Youth 1979 (NLSY79) follows the lives of a sample of men and women born from 1957 to 1964. In their first interview in 1979 they were ages 14 to 22, finishing their education and starting to enter the labor force; today, after 26 rounds of surveying, they are beginning to reach their 60s and preparing for retirement. Over 77 percent of those still alive participated in the last round of interviews. This remarkable retention rate is a result of management policies that seek to maximize survey response in each round and attempt to interview all sample members regardless of their participation in previous rounds. Additionally, the NLSY79 collects data on employment and other domains in an event-history format, which helps maintain the usability of the data of previous wave non-respondents. In this paper, we examine the extent to which attrition leads to bias in the NLSY79.

Attrition bias in longitudinal surveys occurs when sample members either leave the survey or don’t respond to a survey round, and the data not collected are systematically different from the data that are collected from the remaining sample. In this case, model estimates using only respondents will be different than those in the population the sample represents. In this paper, we examine three aspects of attrition within the NLSY79.

We start with an examination of whether survey nonresponse is indeed non-random with respect to demographics, educational attainment, and labor supply variables. We estimate how these measures at age 30 are related to the probability of survey nonresponse 15 or more years later. We then turn to the question of whether attrition leads to biased estimates in models of important economic relationships. It is possible that even if attrition is non-random with respect to demographic variables, estimates of these models are not affected because the basic economic relationships do not vary by demographics. Attrition bias requires both differential attrition and differential relationships between variables among attriters and non-attriters.

Additionally, we examine the issue of mortality. Part of attrition in longitudinal surveys, of course, is a natural process; over time, sample members die, especially in surveys of older people. There is no reason to believe that this causes bias in estimates of models of contemporaneous processes as the deceased sample members just represent those in the general population who have died. However, when estimating models of past processes, this may not be the case. For example, if we use a sample of 55-year-olds to estimate models of wages at 35, we will miss the natural attrition of those who died between the ages of 35 and 55. We analyze the determinants of mortality in the NLSY79 and then estimate how mortality in later rounds changes the estimates of coefficients in models of economic relationships in earlier rounds.

Our analysis builds on earlier work by MaCurdy, Mroz, and Gritz (1998), which examined the effects of attrition in the NLSY79 the survey data up to 1991. They concluded that attrition in the NLSY79 was non-random, but small enough not to affect estimates in models of employment and wages. Over twenty additional years of data are now available. In this paper, we will examine the size and implications of attrition through the 2014 data collection (round 26).

Our paper is also influenced by Michaud, Kapteyn, Smith, and van Soest (2011), which looks at nonresponse in the Health and Retirement Study (HRS). Their work analyzes nonresponse bias through 2004 (7 interviews) for an older cohort that is first surveyed in 1992. The paper’s focus is on temporary vs. permanent attrition. Permanent attrition includes those who have died. Temporary attrition occurs because the HRS attempts to interview respondents who were not interviewed in the previous round, a practice also followed in the NLSY79. The authors find that nonresponse bias in panel models of home ownership and wealth is significantly reduced due to interviewing temporary non-responders in later survey rounds. However, in conditional logit models of labor force participation, attrition bias due to mortality remains.

2. Data and Descriptive Statistics:

The NLSY79 began in 1979 with a nationally-representative sample of 12,686 males and females born between 1957 and 1964. Respondents were interviewed annually through 1994 and biennially afterwards. The NLSY79 dropped two subsamples from the survey over the years, the military oversample in 1985 and the poor white oversample in 1991. As such, we remove those cases from our analysis, leaving us with a sample of 9,964 for this study. All data in Tables 2 and forward use survey weights based on the sample of 9,964 respondents.

Table 1, which uses unweighted data, shows patterns of nonresponse in the NLSY79, by year, by sex, and by race/ethnicity within sex from rounds 14 through 26. The percentage of cases interviewed has fallen slowly and mostly steadily between rounds 14 and 26. Approximately 89 percent of cases were interviewed in round 14 and 71 percent in round 26; note that the deceased are included in the denominator of those top-end numbers. In all of these rounds, women are interviewed at rates two to six percentage points higher than those for men. For instance, in round 20 (2002), 75 percent of men were interviewed compared to 80 percent of women.

Note that the respondents who miss an interview can return to the survey in a later round. Event-history data on employment, education, marriage, and fertility are collected since the date of each respondent’s last interview, thus filling in information from missing interviews (though these responses may be less accurate). Of those not interviewed in round 20, 41 percent participated in at least one subsequent interview, 40 percent among men and 42 percent among women.

The deceased make up an increasing share of those who are not interviewed in a particular round. Between rounds 14 and 21 the percent of the non-interviewed sample who were deceased fluctuated between 15 percent and 19 percent; in the last 5 rounds, the percentage has increased steadily to 27 percent.

The nonresponse patterns differ by race and ethnicity. In later rounds black men and black women were generally interviewed at higher rates than their non-black, non-Hispanic and Hispanic counterparts.¹ For example, in round 26, black men’s interview rates were 3 to 4 percentage points higher than rates of non-black, non-Hispanic men and Hispanic men. In round 26, black women’s interview rates were 8 percentage points higher than those of their non-black, non-Hispanic and Hispanic counterparts. In all rounds shown in Table 1, of those not interviewed, the percent who were deceased is much higher for black men and black women compared to non-black, non-Hispanic and Hispanic men and women. For example, looking again at round 26, 43 percent of black men not interviewed were dead, compared to 29 percent of Hispanic and 24 percent of non-black, non-Hispanic men. About 37 percent of black women interviewed in round 26 were deceased compared to under 20 percent of their non-black, non-Hispanic and Hispanic counterparts.

Table 2, panels a (men) and b (women) compare sample members based on whether they were alive in at the time of Round 26 and then conditional on being alive, by whether they were interviewed. The second and third columns of Table 2 examine those deceased by round 26 and those still alive by round 26. Overall, 9.8 percent of male sample members and 6.0 percent of female sample members were deceased in round 26. Men and women who had died by round 26 had significantly lower Armed Forces Qualifying Test (AFQT) percentile scores. The AFQT is a measure of math and verbal aptitude and is based on four tests from the Armed Services Vocational Aptitude Battery (ASVAB) administered to NLSY79 respondents in 1980. Black men are more likely to have died by round 26, and non-black, non-Hispanic men are less likely to have died by then. Men and women who were deceased by round 26 are more likely to have less than a high school education and less likely to have a bachelor’s degree or more. Characteristics at ages 25, 30, and 35 show that mortality is less likely for men with higher earnings, those who worked 2000 or more hours during the year, and those who were married. The pattern is similar for women. Having had no children by age 35 is associated with higher mortality for men, but lower mortality for women.

Conditional on having been alive in round 26, the next columns compare sample members who were interviewed in round 26 to those who were not, and to those interviewed in round 24 or higher with those who were not interviewed then. The fourth and fifth columns in Table 2 shows that average AFQT scores are similar for those interviewed in round 26 and those not interviewed. Black women are more likely to have been interviewed in round 26, and non-black, non-Hispanic women are less likely. We compare characteristics of the two groups at ages 25, 30, and 35—even though the comparisons will be affected by the fact that we observed a greater proportion of those who continue to participate at the earlier ages. Educational levels are mostly similar for men and women interviewed compared with those not interviewed in round 26. For men, work-related variables--hours and earnings--are similar for the two groups. For women, those with higher hours at age 35 are less likely to have been interviewed in round 26.

The final two columns of Table 2 depict those interviewed in rounds 24, 25, or 26, and those not interviewed in any of those rounds. Several differences are apparent at age 35 and earlier. The distribution of marital status differs based on interview status. A greater percentage of men interviewed are married at age 35 (64 percent vs. 51 percent) and a smaller percentage have never married (22 percent vs. 37 percent). A similar pattern is found for women, although the differentials are smaller. Men and women who did not participate in a recent round have fewer children.

Based on the sample members alive at the time of round 26, Table 3 depicts for men and women separately the probability of being interviewed in round 26 (first two columns) and the probability of being interviewed in round 24 or higher (last two columns). Turning to the first column, black men have a 7 to 8 percentage point higher likelihood of being interviewed in round 26 than their non-black, non-Hispanic and Hispanic counterparts. The rate is about 10 percentage points higher for black women compared to non-black, non-Hispanic or Hispanic women. Men and women with missing information for certain characteristics at age 30 (education level, hours, earnings) have much lower interview rates than those with non-missing information. Women with three or more children at age 30 are more likely to have an interview in round 26 than women with no children (84 percent vs. 73 percent).

The next two columns of Table 3 show the probability of being interviewed in round 24, 25, or 26. Men are less likely to be interviewed in these later rounds (76 percent) than women (84 percent). Black men and black women are more likely to be interviewed in round 24 or higher compared to their non-black, non-Hispanic and Hispanic counterparts. For men and women, the likelihood of being interviewed in the recent rounds mostly falls as the quartile of AFQT score rises. As in the first two columns, those with missing information with respect to education, hours, and earnings at age 30 are much less likely to be interviewed than those who have non-missing information. Women with higher earnings are less likely to be interviewed than those with lower earnings. Men and women with no children have the lowest probability of being interviewed in rounds 24 or higher compared to those with one or more children.

