The impact of having a young child with disabilities on maternal labor supply by race and marital status.
Data from the National Maternal and Infant Health Survey (NMIHS) with the three-year follow-up survey data were used to assess the impact of having a young child with disabilities on maternal labor supply. The oversampling of low birth-weight and African-American births in NMIHS provides a larger proportion of disabled children and a bigger sample of black mothers than available in data used in prior studies. The empirical analyses examined the impact of child disability on maternal employment participation and conditional intensity by race and marital status using alternative measures of disability based on occurrence of specific health conditions as well as activity performance and general development. Exogeneity of activity-based disability measures was tested using health conditions as instruments and instrumental variable models were used to estimate their impacts on maternal labor supply. The results indicated that activity-based measures are endogenous for black single and white married mothers. Model results suggested large reductions in likelihood of employment participation of black single as well as white single and married mothers. Work intensity of employed black single mothers was also substantially reduced by child disability. Overall, no consistent effects of child disability were observed for black married mothers.
Birth defects and child disabilities have been a major focus of research in various academic fields due to the wide array of their economic and psychological impacts at the individual, family, and societal levels. About 1 in every 33 infants in the US is reported to have a birth defect (CDC, 2005). The prevalence of main developmental disabilities among children born with birth defects is at least eight times higher than that among children without birth defects (Decoufle et al., 2001). Although birth defects still form the leading cause of infant mortality in the US, deaths among infants born with birth defects decreased by 48 percent between 1970 and 1997 (Lee et al., 2001). Further, preterm birth and low birth weight (LBW), which are high risk factors for disability, currently affect about 12 and 8 percent of births in the US respectively, and have increased by more than 28 and 16 percent in the past two decades (Arias et al., 2003). At the same time, survival of infants born preterm and at LBW have improved, with a 9 percent decline in infant mortality rate among LBW infants between 1995 and 2001 (Arias et al., 2003), and a 30-75 percent decline in neonatal morality of preterm infants between 1975 and 1994 (Allen et al., 2000). A potential consequence of increase in occurrence of risk factors (prematurity and LBW) combined with improvement in survival rates of infants predisposed to disability is a greater prevalence of developmental disabilities among children. Indeed, about 10-17 percent of children in the US are reported to have some type of developmental disability (Boyle et al., 1994; NCBDDD, 2005).
Child disability has many important implications for public policy. One potential impact is on maternal work effort, including employment participation and work intensity of employed mothers. Children with significant disabilities often require extensive health care and rehabilitation services, which may increase the desire for greater market income or employer-sponsored health insurance to provide the resources that enable obtaining these services. Alternatively, caring for a severely disabled child may also require substantial time inputs--on average about 4 hours of direct care daily (Leonard et al., 1992). Given traditional care giving roles within families, this may increase the value of maternal time inputs in non-market production in the form of care for a disabled child. Thus, the net impact on maternal labor supply is conceptually ambiguous.
Previous Empirical Studies
Several published studies have evaluated the effect of child's health on maternal work effort, and a summary of the main findings is presented below. A more detailed review of earlier studies can be found in Powers (2003). Salkever (1982a) reported that having a disabled child reduced employment participation in white two-parent headed families, but not in white female-headed families and in two-parent and female-headed non-white families. Yet Breslau et al (1982) reported a significant negative interaction term between child disability and indicator for black race in the maternal employment participation model for two-parent families but not for single mothers (disability coefficient was not significant), and in another study of white married mothers, Salkever (1982b) found a weaker evidence for a negative impact on work status, but reported a reduction in work hours and earnings for working mothers. Others have also found that poor child health was associated with a reduction in employment participation of married mothers but not single mothers (Kimmel, 1998). On the contrary, other studies focusing on single mothers reported a significant reduction in employment participation (Salkever, 1990; Wolfe and Hill, 1995) with inconsistent results reported for hours worked. Recent studies have found a significant reduction in both employment participation and work intensity of single mothers with mixed results for married mothers (Powers, 2001; Powers, 2003; Conway et al, 2003).
Powers (2001) considered the possibility that reported child disability is endogenous, and using an instrumental variable (IV) model, found that having a disabled child reduces the employment participation of female heads, but had a smaller impact for two-parent households. The effects of child disability were overestimated when not accounting for its endogeneity. Two other studies also attempted to control for potential endogeneity of child health (Norberg 1998; Corman et al, 2003) with rather inconsistent results, yet these analyses were not stratified by maternal marital status and/or race.
Analyses of potential variation of the impact of child disability by race based on this literature have been inadequate. A common limitation in prior studies is employing data from surveys that have small samples of black mothers, coupled with a small prevalence of disability among children of sampled families, making attempts to effectively address any differential impact by race infeasible. The importance of identifying impacts of child disability by race relates to differences in labor supply by race, which may imply differential impacts of having a disabled child on maternal work effort. Further, the prevalence of child disability may vary by race. In 2002, 11 and 17.5 percent of white and black children respectively were born preterm, and 6.8 and 13.3 percent were born at LBW (Arias et al., 2003).
Mixed results were also reported in terms of effects of child disability by maternal marital status. Differential impacts are expected to be present by marital status with potential interactions between marital status and race, increasing the importance of estimation of the impact of child disability on maternal labor supply by both race and marital status. Findings regarding whether the main effect is on labor force participation or on intensity of work supply also are mixed.
