Thursday, Aug 28
	PSID Guide | Data and Documentation | Publications and Conferences | Child Development Supplement | Search What's New | CDS Guide | Time Diary | Bibliography | Contact Us

Description of the 1997 PSID Child Supplement Weights

October 27, 1999

1. Introduction.

      The University of Michigan Survey Research Center’s Panel Study of Income Dynamics (PSID) is an ongoing longitudinal survey of a representative sample of U.S. families. The study collects data on employment, income, wealth, housing, food expenditures, transfer income, and marital and fertility behavior. The PSID began in 1968 with a national sample of 5,000 families and had grown to include over 8,700 families in 1996 through the formation of new families by children or other sample members of the original 5,000 families. Since no new families were incorporated, new entrants to the United States were not included. The introduction of a national sample of post-1968 immigrants in 1997 (facilitated by a reduction of the core sample) made the PSID sample representative of the U.S. population in 1997. The 1997 PSID Child Development Supplement is an addition to the PSID core data collection designed to provide researchers with comprehensive nationally-representative data about children ages 0-12 and their families. All respondents to the Child Supplement will have been included in the 1997 wave. The majority of respondents will be from long-term PSID families. Others will be from the addition to the core PSID of what we here call the "1997 PSID Immigrant sample." Although the 1997 PSID Child Development Supplement is drawn primarily from the PSID Core families, the Child Supplement includes a group of African-American families with children under age 13 in 1997 that are not included in the 1997 PSID Core. Without these additional families, the Child Supplement family weights would be exactly the same as the 1997 PSID Core family weights. ¹

      In order to construct the family selection weight component for the Child Supplement, the effect of the core reduction on the probabilities of selection for families in each of the six strata must be considered. Table 1 shows the six parts which make up the 1997 PSID Child Development Supplement. Families in Strata 1 and 5 used the 1996 PSID family weights as the preliminary 1997 family weights for the core sample. These two groups were not affected by the 1997 PSID core reduction. Families in Strata 2, 3, and 4 had probabilities of being selected both into the SRC and SEO samples. In the 1997 core reduction, only Black families from the original SEO sample had a chance of retention. Therefore, the 1997 family weights which reflected the joint probabilities of selection had to be recomputed to exclude the SEO probability for non-Blacks and to adjust the SEO probability for Blacks. Families in Stratum 6 are present only in the 1997 PSID Child Development Supplement. They are excluded from the 1997 PSID Core and therefore do not have 1997 Core family weights.

2. Composition of the 1997 PSID Child Development Supplement.

      The PSID consists of two separate samples--a nationally representative sample of U.S. families designed by the University of Michigan Survey Research Center (SRC) and an over sample of low income, mostly African-American, families from the Survey of Economic Opportunity (SEO). In 1997 the full SRC sample was retained plus all the original African-American families from the low-income SEO subsample with children under age 13 in 1997 and a subsample of families without children under 13, a total of 6,792 families.²    The analysis weights constructed for the Child Development Supplement are the product of three factors: 1) a family selection weight which is the inverse of the family’s probability of selection; 2) a post-stratification factor which adjusts the sample family totals to the 1997 CPS estimated totals for forty-eight demographic/geographic cells; and 3) a within family selection weight which is the inverse of the probability of selection of the child from the set of children age 0-12 in the family.

      In order to construct the family selection weight component for the Child Supplement, the effect of the core reduction on the probabilities of selection for families in each of the six strata must be considered. Table 1 shows the six parts which make up the 1997 PSID Child Development Supplement. Families in Strata 1 and 5 used the 1996 PSID family weights as the preliminary 1997 family weights for the core sample. These two groups were not affected by the 1997 PSID core reduction.³    Families in Strata 2, 3, and 4 had probabilities of being selected both into the SRC and SEO samples. In the 1997 core reduction, only Black families from the original SEO sample had a chance of retention. Therefore, the 1997 family weights which reflected the joint probabilities of selection had to be recomputed to exclude the SEO probability for non-Blacks and to adjust the SEO probability for Blacks. Families in Stratum 6 are present only in the 1997 PSID Child Development Supplement. They are excluded from the 1997 PSID Core and therefore do not have 1997 Core family weights.

3. Weighting Adjustment to Integrate Samples and Create Initial Family Weight.

      Most of the Child Supplement families from the 1997 PSID Core Sample and all of the Child Supplement families from the 1997 PSID Immigrant sample had already been assigned family weights which were calculated for the 1997 PSID Core (including the Immigrant) sample. However, cases from Stratum 6 which are present only in the 1997 Child Supplement and not in the Core sample had values of zero for the 1997 PSID Core family weight. In addition, cases which were part of the 1994 PSID recontact effort also had values of zero for the 1997 Core family weight. These recontact cases have no interview data from 1993 and prior years and are unusable for longitudinal analyses using the Core data set, but are usable for the Child Development Supplement which began in 1997. Table 2 shows the number of cases with zero and non-zero weights in each of the six strata.

