U.S. Department of Health and Human Services
Evaluation of the 1989 Child Care Supplement in the National Longitudinal Survey of Youth
Frank L. Mott and Paula Baker
Center for Human Resource ResearchOhio State University
PDF Version: http://aspe.hhs.gov/daltcp/reports/89ccsup.pdf ( PDF pages)
This report was prepared under intergovernmental transfer between the U.S. Department of Health and Human Services, Office of Social Services Policy (now the Office of Disability, Aging and Long-Term Care Policy) and the U.S. Department of Labor, Bureau of Labor Statistics. Additional funds were provided by the National Institute of Child Health and Human Development. For additional information about the study, you may visit the ASPE home page at http://aspe.hhs.gov or contact the Office of the Assistant Secretary for Planning and Evaluation, H.H. Humphrey Building, 200 Independence Avenue, SW, Washington, DC 20201. The e-mail address is: webmaster.ASPE@hhs.gov.
Interpretations and viewpoints contained in this paper do not necessarily represent the official positions or policies of the various agencies. The authors thank Joyce Abma for her outstanding research assistance.
TABLE OF CONTENTS
- SUMMARY AND RECOMMENDATIONS
- 1. Current Child Care Use
- 2. Retrospectively Reported Child Care Arrangements
- LIST OF TABLES
- TABLE 1: Distribution of Number of Survey Week Arrangements and Percent of Total Child Care Hours in Each Arrangement
- TABLE 2: Patterning of Current Primary Arrangement Starting Dates
- TABLE 3: Where Does Child Usually Go After School?
- TABLE 4: Number of Child Care Arrangements Being Used by Women: Current Arrangements Lasting at Least 10 Hours in the Survey Week and Other Arrangements used During Year
- TABLE 5A: Annual Pattern of Starting Dates for Current and Retrospective Childcare Arrangements Lasting 10 Hours or More: Frequencies and Percent Distribution
- TABLE 5B: Monthly Pattern of Starting Dates for Current and Retrospective Childcare Arrangements Lasting 10 Hours or More: Frequencies and Percent Distributions
- TABLE 6: Pattern of Ending Dates for Retrospective Childcare Arrangements
- TABLE 7: Duration of Current and Retrospective Child Care Arrangements
- TABLE 8: Type of Current and Retrospective Care Arrangements: Percent Distribution
- TABLE 9: Main Reason for Beginning Current and Retrospective Child Care Arrangements: Percent Distribution
This report summarizes the results of a special experiment in the collection (from 347 mothers) of detailed current and retrospective child care information carried out in conjunction with the 1989 (11th round) wave of the National Longitudinal Survey of Youth. It addresses a number of child care data quality and evaluation issues and provides some recommendations for subsequent data collection. Interested readers should have access to the 1989 Child Care Supplement in order to maximally be able to interpret the results of this report.
The 1989 (11th wave) round of the National Longitudinal Survey of Work Experience of Youth (NLSY) included a special child care supplement designed to obtain maternal reports of current and usual childcare, and retrospective care information about arrangements used since the last interview. The supplement was completed by 347 mothers in the sample who were all interviewed early in the survey round, primarily during the months of June and July. This special data collection was funded by the Department of Health and Human Services in order to evaluate several data quality issues.
Child care information in the NLSY (and in several other surveys) is typically collected for the primary and secondary arrangement used by the mother for her children. In this "pretest", information was gathered for every child care arrangement lasting at least an hour during the survey week. The intention here is to gauge the extent to which information about only the primary and secondary care arrangement accurately reflects the totality of the household child care arrangement for that time period.
The 1989 special child care supplement included a retrospective child care section, which asked about every child care arrangement used by the mother since the date of the last interview that lasted at least 10 hours in any one week. The intent here was to evaluate, in a preliminary way, the extent to which women are able to retrospectively reconstruct information about prior child care arrangements.
The special child care supplement also contained a variety of questions on other related child care topics of a behavioral and attitudinal nature. The principal reason for including most of these questions was essentially to test their face validity and general utility for helping explain child care behavior; do they include a full range of apparently rational responses and are respondents, for the most part, able to provide an answer? In addition, are the precoded response possibilities, which are essential from a cost perspective in any large scale survey, appropriate?
As indicated above, the child care supplement was completed by 347 women who were interviewed during the first month of the 1989 survey round. These mothers represented a full cross-section of the mothers in the survey. One major advantage of the approach was that over 100 interviewers were involved in this special data collection, with no interviewer collecting information from more than four respondents. Thus, the results are generalizable across a full range of interviewing situations. A potential disadvantage is that the limited interviews per interviewer may have caused some interviewers to be less polished in their administration of the supplement than they would have been had they each conducted a larger number of child care interviews. The reader is cautioned against using the specific statistics in this study to generalize to a full universe of mothers and children. These 347 cases constitute an unweighted sample of black, Hispanic and white mothers aged 24 to 32 who may not fully represent even this particular group--since they reflect interviews with only the first 347 of well over 3,000 mothers to be interviewed in the 1989 survey round. Notwithstanding this caveat, the results probably reflect reasonably well the kinds of results, in terms of quality, which one could anticipate if the full cohort of mothers had been interviewed.
Part I of the child care schedule details all of the survey week child care arrangements reported by the 347 mothers in this survey. Overall, about 85 percent of the mothers indicated that the week preceding the interview was "typical" in terms of the child care they usually receive. About three quarters of the mothers indicated that someone helped them care for their children in the week prior to the interview. This level of actual current child care use is higher than is typically reported in most other studies of mothers with young children; indeed, the 75 percent reported using care is substantially above the rates of current care reported in prior N.L.S.Y. survey rounds. The fact that the data were collected during the summer months while many children are home from school does not account for this high use rate as most women (85 percent) reported last week as being a typical week. Indeed, only a total of 12 mothers indicated that they had to alter their usual arrangement because their children were home from school. More likely, as will be detailed below, the high rate of child care use probably reflects our probing for all arrangements lasting an hour or more, regardless of the respondent's usual activity.