Another common limitation of most previous research is using measures of child health or disability that are potentially endogenous, in that they reflect maternal (and parental) contribution to child health, which is linked to underlying maternal characteristics that also influence her labor supply decisions, such as abilities in household production and preferences for child health and market versus household production. A further potential source of endogeneity is biased reporting of child disability to self justify work decisions (Powers, 2001). Further, several previous studies have combined children of different age groups. As the impact of child disability on maternal work effort may vary by child age, with potentially greater impacts at younger ages due to a greater demand for child and medical care, estimating impacts of child disability in specific main age groups becomes essential.
This paper addresses some of these limitations by using data from the 1988 National Maternal Infant Health Survey (NMIHS) (NCHS 1992) linked to the 1991 Follow-Up Survey (NCHS 1995). The NMIHS provides a national sample of births with an oversample of low birth-weight and African-American infants, which results in a relatively large sample of black mothers with a disabled child by marital status (largest sample of black mothers compared survey data that have previously been used). The detailed health data in NMIHS also permit several definitions of disability to be evaluated (as further described below). We consider two main specifications of child disability to test and account for the potential endogeneity of activity and development based disability measures using an IV model. The analyzed 1991 sample includes single born children between 2 and 4 years of age (average of three years). This sample allows specific estimation of the impact of child disability at a key young age. By the age of 3 years, mothers are expected to have made nearly full adjustments in their labor supply based on child's health. Thus, results obtained from this key age group may provide close estimates for the impact of child disability at older ages. The 1988 data provides data on maternal employment participation during the 12 months prior to the birth of the sampled child. This measure provides a good proxy for unobservable maternal and family fixed effects that may relate to both child disability and maternal labor supply and is included in the labor supply models here [similar measures were included in Corman et al (2003) and Norberg (1998)].
Balanced against these advantages, the NMIHS data provide less detail about sources of income and employment than other survey data that have been typically used for this research (e.g. CPS). However, an analysis of the NMIHS data complements the existing literature by providing a new empirical perspective on the impact of having a young child with disabilities on maternal work effort, including employment participation and conditional work intensity, by race and marital status.
Definition of Child Disability
The National Center on Birth Defects and Developmental Disabilities (NCBDDD) defines developmental disability as the presence of one or more of a set of chronic conditions that are related to physical and/or mental impairments and that are severe enough to cause functional, emotional, and learning limitations (NCBDDD, 2004). Based on this definition, child disability can be empirically measured by the physical or mental impairments and/or by the presence of chronic functional, emotional, and learning limitations that are a consequence of those impairments and that characterize the disability at the activity level of the child. However, limitations in activity are also determined by parental investments in child's health. Further, it can be argued that the occurrence of specific impairments, such as birth defects or chronic health conditions, is more due to exogenous health endowments and is less affected by parental investments. Thus, an empirical specification of disability based on occurrence of specific impairments versus limitations in child activity is less prone to estimation bias due to endogenous demand for child's health. Less error in maternal reporting (due to better recall, clarity of question, and biased reporting) may be expected in the earlier specification as well. Activity based measures may seem more intuitive however for studying the effect of child disability on maternal labor supply, as the effects of impairments on overall health may vary with "unmeasured" severity of impairments and be mediated through medical treatment and care.
In accordance with the NCBDDD's definition, and in light of the above argument, disability is first specified in this study by an indicator (0, 1) for having a health condition or impairment that may limit the physical or cognitive activity of the child or otherwise impede the child's development (Definition I). The NMIHS provides data on several health conditions and impairments that can be used in a condition-based definition of disability. Mothers reported whether they had been told by a health professional that the child had any of these conditions. Table 1 lists the health conditions included in this disability definition. In order to distinguish between functional limitations imposed by these conditions, a sub-specification of disability (Definition I-A) within this condition-based definition is formed based on three indicators for having conditions that would mainly limit sensory, mental, and physical functioning of the child (see Table 1).
A second (activity-based) specification of disability (Definition II) is formed based on 16 Denver developmental items that were part of the 1991 interview. Those items inquire on the child's fine motor, gross motor, language, as well as personal and social development, and thus evaluate disability at the activity level. Each item is scored by 1 if the activity is performed and by 0 otherwise (see the 16 Denver developmental items and scoring in NMIHS documentation; NCHS, 1995). A developmental measure was formed in this study based on quantiles of the summative score for the 16 items (ranging from 0-16). This approach has been previously used for this data (Hummer et al., 2004). We separately measure child's poor development (disability) by indicators for having a score at or below the 10th, 20th and 30th percentiles of the summative developmental score, which corresponded to cutoff scores of 8, 10, and 11 respectively. No differences in percentile cutoff scores were observed between black and white mothers.
In order to account for potential endogeneity of Definition II, we apply an IV model using the three indicators representing health conditions with sensory, mental, and physical limitations (Definition I-A) as instruments for the percentile development indicators, using the simple logic that the occurrence of these conditions is exogenous to parental characteristics that influence their investment in or reporting of child's health (represented by activity performance), yet is also a major predictor of child's development and activity performance [similar to Powers (2001)]. The potential outcome of this approach is not only accounting for endogeneity of the activity based measure (Definition II), yet also exploiting the variation in Definition I that truly impacts child disability.