3. Results:

1. Probability of Participating in Recent Rounds

In this section we examine whether attrition is random with respect to demographics, educational attainment, and labor supply. The sample used in the analyses in this section and the next is conditioned on being alive in round 26.

To explore this question, we estimate how status variables measured at age 30 are related to the likelihood of nonresponse in later rounds of the survey. Equation (1) shows the probability that a sample member responds at interview time t as a function of her educational attainment, labor market behavior, marital status, and number of children at age 30.

where X is a set of individual characteristics including an early test score (AFQT), birth year, sex, and race/ethnicity, and ε is an error term from the extreme value distribution. A are the parameters to be estimated. S30 represents status variables at age 30 and includes highest degree completed, annual hours worked measured near the sample member’s 30^th birthday, annual earnings, marital status, and number of children. We control for the various age-30 status variables in separate regressions, permitting us to examine whether different characteristics are related to future attrition. In a final specification, we control for all status variables simultaneously.

In this analysis, we consider two different measures of I: completion of the most recent interview and completion of at least one of the last three interviews. I takes on a value of 1 if individual i is in that category at time t and 0 if she is not.

Table 4, panels a (men) and b (women), contain estimates from logits of the probability of being interviewed in the round 26 (2014-15) interview. Specifications show the relationship between the likelihood of being interviewed in round 26 and various age 30 characteristics--educational attainment, employment, earnings, marital status, and children, while holding constant race and ethnicity and year of birth. For ease of interpretation, marginal effects are shown in the tables. Data are weighted.

All seven specifications show that black women are more likely to be interviewed in round 26, relative to non-black, non-Hispanic women. Men and women with missing AFQT scores are significantly less likely to be interviewed in round 26, and the same is mostly true for those missing hours of work, and earnings at age 30. Item non-response is associated with subsequent unit non-response. Labor market behavior does not appear to be related to attrition; variables such as hours of work and earnings at age 30 are generally not related to the probability of participating in Round 26. There is no significant effect of marital status at age 30 for either men or women. Finally, number of children does not appear to influence the likelihood of being interviewed in round 26 for men or women. In specifications that control for one set of status variables at a time, men with the highest AFQT scores and those in the highest category of education are more likely (by about 6 to 7 percentage points) to be interviewed in Round 26. Neither of these relationships are statistically significant in specification 7, where all regressors are controlled.

Though the estimates are not shown, we also estimate the probability of being interviewed in round 24, 25, or 26. For the most part, the patterns of signs and significance match those in Table 4. Black women have a much higher probability (in the neighborhood of 10 percentage points) of being interviewed in a later round than non-black, non-Hispanic women. Black men are also more likely to be interviewed (almost 6 percentage points) than non-black, non-Hispanic men. Again, missing AFQT score or other variables measured at age 30 (education, hours, earnings) has a large negative impact on the likelihood of being interviewed in rounds 24, 25, or 26. Aside from having missing information, labor market information at age 30 does not affect the likelihood of being interviewed in round 24 or later. One difference is that for both men and women, those who are married and separated or divorced are more likely to be interviewed in round 24 or later compared to their never married counterparts.

All in all, the effect of the examined variables on continued participation in the study is not great. Missing data at age 30, often caused by spotty participation in early rounds, is the main observed correlate with not being interviewed in recent rounds. Attrition from the NLSY79 appears to be fairly random with respect to basic demographics and status at age 30.

2. Effects of Attrition on Estimates of Labor Market Outcomes

We next look at whether attrition biases estimates of simple models of important economic relationships such as the returns to education. We estimate the effects of education and other characteristics at age 30 on labor force participation, earnings, and family income measured at that age. We permit the estimated relationships to vary based on survey participation by interacting all explanatory variables with an indicator of whether the sample member participated in a given survey round(s). As above, we use two measures of survey participation: (1) completion of the Round 26 interview and (2) completion of at least one interview after Round 23. Our equation is:

where H_i30 is educational attainment measured around age 30, X_i is a set of individual characteristics including an early test score, year of birth, and race and ethnicity, MS_i30 is marital status at age 30, C_i30 is an indicator set equal to one if the respondent had any children at age 30, and NI_i is an indicator that sample member i did not participate in a recent interview. α and β are the parameters to be estimated.

Our interest is whether the estimates vary significantly based on interview status; that is, whether the coefficient estimates on the interaction terms are significantly different from zero. Our assumption is that if the relationship between these characteristics and labor market variables does not change with survey participation, attrition bias is not a significant problem, at least for the labor market outcomes considered.

Tables 5 and 6 show the estimated marginal effects for men and women, respectively. We first look at the relationship between these characteristics and labor force participation (defined as working at least 500 hours per year) at age 30 in columns 1 and 2 of the tables. For men and women, the estimated marginal effects on a few key variables explaining labor force participation are statistically significant, but they do not vary with either measure of survey completion. We do not find attrition bias in the marginal effects. Men who have less than a high school education at age 30 have about a 4 to 5 percentage point lower likelihood of working at age 30 compared to men with a high school diploma. Women with less than a high school diploma are about 7 or so percentage points less likely to participate in the labor force. Being separated or divorced has a positive effect of 7 to 9 percentage points on the likelihood of working at least 500 hours per year at age 30 for women compared to those never married. Having children is associated with a higher likelihood of working for women (but not men).

Columns 3 and 4 show estimates of the log of annual earnings at age 30. Unlike in the earlier tables that show earnings quartiles, here we do not allow for 0 earnings; the sample is limited to those who work at least 500 hours per year at age 30 (that is, those with positive labor force participation in the Columns 1 and 2 dependent variables). For women, the estimates do not vary significantly by whether the sample member participated in recent survey rounds. Among men, there are a couple of instances where the estimated effects are larger for those who have not participated in Round 24 through 26. First, the coefficient on the interaction term between “missing” AFQT score and non-participation is large and positive, implying that those who missed recent interviews and who do not have an AFQT score available have significantly higher earnings. Second, while neither marital status nor having children at age 30 has a significant effect on log of earnings for all men, among those men who missed the three survey rounds, the effects are sizeable and significant. Among men who missed the three previous survey rounds, those who were married at age 30 and those who had children by age 30 have significantly higher earnings than men who had not married by age 30 and who had no children by age 30.

Higher AFQT scores are associated with significantly higher earnings at age 30 for both men and women. Separated or divorced women have much higher earnings than unmarried women, and married men have much higher earnings relative to unmarried men. With respect to education, having less than a high school diploma has a negative effect on earnings for women, relative to their counterparts with a high school diploma. Having a bachelor’s degree or more is associated with significantly higher earnings at age 30 for both men and women.

Columns 5 and 6 show the estimates of log of family income at age 30. Again, we focus on the interaction terms that can show whether inference differs for those who have not participated in the recent rounds. In only one instance do these effects vary for recent wave non-participants. Among men who missed the Round 26 interview or all interviews from Rounds 24 to 26, family income is higher for those who were separated or divorced at age 30.

Black and Hispanic men have much lower family income at age 30 than non-black, non-Hispanic men, but the differentials do not vary based on recent interview completion. Race and ethnicity are not significant predictors of family income for women. Higher AFQT scores are associated with higher family income for both men and women. Married men and women have much higher family income than their counterparts who are not married. Having children is associated with higher family income for men, but not for women.

These relationships were not significantly different among the roughly 20 percent of the sample who did not complete the survey in round 26. For the 10 percent who did not complete a survey after round 23, significant differences are found for men for marital status at age 30 and earnings and family income at age 30. From our perspective, this is the most important finding: the coefficients in the models of these labor market processes do vary in only a couple of instances for those who have not participated recently.

3. Effects of Mortality

This section examines whether mortality in the NLSY79 is non-representative and whether this impacts the relationships between personal characteristics and labor market processes that we examined above. Differential mortality may be informative for evaluating comparisons of the NLSY79 to other long-term surveys, especially those that sample at older ages. To this end, we estimate how status measured at age 30 is related to the likelihood of death in later interview rounds. We use an equation similar to (1) above, though in this analysis, I takes on two values, 0 if the individual is alive in Round 26 and 1 if the individual is dead.

We then estimate how mortality in later rounds affects regression estimates of economic relationships, such as the returns to education on earnings. For instance, we estimate the returns to education around age 30 including interaction terms between all regressors and an indicator for having died by Round 26. Statistically significant interaction terms can be attributed to selective mortality—those who die young are different from those who live longer.