The questions on current child care arrangements were principally designed to assess whether information about a woman's primary and secondary arrangement adequately describes all of the arrangements currently used. For 86 percent of women, two or fewer arrangements represent all the arrangements they are currently using; no woman reported more than five survey week arrangements. Table1, which synthesizes several dimensions of this survey week profile, shows that 88 percent of all survey week arrangements were either a primary or secondary arrangement, and that they accounted for about 91 percent of all child care hours reported during the week. Thus, data collection limited to reports on two arrangements yields approximately 90 percent of the child care time that mothers typically utilize (or at least report about!). It is worth noting, however, that asking for information on three arrangements would increase the report of all arrangements and of all child care hours to about 97 percent. It is suggested that since the addition of third arrangements to data collection incorporates almost all the remaining arrangements, the option of including three rather than two arrangements in the NLSY child care data collection should be considered.
In general, the quality of the reporting with respect to current arrangements appears reasonable. Because this experiment occurred in the summer, and because the sample is essentially an unweighted group of mothers with an over-representation of minority and disadvantaged females, it is difficult to compare the patterning of arrangements in this study with other data. Nonetheless, it does appear that our intensive attempt to pick up all kinds of arrangements, however casual they may be, resulted in a greater tendency for mothers to report non-formal, sometimes non-paid arrangements. This probably reflects the emphasis in the introductory lead in statements encouraging the respondents to tell us about (1) any and all arrangements, (2) all arrangements lasting at least one hour, and (3) arrangements used for non-employment as well as employment activities. The distribution of types of arrangements includes larger proportions of care by relatives and smaller proportions of formal day care arrangements than typically found in other surveys--including earlier NLSY reports. In addition, relatively large proportions of mothers reported their spouse or partner as the caregiver. Also, the proportion of caregivers who are relatives tended to be even higher for secondary and tertiary arrangements and, correspondingly, very few of these arrangements were of a formal nature; and very few of these latter arrangements were in other than a private home.
Parallelling this apparent heavy representation of informal arrangements, only a minority of women reported that they paid for their arrangements. About 40 percent reported that they paid for their first arrangement and about 30 percent for second arrangements. This is well below what is, typically found in the NLSY or other surveys. In addition, the distribution of hours used for the primary arrangement is substantially different from what we have found in prior NLSY survey rounds.
In summary, our emphasis on encouraging women to report on any and all child care arrangements, formal or informal, lasting for as short a time as one hour a week, regardless of employment status, appears to have been relatively successful. In comparison with other available data, the mothers in our sample (1) were much more likely to have reported using child care, (2) were more likely to report non-formal arrangements, particularly relatives, (3) typically made child care arrangements for fewer hours per arrangement and (4) more often than not did not pay for the arrangement. Part of this pattern may reflect the summer season. More importantly, all of the above strongly suggest that question wordings which encourage respondents to report marginal arrangements appear to do so.
For arrangements that lasted ten hours a week or more, a variety of additional arrangement characteristics were collected (see Part I, questions 5I through 5N). This included information about the education and specialized child care training the caregivers may have received, the number of children cared for in the arrangement, the number of adult supervisors per group, and the date (month-year) the arrangement started.1 From the perspective of this particular study, the principal reason for including these questions was to gain insights into whether or not mothers, particularly those with multiple arrangements, are able to provide answers to the questions. (Whether or not the answers are indeed accurate would require independent corroboration, which to some extent is beyond the scope of our evaluation.) For the most part, the respondents were able to provide information about these child care characteristics. This was particularly true for the information about class size and number of child care personnel.2 Mothers were somewhat less able to provide information on individual caregiver characteristics. Of the 234 women who reported a primary child care arrangement, 23 (10 percent) indicated that they did not know how much education the caregiver had. About 5 percent were unable to specify whether or not the caregiver had special early childhood training. The majority of the non-responses were reported for more formal arrangements--where the child was in a day or group care arrangement or was being watched by a non-relative age 15 or over. The non-completion/don't know rates on these items were significantly lower for secondary and tertiary arrangements, undoubtedly reflecting the fact that these arrangements were more likely to involve family members--for whom this information would be known.
In contrast, virtually all the mothers provided information about class size and number of supervisors in the arrangement. About 85 percent of the children's primary arrangements was in groups of 7 or less and three quarters of the arrangements had only one caregiver. The relatively modest size of most arrangements makes the identification of the numbers of children and caregivers relatively easy and also probably quite accurate.
Not surprisingly, the two data items that required the greatest amount of recall and/or subjective evaluation appear to be of the poorest quality. While the start date for many, particularly major, arrangements may in most instances be easily recalled, there are situations where precision in recollection may be quite difficult. For example, if a particular arrangement is used intermittently over a lengthy time period, a respondent may be uncertain regarding which start date (or indeed, which termination date) to provide. A case in point would be a particular relative (e.g., grandmother) who is "available" for child care on a more or less continuing basis, but who is only used intermittently as needed. Unless a precise specification is used (e.g., "When was the very last time you used __________ for child care?") it is virtually impossible to obtain consistent (across respondents) usable information on arrangements of these types. In addition, our minimum cutoff on hours for obtaining dates for the beginning of current arrangements is ten hours. Given this rather modest hours threshold, plus the possibility that a respondent may be trying to reconstruct start dates for multiple arrangements, it is not surprising that recollection can in some instances be uncertain.
Table 2 synthesizes the patterning of reported start dates for the 174 current primary arrangements lasting 10 or more hours a week. A usable start date could not be obtained for only three of these arrangements. The average (median) arrangement as of June 1989 was about nine months, meaning that about half the arrangements started after September 1988. An examination of the annual patterning of the beginning dates for these current arrangements indicates, not surprisingly, that the bulk of the current arrangements are of recent vintage. The reported monthly pattern is analytically somewhat more interesting. If one excludes arrangements that began in 1989, it is readily apparent that the most common months for beginning child care arrangements coincide with late summer-early fall, when children typically begin regular school programs. Indeed, for the pre-1989 period, almost half of all ongoing reported arrangements began either in August or September (the reasonableness of this patterning will be clarified further when the patterning of retrospective arrangements is considered). However, when one examines the pattern for those arrangements which began in 1989, it is apparent that there is a major tendency to report on very recent arrangements--those that began during the survey months. This undoubtedly reflects the reality that at any given point in time some substantial proportion of mothers are changing their child care arrangements and that this tendency may be particularly pronounced at the beginning of the summer, when school has just ended. Clearly, seasonality and duration of current arrangements can be significantly affected by the time of year the survey data are collected. This will be highlighted further in the section on retrospective data collection.