Labor Supply Models
Employment participation is measured by whether the mother had any employment in the past month and conditional work intensity is defined by usual hours worked among those who had a work activity in the past month. The following empirical model is used:
S = [alpha] + [beta] x DC + W[delta] + M[zeta] + X[phi] + [epsilon],
where S represents maternal labor market participation or conditional work intensity, DC is a measure of child disability, W represents wage and other household income, M represents characteristics of the mother (age, education, self-rated health, pre-delivery work status), X represents other covariates (such as number of children in specific age groups), and a represents the error term.
The employment participation model is estimated using a logistic regression, while the conditional intensity model is estimated using a semilog ordinary least squares (OLS) regression specification. A two-stage least squares (2SLS) regression is used for the IV model when Definition II of disability is used. Over-identification restrictions and endogeneity of the activity based disability measure are tested using the Basmann and Hausman tests repesctively (Basmann, 1960; Hausman, 1978; Wooldridge, 2002). OLS models for labor supply are also estimated for comparison to 2SLS. All models are estimated by race (black, white) and marital status (single, married).
A Heckman's selection model (Heckman, 1979) is applied separately for black and white mothers to estimate martial status and race specific predicted wages for both workers and non-workers. To better identify the wage equation, the child disability measure (Definition I), maternal work effort prior to delivery, age of sampled child, indicators for other household income, and AFDC benefits were excluded from the wage equation. Female race-specific labor force participation and unemployment rates at the state level (from the 1990 census data) were only included in the wage equation and excluded from the labor supply models in order to identify the coefficient of the predicted wage variable. Model specifications including and excluding predicted hourly wages were evaluated.
Definition, mean values, and standard deviations of the dependent variables and model are listed in Table 2. The total sample consisted of 2,144 black single mothers, 1,137 black married mothers, 688 white single mothers, and 2,742 white married mothers. Single mothers included those who were widowed, divorced, separated, or never married in 1991. The mean age of the sampled children at the time of the 1991 follow-up survey was approximately three years.
Throughout this paper, we only describe and discuss the results for the disability measures. Table 3 reports the logistic regression results for working at all in the last month, by race and marital status, using Definition I of disability. Using this definition, having a disabled child reduced employment participation only for black single mothers, with about 17% estimated reduction in likelihood of last month employment. The regression coefficients for all disability definitions under alternative estimation models for last month employment participation and conditional work intensity measured as log of weekly hours usually worked are included in Tables 4 and 5 respectively. Results of tests of over-identification for 2SLS and endogeneity of OLS coefficients of activity-based disability measures are highlighted when significant. In order to focus on the disability results, the results for other explanatory variables are not reported and are available from the authors upon request. Estimated coefficients for the child disability variables were virtually unaffected if predicted hourly wage was excluded from the model. Thus, only results of model specifications that included wages are reported here.
The instruments (Definition I-A) predicted significantly the activity-based disability measures (Definition II) for the four maternal groups (all partial F statistics were larger than 10 except for Denver20 and Denver30 in the hour intensity model for white single mothers where the F statistic was about 6). First stage regression results of the IV model are not reported due to space limitations but are available from the authors. Several results in Table 4 are worth highlighting. Having a disabled child using the activity-based measure showed consistently a substantial negative impact on likelihood of employment for white married (about 21-22% reduction) and black single mothers (about 21-25% reduction with larger effect for Denver10) under 2SLS but not OLS, suggesting a potential positive estimation bias when the endogeneity of activity-based child disability was not accounted for (exogeneity of Definition II was rejected under all specifications for these groups).
On the other hand, the OLS and 2SLS coefficients for Denver20 and Denver30 were comparable for white single mothers and the OLS coefficient showed a statistically significant reduction of about 10 percentage points in likelihood of employment. For black married mothers, no significant effects of child disability were observed under any specification. Using disability Definition I-A (sensory, mental, physical) showed generally no statistically significant effects on employment (except for conditions with physical limitations which reduced employment propensity of black single mothers), and the over-identification restrictions for excluding these variables from the employment participation regression when used as instruments for Definition II could only be rejected for black single mothers when predicting Denver20 and Denver 30 at the 0.1 significance level (p=0.09). Finally, health conditions that limit mainly mental activity had larger coefficients in absolute value (statistically insignificant) compared to conditions with a main effect on physical or sensory activity for white mothers and black married mothers, yet the opposite was observed for black single mothers.
Child disability using Definition I showed no statistically significant impact on conditional work intensity measured by usual work hours per week. For black single mothers, the 2SLS model for Definition II showed about 24-29% reduction in weekly work hours (marginally significant largest effect for Denver10) unlike OLS, and the exogeneity assumption of the activity-based disability measures was rejected under all disability specifications. The exogeneity hypothesis for the activity-based disability measures could not be rejected for the other maternal groups, and none of their coefficients were significant. The 2SLS coefficients of Denver10 and 30 were relatively large for white single mothers (but again statistically insignificant). For all maternal groups, OLS and 2SLS Definition II coefficients were generally positive and negative respectively (except for OLS coefficient Denver30 for white married mothers). Another noteworthy result is that when using Definition I-A, only conditions that limit child mental functioning had a negative coefficient for all groups (though generally not statistically significant except for black single mothers).
