DISCLAIMER

This paper reports the results of research and analysis undertaken by Census Bureau Staff. It has undergone a more limited review than official Census Bureau publications. This report is released to inform interested parties of research and to encourage discussion.

In the past few years, research on the well-being of the population has expanded to include the concept of 'at-risk' conditions. Generally, these conditions are thought to be characteristics of the individual, or situations of the context they are a part of, that are believed to create higher likelihoods of undesirable life outcomes (e.g., completing high school, avoiding premarital births), or to impact overall quality of life.

This paper estimates the frequency of three 'personal' and four 'familial' at-risk conditions for the school age population in the United States. The 'personal' conditions are: presence of a disability, ever retained in school, and speaking English less than 'very well'. The 'familial' conditions are: either or both parents absent from the household, at least one foreign-born parent of recent immigration, low family income, and no employed parent. For each of these conditions we estimate levels of occurrence for the total school-age population as well as for age, race, and sex groups. Additional analysis focuses on regional and metropolitan variation. Data for the analysis are taken from the October 1999 Current Population Survey.

The analysis shows that, while a majority (54%) of school-age children has no significant risk factors, a significant minority does. A far larger proportion of children has experienced a familial risk factor (36%) than a personal one (18%). The single most common personal risk factor is being retained in school, while the most common familial factor is not living with both parents.

Additionally, a sizable proportion of children (18%) has more than one risk factor in their life. Substantial variation in the number and kind of risk factors occurs across various demographic groups, with multiple risk factors more frequent for males and blacks. There is little variation across age groups, implying that younger persons have already encountered similar levels of risk factors as the cohort nearly a decade older.

A final analysis on the cross-classification of the various risk factors shows the factors of highest concurrence for individuals.

In the past decade, there has been increased interest in the well-being and general status of children in both the United States and worldwide. Researchers have undertaken a wide array of data collections and analyses focused on investigating the conditions of children and some of the outcomes these children realize as they begin to age into early adulthood. These efforts are underway in the private and public sectors, in governmental and nongovernmental organizations, as well as in academic and policy or applied contexts.

Much of the research and literature related to the well-being of children focuses on the specific detrimental outcomes that may make a child's life more difficult in adulthood. These disadvantageous outcomes are often defined as related to school and education, as in being retained a grade or dropping out of school, but can be expanded to include other outcomes such as a teenage pregnancy. These studies attempt to determine the predictors of these outcomes from a variety of sources.

One source considered for predictors of risk factors is other risk factors. Risk factors are not exclusively outcomes of the personal characteristics of the child or their family, but can be considered as either precursors or outcomes of other risk factors. For example, researchers have studied the effect of having a learning disability on the likelihood of a student being retained (Barnett, Carizio, and Payette, 1996; Roderick 1994), and the effect of grade retention on the likelihood of dropping out (McMillen, Kaufman, and Klein, 1997). These studies usually focus more on the designated risk factors and outcomes of those risk factors than on the characteristics of the children or their environments that can also influence the presence of risk factors in their lives (Bryson, 1997; Annie E. Casey Foundation, 1998). The studies that have linked student background and school outcomes suggest that there is a correlation between the two components (Clark 1990).

Another aspect of research concerning the welfare of children focuses on identifying the various conditions that make a child 'at-risk' of unspecified negative outcomes. These conditions may be related to different aspects of the child's life, such as the child's personal characteristics, their family, their school, or their community (Natriello, McDill, and Pallas, 1990; Annie E. Casey Foundation, 2000). In this area of research, the components of interest are the socioeconomic and personal characteristics of children and the features of their environments that may lead to adverse outcomes later in life, not the outcomes themselves. These analyses attempt to identify the characteristics of students who can be considered 'at-risk' and in need of a possible intervention to avoid negative consequences in their lives.

Despite the volume of research that exists regarding at-risk conditions, the field is still nascent enough that it cannot be firmly concluded that identified risk factors and their antecedents, such as the social and economic backgrounds of the students, actually lead to a less than happy or fulfilling adulthood. Much research remains to both identify the possible factors that place kids at risk, and the causal linkages between these factors and realized outcomes.

One component of this work is being undertaken in the context of identifying and estimating the levels of a large battery of 'indicators' that tell us something about the well-being of children. The volume, America's Children: Selected Indicators of Well-Being, attempts to consolidate a set of federally-collected indicators spread across five 'domains' into one document. Indicators can look both at 'conditions' as well as 'outcomes'. Much current research, primarily due to the limited amount and structure of available data, focuses mainly on the 'conditions' that children are associated with.

This paper focuses on the identification and measurement of a series of at-risk factors for school-age children in the United States. One problem with much of the existing research is that risk factors are often not simultaneously measured, but come from a variety of data collection systems. In the aforementioned federal report on child well-being, the data for the 32 indicators published comes from 16 different data sources. Obviously, the ability to better understand at-risk conditions and their impact on children comes into focus as we are able to assess a number of factors simultaneously and for the same people at the same point in time. In this paper we are able to look at a set of seven different risk factors, representing two different domains of origin - the individual and the family. In undertaking this analysis we intend to demonstrate that large numbers of children are exposed to risk factors, and many children are exposed to many simultaneously. We will also document the kind and level of variation that exist in these factors across different subpopulation segments.