As a final note on the characteristics of current child care givers, it is useful to note that the only data item with a substantial number of responses falling in the "other" category is the question inquiring into the reason the respondent started using a particular child care arrangement. of the 233 women who responded to this item, 77 or 33 percent provided an answer which did not fall into any of the available code categories. In addition, most of the coded responses either have only limited analytical utility or else could have been more appropriately diagnosed through the collection of parallel behavioral information; for example, 44 respondents gave as a reason "lost or stopped their previous caregiver," a response of limited analytical value. Eighty three gave as a reason the fact that they started work or training. Finally, it is important to note that the reason they started a current arrangement may have little to do with why they are currently using that arrangement. It is not clear how the responses to this question can be easily used in any analyses, unless they are more clearly specified for possible connection with other mother activities.
Part 2 of the child care supplement examined the extent to which within-household information about child care hours and payments were separable in households where more than one child utilized a particular caregiver. Our particular interest was in measuring the extent to which hours or costs are indeed separable between children and, additionally, the extent to which mothers indicate that they are unable to provide separate information, not because costs/hours were not separable, but because they were unable to quantify differences. Information was reported for 164 households which included two or more children. The data was collected for the two youngest children in the household.
On the positive side, it appears that respondents are frequently able to separate out child care costs which can be attributed to a particular child even if a particular provider is caring for more than one child in the household, regardless of whether both children are being watched the same or a different numbers of hours. On the downside, it is apparent that the interviewers had considerable difficulty in following the appropriate skip instructions for this section.
In only a very small number of cases were children not reported as being watched all hours by a particular caregiver if that caregiver was watching them at all; and this was true even with respect to secondary or tertiary arrangements. For example, 149 youngest children were reported as being watched by the primary caregiver and in 146 cases, the caregiver watched the child all the hours she worked. For the second youngest child, the same pattern held for 129 and 121 cases respectively.
The interviewers had some difficulty in following the skip pattern instructions regarding the reporting on hours and costs; in a large number of cases they reported information even where it wasn't required. This is a minor problem, however. A bigger problem relates to the fact that many interviewers/mothers did not provide information on whether costs were separate or not (Column D in Child Care Chart BB). This probably reflects a flaw in the child care reporting form as many interviewers/mothers were continuing to follow the hours instructions in Column C--and were only reporting cost information if they had already reported that the hours information was separable; however, it is frequently possible to provide separate cost information even if both children were watched the same number of hours by the same person. However, for a substantial number of children, costs could be separated by child and, in these instances, most of the time separate costs were provided. Across all youngest child arrangements, there are 66 (of 231) instances where a mother indicated that separate costs could be identified, and in 59 of these cases, actual costs were provided.
Thus, in summary, the results suggest that mothers can frequently provide separate hours and/or cost information for each child. However, some of the skip patterns for questions of this type can be quite complex, considerable interviewer training is needed and clear questionnaire instructions provided. In this regard, it is useful to recall that while only experienced NLSY interviewers were selected for this pretest, they had only limited experience with this child care supplement. Their training included a brief self-study manual. Their actual field experience was limited to completing between one and four child care supplements. Thus, there was only limited opportunity to become thoroughly familiar with the questions and the skip patterns.
This relatively heterogeneous (in terms of questions asked) section probes into a variety of issues relating to hypothetical child care interests and availability as well as the flexibility of current and recent arrangements for meeting unusual or emergency situations.
About four out of five mothers indicated that they would remain with their current child care arrangement even if all arrangements were available free of charge. Either this response pattern implies an extraordinarily high level of satisfaction or else individuals may not be sensitive to the full range of possibilities which are available. The question as worded has several limitations. First, a working woman's first preference might be to remain at home with her children (or have her spouse stay home) but it is not clear that this is an intended option given the wording of the question (even though 14 women did indicate this as a preference to their current arrangement). A woman's staying home implies a loss of wages, and it is not clear that "free" child care arrangements implies a wage subsidy to cover lost wages. Second, costs are certainly not the only child care constraint, as a woman might prefer an option which is not viable because of locational constraints (or other reasons). For these reasons, it is not clear how interpretable the responses to questions 7 and 7a really are. However, question 8 (main reason for preferring other arrangement) does suggest that the principal reasons for preferring an alternate arrangement relate to the potential quality of the other arrangement, rather than other arrangement characteristics.
The responses to the questions asked of those not working are consistent with the notion that the dominant reason for non-employment for women at this life cycle stage is voluntary; 71 of 135 indicated that they prefer not to work while their children are young, and only a minority are interested in working at this time. (It is also useful to note that the large majority of respondents was able to provide a codable response to this question.)
In addition, the responses of most of the non-working women regarding what they view as reasonable costs are highly consistent with the actual costs being incurred by all the women who are currently paying for child care. Overall, on an hourly basis, the current child care users indicate a median wage of $1.50 an hour, with 40 percent paying $1.25 an hour or less and 40 percent $2.00 or more. In comparison, the 131 non-working women who responded to the question on "reasonable costs for a full time week" of child care (question 12), gave a median response of $50, with 40 percent indicating $40 or less and 30 percent $55 or more. The modal response was $50. Thus, while this cost distribution was slightly below the actual costs reported as being incurred by those using care, the hypothetical responses were not substantially different from the reported reality.
Finally, with regard to the non-workers, it is useful to note that the large majority specify a relative as the caregiver when they go out for reasons such as entertainment or shopping. There is little tendency to seek formal arrangements. However, there is no hesitancy regarding being able to specify the caregiver which is used for non-employment reasons.
Issues linked with child illness and employment are, not surprisingly, of some importance in evaluating child care/employment considerations for women at this life cycle stage.3 Fully one quarter (52 women) of the employed women had a child who was ill within the past month and 18 of these 52 women lost time from work because of this illness. Thus about two thirds of these employed women were able to make arrangements which permitted them to go to work while their child was ill. Of those who lost time from work, most missed two or fewer days and about one third had to take leave without pay. The important issue here is not the precise quantification of the patterns reported by the experimental sample but rather that (1) the phenomenon, which is of some importance, is analytically non-trivial and (2) women appear to be able to effectively report on this issue.