DISCUSSION AND CONCLUSION
The analysis presented here suggests that having a young child with disabilities may substantially reduce the likelihood of last month employment participation among black single mothers and white single and married mothers. The results were less suggestive for black married mothers in terms of a large negative effect of child disability on labor supply. The analysis also indicates that measures of child disability that are based on maternal report of child's activity performance, though possibly more intuitive than measures based on maternal report of presence of specific health conditions and impairments as diagnosed by health professionals, are likely endogenous for maternal work effort. Ignoring this endogeneity may bias the estimated effects of child disability on maternal labor supply.
A positive bias was suggested in the OLS coefficients of the activity based measures for black single and white married mothers. This potential bias can be due to unobserved maternal characteristics, such as efficiency or ability in household production, which may improve child's activity performance through better care (i.e. lower disability based on activity limitations) yet also increase maternal propensity to engage in household production instead of market employment. It may also reflect a reverse adverse effect of maternal employment on child's activity performance, due to a smaller time investment in child's development. Maternal reporting of child's activity performance may be biased as well and the bias may be correlated with her work behavior. For instance, working mothers may observe less frequently their children and may underestimate the child's activity performance (i.e. overestimate activity limitations). While a positive bias was observed in OLS in this study, the bias direction may be sample dependent and may also vary with the health measure used [opposite results observed in Powers (2001)].
Measuring child disability by the presence of health conditions and impairments that are well known to adversely affect child's development showed generally no effect on maternal work participation except for black single mothers. While this measure is likely more exogenous for maternal work effort than activity-based measures of disability, it may form a less intuitive measure for studying the effects of child disability on maternal behaviors. The effects of health conditions on child's development are potentially mediated by household and medical care and are expected to vary by severity of conditions. Thus, this result might be interpreted as an attenuated effect due to potential measurement error and heterogeneity in the disability variable (i.e. a less valid measure of disability). The observed effect of this disability measure among black single mothers might be in part due to a greater severity of health conditions and lower access to medical care.
A reduction in work intensity of black single mothers was suggested based on the 2SLS results but not for other groups. The 2SLS coefficients for white single mothers had a negative sign and were relatively large but the sample size was about half that of black single mothers. The suggested lack of an adverse effect of child disability on work intensity for married mothers might be due to a potential larger fixed-cost versus variable-cost component of total childcare costs and/or a smaller impact of child disability on variable costs compared to single mothers. Single mothers might face a more restricted access to childcare resources, and potentially a larger variable-cost versus fixed-cost component of childcare total costs as well as a potentially stronger impact of child disability on variable costs of childcare. In the same notion, the suggested negative impact for all maternal groups of health conditions that mainly limit mental functioning on weekly hours compared to other conditions (again though generally not statistically significant) might be due to a larger variable-cost component for these children.
Of course, these conclusions are subject to a number of qualifications. Except for the number of children born at LBW before the sampled child, no data is available on other indicators of health or disability of other children in the family. Family specific biologic and environmental characteristics may induce correlations in disability status and health conditions of children. This also indicates that the used IV instruments may also be correlated with other children's health. Further, the exogeneity of some of the health conditions might be questioned as these might also be affected by parental investment. The Denver quantiles are the only activity-based measures of disability that were used in this study. Quantitative (and even qualitative) results can be highly sensitive to the measurement of child disability. These are issues that will need to be addressed in future research using samples with adequate health data for all children and alternative measures of child disability. Further research is also needed to investigate potential differences in variable versus fixed cost components of total childcare costs by race and marital status and the need to design childcare policies accordingly.
Allen, M. C., Alexander, G. R., Tompkins, M. E., and Hulsey, T. C. (2000). "Racial Differences in Temporal Changes in Newborn Viability and Survival by Gestational Age." Paediatric and Perinatal Epidemiology, 14(2), 152-158.
Arias, E., MacDorman, M. F., Strobino, D. M., and Guyer, B. (2003). "Annual Summary of Vital Statistics--2002." Pediatrics, 112(6 Pt 1), 1215-1230.
Basmann, R. (1960). "On Finite Sample Distributions of Generalized Classical Linear Identifiability Test Statistics." Journal of the American Statistical Association, 55, 650-659.
Boyle, C. A., Decoufle, P., and Yeargin-Allsopp, M. (1994). "Prevalence and Health Impact of Developmental Disabilities in US Children." Pediatrics, 93(3), 399-403.
Breslau, N., Salkever, D., and Staruch, K. S. (1982). "Women's Labor Force Activity and Responsibilities for Disabled Dependents: A Study of Families with Disabled Children." Journal of Health and Social Behavior, 23(2), 169-183.
Centers for Disease Control and Prevention (CDC) (2005). Birth Defects: Frequently Asked Questions (FAQs). Retrieved 15 July,, 2005, from http://www.cdc.gov/ncbddd/bd/faq1.htm#Common BD
Corman, H., Reichman, N. E., and Noonan, K. (2003). "Mothers' and Fathers' Labor Supply in Fragile Families: The Role of Child Health." National Bureau of Economic Research, Inc, NBER Working Papers: 9918.
Decoufle, P., Boyle, C. A., Paulozzi, L. J., and Lary, J. M. (2001). "Increased Risk for Developmental Disabilities in Children Who Have Major Birth Defects: A Population-Based Study." Pediatrics, 108(3), 728-734.