      As a first step in the assignment of family weights to the cases that had zero weights for the 1997 Core family weight, sixteen weighting cells were formed by crossing the strata defined above (collapsed from six to four)⁵ by a two-category family type (two-parent or other) by a two category age of family head (<30 or 30 and older). The algorithm used for assigning family weights to these "zero weight" cases was initially to assign to them the mean weight of the cases in the weight domain cell to which they belong and then to scale the sum of the weights in the weight domain cell back to the total of the original non-zero weights. The total of the non-zero weights in the 1997 Core data set was matched to 1997 CPS totals for the cells (in 000s). Table 3 shows the grouping of cases into weight domain cells and the sums of the Core family weights in each domain. The Adjustment Factor was the factor used to scale back the family weight of each case in the cell (both the positive weight cases and the cases which originally had zero weights and were assigned weights equal to the cell means) so that the sum of weights in the cell was equal to the 1997 Core family weight total.

4. Poststratification to 1997 CPS Totals.

      The PSID Child Development Supplement famiies were assigned to forty-eight poststratification cells formed by crossing the following four variables: (1) race of head (Non-Black / Black); (2) education level of head (< high school graduation / high school graduation / some post-secondary education); (3) MSA status (MSA / non-MSA); (4) Census Region (Northeast / Midwest / South / West). If there were fewer than 15 families in a cell, adjacent cells were collapsed across the Census Region dimension.

      Corresponding weighted totals for each poststratification cell were calculated for the PSID Child Development Supplement families and for the 1997 CPS households.⁷   The weight used for the PSID Child Supplement totals was the adjusted family weight described in Part 3.

      The 1997 CPS March Supplement household weight (divided by 1000) was used for the CPS totals. Only CPS households which included a child 0 - 12 were included in the poststratification totals. Poststratification factors were computed by dividing the CPS total by the PSID Child Supplement total for each cell. Table 4 shows the poststratification cells, the number of Child Supplement families, the weighted CPS total, the weighted Child Supplement total, and the poststratification factor.

5. Within Family Child Selection Weight Factor.

      Interviewers collected data about each age-eligible child in the PSID Child Development Supplement families up to a maximum of two children per family. If there were more than two age-eligible children in a family, two children were selected using a random procedure. The probability of selection of a child is the number of children selected divided by the number of eligible children in the family; and the within family selection weight is the inverse of this probability. The denominator used to construct the PSID child selection weight factor was the number of children interviewed in the family instead of the number of children selected. This weight factor, therefore, incorporates a child-level nonresponse adjustment. For example if there were three eligible children in the family and two were selected and one interviewed, the nonresponse adjusted selection weight factor is 3/1 = 3.0. If both children were interviewed the selection weight factor would be 3/2 = 1.5. Table 5 shows the distribution of PSID Child Development Supplement families by the number of age eligible children and the number of children interviewed.

Table 6: Weighted Distribution of Race by Sex by 2-Year Age Group
for the 1997 PSID Child Development Supplement
and the 1997 CPS March Supplement.

Race	Sex	Age	PSID Child unwtd.	PSID Child unwtd. %	PSID Child    wtd.	PSID Child wtd. %	1997 CPS March Suppl. (000s)	1997 CPS March Suppl. %
Non-Black	Male	0-1	159	4.5	3,115.8	6.2	3,354.9	6.5
Non-Black	Male	2-3	140	3.9	2,617.5	5.2	3,415.5	6.6
Non-Black	Male	4-5	168	4.7	3,190.7	6.3	3,508.9	6.8
Non-Black	Male	6-7	164	4.6	3,604.8	7.1	3,568.8	6.9
Non-Black	Male	8-9	159	4.5	3,286.0	6.5	3,375.6	6.5
Non-Black	Male	10-11	180	5.1	3,637.8	7.2	3,378.9	6.5
Non-Black	Male	12	75	2.1	1,573.8	3.1	1,730.1	3.3
Non-Black	Female	0-1	150	4.2	2,909.5	5.7	3,213.2	6.2
Non-Black	Female	2-3	161	4.5	3,337.8	6.6	3,264.6	6.3
Non-Black	Female	4-5	156	4.4	3,044.6	6.0	3,353.5	6.5
Non-Black	Female	6-7	148	4.2	2,834.5	5.6	3,368.5	6.5
Non-Black	Female	8-9	172	4.8	3,656.2	7.0	3,154.6	6.1
Non-Black	Female	10-11	161	4.5	3,345.7	6.6	3,233.4	6.2
Non-Black	Female	12	112	3.1	2,509.4	5.0	1620.3	3.1
Black	Male	0-1	78	2.2	460.4	0.9	609.7	1.2
Black	Male	2-3	124	3.5	638.0	1.3	637.1	1.2
Black	Male	4-5	135	3.8	694.0	1.4	686.2	1.3
Black	Male	6-7	114	3.2	1054.2	2.1	687.8	1.3
Black	Male	8-9	106	3.0	548.9	1.1	660.8	1.3
Black	Male	10-11	154	4.3	921.5	1.8	643.1	1.2
Black	Male	12	69	1.9	362.5	0.7	309.3	0.6
Black	Female	0-1	85	2.4	420.2	0.8	594.5	1.1
Black	Female	2-3	93	2.6	338.0	0.7	620.2	1.2
Black	Female	4-5	105	2.9	481.6	1.0	665.7	1.3
Black	Female	6-7	101	2.8	626.0	1.2	670.4	1.3
Black	Female	8-9	110	3.1	621.9	1.2	641.0	1.2
Black	Female	10-11	109	3.1	512.6	1.0	625.1	1.2
Black	Female	12	75	2.1	379.3	0.7	294.8	0.6
TOTAL			3563	100.0	50,633.2	100.0	51,886.5	100.0