Tables 7 (men) and 8 (women) contain estimates from logits of the probability the respondent dies by the round 26 interview. As in prior tables, we hold constant background characteristics of race and ethnicity and year of birth, and then in subsequent columns include sets of variables relating to cognition, education, employment, marriage, and children at age 30, one set at a time. Turning to the first column, Hispanic men and women are no more likely to die by round 26 than non-black, non-Hispanic men and women. Black men are 4 percentage points more likely to die by round 26 than non-black, non-Hispanic men, and black women are 2 percentage points more likely to die by round 26 than non-black, non-Hispanic women. The second column shows that higher AFQT scores are associated with a lower likelihood of mortality by round 26; for example, men and women with the highest quartile of AFQT scores are approximately 7 percentage points less likely to die by round 26 than those with the lowest quartile of AFQT scores.

The third column in Tables 7 and 8 includes background characteristics (race, ethnicity, and year of birth), and adds educational attainment at age 30. The results point to differentials in mortality associated with educational attainment. For men, having a bachelor’s degree or more is associated with over a 6 percentage point decrease in mortality relative to those with a high school education, and having less than a high school education is associated with about a 4 percentage point increase in mortality. For women, having some college or a bachelor’s degree or more is associated with over a 2 percentage point decrease in mortality compared with those with a high school diploma, and those with less than a high school diploma have almost a 4 percentage point higher likelihood of mortality.

In the fourth column, hours of work during the year turned 30 replace educational attainment. Men working less than 2000 hours annually in the year that they turn 30 are more likely to die by Round 26 compared to men who work at least 2000 hours. Women whose hours are missing as well as those who did not work are more likely to have died by round 26, compared with women who work more than 2000 hours in the year they turn 30. Column (5) shows that women who have the highest quartile of earnings have a 7 percentage point lower likelihood of dying by round 26 relative to women with the lowest quartile of earnings. Column (6) focuses on the effects of marital status and children at age 30 on the likelihood of dying by round 26. For both men and women, being separated or divorced decreases the likelihood of dying by age 30 compared with those who have never married.

To examine the effects of nonresponse due to mortality on estimations of the returns to education and other labor market relationships we turn to Table 9. In these estimates, there is no case in which the interaction terms between the explanatory variables and an indicator that the sample member has died by Round 26 is statistically significant. While Tables 7 and 8 indicate that mortality is highly selective, at the time of Round 26, the level of mortality is still small enough to not affect estimates of labor supply, earnings, or income.

4. Summary:

After 26 rounds of data collection, attrition in the NLSY79 remains remarkably low. Over 77 percent of those still living participated in the round 26 interview in 2014-15. In logits examining the probability of participating in later rounds, attrition from the NLSY79 appears to be fairly random with respect to basic demographics and status at age 30. Estimates of labor force participation, earnings, and family income measured around age 30 show no evidence that attrition biases relationships between these labor market outcomes and educational attainment or other characteristics for those who were not interviewed in round 26. However, for the 10 percent of respondents who did not complete a survey after round 23, significant differences are found for men for marital status at age 30 and earnings and family income at age 30. We have not found evidence that mortality by round 26 has biased any of these economic relationships.

References

MaCurdy, Thomas, Thomas Mroz, and R. Mark Gritz, 1998, “An Evaluation of the National Longitudinal Survey of Youth,” Journal of Human Resources, vol. 33(2): 345-436.

Michaud, Pierre-Carl, Arie Kapteyn, James P. Smith, and Arthur van Soest, 2011, “Temporary and Permanent Unit Non-response in Follow-up Interviews of the Health and Retirement Study,” Longitudinal and Life Course Studies, vol. 2(2): 145-69.

Table 1: Nonresponse Patterns in the NLSY79

	Round 14/ ’92	Round 15/’93	Round 16/’94	Round 17/’96	Round 18/’98	Round 19/’00	Round 20/’02	Round 21/’04	Round 22/’06	Round 23/’08	Round 24/’10	Round 25/’12	Round 26/’14
Full Sample
Percent interviewed	0.905	0.904	0.892	0.867	0.842	0.806	0.775	0.769	0.768	0.779	0.759	0.733	0.710
Percent of interviewed who had missed a previous interview (Returnees)	0.180	0.191	0.195	0.197	0.207	0.214	0.223	0.251	0.276	0.303	0.312	0.313	0.323
Percent not interviewed	0.095	0.096	0.108	0.133	0.158	0.194	0.225	0.231	0.232	0.221	0.241	0.267	0.290
Percent of non-interviewed…
Deceased	0.165	0.186	0.190	0.183	0.176	0.162	0.154	0.173	0.197	0.228	0.239	0.259	0.273
Missing their first interview	0.132	0.110	0.129	0.164	0.173	0.183	0.137	0.116	0.086	0.059	0.083	0.073	0.078
Returned at later interview	0.451	0.393	0.377	0.403	0.412	0.418	0.409	0.359	0.293	0.197	0.155	0.111	0.000
Had not returned by Round 26	0.549	0.607	0.623	0.597	0.588	0.582	0.591	0.641	0.707	0.803	0.845	0.889	1.000
Men
Percent interviewed	0.892	0.889	0.878	0.851	0.816	0.780	0.750	0.732	0.744	0.753	0.730	0.702	0.678
Percent of interviewed who had missed a previous interview (Returnees)	0.212	0.224	0.233	0.235	0.242	0.255	0.262	0.289	0.327	0.353	0.361	0.364	0.376
Percent not interviewed	0.108	0.111	0.122	0.149	0.184	0.220	0.250	0.268	0.256	0.247	0.270	0.298	0.322
Percent of non-interviewed…
Deceased	0.205	0.233	0.239	0.234	0.213	0.198	0.191	0.201	0.237	0.268	0.274	0.290	0.305
Missing their first interview	0.129	0.118	0.131	0.155	0.174	0.169	0.111	0.124	0.076	0.056	0.078	0.067	0.072
Returned at later interview	0.441	0.386	0.351	0.373	0.410	0.413	0.400	0.377	0.284	0.197	0.162	0.115	0.000
Had not returned by Round 26	0.559	0.614	0.649	0.627	0.590	0.587	0.600	0.623	0.716	0.803	0.838	0.885	1.000
Non-Black, non-Hispanic Men
Percent interviewed	0.894	0.890	0.881	0.855	0.823	0.790	0.760	0.740	0.726	0.743	0.726	0.695	0.668
Percent of interviewed who had missed a previous interview (Returnees)	0.178	0.187	0.193	0.192	0.199	0.207	0.208	0.223	0.237	0.271	0.280	0.281	0.294
Percent not interviewed	0.106	0.110	0.119	0.145	0.177	0.210	0.240	0.260	0.274	0.257	0.274	0.305	0.332
Percent of non-interviewed…
Deceased	0.165	0.176	0.180	0.175	0.166	0.152	0.144	0.156	0.171	0.198	0.204	0.230	0.244
Missing their first interview	0.109	0.104	0.107	0.140	0.175	0.155	0.104	0.104	0.078	0.046	0.070	0.075	0.084
Returned at later interview	0.395	0.338	0.300	0.329	0.357	0.348	0.331	0.301	0.266	0.175	0.139	0.095	0.000
Had not returned by Round 26	0.605	0.662	0.700	0.671	0.643	0.652	0.669	0.699	0.734	0.825	0.861	0.905	1.000

Black Men
Percent interviewed	0.888	0.896	0.883	0.861	0.810	0.774	0.759	0.732	0.770	0.776	0.751	0.722	0.703
Percent of interviewed who had missed a previous interview (Returnees)	0.233	0.251	0.261	0.267	0.268	0.286	0.311	0.337	0.391	0.413	0.422	0.424	0.434
Percent not interviewed	0.112	0.104	0.117	0.139	0.190	0.226	0.241	0.268	0.230	0.224	0.249	0.278	0.297
Percent of non-interviewed…
Deceased	0.287	0.365	0.374	0.368	0.301	0.285	0.296	0.298	0.380	0.418	0.417	0.416	0.429
Missing their first interview	0.158	0.094	0.162	0.151	0.201	0.180	0.122	0.142	0.069	0.061	0.084	0.066	0.060
Returned at later interview	0.509	0.428	0.391	0.406	0.484	0.483	0.451	0.452	0.303	0.202	0.172	0.128	0.000
Had not returned by Round 26	0.491	0.572	0.609	0.594	0.516	0.517	0.549	0.548	0.697	0.798	0.828	0.872	1.000
Hispanic Men
Percent interviewed	0.893	0.876	0.864	0.826	0.807	0.764	0.714	0.713	0.751	0.742	0.709	0.687	0.663
Percent of interviewed who had missed a previous interview (Returnees)	0.267	0.278	0.291	0.299	0.314	0.332	0.330	0.388	0.446	0.464	0.476	0.481	0.492
Percent not interviewed	0.107	0.124	0.136	0.174	0.193	0.236	0.286	0.287	0.249	0.258	0.291	0.313	0.337
Percent of non-interviewed…
Deceased	0.171	0.189	0.188	0.193	0.190	0.177	0.157	0.163	0.221	0.241	0.253	0.267	0.290
Missing their first interview	0.133	0.180	0.143	0.193	0.132	0.185	0.110	0.145	0.082	0.071	0.088	0.049	0.060
Returned at later interview	0.448	0.443	0.414	0.427	0.418	0.457	0.480	0.447	0.307	0.245	0.204	0.147	0.000
Had not returned by Round 26	0.552	0.557	0.586	0.573	0.582	0.543	0.520	0.553	0.693	0.755	0.796	0.853	1.000
Women
Percent interviewed	0.918	0.920	0.907	0.883	0.870	0.832	0.800	0.806	0.793	0.804	0.789	0.764	0.742
Percent of interviewed who had missed a previous interview (Returnees)	0.148	0.159	0.159	0.159	0.173	0.176	0.185	0.217	0.229	0.256	0.264	0.266	0.274
Percent not interviewed	0.082	0.080	0.093	0.117	0.130	0.168	0.200	0.194	0.207	0.196	0.211	0.236	0.258
Percent of non-interviewed…
Deceased	0.111	0.119	0.126	0.117	0.121	0.114	0.108	0.135	0.147	0.177	0.193	0.218	0.232
Missing their first interview	0.135	0.099	0.126	0.176	0.171	0.202	0.169	0.106	0.098	0.064	0.090	0.081	0.085
Returned at later interview	0.466	0.404	0.410	0.441	0.414	0.425	0.422	0.332	0.303	0.197	0.147	0.106	0.000
Had not returned by Round 26	0.534	0.596	0.590	0.559	0.586	0.575	0.578	0.668	0.697	0.803	0.853	0.894	1.000