In contrast, only very small numbers report other difficulties with their child care arrangements over the last month. Only 5 of 213 reported difficulty with their regular arrangement and only 10 reported that they were late to work or had to leave early. Indeed, only 18 of 216 (about 8 percent) indicated that over the last 12 months, "problems with the cost, availability or quality of child care influenced their employment in any way." This very modest percentage reporting problems would appear to run counter to some popular perceptions. On the other hand, it is consistent with the earlier stated result that few of the women would change arrangements even if costs were not a constraint. Thus, either these results reflect reality or else there is something in the wording of the questions which affects the woman's frame of reference in such a way as to cause women to not consider a full range of options. Interviewer notes indicate that while some mothers did not actually miss work, they did occasionally take children to work when their caregiver was unavailable.
The final child care dimension considered in Part 3 of the interview schedule examined the issue of potential availability of relatives for child care assistance for those women who were currently employed as of the survey date. Somewhat surprisingly, the vast majority--190 of 213 working women indicated that they had a relative living in the vicinity. The respondents were in all instances able to identify the work status of these relatives. Only a modest percentage of these relatives, about 25 percent, are currently helping with child care on a regularly scheduled basis. The high percentage of mothers who have a relative (blood or through marriage) nearby may partly reflect the fact that the sample is not fully representative of all mothers; that it disproportionally includes women who married at younger ages and have limited education. It is probable that these younger marrieds are less likely than other women to live far away from their familial neighborhood/environment.4
As a general comment regarding all of the attitudinal and behavioral items in this section of the questionnaire, there is no evidence of unwillingness or inability to answer questions. The cost-related information was provided on request, although, of course, its accuracy (regarding actual amounts) can be questioned. Generally speaking, the items on reasons for not working or child care preference were answered fairly readily and, as importantly, the pre-determined code categories were generally appropriate; the number of responses which needed to be coded into the "other" category was generally modest.
Because this pretest took place during the months of June and July, when most school age children are out of school for the summer, the pretest included a few essentially retrospective questions probing into the usual arrangements made for school age children after school. The bulk of the school age children in this study are between the ages of five and nine, and thus either in kindergarten or in the earlier elementary school grades. Of the 222 youngest school children, 196 were between the ages of five and nine; 102 of their next oldest siblings ("second youngest school age child") were also considered in this evaluation.
As may be noted from Table 3, the vast majority of the school age children follow a traditional post-school child care mode--they apparently go home and are supervised in the home by someone who is age 14 or older. Of those who don't go home, the largest proportion are in some other fairly traditional child care arrangement--cared for by other relatives, a baby sitter, or in a child care center. Very few mothers reported latch key arrangements. There are also very few reports of multiple arrangements. Apparently, the vast majority of the children are in the same arrangement every day. It is possible that if the children in this study were somewhat older, more latch key and other non-traditional arrangements would have been reported.5 It is also possible that some parents might have been reluctant to report about latch key or other "socially less acceptable" arrangements. What the data suggest is that analysis of the incidence and consequences of latch key arrangements will in all likelihood not be feasible unless one has a very large sample of children, given the rarity of reporting of this arrangement.
The results from the retrospective data collection in the child care supplement clarified a variety of issues regarding the efficacy of collecting child care information in a quasi-event history format. Some useful data were collected, and some significant problems are apparent. The problems reflect limitations which are inherent to retrospective data collection, imprecision in some of the child care concepts and oversights (inconsistencies) in how we defined current and retrospective arrangements.
The major conceptual problem relates to how one defines an arrangement--either current or retrospective. It is apparent from both the coded responses as well as marginal notes in the interview schedules, that the concept of a continuous child care arrangement is extremely ambiguous. At one extreme, a mother may use a particular arrangement continuously, for a large number of hours each week with no intermittent interruptions. At the other extreme, one has situations where a "caregiver" is essentially continuously "on call," but only used intermittently. A grandmother living nearby (or even the other parent) may fall into this category--as might a more formal caregiving arrangement. This poses a major problem when one attempts to obtain "beginning" and "ending" dates for a particular arrangement. With regard to a current arrangement, one must consciously make a judgement as to what starting date one wishes. Does one want the date that caregiver was first utilized by the respondent? Alternatively, one could focus on the last date that the caregiver began assisting the respondent. Or, perhaps most reasonably, one can focus on a start date that allows the caregiver to have had modest interruptions (2 weeks?). This last option may be most appropriate in many instances, but perhaps not always easy to quantify, because of recall problems.
From an event history perspective, this issue becomes inherently even more complex if one chooses to consider several child care episodes with the same child care personnel as distinct arrangements. For example, a respondent may have used the same arrangement several times over the past year, with distinct time breaks between use. The gaps between use may or may not have been filled by other arrangements.
Somewhat parallel to this problem of specifying beginning and ending dates for arrangements, is the problem of defining hours for arrangements. It is likely that respondents are better able to recall arrangements which not only lasted longer, but which were used for a greater number of hours within a particular time span. In the extreme case, an arrangement that lasted 40 hours in a given week will be recalled more readily than a three hour babysitting arrangement. (In addition, arguably, a paid arrangement will perhaps be more readily recalled than a non-paying arrangement, unless specific probes for non-paying arrangements are utilized). Thus, in a retrospective context, unless one is probing for every hour of care utilized in a time frame, one needs to specify an hour minimum for a given arrangement. This is inherently complex, particularly if a respondent used the same arrangement for more than one distinct period during the year, and used that arrangement for a different number of hours each time.
In our case, we limited retrospective data collection to arrangements which lasted at least ten hours in any one week during the year. In addition, we failed to clarify whether or not using the same caregiver in two distinctly different arrangements constituted one or multiple arrangements. It is apparent from our results that many respondents were uncertain as to whether or not to count the same arrangement more than once, even if that arrangement was used intermittently over the course of the year with substantial time breaks between usage. In the extreme case, a respondent who may have use a particular arrangement off and on for a particular child over a period of many years may have given the child's birth as the start date, and also indicate that she is still (on occasion) using that arrangement. Because of these serious constraints, the data we have collected do not lend themselves to developing an annual week by week or month by month child care use profile; apparent "stop" and "start" dates encompass substantial time periods during which the arrangement may not have been used.