6. Weights for Supplementary Questionnaire Data

      The 1997 PSID Child Development Supplement collected reliable age-graded assessments of the cognitive, behavioral, and health status of children age 0-12 from several sources including the primary caregiver, a second caregiver, an absent parent, a teacher, a school administrator, and the child. The Primary Caregiver is the main respondent and is usually the child’s mother. If the mother is not living with the child, the primary caregiver could be the father, legal guardian or person who knows most about the child’s activities. If the primary caregiver was interviewed about two children, the interviewer filled out a separate questionnaire for each child.

      These Primary Caregiver/Child interviews form the core data collection of the Supplement. In order to be included in the PSID Child Development Supplement data set, data about the child must have been obtained from the primary caregiver. The response rate for the primary caregiver interviews about each child was 88 percent. Response rates for the auxiliary questionnaires were lower. For example, although the response rate for the Primary Caregiver/Child interview was 88 percent, the response rate for the Primary Caregiver completion of the household questionnaire was 63 percent. In order to construct weights for analyzing these supplementary data sets, a nonresponse adjustment was constructed by multiplying each weight by the ratio of the sum of weights for the total eligible by the sum of weights for the total interviewed. This nonresponse adjustment was computed separately for each poststratification cell (as defined in Table 4: race of head / education of head / MSA status / and Census Region).

      The following example shows how these weights were calculated. The poststratification cell for Non-Black heads who did not complete high school and live in an MSA in the Northeast has a weighted total of 658.138 for the total set of Primary Caregiver/Child respondents. Of these, all were eligible to complete the Primary Caregiver household book, but the weighted total for Primary Caregiver/Child respondents in this poststratification cell who completed this questionnaire was 345.861. Therefore, the nonresponse adjustment factor for the Primary Caregiver/Child weights in this cell was 658.138/345.861 = 1.903.

      Not all Child Development Supplement families had Other Caregivers. The nonresponse adjustments for the Other Caregiver weights were created by dividing the total weight for the eligible cases in each poststratification cell by the total weight for the respondents. For example, the weighted total for cases with an Other Caregiver in the cell which includes Non-Black heads who did not complete high school and who live in an MSA in the Northeast is 541.369. The weighted sum of Other Caregiver respondents in this cell is 300.362. Therefore the nonresponse adjustment factor for this cell is 541.369/300.362 = 1.802.

      Special weights were computed for the Primary Caregiver Household Book data (PCGHHBWT), the Other Caregiver Child Book data (OTCGCHWT), and the Other Caregiver Household Book data (OTCGHHWT). These weights, like the Primary Caregiver/Child weight (CH97PRWT), should be used for analyses involving child-level data or data involving the relationship of the child with a caregiver or with family characteristics. The family level weight (CH97HHWT) should be used for analyses at the family level. The special weights described above (PCGHHBWT, OTCGCHWT, and OTCGHHWT) can be converted to family level weights by dividing them by the within family subselection weight (SUBSELWT). Table 7 shows for each supplemental weight the weighted totals for eligible and responding cases in each poststratification cell and the corresponding nonresponse adjustment factor.

      If an analysis requires both a primary and "other" caregiver for a child (i.e. only children who have both a primary and other caregiver are included), the nonresponse adjusted weight for the least often measured data type should be used. This would be the Other Caregiver Child Book (OTCGCHWT) or the Other Caregiver Household Book (OTCGHHWT) weight if data from the Household Booklet is used.