Hausman, J. (1978). "Specification Tests in Econometrics." Econometrica, 46(6), 1251-1271.
Heckman, J. (1979). "Sample Selection Bias as a Specification Error." Econometrica, 47(1), 153-161.
Hummer, R. A., Hack, K. A., and Raley, R. K. (2004). "Retrospective Reports of Pregnancy Wantedness and Child Well-Being in the United States." Journal of Family Issues, 25(3), 404-428.
Kimmel, J. (1998). "Child Care Costs as a Barrier to Employment for Single and Married Mothers." Review of Economics and Statistics, 80(2), 287-299.
Lee, K., Khoshnood, B., Chen, L., Wall, S. N., Cromie, W. J., and Mittendorf, R. L. (2001). "Infant Mortality from Congenital Malformations in the United States, 1970-1997." Obstetrics and Gynecology, 98(4), 620-627.
Leonard, B., Brust, J. D., and Sapienza, J. J. (1992). "Financial and Time costs to Parents of Severely Disabled Children." Public Health Reports, 107(3), 302-312.
National Center on Birth Defects and Developmental Disabilities (NCBDDD) (2004). "What are Developmental Disabilities?" Retrieved 15 July, 2005, from http://www.cdc.gov/ncbddd/dd/
National Center on Birth Defects and Developmental Disabilities (NCBDDD) (2005). Metropolitan Atlanta Developmental Disabilities Study. Retrieved 15 July, 2005, from http://www.cdc.gov/ncbddd/dd/ddres.htm
Norberg, K. (1998). "The Effects of Daycare Reconsidered". National Bureau of Economic Research, Inc, NBER Working Papers: 6769.
Powers, E. T. (2001). "New Estimates of the Impact of Child Disability on Maternal Employment." American Economic Review, 91(2), 135-139.
Powers, E. T. (2003). "Children's Health and Maternal Work Activity: Estimates under Alternative Disability Definitions." Journal of Human Resources, 38(3), 522-556.
Salkever, D. S. (1982a). "Children's Health Problems and Maternal Work Status." Journal of Human Resources, 17(1), 94-109.
Salkever, D. S. (1982b). Children's Health Problems: Implications for Parental Labor Supply and Earnings. In V. R. Fuchs (Ed.), Economic Aspects of Health (pp. 221-251). Chicago: University of Chicago Press.
Salkever, D. S. (1990). Child Health and Other Determinants of Single Mothers' Labor Supply and Earnings. In I. Sirageldin, A. Sorkon and R. Frank (Eds.), Research in Human Capital and Development (Vol. 6, pp. 147-181). Greenwich: JAI Press.
Wolfe, B. L., and Hill, S. C. (1995). "The Effect of Health on the Work Effort of Single Mothers." Journal of Human Resources, 30(1), 42-62.
U.S. Dept. of Health and Human Services, National Center for Health Statistics. National Maternal and Infant Health Survey, 1988 [Computer file]. Hyattsville, MD: U.S. Dept. of Health and Human Services, National Center for Health Statistics [producer], 1991. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 1992.
U.S. Dept. of Health and Human Services, National Center for Health Statistics. National Maternal and Infant Health Survey, 1988: Longitudinal Follow-Up, 1991 [Computer file]. Hyattsville, MD: U.S. Dept. of Health and Human Services, National Center for Health Statistics [producer], 1993. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 1995.
Wooldridge, J. (2002). Econometric Analysis of Cross Section and Panel Data. Cambridge and London: MIT Press.
GEORGE L. WEHBY
University of Iowa
ROBERT L. OHSFELDT
Texas A&M University
Table 1 Included Child Health Conditions Main Functional Limitation of Health Condition (a) Health Condition (b) Sensory Mental Physical Deafness/other hearing problems X Delayed speech problem X Sight problems X Epilepsy or convulsions or X seizures (without fever) Brain Problems X Developmentally delayed or X mentally retarded Neuromuscular problems X X Cerebral palsy X X Genetic disorders X X X Asthma X Chronic respiratory lung X problem Eating or swallowing problems X Urology/kidney problems X Problems with feet/toes: pigeon- X toed, clubfoot, etc ... Gastrointestinal/rectal problems X Problems with upper respiratory X systems: adenoids, tonsils, ears Sickle cell anemia X Chronic heart condition X Spina bifida X Other serious chronic problems X Note: Table 1 lists the health conditions used in disability Definitions I and I-A in this study. Further detailed description of some conditions are available in the NMIHS codebook files. (a) According to Definition I of disability, a child was considered disabled if any of those health conditions was present and not disabled when none of them was present. (b) Based on Definition I-A of disability, three indicators were included for having conditions that limit sensory, mental, and physical functioning. An "X" indicates the category of limitation that the health condition