Attrition Patterns Manajemen | Fakultas Ekonomi Universitas Maritim Raja Ali Haji 536.full

and the children’s scores on developmental assessments. This is done by comparing the full sample with the sample of respondents who have never missed an interview. Section III examines how attrition affects estimates of the relationship between fam- ily characteristics and the children’s assessment scores. Lastly, Section IV summa- rizes and discusses the Žndings.

II. Attrition Patterns

The analysis of the NLSY79 and NLSY79-C in this study is based on 4,926 women and their 9,460 children whose births are recorded through the 2000 interview. 2 Analysis is done separately for the full sample, that is, all 4926 women and their children, and for the cross-sectional sample which omits the women in the Black and Hispanic oversamples and their children. The data are unweighted. 3 Table 1 describes the attrition patterns for the women of the NLSY79 by present- ing the rates of attrition and Žrst-attrition for the full sample and the cross-section. Attrition has been relatively low among the women of the NLSY79, although it has increased over time. The attrition rates range from 3 percent to 4 percent in the early years of the survey and from 12 percent to 17 percent in the recent biennial survey- years with the patterns being quite similar for the full-sample of women and the cross-section. Until the most recent interview, initial attrition was greatest in the second year of the survey when nearly 4 percent of the full-sample and cross-sec- tional women attrited. In the subsequent years, Žrst attrition is generally between 1 percent and 2 percent until the 1996 interview. The patterns displayed in this table show the extent to which attrition has affected the sample of women in the NLSY79, but do not provide information on whether attrition has caused the sample to be nonrepresentative. Table 2 compares all person-year observations of the women in the NLSY79 with the annual observations from only those women who have never missed an interview continuous subsample to assess whether attrition is nonrepresentative along se- lected characteristics that describe the women’s fertility decisions, marital status and economic status. Fitzgerald, Gottschalk, and MofŽtt 1998 point out that the appro- priate comparison is between the full sample and the nonattritors, not between the attritors and nonattritors where attrition may lead to bias in both samples. Table 2 is organized into two panels. Panel A presents the statistics for the full sample, and Panel B the statistics for the cross-section. In each of the panels, Col- umns 1-4 present the descriptive statistics for all person-year observations and Col- umns 5-8 for the person-year observations of women in the continuous subsample. 2. Because they have been dropped from the NLSY79, the samples used in this paper exclude women from the armed forces supplement and the supplement of economically disadvantaged whites and their children. 3. The analysis in this paper on the women in the NLSY79 also was done using the full and cross-sectional samples weighted by the NLSY79-provided weights and the MaCurdy-Mroz-Gritz cross-sectional weights. The analysis on children’s descriptive statistics Table 4 for the full sample also was conducted using the NLSY79-C weights. The use of weights does not affect the key Žndings so long as the cross-section and the supplemental samples are examined separately, though the levels for some variables family in- come, AFQT are affected. Table 1 Attrition Rates for the NLSY79 Women Attrition First Attrition Full Cross- Full Cross- Year Sample Section Sample Section 1980 0.038 0.039 0.038 0.039 1981 0.033 0.033 0.015 