      If an analysis uses the union of the data sets, a Primary Caregiver for single caregiver families and both Primary and Other booklets when both primary and other caregivers are present, a combined weight should be used. For primary caregiver only families, the weight for the least often measured primary caregiver booklet type should be used. For children in families with both a primary and other caregiver, the weight for the least often measured other caregiver data type should be used.

Table 7: Nonresponse Adjustments for Supplemental Data Types

Race of Head	Educ. Of Head	MSA Status	Census Region	n	PCG Child Intvw. (wtd)	PCG HH Intvw. (wtd)	PCG HH Non-resp. Factor	Children with OCG (wtd)	OCG Child Intvw. (wtd)	OCG Child Non-resp. Factor	OCG HH Intvw. (wtd)	OCG HH Non-resp. Factor
Non-Black	< HS Grad.	MSA	Northeast	27	658.138	345.861	1.903	541.369	300.362	1.802	300.362	1.802
Non-Black	< HS Grad.	MSA	Midwest	53	713.118	421.504	1.692	603.633	354.035	1.705	304.712	1.981
Non-Black	< HS Grad.	MSA	South	90	1218.322	792.241	1.538	957.779	554.833	1.726	552.132	1.735
Non-Black	< HS Grad.	MSA	West	118	2602.995	1431.954	1.818	2357.754	1151.630	2.048	1071.084	2.201
Non-Black	< HS Grad.	Non-MSA	Northeast & Midwest	50	699.046	437.565	1.598	517.778	168.858	3.066	168.858	3.066
Non-Black	< HS Grad.	Non-MSA	South & West	71	1726.294	1051.209	1.642	1415.855	707.544	2.001	787.115	1.799
Non-Black	HS Grad.	MSA	Northeast	96	1718.821	1219.663	1.409	1439.407	888.557	1.620	860.546	1.673
Non-Black	HS Grad.	MSA	Midwest	100	2133.720	1404.816	1.519	1824.145	1001.955	1.820	915.147	1.993
Non-Black	HS Grad.	MSA	South	98	2185.183	1258.495	1.736	1826.004	1063.820	1.716	984.969	1.854
Non-Black	HS Grad.	MSA	West	117	1808.144	1285.586	1.406	1579.060	1020.653	1.547	924.934	1.707
Non-Black	HS Grad.	Non-MSA	Northeast & Midwest	136	2553.921	1548.351	1.649	2280.285	1258.028	1.812	1231.949	1.851
Non-Black	HS Grad.	Non-MSA	South	88	1900.237	1270.732	1.495	1782.392	1039.094	1.715	1027.828	1.734
Non-Black	HS Grad.	Non-MSA	West	17	590.906	418.360	1.412	391.777	255.737	1.532	255.737	1.532
Non-Black	> HS Grad.	MSA	Northeast	201	3255.257	2286.059	1.424	3070.207	1675.123	1.833	1738.888	1.766
Non-Black	> HS Grad.	MSA	Midwest	207	3490.594	2254.057	1.548	3151.348	1976.541	1.594	1852.407	1.701
Non-Black	> HS Grad.	MSA	South	209	4109.579	3056.855	1.344	3875.160	2377.241	1.630	2382.740	1.626
Non-Black	> HS Grad.	MSA	West	192	4738.528	3459.371	1.370	4315.284	2365.252	1.824	2322.972	1.858
Non-Black	> HS Grad.	Non-MSA	Northeast	32	1162.528	814.306	1.428	891.945	594.086	1.501	616.721	1.446
Non-Black	> HS Grad.	Non-MSA	Midwest	99	1944.867	1133.928	1.715	1739.554	995.306	1.748	984.126	1.768
Non-Black	> HS Grad.	Non-MSA	South	67	2088.320	1267.984	1.647	1585.032	715.052	2.217	79.709	2.033
Non-Black	> HS Grad.	Non-MSA	West	34	1219.224	1065.745	1.144	1103.815	922.581	1.196	922.581	1.196
Black	< HS Grad.	MSA	Northeast & Midwest	99	654.032	394.504	1.658	294.808	67.428	4.372	81.700	3.608
Black	< HS Grad.	MSA	South & West	205	724.320	438.965	1.650	242.926	49.426	4.915	54.738	4.438
Black	< HS Grad.	Non-MSA	Northeast Midwest South, West	88	509.976	366.326	1.392	366.384	192.260	1.906	195.774	1.871
Black	HS Grad.	MSA	Northeast	46	543.021	419.972	1.293	152.920	83.202	1.838	83.202	1.838
Black	HS Grad.	MSA	Midwest	91	507.152	347.535	1.459	219.260	101.431	2.162	96.012	2.284
Black	HS Grad.	MSA	South	324	1112.435	515.470	2.158	598.945	189.162	3.166	174.481	3.433
Black	HS Grad.	MSA	West	46	145.991	69.747	2.093