Non-Black, non-Hispanic Women
Percent interviewed	0.923	0.921	0.906	0.879	0.872	0.834	0.807	0.801	0.787	0.792	0.779	0.747	0.719
Percent of interviewed who had missed a previous interview (Returnees)	0.133	0.140	0.136	0.134	0.150	0.148	0.156	0.178	0.192	0.216	0.230	0.227	0.233
Percent not interviewed	0.077	0.079	0.094	0.121	0.128	0.166	0.193	0.199	0.213	0.208	0.221	0.253	0.281
Percent of non-interviewed…
Deceased	0.089	0.097	0.107	0.097	0.107	0.092	0.090	0.111	0.121	0.141	0.155	0.181	0.191
Missing their first interview	0.157	0.097	0.107	0.173	0.151	0.191	0.148	0.119	0.100	0.076	0.095	0.089	0.090
Returned at later interview	0.408	0.359	0.376	0.423	0.366	0.370	0.344	0.285	0.256	0.171	0.113	0.089	0.000
Had not returned by Round 26	0.592	0.641	0.624	0.577	0.634	0.630	0.656	0.715	0.744	0.829	0.887	0.911	1.000
Black Women
Percent interviewed	0.911	0.919	0.917	0.900	0.882	0.852	0.819	0.833	0.817	0.839	0.821	0.805	0.797
Percent of interviewed who had missed a previous interview (Returnees)	0.125	0.141	0.151	0.153	0.164	0.175	0.184	0.220	0.228	0.259	0.263	0.269	0.285
Percent not interviewed	0.089	0.081	0.083	0.100	0.118	0.148	0.181	0.167	0.183	0.161	0.179	0.195	0.203
Percent of non-interviewed…
Deceased	0.137	0.150	0.156	0.156	0.160	0.160	0.154	0.198	0.215	0.277	0.299	0.333	0.370
Missing their first interview	0.137	0.100	0.123	0.170	0.217	0.228	0.191	0.109	0.104	0.059	0.091	0.087	0.090
Returned at later interview	0.573	0.525	0.500	0.490	0.497	0.507	0.506	0.405	0.385	0.248	0.220	0.142	0.000
Had not returned by Round 26	0.427	0.475	0.500	0.510	0.503	0.493	0.494	0.595	0.615	0.752	0.780	0.858	1.000
Hispanic Women
Percent interviewed	0.914	0.919	0.893	0.864	0.847	0.800	0.755	0.781	0.768	0.784	0.764	0.748	0.716
Percent of interviewed who had missed a previous interview (Returnees)	0.219	0.232	0.230	0.234	0.247	0.251	0.268	0.311	0.325	0.350	0.358	0.362	0.359
Percent not interviewed	0.086	0.081	0.107	0.136	0.153	0.200	0.245	0.219	0.232	0.216	0.236	0.252	0.284
Percent of non-interviewed…
Deceased	0.119	0.127	0.133	0.120	0.107	0.107	0.092	0.116	0.128	0.151	0.165	0.178	0.187
Missing their first interview	0.083	0.101	0.171	0.188	0.160	0.194	0.188	0.070	0.084	0.042	0.078	0.053	0.065
Returned at later interview	0.429	0.329	0.381	0.429	0.420	0.449	0.483	0.358	0.317	0.203	0.147	0.105	0.000
Had not returned by Round 26	0.571	0.671	0.619	0.571	0.580	0.551	0.517	0.642	0.683	0.797	0.853	0.895	1.000

Table 2: Characteristics Based on Interview History

Panel A: Men

	All	Deceased by Round 26	Alive in Round 26	Interviewed in Round 26*	Not interviewed in Round 26*	Interviewed in Round 24 to 26*	Not interviewed in Rounds 24 to 26*
Characteristics
AFQT score	50.13	40.42	51.10	50.60	49.23	50.05	50.60
Race/ethnicity
Black	0.14	0.20	0.13	0.15	0.13	0.15	0.13
Hispanic	0.07	0.07	0.07	0.07	0.07	0.07	0.07
Non-black, non-Hispanic	0.79	0.73	0.80	0.79	0.81	0.79	0.81
Characteristics at age 25
Education level
Less than high school	0.14	0.24	0.13	0.13	0.16	0.14	0.15
High school diploma	0.45	0.45	0.45	0.44	0.45	0.45	0.42
Some college	0.24	0.23	0.20	0.20	0.20	0.20	0.21
Bachelor's degree or more	0.21	0.10	0.22	0.22	0.19	0.21	0.22
Earnings, quartile
Lowest	0.18	0.24	0.17	0.17	0.18	0.18	0.18
2nd	0.23	0.28	0.22	0.23	0.21	0.23	0.12
3rd	0.25	0.23	0.26	0.25	0.27	0.25	0.26
Highest	0.34	0.25	0.35	0.35	0.33	0.34	0.35
Hours Worked
zero	0.10	0.11	0.10	0.10	0.10	0.10	0.10
less than 500	0.04	0.06	0.04	0.04	0.05	0.04	0.05
500 to less than 1000	0.06	0.06	0.06	0.06	0.06	0.06	0.05
1000 to less than 2000	0.23	0.28	0.22	0.23	0.23	0.23	0.22
2000 or more	0.58	0.49	0.58	0.58	0.57	0.58	0.57
Marital status
Never Married	0.56	0.61	0.55	0.56	0.55	0.55	0.60
Married	0.38	0.30	0.39	0.39	0.37	0.39	0.33
Separated, Divorced, Widowed	0.06	0.09	0.06	0.05	0.08	0.06	0.07
Number of Children
None	0.69	0.63	0.70	0.69	0.69	0.68	0.73
One	0.19	0.20	0.19	0.19	0.18	0.20	0.16
Two	0.09	0.12	0.09	0.09	0.09	0.09	0.08
Three or more	0.03	0.04	0.03	0.03	0.03	0.03	0.03
Characteristics at age 30
Education level
Less than high school	0.13	0.21	0.12	0.13	0.15	0.13	0.14
High school diploma	0.44	0.47	0.43	0.43	0.45	0.44	0.44
Some college	0.21	0.20	0.21	0.21	0.20	0.21	0.20
Bachelor's degree or more	0.23	0.11	0.24	0.24	0.19	0.23	0.23
Earnings, quartile
Lowest	0.15	0.22	0.15	0.15	0.16	0.15	0.16
2nd	0.21	0.29	0.21	0.21	0.21	0.21	0.21
3rd	0.27	0.22	0.27	0.28	0.25	0.28	0.24
Highest	0.36	0.27	0.37	0.36	0.37	0.36	0.39
Hours Worked