One other problem, which reflected an oversight on our part in the construction of the interview schedule, intensified the difficulty of defining the duration of arrangements. Data were collected for all current arrangements lasting at least one hour in the past week. However, the retrospective data collection was limited to arrangements lasting at least ten hours in any week. This made it difficult to attempt to link the current and retrospective arrangements. For example, a mother might have reported on a survey week arrangement which lasted four hours. She then reported once again on that same arrangement in the retrospective child care history, because there was some other time during the year when she used that arrangement at least ten hours. We had not allowed for this option in the interview schedule, so it created confusion for some interviewers and respondents. The introductory paragraph in the retrospective section asked the respondent "not counting your current arrangement(s), have you used any other child care for at least 10 hours in any one week since the last interviews" Because of this oversight, quite a bit of hand editing was necessary in order to eliminate arrangement duplication. Future data collection could minimize problems of this type by attempting to use parallel definitions of child care for current and retrospective arrangements and, most importantly, assigning I.D.s to each arrangement. The latter procedure, if properly done, would permit researchers to program the child care information in whatever way is deemed most appropriate. Using IDs would allow one to remove duplication, if one desires, to compare arrangements over the full retrospective time period, defining the same arrangements which are temporally dispersed as the same or different arrangements), depending on one's inclination.
One final issue regarding the time frame of the retrospective section warrants mention. The retrospective section of the questionnaire collected information on arrangements used since the date of last interview. From a post-hoc perspective, this was ill advised, for methodological and perhaps substantive reasons. A preferable option would have been to collect the retrospective child care information for a full twelve or thirteen month period for all respondents, beginning the event history on June or July 1, 1988. Using a last interview starting point resulted in a gradual diminution of the level of child care use as one moved backwards into the summer months of 1988. We could certainly partially resolve this issue by translating our levels of start and stop dates into proportions. However, this would not resolve the problem of selective bias due to certain kinds of women (e.g., employed? home at certain hours) perhaps being interviewed earlier or later in the summer. This, of course, is only an issue in a continuing longitudinal survey such as this, where there are major reasons for linking with a prior survey date (e.g., to link child care histories with employment history).
Notwithstanding the above issues, the retrospective child care data collection provided some useful analytical results--beyond the insights already suggested. Substantive differences (and similarities) between current and retrospective child care arrangements are highlighted in this section, as they help clarify biases inherent in the essentially different methods used to collect current and retrospective data.
Table 4 summarizes the information on current and retrospective child care arrangements. It includes distributions of the number of arrangements currently used by mothers and the number of additional arrangements used during the preceding year. Most of the comparisons of current and retrospective patterns will essentially contrast distributions limited to current arrangements lasting ten hours or more during the survey week with distributions for retrospective arrangements which were used for ten hours in at least one week during the year. Thus, the comparisons are only approximately comparable. When arrangements of less than ten hours are excluded from the current arrangement distribution, only about 56 percent of women report having any current childcare assistance compared with about 75 percent when all arrangements are included. Thus, eliminating arrangements of fewer hours makes the percentage using care much more comparable with current usage statistics reported in other data sets, including earlier NLSY rounds. The largest majority (about 73 percent) of mothers using arrangements lasting ten hours or more have only one such arrangement and about 94 percent of these mothers using care report only one or two such arrangements.
Excluding all current arrangements, only about 31 percent of all the mothers reported on an additional arrangement over the preceding year which lasted at least ten hours in any one week. Thus, if these data are at all reliable, it may be concluded that the bulk of the more extensive arrangements which a woman has during a given year can at least be identified as ongoing as of one point in time. In addition, the intensity of involvement with the retrospective arrangements is somewhat less certain than is the involvement with the current arrangements, reflecting the ambiguity regarding the hours the retrospective arrangements are/were being used. In this regard, it is important to note that of the 109 women who report at least one retrospective arrangement, 17 (16 percent) indicate that that arrangement is still ongoing as of the survey week, but that they are not using that arrangement for any hours at this time. This highlights the issue alluded to above--that there are a significant number of child care users who define the duration of arrangements (including beginning and ending dates) in terms of ready availability of the caregiver, not just in terms of current usage. These caregivers, who frequently, although not always, are family members, pose a significant definitional problem in defining meaningful child care event histories.
Table 5A and Table 5B describe the temporal patterning of current and retrospective child care arrangements, both in terms of seasonality and longevity. A comparison of the starting dates for the ongoing as well as retrospective child care arrangements suggests some important similarities as well as differences. With regard to the monthly patterning of start dates, child care arrangement "starts" in 1989 are separated from the results for preceding years because the interviews were all carried out in June and July. Most of the current (ongoing) arrangements which started in 1989 began in May or June, many within just a few weeks of the survey date. This trend partly reflects the fact that some parents are beginning summer arrangements for school and preschool age children to cover the out-of-school summer months. Part of the high reporting level for May and June also probably reflects an above average tendency to recall recent events. When one examines the pre-1989 distribution of starting dates for current arrangements (focusing on the January through June distribution for all preceding years), it is apparent that May and June are above average months for beginning child care arrangements. This same pattern also exists when one examines retrospective arrangements--those which for the most part have been terminated prior to the 1989 survey week.
The start dates for pre-1989 current arrangements as well as for retrospective arrangements both show a pronounced pattern of arrangements beginning in August and September, coinciding with the beginning of the school year. Thus, the patterning of the starting dates for both current and (recent) past arrangements both appear to have a high face validity in that the most prevalent beginning dates coincide with the months when parents are most likely to be making new arrangements.
The patterning of "ending dates" for retrospective arrangements, described in Table 6, is harder to interpret. Part of the reason for the very low percentage of arrangements ending in the early calendar year months is that these months occurred prior to the preceding interview date for some Respondents thus their arrangement ending date would not have fallen within the time reference period being investigated. The high proportion of arrangement ending dates in the Spring of 1989 undoubtedly is closely linked with the ending of the school year, a transition point for many child care arrangements.