0.016 1982 0.039 0.037 0.014 0.011 1983 0.032 0.031 0.010 0.008 1984 0.041 0.043 0.015 0.016 1985 0.052 0.053 0.020 0.018 1986 0.070 0.068 0.024 0.022 1987 0.081 0.080 0.024 0.023 1988 0.090 0.091 0.028 0.029 1989 0.073 0.076 0.011 0.010 1990 0.089 0.088 0.019 0.018 1991 0.080 0.079 0.011 0.010 1992 0.082 0.081 0.014 0.015 1993 0.080 0.082 0.010 0.010 1994 0.093 0.096 0.015 0.013 1996 0.118 0.120 0.027 0.024 1998 0.130 0.131 0.030 0.028 2000 0.167 0.166 0.047 0.044 Number of observations 4,926 3,108 Notes: Attrition rates are deŽned as the number of noninterviews out of the women in the sample. Rates of Žrst attrition are deŽned as the number of women not interviewed in a given year out of the women who have continuously remained in the sample throughout that year. The descriptive statistics are grouped based on the year of observation. In the Žrst and Žfth columns the observations across all years 1986, 1988, 1990, 1992, 1994, 1996, 1998, and 2000 are pooled, in the second and sixth columns the observations from 1986, 1988, and 1990 are pooled, in the third and seventh columns those obser- vations from 1992, 1994, and 1996 are pooled, and in the fourth and eight columns observations from 1998 and 2000 are pooled. In both the full sample and the cross-section, the members of the continuous sub- sample have family incomes that are slightly larger than those for the entire sample, implying that the distribution of family income for attritors lies below that of nonat- tritors. In each of the samples, however, the marital status of the women, the presence of children, the incidence of poverty, and highest grade completed appear to be largely unaffected by the exclusion of the attritors. Table 2 shows that once the NLSY79-C begins, the effects of attrition on the sample of women in the NLSY79 are small. However, this comparison neglects the women who permanently attrite from the NLSY79 prior to the 1986 interview. The Journa l of Hum an Re so ur ce s All Annual Observations Annual observations, Continuous Subsample All 1986, 1992, 1998 All 1986, 1992, 1998 Child 1988, and 1994, and and Child 1988, and 1994, and and Years 1990 1996 2000 Years 1990 1996 2000 Age Mean 32 27 33 38 32 27 33 38 Median 32 27 33 38 32 27 33 38 10 percent 25 23 29 35 25 23 29 35 90 percent 39 31 37 41 39 31 37 41 Any children 0.70 0.60 0.75 0.78 0.70 0.60 0.75 0.79 Number of children Mean 1.44 1.14 1.60 1.69 1.47 1.15 1.63 1.71 Median 1 1 2 2 1 1 2 2 10 percent 90 percent 3 3 3 3 3 3 3 3 Marital status Never married 0.26 0.36 0.23 0.17 0.26 0.36 0.23 0.18 Married 0.54 0.50 0.56 0.58 0.55 0.50 0.57 0.59 Separated 0.06 0.05 0.07 0.07 0.06 0.05 0.07 0.06 Divorced 0.12 0.09 0.13 0.16 0.12 0.08 0.13 0.16 Widowed 0.02 0.00 0.01 0.03 0.01 0.01 0.00 0.01 Family income Mean 29,790 23,277 34,607 32,970 30,153 23,697 34,475 33,599 Median 22,007 19,355 22,310 26,224 22,678 19,756 22,966 27,011 10 percent 5,437 5,139 5,577 6,002 5,790 5,323 6,063 6,603 90 percent 52,399 45,806 51,969 62,617 53,072 46,524 52,148 62,928 Poverty status 0.19 0.21 0.18 0.15 0.18 0.21 0.17 0.14 Highest grade completed Mean 13 13 13 13 13 13 13 13 Median 12 12 12 12 12 12 12 12 10 percent 11 10 11 11 11 11 11 12 90 percent 16 16 16 16 16 16 16 16 Notes: “All child years” includes years 1986, 1988, 1990, 1992, 1994, 1996, 1998, and 2000. indicates that some observations are missing values for these variables. A u ghinba ugh 541 All Annual Observations Annual observations, Continuous Subsample All 1986, 1992, 1998 All 1986, 1992, 1998 Child 1988, and 1994, and and Child 1988, and 1994, and and Years 1990 1996 2000 Years 1990 1996 2000 Age Mean 32 27 33 38 32 27 33 38 Median 32 27 33 38 32 27 33 38 10 percent 25 23 29 35 25 23 29 35 90 