zero	0.07	0.11	0.07	0.07	0.07	0.07	0.07
less than 500	0.03	0.06	0.03	0.03	0.03	0.03	0.03
500 to less than 1000	0.04	0.06	0.04	0.04	0.05	0.04	0.04
1000 to less than 2000	0.16	0.19	0.15	0.16	0.14	0.16	0.13
2000 or more	0.70	0.58	0.71	0.70	0.71	0.70	0.72
Marital status
Never Married	0.34	0.49	0.33	0.32	0.39	0.32	0.43
Married	0.55	0.38	0.56	0.57	0.50	0.57	0.47
Separated, Divorced, Widowed	0.11	0.12	0.11	0.11	0.11	0.11	0.10
Number of Children
None	0.48	0.53	0.47	0.46	0.50	0.45	0.55
One	0.24	0.18	0.24	0.24	0.23	0.24	0.22
Two	0.19	0.20	0.19	0.20	0.18	0.20	0.15
Three or more	0.10	0.09	0.10	0.10	0.09	0.10	0.08
Characteristics at age 35
Education level
Less than high school	0.12	0.19	0.11	0.11	0.13	0.11	0.12
High school diploma	0.44	0.52	0.43	0.42	0.49	0.44	0.47
Some college	0.21	0.18	0.21	0.21	0.25	0.21	0.18
Bachelor's degree or more	0.24	0.11	0.25	0.26	0.19	0.24	0.23
Earnings, quartile
Lowest	0.15	0.32	0.14	0.14	0.19	0.15	0.19
2nd	0.19	0.23	0.19	0.19	0.19	0.19	0.21
3rd	0.29	0.25	0.30	0.30	0.28	0.30	0.26
Highest	0.36	0.21	0.38	0.37	0.34	0.37	0.34
Hours Worked
zero	0.08	0.16	0.07	0.07	0.09	0.07	0.09
less than 500	0.02	0.04	0.02	0.02	0.02	0.02	0.02
500 to less than 1000	0.03	0.08	0.02	0.02	0.03	0.02	0.04
1000 to less than 2000	0.11	0.14	0.11	0.11	0.12	0.11	0.12
2000 or more	0.76	0.58	0.78	0.77	0.74	0.77	0.73
Marital status
Never Married	0.25	0.42	0.23	0.22	0.31	0.22	0.37
Married	0.61	0.40	0.63	0.64	0.54	0.64	0.51
Separated, Divorced, Widowed	0.14	0.18	0.14	0.14	0.14	0.14	0.13
Number of Children
None	0.35	0.48	0.34	0.32	0.42	0.31	0.47
One	0.21	0.17	0.21	0.21	0.19	0.21	0.19
Two	0.27	0.22	0.27	0.29	0.24	0.29	0.21
Three or more	0.17	0.13	0.17	0.18	0.15	0.19	0.13
Sample Size	5023	494	4529	3405	1618	3888	1135

Notes: * based on sample alive in Round 26. Data are weighted.

Table 2: Characteristics Based on Interview History

Panel B: Women

	All	Deceased by Round 26	Alive in Round 26	Interviewed in Round 26*	Not interviewed in Round 26*	Interviewed in Round 24 to 26*	Not interviewed in Rounds 24 to 26*
Characteristics
AFQT score	49.26	38.35	49.91	49.00	49.80	48.71	51.86
Race/ethnicity
Black	0.14	0.18	0.14	0.15	0.10	0.15	0.08
Hispanic	0.06	0.06	0.06	0.06	0.07	0.06	0.07
Non-black, non-Hispanic	0.80	0.76	0.80	0.78	0.83	0.79	0.85
Characteristics at age 25
Education level
Less than high school	0.12	0.23	0.12	0.12	0.12	0.13	0.10
High school diploma	0.45	0.48	0.45	0.45	0.44	0.45	0.43
Some college	0.23	0.16	0.23	0.23	0.26	0.23	0.25
Bachelor's degree or more	0.2	0.12	0.20	0.20	0.19	0.20	0.21
Earnings, quartile
Lowest	0.32	0.47	0.32	0.32	0.35	0.32	0.34
2nd	0.28	0.27	0.28	0.29	0.25	0.28	0.24
3rd	0.25	0.2	0.25	0.24	0.25	0.24	0.27
Highest	0.15	0.07	0.16	0.15	0.15	0.15	0.14
Hours Worked
zero	0.16	0.22	0.16	0.16	0.17	0.16	0.19
less than 500	0.08	0.11	0.08	0.08	0.09	0.08	0.08
500 to less than 1000	0.09	0.11	0.09	0.09	0.09	0.09	0.08
1000 to less than 2000	0.30	0.28	0.30	0.30	0.28	0.30	0.29
2000 or more	0.37	0.28	0.37	0.37	0.37	0.37	0.36
Marital status
Never Married	0.39	0.45	0.38	0.38	0.41	0.38	0.43
Married	0.52	0.43	0.53	0.53	0.50	0.53	0.49
Separated, Divorced, Widowed	0.09	0.13	0.09	0.09	0.09	0.09	0.08
Number of Children
None	0.54	0.49	0.54	0.52	0.58	0.52	0.61
One	0.23	0.27	0.23	0.23	0.23	0.23	0.21
Two	0.17	0.19	0.17	0.18	0.14	0.17	0.14
Three or more	0.06	0.06	0.06	0.07	0.04	0.07	0.04
Characteristics at age 30
Education level
Less than high school	0.10	0.21	0.10	0.10	0.11	0.10	0.10
High school diploma	0.43	0.46	0.43	0.43	0.43	0.40	0.44
Some college	0.24	0.16	0.24	0.24	0.24	0.25	0.24
Bachelor's degree or more	0.22	0.12	0.23	0.23	0.23	0.26	0.22
Earnings, quartile
Lowest	0.35	0.46	0.34	0.34	0.37	0.40	0.34
2nd	0.29	0.32	0.29	0.30	0.27	0.25	0.30
3rd	0.23	0.18	0.23	0.24	0.20	0.19	0.23
Highest	0.14	0.04	0.14	0.13	0.16	0.16	0.13

Hours Worked
zero	0.19	0.28	0.18	0.18	0.22	0.18	0.22
less than 500	0.07	0.1	0.07	0.08	0.07	0.08	0.07
500 to less than 1000	0.08	0.08	0.08	0.08	0.07	0.08	0.07
1000 to less than 2000	0.25	0.21	0.25	0.26	0.23	0.25	0.23
2000 or more	0.41	0.34	0.42	0.41	0.42	0.41	0.40
Marital status
Never Married	0.23	0.37	0.22	0.22	0.25	0.22	0.29
Married	0.64	0.45	0.65	0.64	0.63	0.64	0.61
Separated, Divorced, Widowed	0.13	0.18	0.13	0.14	0.12	0.14	0.11
Number of Children
None	0.35	0.39	0.34	0.32	0.40	0.32	0.45
One	0.22	0.18	0.23	0.22	0.22	0.23	0.21
Two	0.27	0.25	0.27	0.28	0.25	0.28	0.23
Three or more	0.16	0.19	0.16	0.18	0.13	0.17	0.11
Characteristics at age 35
Education level
Less than high school	0.09	0.18	0.09	0.09	0.10	0.09	0.06
High school diploma	0.42	0.50	0.41	0.41	0.42	0.42	0.37
Some college	0.25	0.17	0.26	0.26	0.25	0.25	0.27
Bachelor's degree or more	0.24	0.15	0.24	0.24	0.23	0.24	0.29
Earnings, quartile
Lowest	0.34	0.45	0.34	0.33	0.39	0.34	0.38
2nd	0.31	0.38	0.31	0.32	0.29	0.31	0.29
3rd	0.21	0.13	0.21	0.21	0.18	0.21	0.18
Highest	0.14	0.04	0.14	0.13	0.15	0.13	0.15
Hours Worked
zero	0.18	0.23	0.18	0.17	0.22	0.17	0.22
less than 500	0.07	0.06	0.07	0.07	0.07	0.07	0.07
500 to less than 1000	0.07	0.09	0.07	0.07	0.07	0.07	0.06
1000 to less than 2000	0.25	0.30	0.24	0.24	0.27	0.24	0.26
2000 or more	0.43	0.32	0.44	0.45	0.38	0.44	0.39
Marital status
Never Married	0.17	0.32	0.16	0.16	0.20	0.18	0.23
Married	0.67	0.46	0.69	0.68	0.66	0.68	0.64
Separated, Divorced, Widowed	0.16	0.22	0.15	0.16	0.14	0.17	0.12
Number of Children
None	0.24	0.35	0.23	0.22	0.30	0.21	0.36
One	0.20	0.16	0.20	0.19	0.20	0.20	0.20
Two	0.33	0.24	0.33	0.33	0.31	0.34	0.28
Three or more	0.24	0.25	0.24	0.26	0.20	0.26	0.17
Sample Size	4941	296	4645	3666	1275	4050	891

Notes: *based on sample alive in Round 26. Data are weighted.