Focusing briefly on the annual patterns suggested by Table 5A, Table 5B, and Table 6, it appears that the arrangements which respondents report as ongoing for ten hours or more as of the survey date are indeed more likely to typify the more permanent arrangements a respondent uses than do the other retrospective arrangements reported. It is of some interest to note that almost 40 percent of all current arrangements (used for at least ten hours in the survey week) are reported as having started in 1987 or earlier, compared with 15 percent for the retrospective arrangements. This probably reflects the fact that arrangements which are considered more satisfactory are probably more likely to be sustained for a longer time period--up to the present time. In addition, one's recollection about the history of recent or ongoing activities is probably better than recall about historic events. Third, many parents probably sustain a major arrangement on as continuing a basis as possible and then supplement that arrangement as needed. Analytically, this patterning has at least two implications. One needs to be cautious about assuming that lengthy ongoing arrangements are being used continuously, that from an event history perspective, one can assume a continuity of use between stop and start dates. This will be clarified with some of the subsequent tabular materials.
As importantly, one must be cautious when interpreting quality dimensions of child care when examining a cross-section. If one can (reasonably) assume that the current arrangements, which on average can be documented as much longer lasting than terminated (retrospective) arrangements, are more satisfactory to the respondent, making judgments about satisfaction or quality from the cross-section must be done cautiously. The less satisfactory arrangements are more likely to have been discontinued and thus less likely to appear at any one point in time!
Similarly, if one is focusing on the length of child care arrangements and how these arrangements are linked with other family characteristics, one needs to be doubly cautious, as both issues of seasonality and continuity are important. These phenomena are highlighted in Table 7, which describes the duration, in months, of current (ongoing) as well as retrospective child care arrangements. It may be noted that a substantial proportion of current arrangements are very recent, coinciding with the fact that the interviews occurred in late spring-early summer when respondents are beginning summer arrangements. It should also be noted that a secondary "heaping" occurs at nine and ten months, coinciding with the beginning of the preceding school year. Thus, at any point in time the "shape" (and probably the nature) of the duration distribution will be extremely sensitive to the time of year the interviews are completed. For example, interviewing carried out in September or October will result in a duration distribution which includes a particularly large proportion of new arrangements (and probably a secondary heaping at 11 and 12 months). It may also be noted that the duration distribution of retrospective arrangements has a somewhat similar shape, suggesting that even most arrangements which have been terminated are driven by the same seasonal factors. However, in this case, the disproportionate heaping of durations up to three months is consistent with the fact that respondents are more likely to recollect retrospective shorter term arrangements from recent time periods. In addition, it is important to highlight one other fact about the retrospective durations which is not immediately evident from Table 7; almost 50 percent of the retrospective arrangements of more than 12 months duration fall in the category of "ongoing" arrangements (see note to Table 7), which are not being utilized for any hours in the survey week but which the respondent undoubtedly considers as available if needed. Thus, not only are retrospective arrangements less likely to be long term (as described by the distributions in Table 7), but those which are long term are probably much more likely to represent non-continuous arrangements than are comparable long term current arrangements.
Table 8 synthesizes information about the types of arrangements which predominate in ongoing and retrospective child care situations. The results summarized in this table demonstrate (1) the difficulties inherent in gathering comprehensive data for retrospective periods and (2) potential analytical biases which may exist if one treats cross-sectional data as typical of all child care arrangements.
The first point to note from Table 8 is that the nature of the current arrangements is, not surprisingly, sensitive to the number of hours that the child care arrangement is used; arrangements lasting more hours are less likely to be care by the "other parent" or a relative, and (not reported in table) more likely to be paid arrangements. Thus, the more one probes for casual, shorter term arrangements (as we have done in this pretest), the more accurately the mix of arrangements portrays the reality of current child care. Because of our probes for arrangements (1) lasting at least one hour during the week, (2) not necessarily linked with employment, and (3) of a less formal nature, it appears that this experiment picked up a fuller representation of ongoing child care arrangements than most other surveys have accomplished, as well as a larger proportion of all women using child care. (As a cautionary note, part of this differing level and mix also undoubtedly represents a seasonality factor, that the respondents were mostly interviewed in June; however, one of our comparisons is with 1986 NLSY child care data, which also were disproportionately collected in the summer.)
In contrast, the retrospective data on child care which was limited to arrangements lasting at least ten hours in any one week provided very different results, more consistent with what has traditionally been available. Respondents who reported retrospective arrangements were much less likely to list parents and grandparents, and much more likely to mention non-relatives and, to some extent, daycare centers. In short, truncating the data collection on the basis of hours shifts the child care distribution rather dramatically away from informal and towards formal (paid) arrangements. From the perspective of data collection, it is clear that the representation of child care arrangements will be very different for current and retrospective periods unless one maximally strives to use comparable definitions of child care for both, a difficult task at best. Analytically, the implications are of equal importance. Any examination of either determinants or consequences of child care in a longitudinal context will be difficult to interpret unless (1) standardized definitions are used and, more importantly, (2) these definitions are as comprehensive as possible. For example, evaluating the child development consequences of child care will probably be very sensitive to the extensiveness of the arrangements included in the child care definition. Thus, child care use defined in a more limited manner--including comprehensively only extensive arrangements of a more formal nature and shortchanging shorter term (but perhaps diverse) arrangements which are less formal--may have very differing implications for measuring impacts on childrens' development. Indeed, one might argue that, from a child development analytical perspective, it may be counter-productive to collect only selective, longer term child care input information. Analysis based on this limited data could well produce erroneous results leading to inappropriate policy conclusions.