percent 39 31 37 41 39 31 37 41 Any children 0.67 0.55 0.73 0.77 0.68 0.56 0.74 0.78 Number of children Mean 1.33 1.01 1.49 1.60 1.37 1.03 1.52 1.63 Median 1 1 2 2 1 1 2 2 10 percent 90 percent 3 3 3 3 3 3 3 3 Marital status Never married 0.22 0.31 0.17 0.13 0.21 0.31 0.17 0.12 Married 0.61 0.55 0.64 0.66 0.62 0.55 0.65 0.67 Separated 0.05 0.04 0.05 0.05 0.04 0.04 0.05 0.04 Divorced 0.12 0.09 0.13 0.16 0.12 0.09 0.13 0.15 Widowed 0.00 0.00 0.01 0.00 0.01 0.01 0.00 0.01 Family income Mean 33,416 25,582 39,963 36,155 33,689 25,990 39,534 36,854 Median 24,913 21,774 25,744 29,841 25,543 22,305 26,247 30,612 10 percent 6,530 5,916 6,920 7,477 6,996 6,162 7,489 8,163 90 percent 55,701 48,387 55,363 65,541 56,055 49,257 55,433 65,660 Poverty status 0.14 0.16 0.13 0.11 0.13 0.15 0.11 0.11 Highest grade completed Mean 13 13 13 13 13 13 13 14 Median 12 12 12 12 12 12 12 13 10 percent 11 11 12 12 12 11 12 12 90 percent 16 16 16 17 16 16 16 17 Notes: “All child years” includes years 1986, 1988, 1990, 1992, 1994, 1996, 1998, and 2000. indicates that some observations are missing values for these variables. Next, I describe the attrition patterns for the children of the NLSY79 women. The samples of children examined in this paper are limited to children born in 1994 or earlier so that the children will be old enough to have been assessed and to have had at least one chance to attrite after assessment. Because the children are no longer eligible for assessment upon turning 15, they are dropped from the samples at age 15. 4 There are two difŽculties in examining the attrition patterns for the children. First, because children are being born into the sample as the survey continues causing the sample to change over time, the patterns must be described separately for groups of children based on their year of birth. Second, some of the children who are, by deŽni- tion, part of the sample have never been identiŽed, much less assessed. For instance, there would be no information collected about a child born in 1991 whose mother attrited in 1989 and never returned to the survey. Using information on fertility through the 2000 interview for those women who are interviewed in that round, the expected number of missed births is approximately 150, with an upper bound of 330. 5 Table 3 shows the fraction of children about whom supplemental information was collected 1 out of the children known to have been born by the start of the interview period and 2 out of the children whose mothers were interviewed. Because inter- viewers are instructed that interviewing the main youth is the highest priority, assess- ing a child generally accompanies the collection of his mother’s interview. Conse- quently, having a mother who is interviewed substantially increases the odds that information on the child is collected. There are few cases where the child assessment occurs, but the mother is not interviewed. Excluding the 2000 interview —at each interview, supplemental information is not collected for 10 to 20 percent of all known children, and is missing for between 5 and 10 percent of the children whose mothers are interviewed. 6 Table 4 attempts to assess the effects of attrition in the child supplement since its beginning in 1986. This table compares the descriptive statistics for the sample of all child-year observations and the subsample of child-year observations for the chil- dren for whom supplemental information is collected in all eligible years the conti- nuous subsample. This table is organized in the same way as Table 2. The statistics focus on the items that describe the child’s family background and the resources avail- able to the child, such as, mother’s marital status, mother’s highest