was considered to primarily impose. Table 2 Definition, means, and standard deviation of the variables Variable/Definition Sample of Mothers White White Single Married Independent Variables Disability I / Dummy variable 0.29 0.23 = 1 if child is disabled (0.46) (0.42) according to Definition I Sensory / Dummy variable = 0.12 0.07 1 if child has a condition that (0.32) (0.26) limits sensory functioning (Disability definition I-A) Mental / Dummy variable = 1 0.04 0.03 if child has a condition that (0.2) (0.18) limits mental functioning (Disability definition I-A) Physical / Dummy variable = 0.23 0.19 1 if child has a condition that (0.42) (0.39) limits physical functioning (Disability definition I-A) Denver10 / Dummy variable 0.1 0.11 = 1 if child's Denver (0.31) (0.31) development score is at or below 10th percentile score of 8 (Disability Definition II) Denver20 / Dummy variable 0.24 0.25 = 1 if child's Denver (0.43) (0.43) development score is at or below 20th percentile of 10 (Disability Definition II) Denver30 / Dummy variable 0.35 0.36 = 1 if child's Denver (0.48) (0.48) development score is at or below 30th percentile of 11 (Disability Definition II) Hourly wage (a) / Predicted 7.1 11.7 hourly wage based on (2.2) (3.7) selection model Work prior to delivery / 0.69 0.73 Dummy Variable =1 if (0.46) (0.44) mother was employed during 12 months preceding birth of sampled child Maternal age (years) 27.1 30.4 (5.6) (5.2) Maternal age squared (years 764.4 951.7 squared) (331.4) (323.8) Hispanic / Dummy variable 0.19 0.11 =1 if mother is of a Hispanic (0.39) (0.32) origin Very good maternal health / 0.55 0.75 Dummy variable =1 if mother (0.5) (0.44) reported her health to be excellent or very good based on 5-category Likert scale Fair/poor maternal health / 0.13 0.05 Dummy variable =1 if mother (0.33) (0.23) reported fair or poor health Schooling years completed by 11.8 13.2 the mother (2.2) (2.4) Schooling years squared 144.4 180.9 (49.5) (60.8) Age of sampled child in 36.1 34.2 months (4.7) (4.1) Number of children in 1.53 1.56 household between the age of (0.75) (0.66) 0 and 5 years Number of children between 0.37 0.47 the age of 6 and 12 years (0.66) (0.74) Number of children between 0.13 0.1 the age of 13 and 18 years (0.38) (0.38) Total number of pregnancies 0.5 0.47 post sampled child (0.79) (0.66) Number of children born at 0.09 0.06 LBW prior to sampled child (0.35) (0.28) Low other household income (b) 0.47 0.49 (0.5) (0.5) AFDC 1991 / Maximum 418.0 408.7 AFDC benefits in 1991 for a (168.9) (163.8) family of three (state level) Female race-specific 5.1 5.1 unemployment rates in 1990 (0.8) (0.8) (census data, state level) Female race-specific labor 56.2 56.4 force participation rates (3.3) (3.4) (census data, state level) Dependent Variables Dummy variable =1 if the 0.48 0.55 mother worked at all in the (0.5) (0.5) past month Weekly hours usually worked 33.7 30.9 by mothers who worked in (10.5) (12.1) the past month (c) Sample Size (d) 688 2742 Variable/Definition Sample of Mothers Black Black Single Married Independent Variables Disability I / Dummy variable 0.32 0.25 = 1 if child is disabled (0.47) (0.43) according to Definition I Sensory / Dummy variable = 0.1 0.06 1 if child has a condition that (0.3) (0.24) limits sensory functioning (Disability definition I-A) Mental / Dummy variable = 1 0.04 0.02 if child has a condition that (0.2) (0.15) limits mental functioning (Disability definition I-A) Physical / Dummy variable = 0.26 0.22 1 if child has a condition that (0.44) (0.41) limits physical functioning (Disability definition I-A) Denver10 / Dummy variable 0.11 0.11 = 1 if child's Denver (0.31) (0.31) development score is at or below 10th percentile score of 8 (Disability Definition II) Denver20 / Dummy variable 0.24 0.21 = 1 if child's Denver (0.43) (0.41) development score is at or below 20th percentile of 10 (Disability Definition II) Denver30 / Dummy variable 0.35 0.32 = 1 if child's Denver (0.48) (0.47) development score is at or below 30th percentile of 11 (Disability Definition II) Hourly wage (a) / Predicted 7.6 8.9 hourly wage based on (3.1) (2.8) selection model Work prior to delivery / 0.54 0.76 Dummy Variable =1 if (0.5) (0.43) mother was employed during 12 months preceding birth of sampled child Maternal age (years) 26.0 29.4 (5.4) (5.5) Maternal age squared (years 705.1 894.6 squared) (307.0) (335.9) Hispanic / Dummy variable 0.01 0.04 =1 if mother is of a Hispanic (0.11) (0.2) origin Very good maternal health / 0.53 0.59 Dummy variable =1 if mother (0.5) (0.49) reported her health to be excellent or very good based