A final analytical comment relates to the quality and utility of retrospective attitudinal child care information which can be collected in a large scale survey of this type. Table 9 synthesizes the reasons which respondents gave for beginning their current and retrospective child care arrangements. Nor surprisingly, the dominant reason given for starting an arrangement was employment related--a job was started, or a previous caregiver had ceased to provide assistance. In addition, relatively substantial proportions of women gave idiosyncratic reasons ("other") which could not readily be coded in any coherent manner. It is suggested that in a longitudinal survey which attempts to collect event history information on a variety of employment, family, and education characteristics, retrospective, even relatively "hard" attitudinal information of this kind is relatively worthless. More insightful information on the linkage between employment and child care can be gained by comparing behavioral event-history information on employment or schooling and child care, comparing beginning and ending points of different events, if this information is carefully collected. The fact that an arrangement was started because a prior one had ended is also not particularly useful analytically unless one also has comprehensive information about the reason for the change in arrangements, something which may not be easy to determine retrospectively. Analytically, in a longitudinal context, a more fruitful approach is to gather comprehensive objective and subjective information about current arrangements, and then wait and see what the implications are for subsequent change or continuity.
As a final note on the apparent validity of the retrospective child care information, it is useful to note that there is virtually no evidence of respondents being unable to recall the characteristics of retrospective arrangements. While we have no external cross-checks on the quality of the reports, it should be noted that virtually all of the mothers provided responses to the questions on the number of hours per week retrospective arrangements lasted, the number of children cared for in the arrangement, the child to adult ratio, educational attainment of the caregivers, costs of the arrangement, as well as information about early childhood training received by the caregiver. It is particularly interesting to note that a "yes" or "no" answer (26 yes and 82 no) regarding early childhood training was given by all except one parent for the primary retrospective arrangement reported. whether the "nos" in some instances really represent "don't knows" cannot be answered. In retrospect, the information on the number of children in an arrangement number of caregivers, and characteristics of caregivers were probably relatively easy to recall because the vast majority of children were in arrangements which included three or fewer children and either one or two adults.
1. Current Child Care Use
Overall, about three quarters of the mothers had a child care arrangement lasting at least one hour during the survey week, a proportion considerably higher than has typically been reported in other studies. Eighty six percent of the care users reported that they used one or two arrangements during the week. Ninety seven percent of care users report no more than three arrangements in the survey week.
The survey week arrangements reported on appear to include a larger proportion of informal (e.g., care by spouse and other relatives) and a smaller proportion of formal arrangements than has typically been found in other studies. Parallelling this result, it was found that a substantial majority of women do not make any cash payments for their current arrangements.
It is likely that the apparently more comprehensive nature of current child care reported in this experiment reflects at least in part our emphasis on encouraging women to report on any and all child care arrangements, formal or informal, lasting for as short a time as one hour a week, regardless of the employment orientation of the mother. In addition, we put no constraint on the number of arrangements which a mother could report on. In this regard, we have strong empirical evidence that allowing for three arrangements results in almost complete coverage.
Based on the above, we recommend (for the NLSY or other comparable surveys) that a comprehensive introductory paragraph (similar to that used in this study) establishing the broad scope of the intended data collection be utilized and that data be collected on primary, secondary, and tertiary arrangements. Our evidence suggests that this should provide reasonably comprehensive coverage regarding all ongoing child care activities.
The relatively low proportion reporting that their arrangements are paid for suggests the likelihood that other arrangements of a quid pro quo nature may be operative for some significant proportion of all arrangements. This could include the provision of exchange services or other benefits (such as food and/or housing) to caregivers. Information on arrangements of this kind could greatly enhance analyses focusing on either social-psychological or economic exchanges for child care services. For this reason, it is recommended that subsequent data collection consider the inclusion of at least a preliminary attempt to measure such less easily quantifiable costs of child care arrangements.
Several sections of the child care interview schedule asked a variety of questions about the characteristics of current (and retrospective) arrangements, including questions on costs, hours, caregiver education, the number of children included in the arrangement, and the number of caregivers in the arrangement. In addition, where a caregiver was used for more than one child, questions were asked regarding the hours each (of the two youngest children) was watched and the costs associated with watching each child. With only minor exceptions (e.g., the education/specialized training of caregivers in formal arrangements), the respondents were able to provide what appears to be reasonable responses. In addition, the results support the notion that mothers frequently are able to provide separate cost estimates where more than one child is being watched in a particular arrangement.
Similarly, the respondents were able to provide important and apparently reasonable information about the extent to which they needed to absent themselves from work to meet child health-related exigencies in the home or to cover for a "snafu" in their child care arrangement. The results suggest sufficient response variability to warrant collecting this information on a larger scale.
The results of the attitudinal data collection, particularly regarding hypothetical child care needs, are somewhat more ambiguous. The respondents expressed a high level of satisfaction regarding their current child care arrangements and only a small proportion expressed a willingness to change arrangements even if their ideal were available "cost-free." It may be that this high level of satisfaction reflects the possibility that the questions were not sufficiently broad. For example, a woman might, in an ideal situation, prefer to remain home with her children, but the foregone cost-giving up real wages--may have been viewed as beyond the scope of the question. It is suggested that the question, as worded, perhaps has limited analytical potential, and that future data collection attempts in this area consider several broader concepts; in particular, specifically, what would it take to get a mother to stay home with her child, if this is truly the option she would prefer?
2. Retrospectively Reported Child Care Arrangements
In general, the seasonal patterning of current and retrospectively reported arrangements appear reasonable, and are comparable to each other; the beginning and end of the summer show above average tendencies to begin new child care arrangements. Thus, analysis of child care duration, and its linkage with other explanatory factors, needs to be sensitive to the month the data are collected. Data collection during the summer or early fall results in child care duration characteristics which will include a high proportion of short term arrangements.
It appears that reports on retrospective arrangements tend to under-report the more casual, shorter term arrangements and overemphasize more formal arrangements. This tendency, of course, at least partly reflects our definitional constraints on the question on retrospective child care--that the arrangement must have been used for at least ten hours in a week at some time during the year. The retrospective reports also appear to include a substantial number of arrangements of an intermittent nature, used now and then but not on a continuing basis. Indeed, the longer term retrospectively reported arrangements are more likely to fall in this category. In general, however, retrospective arrangements are less likely to be long term, suggesting a "selection bias;" preferred arrangements are more likely to be continued into the present than less preferred arrangements. This suggests the potential for significant analytical bias if one uses current (ongoing) child care arrangements as proxies for all arrangements when examining child care quality issues.