grade completed, average family income during the child’s Žrst three years, and maternal employment during the child’s Žrst year, as well as, the child’s standard scores on the Peabody Pictorial Vocabulary Test PPVT and the Behavior Problems Index BPI. For the full sample Table 4a and the cross-section Table 4b, the sample of all observation years and the continuous subsample are similar along the dimensions presented. Although children in the continuous samples score higher on the PPVT, 4. Beginning in 1994, a young adult interview was conducted for those children 15 and older. The adminis- tration of the young adult interview is not linked to the interview of the mother. 5. The estimates of the number of children who are entirely omitted from the sample are calculated using information about the fertility of women who are interviewed in 2000. The mean number or the mean plus one standard deviation of subsequent births for women of the same parity in the last survey that the attriting woman was interviewed is used as the number of missed births for each attriting woman. 6. Because of funding issues, a portion of the children were not eligible for interview in 2000. The ineligi- bles were randomly selected. In the 2002 interview, the entire sample of children was restored. A u ghinba ugh 543 1986 1988 1990 1992 1994 1996 1998 2000 Number of Mother Mother Mother Mother Mother Mother Mother Mother observations All Interview All Interview All Interview All Interview All Interview All Interview All Interview All Interview Born Prior to 1986 Interview Period Full sample 4,533 0.90 0.95 0.88 0.95 0.83 0.89 0.83 0.88 0.83 0.89 0.77 0.85 0.80 0.89 — — Cross-section 2,462 0.90 0.95 0.89 0.95 0.83 0.89 0.83 0.88 0.83 0.89 0.78 0.84 0.79 0.88 — — Born After Start of 1986 Interview Period and Prior to Start of 1988 Interview Period Full sample 1,199 — — 0.90 0.96 0.88 0.95 0.89 0.94 0.87 0.94 0.83 0.91 0.83 0.92 0.62 0.72 Cross-section 730 — — 0.90 0.97 0.89 0.95 0.89 0.95 0.88 0.95 0.83 0.92 0.82 0.91 0.72 0.85 Born After Start of 1988 Interview Period and Prior to Start of 1990 Interview Period Full sample 1,010 — — — — 0.90 0.95 0.91 0.95 0.89 0.95 0.84 0.92 0.84 0.91 0.65 0.74 Cross-section 638 — — — — 0.91 0.96 0.91 0.95 0.90 0.96 0.82 0.92 0.83 0.91 0.75 0.85 Born After Start of 1990 Interview Period and Prior to Start of 1992 Interview Period Full sample 806 — — — — — — 0.92 0.95 0.92 0.96 0.87 0.93 0.85 0.92 0.66 0.74 Cross-section 520 — — — — — — 0.92 0.95 0.92 0.96 0.88 0.93 0.84 0.90 0.75 0.86 Born After Start of 1992 Interview Period and Prior to Start of 1994 Interview Period Full sample 779 — — — — — — — — 0.93 0.96 0.89 0.94 0.87 0.93 0.68 0.75 Cross-section 477 — — — — — — — — 0.94 0.97 0.90 0.95 0.87 0.94 0.80 0.87 Born After Start of 1994 Interview Period and Prior to Start of 1996 Interview Period Full sample 481 — — — — — — — — — — 0.88 0.92 0.90 0.95 0.67 0.74 Cross-section 310 — — — — — — — — — — 0.89 0.93 0.92 0.97 0.76 0.84 Notes: In 2000, about 38 percent of the Black and Hispanic oversamples were not eligible for interview because of funding constraints. In calculating the fraction of children interviewed in 2000, the ineligible have been removed from the denominator. The ineligibles were randomly selected. The Journa l of Hum an Re so ur ce s Table 4a Descriptive Statistics for Person-Year Observations of Children, Children Born in 1994 or Earlier: Full Sample All Annual Observations Annual observations, Continuous sample All 1986, 1992, 1998 All 1986, 1992, 1998 Child 1988, and 1994, and and Child 1988, and 1994, and and Years 1990 1996 2000 Years 1990 1996 2000 Child’s age in months Mean 88 69 94 117 92 73 98 118 Median 87 63 94 120 92 68 101 120 10 