on 5-category Likert scale Fair/poor maternal health / 0.14 0.1 Dummy variable =1 if mother (0.34) (0.3) reported fair or poor health Schooling years completed by 11.9 13.2 the mother (1.7) (2.0) Schooling years squared 144.2 177.3 (39.0) (53.6) Age of sampled child in 36.2 35.5 months (4.7) (4.6) Number of children in 1.76 1.61 household between the age of (0.88) (0.77) 0 and 5 years Number of children between 0.56 0.62 the age of 6 and 12 years (0.82) (0.79) Number of children between 0.24 0.18 the age of 13 and 18 years (0.55) (0.48) Total number of pregnancies 0.68 0.54 post sampled child (0.85) (0.8) Number of children born at 0.16 0.15 LBW prior to sampled child (0.47) (0.45) Low other household income (b) 0.47 0.5 (0.5) (0.5) AFDC 1991 / Maximum 357.9 368.5 AFDC benefits in 1991 for a (168.4) (180.1) family of three (state level) Female race-specific 12.8 12.5 unemployment rates in 1990 (3.0) (2.8) (census data, state level) Female race-specific labor 58.8 59.1 force participation rates (4.1) (4.1) (census data, state level) Dependent Variables Dummy variable =1 if the 0.37 0.64 mother worked at all in the (0.48) (0.48) past month Weekly hours usually worked 34.2 35.2 by mothers who worked in (9.3) (8.8) the past month (c) Sample Size (d) 2144 1137 (a) Both hours worked per week and income earned in past month were categorized in the NMIHS dataset, and the range midpoints were used in this study. Two observations in the white married sample with extreme wage values ($50 and $115) were excluded from the Heckman's model for wage prediction as these observations were contributing to having negative predicted wages for about quarter of the sample. (b) Other monthly income was calculated by subtracting reported total maternal income in the past month from reported total household income in the past 12 months divided by 12. Income was recoded into income categories in the NMIHS dataset so midpoints of the category ranges were used, which in part resulted in a nontrivial proportion of the sample having negative values. Therefore, a dummy variable (0,1) for other income below median value was used to indicate low other household income (i.e. negative values were considered as zero income). The median was $216.1, $2091.1, $174.5, and $1207.8 respectively for white single, white married, black single, and black married mothers. (c) Number of hours worked were recoded within categories in the survey file. The midpoints of those categories were used. Respondents with seasonal and variable jobs were asked to report the number of hours that they worked in the last month they worked. Weekly hours were calculated from these responses as well. (d) The sample sizes reported here include cases with non-missing data for Disability I, other covariates, and employment in past month. The final sample sizes in models using other disability definitions varied slightly due to different missing data patterns for these variables. Table 3 Logistic Regression Coefficients for Last Month Employment Using Disability Definition I Sample of Mothers Variable White White Single Married (N=688) (N=2742) Intercept 2.53 3.77 (3.69) (2.81) Disability I 0.002 -0.06 (0.2) -0.1 Hourly wage 0.32 0.69 *** (0.27) (0.21) Work prior to 1.56 *** 1.45 *** delivery (0.22) (0.1) Maternal age -0.35 -0.53 ** (0.3) (0.23) Maternal age 0.01 0.005 * squared (0.004) (0.003) Hispanic 0.09 1.16 *** (0.28) (0.37) Very good maternal -0.03 -0.19 health (0.21) (0.13) Fair/poor maternal 0.06 -0.04 health (0.32) (0.22) Schooling years 0.28 0.57 *** (0.29) (0.22) Schooling years -0.01 -0.04 *** squared (0.02) (0.02) Age of child -0.005 0.01 (0.02) (0.01) Children between 0 -0.45 ** -0.79 *** and 5 (0.18) (0.17) Children between 6 0.13 -0.92 *** and 12 (0.16) (0.24) Children between 13 0.08 0.67 *** and 18 (0.24) (0.23) Total pregnancies -0.23 * -0.45 *** (0.14) (0.07) LBW children 0.07 -0.05 -0.27 -0.16 Low other 0.12 0.75 *** household income (0.18) (0.09) AFDC 1991 -0.002 *** -0.0002 (0.0010) (0.0003) Mc Fadden [R.sup.2] 0.18 0.13 Sample of Mothers Variable Black Single Black (N=2144) Married (N=1137) Intercept -0.8 -6.3 *** (1.58) (2.36) Disability I -0.38 *** 0.15 (0.11) (0.16) Hourly wage 0.16 *** 0.08 (0.06) (0.18) Work prior to 1.14 *** 1.33 *** delivery (0.11) (0.16) Maternal age -0.03 0.28 ** (0.08) (0.13) Maternal age 0.0003 -0.005 ** squared (0.002) (0.002) Hispanic -3.0 ** -0.11 (1.45) (0.44) Very good maternal -0.17 0.08 health (0.14) (0.16) Fair/poor maternal -0.41 ** -0.31 health (0.18) (0.37) Schooling years -0.26 0.12 (0.18) (0.31) Schooling years 0.02 *** -0.002 squared (0.01) (0.02) Age