Table 3: Probability of Being Interviewed

	Round 26		Round 24 to 26
	Males	Females	Males	Females
Total	73.4	77.1	76.2	84.0
Race/ethnicity
Black	80.3	86.1	88.1	90.8
Hispanic	72.6	75.8	81.8	82.8
Non-black, non-Hispanic	72.3	75.7	80.2	83.0
AFQT score, by quartile
Lowest	78.3	82.6	87.6	89.6
2nd	73.7	78.6	81.4	86.3
3rd	72.2	77.2	81.8	83.9
Highest	75.4	76.5	80.5	82.0
AFQT score not available	51.1	49.9	62.1	59.8
Characteristics at age 30
Education level
missing	20.7	19.8	24.7	22.9
Less than high school	77.6	83.6	87.3	91.3
High school diploma	75.8	80.9	85.2	88.8
Some college	77.0	79.2	85.5	86.4
Bachelor's degree or more	77.9	79.0	83.5	84.9
Hours worked
None	81.7	80.2	88.7	88.3
Greater than 0, but less than 500	81.0	84.7	89.1	89.7
500 to less than 1000	77.1	83.9	89.7	89.4
1000 to less than 2000	82.7	83.3	90.8	89.3
2000 or more	77.1	80.7	85.5	88.5
Missing	34.5	31.5	39.6	37.3
Earnings (1984$), by quartile
Lowest	78.1	80.1	86.7	87.1
2nd	79.4	83.0	87.8	90.0
3rd	77.7	83.0	85.9	89.3
Highest	75.3	73.6	83.4	83.0
Missing	23.6	20.1	29.1	23.1
Marital status
Never Married	72.0	76.6	78.1	81.8
Married	74.2	76.3	82.6	83.5
Separated, Divorced, Widowed	73.2	81.6	85.1	89.9
Missing	74.2	79.1	84.7	87.8
Number of Children
None	72.1	72.9	78.6	79.0
1 child	72.9	76.4	81.5	84.7
2 children	75.8	78.9	85.7	86.2
3 or more children	75.8	83.9	86.4	90.2
Sample Size	3405	3666	3888	4050
Notes: Data are weighted.

Table 4. Marginal Effects, Logit of Probability Interviewed in Round 26 (2014) NLSY79

Panel A: Men

Characteristics	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Race/ethnicity
Hispanic	-0.007	0.011	0.004	-0.002	-0.003	-0.002	-0.003
	(0.019)	(0.020)	(0.020)	(0.019)	(0.019)	(0.020)	(0.020)
Black	0.025	0.042	0.035	0.028	0.028	0.035	0.059
	(0.017)	(0.019)	(0.017)	(0.017)	(0.017)	(0.018)	(0.020)
Quartile of AFQT
Missing		-0.074					-0.107
		(0.035)					(0.033)
2nd		-0.014					-0.018
		(0.022)					(0.023)
3rd		0.001					-0.022
		(0.023)					(0.025)
4th		0.069					0.020
		(0.025)					(0.030)
Education at Age 30
Missing			-0.030				0.017
			(0.154)				(0.136)
Less than high school			-0.019				0.010
			(0.023)				(0.025)
Some college			0.020				0.015
			(0.021)				(0.021)
Bachelor's degree or more			0.061				0.014
			(0.022)				(0.024)
Hours of Work at Age 30
Missing				-0.231			0.165
				(0.030)			(0.089)
Zero				-0.005			0.062
				(0.030)			(0.042)
More than 0 up to 500 hours				-0.047			0.054
				(0.045)			(0.052)
500 hours up to 1000 hours				-0.038			-0.003
				(0.038)			(0.041)
1000 to less than 2000				0.026			0.054
				(0.023)			(0.025)
Earnings at age 30, by quartile
Missing					-0.161		-0.165
					(0.061)		(0.057)
2nd					0.016		0.031
					(0.030)		(0.033)
3rd					0.036		0.033
					(0.030)		(0.004)
Highest					0.020		0.004
					(0.029)		(0.034)
Did not work					0.013		---
					(0.037)
Marital Status at Age 30

Married						-0.019	-0.018
						(0.028)	(0.027)
Separated or divorced						0.034	0.012
						(0.020)	(0.020)
Missing						0.025	-0.006
						(0.058)	(0.054)
Number of Children at Age 30
No children						0.020	0.018
						(0.021)	(0.022)
2 children						-0.008	-0.008
						(0.024)	(0.022)
3 or more children						-0.003	-0.004
						(0.028)	(0.026)

Note: Sample size is 4529. Data are weighted. Standard errors are in parentheses. All specifications control for year of birth.

Table 4. Marginal Effects, Logit of Probability Interviewed in Round 26 (2014) NLSY79

Panel B: Women

Characteristics	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Race/ethnicity
Hispanic	0.013	0.026	0.017	0.014	0.013	0.006	0.020
	(0.018)	(0.019)	(0.018)	(0.017)	(0.017)	(0.018)	(0.019)
Black	0.086	0.096	0.089	0.085	0.089	0.074	0.090
	(0.017)	(0.019)	(0.017)	(0.017)	(0.017)	(0.018)	(0.020)
Quartile of AFQT
Missing		-0.122					-0.089
		(0.035)					(0.035)
2nd		0.027					0.024
		(0.021)					(0.022)
3rd		0.024					0.034
		(0.023)					(0.024)
4th		0.034					0.040
		(0.024)					(0.028)
Education at Age 30
Missing			0.075				0.120
			(0.200)				(0.171)
Less than high school			-0.023				-0.014
			(0.025)				(0.026)
Some college			-0.006				-0.012
			(0.019)				(0.019)
Bachelor's degree or more			0.001				0.016
			(0.020)				(0.024)
Hours of Work at Age 30
Missing				-0.263			-0.033
				(0.030)			(0.057)
Zero				-0.033			-0.032
				(0.020)			(0.032)
More than 0 up to 500 hours				0.018			0.009
				(0.031)			(0.036)
500 hours up to 1000 hours				0.021			0.018
				(0.031)			(0.035)
1000 to less than 2000				0.024			0.023
				(0.020)			(0.020)
Earnings at age 30, by quartile
Missing					-0.233		-0.214
					(0.063)		(0.063)
2nd					0.009		0.010
					(0.023)		(0.027)
3rd					0.025		0.034
					(0.024)		(0.031)
Highest					-0.043		-0.039
					(0.027)		(0.034)
Did not work					-0.042		---
					(0.025)
Marital Status at Age 30

Married						-0.012	-0.024
						(0.027)	(0.021)
Separated or divorced						-0.023	-0.017
						(0.021)	(0.027)
Missing						-0.033	-0.012
						(0.051)	(0.051)
Number of Children at Age 30
No children						-0.019	-0.020
						(0.022)	(0.021)
2 children						0.005	0.018
						(0.021)	(0.021)
3 or more children						0.038	0.057
						(0.024)	(0.024)

Note: Sample size is 4645. Data are weighted. Standard errors are in parentheses. All specifications control for year of birth.

Table 5: Estimates of Economic Outcomes, Marginal Effects, Men
	Labor Force Participation		Ln (Earnings)		Ln(Family Income)
	Interactions: Not interviewed in Round 26	Interactions: Not interviewed in Rounds 24 to 26	Interactions: Not interviewed in Round 26	Interactions: Not interviewed in Rounds 24 to 26	Interactions: Not interviewed in Round 26	Interactions: Not interviewed in Rounds 24 to 26
Variable
Main
Race/ethnicity
Hispanic	-0.047	-0.049	-0.074	-0.013	-0.214	-0.193
	(0.015)	(0.014)	(0.128)	(0.115)	(0.095)	(0.090)
Black	-0.082	-0.089	-0.051	-0.081	-0.456	-0.466
	(0.012)	(0.011)	(0.105)	(0.098)	(0.089)	(0.081)
Quartile of AFQT
Missing	-0.037	-0.035	-0.321	-0.388	0.094	0.120
	(0.025)	(0.022)	(0.304)	(0.299)	(0.182)	(0.159)
2nd	0.034	0.022	0.131	0.095	0.313	0.255
	(0.015)	(0.015)	(0.148)	(0.137)	(0.089)	(0.091)
3rd	0.017	0.014	0.513	0.490	0.363	0.338
	(0.018)	(0.017)	(0.144)	(0.128)	(0.086)	(0.079)
4th	0.004	0.000	0.632	0.565	0.410	0.385
	(0.021)	(0.019)	(0.157)	(0.147)	(0.096)	(0.088)
Marital Status at Age 30
Separated or divorced	-0.024	-0.017	0.112	0.085	-0.155	-0.102
	(0.019)	(0.018)	(0.154)	(0.150)	(0.153)	(0.134)
Married	0.022	0.018	0.321	0.299	0.740	0.708
	(0.013)	(0.012)	(0.118)	(0.112)	(0.067)	(0.062)
Missing	0.103	0.110	0.128	-0.162	0.567	0.589
	(0.031)	(0.030)	(0.326)	(0.387)	(0.151)	(0.140)
Any children at Age 30	0.009	0.011	-0.191	-0.193	0.224	0.203
	(0.013)	(0.012)	(0.109)	(0.105)	(0.058)	(0.053)
Education at age 30
Missing	-0.020	-0.029	-0.935	-0.775	-1.241	-1.259
	(0.070)	(0.071)	(1.251)	(1.321)	(0.969)	(0.968)
Less than high school	-0.049	-0.051	-0.025	-0.096	-0.544	-0.593
	(0.016)	(0.015)	(0.139)	(0.133)	(0.121)	(0.124)
Some college	-0.013	-0.012	-0.004	-0.045	0.115	0.114
	(0.014)	(0.013)	(0.132)	(0.127)	(0.066)	(0.061)
Bachelor's degree or higher	0.038	0.029	0.470	0.478	0.434	0.422
	(0.019)	(0.018)	(0.114)	(0.111)	(0.066)	(0.061)
Dummy variable for non-participation	-0.020	-0.079	-0.121	-0.525	0.006	-0.314
	(0.035)	(0.038)	(0.340)	(0.443)	(0.216)	(0.272)
Interaction Terms
Race/ethnicity
Hispanic	0.025	0.053	0.226	-0.044	0.005	-0.069