percent 22 13 26 68 24 14 28 69 90 percent 155 131 159 163 158 137 162 164 Number of children in family Mean 2.52 2.36 2.60 2.66 2.54 2.39 2.62 2.66 Median 2 2 2 2 2 2 2 2 10 percent 1 1 1 1 1 1 1 1 90 percent 4 4 4 4 4 4 4 4 Mother’s marital status Never married 0.16 0.21 0.14 0.10 0.16 0.21 0.14 0.10 Married 0.63 0.61 0.64 0.67 0.64 0.61 0.65 0.69 Separated 0.08 0.08 0.08 0.07 0.08 0.08 0.08 0.07 Divorced 0.12 0.10 0.12 0.14 0.11 0.10 0.12 0.14 Widowed 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 Mother’s highest grade Mean 12 12 13 13 13 12 13 13 Median 12 12 12 12 12 12 12 13 10 percent 10 9 10 11 10 9 10 12 90 percent 16 14 16 16 16 14 16 16 A u ghinba ugh 545 Median 20,945 16,729 21,453 29,283 21,371 16,945 22,026 30,012 10 percent 5,459 4,919 5,706 7,297 5,692 5,076 6,035 7,868 90 percent 49,827 40,323 50,519 66,791 49,844 40,363 49,869 66,044 Hours worked in Žrst year Mean 688 565 720 840 691 567 710 857 Median 62 132 463 105 144 520 10 percent 90 percent 2,080 1,880 2,080 2,080 2,080 1,894 2,80 2,080 Worked in Žrst year 0.51 0.46 0.53 0.58 0.52 0.47 0.53 0.59 Child’s PPVT score Mean 89 85 90 94 90 86 91 95 Median 90 87 91 95 91 88 92 97 10 percent 64 60 67 68 65 61 68 70 90 percent 113 109 113 118 114 109 114 120 Child’s BPI score Mean 106 108 106 104 106 108 107 103 Median 105 107 105 103 106 107 106 103 10 percent 88 90 88 84 88 90 89 84 90 percent 126 126 126 123 125 126 126 122 Number of observations 39,607 15,125 17,673 6,809 28,902 10,941 12,653 5,308 Notes: “All child years” includes years 1986, 1988, 1990, 1992, 1994, 1996, 1998, and 2000. indicates that some observations are missing values for these variables. PPVT and BPI scores are standardized for age of the child. The Journa l of Hum an Re so ur ce s Table 4b Descriptive Statistics for Person-Year Observations of Children, Children Born in 1994 or Earlier: Cross-section All Annual Observations Annual observations, Continuous sample All 1986, 1992, 1998 All 1986, 1992, 1998 Child 1988, and 1994, and and Child 1988, and 1994, and and Years 1990 1996 2000 Years 1990 1996 2000 Child’s age in months Mean 86 65 90 116 88 67 92 116 Median 85 59 89 118 87 62 91 118 10 percent 21 11 24 68 22 12 24 67 90 percent 153 127 156 162 155 131 158 163 Number of children in family Mean 2.41 2.26 2.47 2.54 2.43 2.29 2.48 2.54 Median 2 2 2 2 2 2 2 2 10 percent 1 1 1 1 1 1 1 1 90 percent 4 4 4 4 4 4 4 4 Mother’s marital status Never married 0.09 0.13 0.08 0.06 0.08 0.13 0.08 0.06 Married 0.73 0.70 0.74 0.75 0.74 0.71 0.75 0.76 Separated 0.06 0.06 0.06 0.06 0.05 0.06 0.05 0.05 Divorced 0.12 0.10 0.12 0.13 0.11 0.10 0.11 0.13 Widowed 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 Mother’s highest grade Mean 12 12 13 14 13 12 13 14 Median 12 12 12 12 12 12 12 13 10 percent 10 10 11 12 11 10 11 12 90 percent 16 15 16 16 16 15 16 16 A u ghinba ugh 547 Family income Mean 32,631 23,274 37,799 38,648 32,181 23,240 36,216 38,784 Median 24,930 20,446 25,949 33,013 25,234 20,968 26,057 33,206 10 percent 6,308 5,576 6,780 8,403 6,605 10,595 7,242 8,972 90 percent 53,772 43,306 54,740 67,098 53,084 42,831 53,287 66,231 Hours worked in Žrst year Mean 715 589 743 859 735 603 753 885 Median 221 282 560 285 15 332 626 10 percent 90 percent 2,080 1,920 2,080 2,080 2,080 1,950 2,080 2,080 Worked in Žrst year 0.54 0.49 0.56 0.60 0.56 0.50 0.57 0.62 Child’s PPVT score Mean 95 91 95 98 95 92 96 99 Median 96 93 96 99 96 94 96 99 10 percent 71 67 74 76 72 68 73 77 90 percent 116 112 116 122 117 113 117 123 Child’s BPI score Mean 106 108 106 103 106 108 106 103 Median 105 107 105 102 105 108 105 102 10 percent 86 90 87 84 88 91 88 84 90 percent 124 126 124 121 124 126 124 121 Number of observations 23,585 8,489 10,575 4,521 18,419 