of child 0.01 0.02 (0.01) (0.02) Children between 0 -0.39 *** -0.37 *** and 5 (0.08) (0.11) Children between 6 -0.15 ** -0.2 ** and 12 (0.07) (0.1) Children between 13 0.07 -0.03 and 18 (0.1) (0.15) Total pregnancies -0.47 *** -0.36 *** (0.08) (0.10) LBW children -0.25* 0.17 -0.13 -0.16 Low other 0.29 *** 0.34 ** household income (0.1) (0.15) AFDC 1991 -0.001 *** -0.0003 (0.0003) (0.0004) Mc Fadden [R.sup.2] 0.19 0.15 Note: *, **, and *** indicate significant coefficients at p<0.1, p<0.05, and p<0.01 respectively. Standard errors of coefficients are in parentheses. Table 4 Regression Coefficients for Child Disability in Last Month Employment Models Sample of Mothers Disability Definition White White (Analytical Model) Single Married (N=688) (N=2742) Disability I (Logistic) 0.002 -0.06 (0.2) (0.1) Disability I-A (Logistic) Sensory -0.04 -0.28 (0.32) (0.18) Mental -0.54 -0.41 (0.51) (0.27) Physical 0.11 0.08 (0.23) (0.12) Denver10 (Logistic) -0.31 -0.07 (0.30) (0.14) Denver10 (2SLS) -0.12 -0.22 ** (0.18) (0.09) Denver10 (OLS) -0.06 -0.01 (a) (0.06) (0.03) Denver20 (Logistic) -0.51 ** -0.17 (0.23) (0.10) Denver20 (2SLS) -0.1 -0.21 ** (0.18) (0.09) Denver20 (OLS) -0.1 ** -0.04 * (a) (0.04) (0.02) Denver30 (Logistic) -0.46 ** -0.09 (0.21) (0.1) Denver30 (2SLS) -0.1 -0.21 ** (0.16) (0.09) Denver30 (OLS) -0.09 ** -0.02 (a) (0.04) (0.02) Sample of Mothers Disability Definition Black Single Black (Analytical Model) (N=2144) Married (N=1137) Disability I (Logistic) -0.38 *** 0.15 (0.11) (0.16) Disability I-A (Logistic) Sensory -0.29 -0.02 (0.19) (0.32) Mental -0.07 -0.25 (0.30) (0.51) Physical -0.32 ** 0.17 (0.12) (0.18) Denver10 (Logistic) -0.03 -0.35 (0.18) (0.23) Denver10 (2SLS) -0.25 ** -0.06 (0.1) (0.15) Denver10 (OLS) 0.003 (b) -0.07 (0.03) (0.04) Denver20 (Logistic) -0.15 -0.01 (0.13) (0.17) Denver20 (2SLS) -0.2 ** (c) -0.04 (0.08) (0.13) Denver20 (OLS) -0.02 (a) -0.003 (0.02) (0.03) Denver30 (Logistic) -0.15 0.09 (0.11) (0.16) Denver30 (2SLS) -0.21 ** (c) -0.05 (0.09) (0.14) Denver30 (OLS) -0.02 (a) 0.02 (0.02) (0.03) Note: Sample sizes are reported for disability Definition I. Sample sizes for other disability specifications varied slightly due to different missing data patterns. *, **, and *** indicate significant coefficients at p<0.1, p<0.05, and p<0.01 respectively. Standard errors of coefficients are in parentheses. (a) The exogeneity hypothesis was rejected at p<0.05 with a potential positive bias in OLS coefficient. (b) The exogeneity hypothesis was rejected at p<0.01 with a potential positive bias in OLS coefficient. (c) The over-identification test statistic was significant at p=0.09. Table 5 Regression Coefficients for Child Disability in Semilog Model of Weekly Hours Worked among Those who worked in Past Month Sample of Mothers Disability White Single White Definition (N=324) Married (Analytical Model) (N=1498) Disability I (OLS) 0.06 0.03 (0.06) (0.04) Disability I-A (OLS) Sensory -0.07 0.02 (0.1) (0.07) Mental -0.13 -0.13 (0.17) (0.1) Physical 0.06 0.03 (0.07) (0.04) Denver10 (2SLS) -0.23 -0.14 (0.31) (0.20) Denver10 (OLS) 0.07 0.09 (0.1) (0.05) Denver20 (2SLS) -0.08 -0.11 (0.31) (0.18) Denver20 (OLS) 0.04 0.01 (0.07) (0.04) Denver30 (2SLS) -0.22 -0.07 (0.27) (0.18) Denver30 (OLS) 0.04 -0.01 (0.07) (0.04) Sample of Mothers Disability Black Single Black Definition (N=786) Married (Analytical Model) (N=718) Disability I (OLS) -0.02 0.01 (0.04) (0.03) Disability I-A (OLS) Sensory -0.02 0.001 (0.07) (0.07) Mental -0.18 * -0.05 (0.1) (0.11) Physical -0.01 0.03 (0.04) (0.04) Denver10 (2SLS) -0.29 * -0.05 (0.17) (0.17) Denver10 (OLS) 0.09 (a) 0.001 (0.06) (0.05) Denver20 (2SLS) -0.24 * -0.03 (0.13) (0.16) Denver20 (OLS) 0.04 (a) 0.04 (0.04) (0.04) Denver30 (2SLS) -0.25 * -0.05 (0.15) (0.15) Denver30 (OLS) 0.02b 0.05 (0.04) (0.03) Note: Sample sizes are reported for disability Definition I. Sample sizes for other disability specifications varied slightly due to different missing data patterns. * indicates significant coefficients at p<0.1. Standard errors of coefficients are in parentheses. (a) The exogeneity hypothesis was rejected at p<0.05. (b) The exogeneity hypothesis was rejected at p<0.1.
|Printer friendly Cite/link Email Feedback|
|Author:||Wehby, George L.; Ohsfeldt, Robert L.|
|Publication:||Journal of Health and Human Services Administration|
|Date:||Dec 22, 2007|
|Previous Article:||How well are we doing addressing disability in America? Examining the status of adults with chronic disabling conditions, 1995 and 2005.|
|Next Article:||The road to empowerment: a historical perspective on the medicalization of disability.|