	(0.026)	(0.031)	(0.223)	(0.277)	(0.178)	(0.212)
Black	-0.007	0.029	0.067	0.221	0.097	0.170
	(0.024)	(0.028)	(0.223)	(0.274)	(0.164)	(0.196)
Quartile of AFQT
Missing	-0.006	-0.002	0.506	1.028	0.124	0.121
	(0.039)	(0.043)	(0.481)	(0.386)	(0.231)	(0.245)
2nd	-0.017	0.036	-0.216	-0.197	-0.283	-0.119
	(0.029)	(0.034)	(0.277)	(0.336)	(0.209)	(0.220)
3rd	0.025	0.060	-0.312	-0.399	-0.005	0.136
	(0.035)	(0.044)	(0.257)	(0.343)	(0.150)	(0.175)
4th	-0.009	0.012	-0.047	0.281	-0.145	-0.066
	(0.041)	(0.050)	(0.317)	(0.373)	(0.174)	(0.202)
Marital Status at Age 30
Separated or divorced	0.050	0.041	0.318	0.699	0.610	0.720
	(0.036)	(0.044)	(0.314)	(0.356)	(0.236)	(0.266)
Married	0.000	0.021	0.252	0.512	0.116	0.322
	(0.025)	(0.027)	(0.242)	(0.280)	(0.154)	(0.195)
Missing	-0.067	-0.102	-1.444	-0.493	0.156	0.094
	(0.062)	(0.070)	(1.239)	(1.441)	(0.268)	(0.299)
Any children at Age 30	0.027	0.037	0.334	0.559	-0.042	0.062
	(0.024)	(0.026)	(0.220)	(0.251)	(0.113)	(0.139)
Education at age 30
Less than high school	0.020	0.040	-0.267	-0.008	-0.127	0.063
	(0.029)	(0.032)	(0.251)	(0.272)	(0.244)	(0.224)
Some college	0.012	0.009	-0.315	-0.194	-0.056	-0.091
	(0.027)	(0.032)	(0.298)	(0.362)	(0.165)	(0.228)
Bachelor's degree or higher	0.016	0.088	-0.355	-0.507	-0.004	0.049
	(0.041)	(0.056)	(0.271)	(0.321)	(0.114)	(0.136)

Note: Sample size is 4657 for labor force participation, 4020 for earnings, and 4272 for family income. Samples condition on having an interview around age 30. Data are weighted. Standard errors are in parentheses. All specifications control for year of birth.

Table 6: Estimates of Economic Outcomes, Marginal Effects, Women
	Labor Force Participation		Ln (Earnings)		Ln(Family Income)
	Interactions: Not interviewed in Round 26	Interactions: Not interviewed in Rounds 24 to 26	Interactions: Not interviewed in Round 26	Interactions: Not interviewed in Rounds 24 to 26	Interactions: Not interviewed in Round 26	Interactions: Not interviewed in Rounds 24 to 26
Variable
Main
Race/ethnicity
Hispanic	-0.006	-0.001	0.159	0.135	0.009	-0.046
	(0.021)	(0.020)	(0.129)	(0.120)	(0.080)	(0.081)
Black	-0.001	0.003	0.269	0.178	-0.120	-0.107
	0.020	0.019	0.120	0.114	0.080	0.080
Quartile of AFQT
Missing	-0.063	-0.053	0.036	0.098	-0.057	0.108
	(0.044)	(0.041)	(0.378)	(0.321)	(0.193)	(0.171)
2nd	0.047	0.048	0.549	0.526	0.311	0.322
	(0.022)	(0.021)	(0.163)	(0.152)	(0.099)	(0.097)
3rd	0.058	0.059	0.629	0.589	0.529	0.550
	(0.026)	(0.025)	(0.178)	(0.169)	(0.092)	(0.093)
4th	0.057	0.062	0.584	0.546	0.502	0.537
	(0.030)	(0.029)	(0.198)	(0.187)	(0.101)	(0.101)
Marital Status at Age 30
Separated or divorced	0.067	0.086	0.295	0.227	-0.002	0.039
	(0.029)	(0.028)	(0.149)	(0.139)	(0.117)	(0.113)
Married	0.009	0.012	0.091	0.045	0.713	0.722
	(0.022)	(0.021)	(0.121)	(0.114)	(0.080)	(0.081)
Missing	0.085	0.058	-0.791	-0.687	-0.120	-0.025
	(0.056)	(0.050)	(0.619)	(0.573)	(0.404)	(0.358)
Any children at Age 30	0.238	0.247	0.545	0.531	0.149	0.167
	(0.023)	(0.022)	(0.119)	(0.110)	(0.071)	(0.070)
Education at age 30
Missing	-0.226	-0.230	0.756	0.689	0.169	0.245
	(0.132)	(0.133)	(0.870)	(0.880)	(0.391)	(0.354)
Less than high school	-0.075	-0.068	-0.778	-0.621	-0.578	-0.478
	(0.025)	(0.024)	(0.251)	(0.224)	(0.134)	(0.124)
Some college	0.040	0.047	0.177	0.162	0.159	0.150
	(0.022)	(0.021)	(0.119)	(0.112)	(0.061)	(0.063)
Bachelor's degree or higher	0.021	0.025	0.355	0.355	0.475	0.473
	(0.026)	(0.025)	(0.140)	(0.130)	(0.064)	(0.061)
Dummy variable for non-participation	-0.084	-0.030	-0.471	-1.805	-0.646	-0.789
	(0.062)	(0.075)	(0.457)	(0.705)	(0.343)	(0.407)
Interaction Terms
Race/ethnicity
Hispanic	0.100	0.115	-0.023	0.510	-0.039	0.227
	(0.045)	(0.056)	(0.339)	(0.492)	(0.238)	(0.302)

Black	0.049	0.034	-0.045	0.594	0.312	0.365
	(0.044)	(0.053)	(0.377)	(0.530)	(0.221)	(0.295)
Quartile of AFQT
Missing	0.125	0.139	0.380	0.817	0.334	-0.069
	(0.077)	(0.085)	(0.697)	(0.896)	(0.489)	(0.664)
2nd	0.057	0.078	0.315	0.849	0.498	0.740
	(0.049)	(0.059)	(0.441)	(0.647)	(0.259)	(0.337)
3rd	0.026	0.033	0.057	0.565	0.207	0.239
	(0.055)	(0.065)	(0.553)	(0.800)	(0.283)	(0.379)
4th	0.043	0.044	0.358	1.061	0.530	0.649
	(0.066)	(0.076)	(0.588)	(0.833)	(0.287)	(0.380)
Marital Status at Age 30
Separated or divorced	0.073	-0.038	-0.016	0.417	0.322	0.212
	(0.065)	(0.076)	(0.383)	(0.524)	(0.308)	(0.405)
Married	0.017	-0.012	-0.245	-0.082	0.431	0.609
	(0.049)	(0.058)	(0.283)	(0.354)	(0.229)	(0.251)
Missing	-0.144	-0.067	1.653	1.693	0.974	1.211
	(0.105)	(0.138)	(0.749)	(0.892)	(0.523)	(0.613)
Any children at Age 30	0.019	-0.039	0.063	0.328	0.242	0.283
	(0.046)	(0.051)	(0.280)	(0.382)	(0.165)	(0.180)
Education at age 30
Less than high school	0.006	-0.070	0.531	-0.198	0.329	-0.101
	(0.053)	(0.065)	(0.586)	(0.907)	(0.345)	(0.516)
Some college	-0.009	-0.062	-0.066	0.083	-0.442	-0.592
	(0.045)	(0.053)	(0.351)	(0.477)	(0.208)	(0.251)
Bachelor's degree or higher	0.037	0.026	0.480	0.694	-0.097	-0.186
	(0.057)	(0.064)	(0.401)	(0.538)	(0.153)	(0.188)

Note: Sample sizes are 4650 for labor force participation, 3352 for earnings, and 4365 for family income. Data are weighted. Samples condition on having an interview around age 30. Standard errors are in parentheses. All specifications control for year of birth.