From an event history perspective, it is unclear whether the retrospective event history data collected in this experiment is of sufficient quality to have significant analytical value. Within the context of a survey such as the NLSY, to be of significant analytical value, the child care event dating must be of sufficient precision so that it can be matched with parallel event history information on, for example, employment, education, and family experiences. For the reasons described in this report, it is not likely that the retrospective child care data parallel in quality the other event history data.
It is likely that retrospective information of sufficient quality could be collected, but the time and monetary costs of doing so should not be underestimated. The data collection would require the following; beginning at a fixed historical point (preferably matched with other events, using appropriate calendars), one would need to move forward week by week (or at least month by month) collecting the following information: (1) child care use/non use for each week/month interval including the precise identification of the arrangement, the number of hours per unit time the arrangement is being utilized, as well as any complementary information desired. Whatever definitions are used for the collection of retrospective child care information should parallel the definitions used in collecting current child care data. In addition, each unique arrangement should have an identification number. If an arrangement is repeated at a different point in time, it could then be readily identified and linked. At the point where any arrangement is terminated for any length of time, an ending date would be obtained. At the point where any arrangement is initiated or re-initiated, a beginning date would be obtained. Ideally, the collection of these event history data would parallel equally precise data collection on other activities which tend to be linked with child care use--such as employment, education, or family events. The instructions relating to this child care data collection should ask respondents to describe (1) child care used for as little as one hour a week, and (2) arrangements of a formal or informal nature, including any and all care by relatives. The over-riding emphasis must be on actual hours that a child care arrangement was used in a given period, not simply access to a possible arrangement. It should be apparent that careful data collection of this type can be time consuming and expensive. However, any lesser efforts may well result in the collection of event history data which, on the surface, appears usable, but in reality is severely limiting from a longitudinal analytical perspective.
From the perspective of examining the consequences of child care for subsequent child development, the definitions of child care used may be of critical importance. Child care use defined in a limited manner, focusing on longer term, more formal arrangements may affect in major ways analytical results in this area of research; the results may be of value, but generalizations regarding the costs or benefits of child care will perhaps be less feasible and will require more extensive qualifications.
It is useful to note that many interviewers ignored the skip instruction regarding 10 hours and asked the questions for all arrangements, regardless of the actual number of hours the arrangement was utilized. For example, 232 interviews reported on the number of years of education of the child care giver, even though only 170 primary arrangements were reported as lasting ten hours or more.
It is worth noting that a number of parents mentioned that there was variable group size over the course of a day. Our questions did not allow for this option.
Only a limited number of questions focused directly on unemployed mothers because there were clearly only going to be a limited number of unemployed mothers in the study.
Several respondents indicated in marginal notes that they would have reported male relatives if the opportunity to do so had been there.
It is perhaps useful to note that many parents parenthetically mentioned that their children spent some time unsupervised but responded "no" to the latch key questions. The wording of our questions did not allow for this option.
|TABLE 1. Distribution of Number of Survey Week Arrangements and Percent of Total Child Care Hours in Each Arrangement|
|Distribution of Arrangement||Percent of Hours in Arrangement|
|Arrangement No. 1||61.2||70.5|
|Arrangement No. 2||27.0||20.7|
|Arrangement No. 3||8.9||5.9|
|Arrangement No. 4||2.4||2.0|
|Arrangement No. 5||0.5||0.8|
|TABLE 2. Patterning of Current Primary Arrangement Starting Dates(Arrangement Currently Lasting 10 or More Hours Weekly)|
|Annual Pattern||Monthly Pattern|
|1989 (6 months)||60||December||4|
|TABLE 3. Where Does Child Usually Go After School?|
|After School Care/Latch Key Program||8||2|
|Babysitter Outside Home||19||4|
|Day Care Home||2||2|
|Child Care Center||6||4|
|TABLE 4. Number of Child Care Arrangements Being Used by Women: Current Arrangements Lasting at Least 10 Hours in the Survey Week and Other (Retrospective) Arrangements used During Year|
|Number of Arrangements||Current Arrangements||Retrospective Arrangements|
|Frequency||Percent Dist.||Frequency||Percent Dist.|
|NOTE: Current arrangements in this table include all arrangements used at least 10 hours in the survey week; retrospective arrangements include all arrangements used at least 10 hours in any week during the year. This includes 25 retrospective arrangements which are "ongoing" as of the survey week even though the arrangement was not used during that week.|
|TABLE 5A. Annual Pattern of Starting Dates for Current and Retrospective Childcare Arrangements Lasting 10 Hours or More: Frequencies and Percent Distribution|
|Current Arrangements||Retrospective Arrangements||Current Arrangements||Retrospective Arrangements|
|1989 (6 months)||78||43||32.4||25.9|
|TABLE 5B. Monthly Pattern of Starting Dates for Current and Retrospective Childcare Arrangements Lasting 10 Hours or More: Frequencies and Percent Distributions|
|1989||Pre 1989||1989||Pre 1989||1989||Pre 1989||1989||Pre 1989|
|( ) = Distribution limited to January through June.|
|TABLE 6. Pattern of Ending Dates for Retrospective Childcare Arrangements|
|1989 (6 months)||106|
|Monthly Pattern||Frequency||Percent Distribution|
|TABLE 7. Duration (In Months) of Current (10 or More Hours) and Retrospective Child Care Arrangements|
|Interval in Months||Current Arrangements||Retrospective Arrangements|
|Frequency||Percent Dist.||Frequency||Percent Dist.|
|Less than one||42||17.2||31||19.0|
|36 and over||49||20.1||12||7.4|
|NOTE: The retrospective arrangements include 25 arrangements which are "ongoing" but which were not used for any hours in the survey week. Current arrangements are limited to those being used at least 10 hours in the survey week.|
|TABLE 8. Type of Current and Retrospective Care Arrangements: Percent Distribution|
|Current Arrangements||Retrospective Arrangements|
|Less Than 10 Hours||10 Hours or More||Total|
|Day Care Center||3.6||13.8||10.5||13.7|
|TABLE 9. Main Reason for Beginning Current and Retrospective Child Care Arrangements: Percent Distribution|
|Current Arrangements||Retrospective Arrangements|
|Child Needs Playmate||2.5||2.0|
|Stopped Previous Caregiver||19.4||19.2|
|NOTE: Includes all primary and secondary current and retrospective arrangements.|