6,359 8,232 3,828 Notes: “All child years” includes years 1986, 1988, 1990, 1992, 1994, 1996, 1998, and 2000. indicates that some observations are missing values for these variables. PPVT and BPI scores are standardized for age of the child. Table 5 Means for Children’s Mothers 1979: All Known Children, and Children who were Continuously, Sometimes, and Never in the Child Supplement Continuously Sometimes in Never in in NLSY79-C NLSY79-C NLSY79-C All Sample Sample Sample Full Sample Age 17.87 17.93 17.73 18.10 Any children 0.23 0.24 0.20 0.31 Number of children 0.33 0.35 0.27 0.49 Married 0.18 0.19 0.16 0.28 Divorced 0.01 0.01 0.01 0.01 Separated 0.02 0.02 0.02 0.04 Widowed 0.00 0.00 Never married 0.79 0.78 0.81 0.68 Highest grade completed 10.32 10.40 10.21 10.03 AFQT score 33.21 35.60 28.83 30.53 Number of observations 8,457 5,256 2,846 355 Cross-Section Age 17.89 17.91 17.81 18.01 Any children 0.19 0.19 0.19 0.25 Number of children 0.27 0.27 0.26 0.37 Married 0.19 0.19 0.18 0.30 Divorced 0.01 0.01 0.01 0.00 Separated 0.02 0.02 0.03 0.02 Widowed Never married 0.78 0.78 0.78 0.67 Highest grade completed 10.54 10.61 10.41 10.23 AFQT score 41.79 44.11 36.36 39.36 Number of observations 4,950 3,322 1,424 204 Notes: Sample is composed mothers of children born in 1994 or earlier. indicates that some observations are missing values for these variables. most characteristics of the children and their families are affected little by whether attritors are included or excluded. In general, the comparisons in Table 4 imply that small differences exist between attritors and nonattritors. However, the comparisons neglect the group of children for whom supplemental information is never collected; they are potentially different from those for whom supplemental information is col- lected at least once. Table 5 explores the differences between mean 1979 characteristics of the mothers of the children who are and are not included in the child supplement. This information is presented for all known children and three subsamples: 1 children continuously in the sample, 2 children for whom supplemental information is sometimes col- lected, and 3 children for whom supplemental information is never collected. The comparisons are done separately for the full and cross-sectional samples. Although the mothers of children in the continuous sample score higher on the AFQT than the mothers of children who are either sometimes or never in the NLSY-C, along most other dimensions they look similar to the mothers of the children for whom supple- mental information is sometimes collected. The mothers of the children for whom supplemental information is never collected look different in that they are the most likely to have had a child and the least likely to have never been married at the 1979 interview. Overall, children who are surveyed in some of the child supplement years in which they are eligible are similar to those who are surveyed in all eligible years. The mothers of these two groups of children also appear similar. However, the children who are not interviewed in any of the mother-child supplement years and their moth- ers are different. These mothers more often experienced an early marriage or birth. Attrition from the NLSY79 prior to the start of the child supplement appears nonran- dom with respect to women’s fertility through 1979 and marital status at that time. Evaluating the effects that this attrition has on the sample of children is made difŽcult because there is no information on children whose mothers attrite before their births, but the fact that there are not many of these children minimizes the chance that their absence will have large consequences.

III. Nature of Attrition