Table 7: Marginal Effects, Probability of Being Deceased at Round 26 (2014), Men

				(1)		(2)			(3)			(4)		(5)				(6)
Race/ethnicity
Hispanic				0.006		-0.011			-0.005			-0.001		0.001				0.002
				(0.011)		(0.012)			(0.011)			(0.011)		(0.011)				(0.011)
Black				0.04		0.017			0.032			0.031		0.032				0.025
				(0.008)		(0.010)			(0.009)			(0.009)		(0.009)				(0.010)
Quartile of AFQT
Missing						-0.037
						(0.020)
2nd						-0.019
						(0.013)
3rd						-0.045
						(0.014)
4th						-0.072
						(0.016)
Education at Age 30
Missing									0.116
									(0.013)
Less than high school									0.039
									(0.013)
Some college									-0.009
									(0.014)
Bachelor's degree or more									-0.064
									(0.018)
Hours of Work at Age 30
Missing												0.110
												(0.012)
Zero												0.047
												(0.016)
More than 0 up to 500 hours							0.084
										(0.025)
500 hours up to 1000 hours							0.046
										(0.023)
1000 to less than 2000										0.033
										(0.015)
Earnings at Age 30, by quartile
Missing													0.045
													(0.022)
2nd													0.013
													(0.018)
3rd													-0.032
													(0.019)
Highest													-0.034

									(0.018)
Did not work									0.017
									(0.020)
Marital Status at Age 30
Married											-0.027
											(0.016)
Separated or divorced											-0.069
											(0.012)
Missing											-0.072
											(0.038)
Number of Children at Age 30
No children											0.009
											(0.014)
2 children											0.033
											(0.015)
3 or more children											0.019
											(0.018)

Note: Sample size is 5023. Data are weighted. Standard errors are in parentheses. All specifications control for year of birth.

Table 8: Marginal Effects, Probability of Being Deceased at Round 26 (2014), Women

	(1)	(2)	(3)	(4)	(5)	(6)
Race/ethnicity
Hispanic	-0.003	-0.025	-0.013	-0.006	-0.007	-0.008
	(0.010)	(0.011)	(0.010)	(0.010)	(0.010)	(0.010)
Black	0.019	-0.006	0.015	0.018	0.016	0.004
	(0.007)	(0.009)	(0.007)	(0.007)	(0.007)	(0.008)
Quartile of AFQT
Missing		-0.014
		(0.015)
2nd		-0.044
		(0.011)
3rd		-0.057
		(0.013)
4th		-0.067
		(0.014)
Education at Age 30
Missing			0.057
			(0.012)
Less than high school			0.036
			(0.011)
Some college			-0.022
			(0.012)
Bachelor's degree or more			-0.024
			(0.013)
Hours of Work at Age 30
Missing				0.063
				(0.012)
Zero				0.032
				(0.011)
More than 0 up to 500 hours				0.026
				(0.016)
500 hours up to 1000 hours				0.011
				(0.016)
1000 to less than 2000				0.001
				(0.012)
Earnings at Age 30, by quartile
Missing					0.022
					(0.020)
2nd					-0.003
					(0.012)
3rd					-0.022
					(0.014)
Highest					-0.073

					(0.023)
Did not work					0.012
					(0.012)
Marital Status at Age 30
Married						-0.012
						(0.012)
Separated or divorced						-0.049
						(0.011)
Missing						-0.011
						(0.022)
Number of Children at Age 30
No children						0.005
						(0.011)
2 children						0.011
						(0.012)
3 or more children						0.021
						(0.012)

Note: Sample size is 4941. Data are weighted. Standard errors are in parentheses. All specifications control for year of birth.

Table 9: Estimates of Economic Outcomes with Interactions with Indicator for Whether Deceased at Round 26, Marginal Effects
	Labor force participation		Ln(Earnings)			Ln(Family income)
	Men	Women	Men		Women	Men	Women
Variable
Main
Race/ethnicity
Hispanic	-0.043	0.013	-0.008	0.159		-0.204	0.014
	(0.013)	(0.019)	(0.110)	(0.129)		(0.081)	(0.081)
Black	-0.085	0.015	0.014	0.269		-0.443	-0.072
	(0.011)	(0.018)	(0.098)	(0.120)		(0.077)	(0.080)
Quartile of AFQT
Missing	-0.043	-0.024	-0.088	0.036		0.143	0.205
	(0.020)	(0.037)	(0.245)	(0.378)		(0.136)	(0.169)
2nd	0.026	0.057	0.118	0.549		0.284	0.398
	(0.014)	(0.020)	(0.134)	(0.163)		(0.078)	(0.096)
3rd	0.024	0.062	0.445	0.629		0.376	0.559
	(0.016)	(0.024)	(0.130)	(0.178)		(0.072)	(0.094)
4th	0.004	0.068	0.682	0.584		0.388	0.589
	(0.018)	(0.028)	(0.140)	(0.198)		(0.081)	(0.099)
Marital Status at Age 30
Separated or divorced	-0.010	0.075	0.177	0.295		-0.014	0.049
	(0.017)	(0.027)	(0.145)	(0.149)		(0.122)	(0.115)
Married	0.021	0.009	0.372	0.091		0.732	0.788
	(0.012)	(0.020)	(0.111)	(0.121)		(0.061)	(0.077)
Missing	0.082	0.056	-0.288	-0.791		0.528	0.110
	(0.030)	(0.050)	(0.420)	(0.619)		(0.128)	(0.345)
Any children at Age 30	0.010	0.243	-0.112	0.545		0.200	0.231
	(0.012)	(0.020)	(0.101)	(0.119)		(0.050)	(0.063)
Education at age 30
Missing	-0.010	-0.205	-0.659	0.756		-1.218	0.694
	(0.068)	(0.121)	(1.226)	(0.870)		(0.978)	(0.443)
Less than high school	-0.038	-0.074	-0.067	-0.778		-0.513	-0.582
	(0.014)	(0.023)	(0.125)	(0.251)		(0.099)	(0.137)
Some college	-0.009	0.039	-0.097	0.177		0.135	0.102
	(0.013)	(0.019)	(0.126)	(0.119)		(0.061)	(0.061)
Bachelor's degree or higher	0.036	0.025	0.360	0.355		0.433	0.447
	(0.017)	(0.024)	(0.109)	(0.140)		(0.056)	(0.059)
Dummy variable for non-participation	-0.077	-0.076	0.281	-0.471		-0.247	-0.347
	(0.054)	(0.098)	(0.338)	(0.457)		(0.430)	(0.448)
Interaction Terms
Race/ethnicity
Hispanic	0.020	0.065	-0.240	-0.023		-0.176	-0.674
	(0.043)	(0.081)	(0.369)	(0.339)		(0.436)	(0.483)
Black	0.013	-0.067	-0.555	-0.045		0.124	0.296

	(0.038)	(0.068)	(0.338)	(0.377)	(0.296)	(0.380)
Quartile of AFQT
Missing	0.008	-0.012	0.271	0.380	0.196	-2.791
	(0.067)	(0.155)	(0.364)	(0.697)	(0.333)	(1.766)
2nd	0.039	0.017	-0.720	0.315	-0.605	0.381
	(0.044)	(0.084)	(0.387)	(0.441)	(0.474)	(0.406)
3rd	0.000	-0.007	-0.177	0.057	-0.041	0.077
	(0.057)	(0.103)	(0.315)	(0.553)	(0.336)	(0.456)
4th	-0.036	-0.089	-0.832	0.358	-0.117	0.519
	(0.063)	(0.133)	(0.658)	(0.588)	(0.417)	(0.566)
Marital Status at Age 30
Separated or divorced	0.002	0.090	0.545	-0.016	0.699	0.092
	(0.049)	(0.101)	(0.368)	(0.383)	(0.462)	(0.482)
Married	-0.012	0.011	0.371	-0.245	0.448	0.056
	(0.036)	(0.077)	(0.283)	(0.283)	(0.327)	(0.424)
Missing	-0.062	-0.069	1.463	1.653	1.171	-0.452
	(0.097)	(0.149)	(0.586)	(0.749)	(0.402)	(0.625)
Any children at Age 30	0.070	0.004	0.241	0.063	0.224	-0.174
	(0.036)	(0.079)	(0.253)	(0.280)	(0.257)	(0.450)
Education at age 30
Less than high school	-0.002	-0.001	-0.391	0.531	-0.476	1.208
	(0.043)	(0.085)	(0.319)	(0.586)	(0.511)	(0.414)
Some college	-0.004	-0.042	-0.065	-0.066	-0.620	-0.918
	(0.040)	(0.088)	(0.461)	(0.351)	(0.436)	(0.506)
Bachelor's degree or higher	0.150	0.090	0.131	0.480	-0.010	0.135
	(0.059)	(0.121)	(0.479)	(0.401)	(0.241)	(0.389)

Note: Sample sizes for men are 4657 for labor force participation, 4020 for earnings, and 4272 for family income. Sample sizes are 4650 for labor force participation, 3352 for earnings, and 4365 for family income. Samples condition on having an interview around age 30 and thus on being alive around age 30. Data are weighted. Standard errors are in parentheses. All specifications control for year of birth.

1 Race and ethnicity is divided into three mutually exclusive and exhaustive categories: (1) Black, Non-Hispanic, (2) Hispanic, and (3) Non-Black, Non-Hispanic.

3