The data for this research come from the October 1999 Current Population Survey (CPS). The CPS is a monthly, nationally-representative survey of the U.S. civilian noninstitutionalized household population, conducted by personal interviews. Each month about 50,000 households are contacted and the residents are interviewed on a number of topics. The U.S. Census Bureau conducts the CPS for the Bureau of Labor Statistics, and the fundamental purpose of the monthly CPS interview is to assess employment conditions and to determine the national unemployment rate of the population.

The CPS has been a data collection vehicle now for over half a century. However, over its life the CPS has become much more than simply the source of the unemployment rate. In most months, additional supplemental surveys are added to the CPS, collecting information on a wide variety of topics. Some of these supplemental collections are one-time surveys only, while others have become a part of the routine CPS data collection operation. Perhaps the best known of these supplements is the annual March "Demographic" Supplement, which concentrates primarily on the collection of detailed information regarding earnings and in-kind financial resources. For many years as well, data on school enrollment have been collected as part of the October CPS, in a supplement sponsored by the Census Bureau and the National Center for Education statistics. In some years this supplement itself has focused on basic school enrollment issues, while in other years additional special topics are added to the supplement depending on sponsor needs.

In the October 1999 supplement, additional questions were asked which allow us to identify certain characteristics of the students. These factors include English-speaking ability; the presence of personal disability; and whether the child has ever been retained in school (failed a grade). Each of these, in some respect, might be viewed as a personal at-risk factor. Children who do not speak English may receive little or no attention in their native language, or the language spoken most of the time in their home. The self-report of English-speaking ability has been shown to be highly correlated with actual English functioning and understanding (Kominski, 1989). Disability constitutes another potential risk factor for children. While some disabilities actually do restrict the child from attending a 'conventional' school with other non-disabled children, many children possess disabilities that do not restrict their attendance, but still may act as a hindrance in a variety of ways, including involvement in the full range of school and school based social activities. Finally, the condition of having failed a grade, or being retained, may be seen as both an 'outcome' and an at-risk 'condition'. Once given the message that he or she is less capable than other children, a child may start to alter their image of self worth and become disengaged from school (Roderick 1995). Each of these three conditions then, identify possible 'personal' at-risk conditions that an individual might have.

At-risk conditions may come from sources other than the individual themselves. For example, institutions and social context may also bring risk conditions with them. In addition to these three personal conditions in the October CPS, we are able to identify four other 'familial' conditions that may define risk for the child. These are identified as: the absence of either or both parents from the household the child resides in; at least one foreign-born parent of recent immigration; low family income; and the absence of any employed parent or guardian in the household. While more detailed economic information for some of these factors is available from the March CPS datafile (e.g., the actual poverty level and more detailed information on earnings and family income), using the October file allows us to examine the concurrent occurrence of both sets of risk factors.

Together, these three 'personal' and four 'familial' at-risk conditions are estimated for the U.S. school age population (persons ages 5-17).

Table 1 shows the results of the tabulations of the seven at-risk conditions for the school age population across a set of demographic and geographic variables. For each panel identifying a risk factor, a highlighted line shows the fundamental factor with which the greatest amount risk might be presumed to occur. The specific proportions associated with each risk condition for the total school-age population are:

- At least one disability - 7.6%
- Retained in grade at least once - 8.1%
- Speaks English less than 'very well' - 4.9%
- Does not live with both parents - 30.8%
- Either parent emigrated in past 5 years - 2.3%
- Family income below $10,000 - 8.5%
- Neither parent/guardian employed - 10.5%

As can be seen, for any of the individual factors (except living with both parents), the overall risk is relatively low.

Across the various subgroups identified in the table there are several notable points of variation. By gender, for example, there are strong points of divergence in the factors of disability and retention, where males tend to have higher levels than females. But for the most part, there do not tend to be strong differences between boys and girls in the levels of risk they are experiencing.

For the race and ethnicity groups identified¹ there are more sizable differences. Black children, for example, have much higher levels of grade retention, low family income and likelihood of living with fewer than 2 parents - 63% of all Black children did not live with both of their parents. Both Hispanic and Asian and Pacific Islander children had high levels of speaking English less than "very well" and a parent who had emigrated within the past 5 years. Nearly a quarter of Hispanic children were reported to speak English less than "very well".

Dividing the child population into three similarly sized age groups (5-9, 10-13, 14-17), there are relatively few differences in the levels of risk. Grade retention, as one might expect intuitively, does rise with age - the longer one is in school, the greater the exposure to having failed a grade. More potentially problematic, however, is the large number of young children (5-9) who are reported to speak English less than "very well". These children may represent a subpopulation of special need that some schools are ill-prepared to deal with.

One aspect of variation across age that is also notable is the extent to which there is little or no variation of many risk factors across the age groups - in other words, for many of the factors, very young children have about the same level of risk as do children who are a decade older. It is not clear if this pattern foreshadows overall rises in the rates of risk that may be occurring over time, since we do not have longitudinal measures to see how much change occurred for the older children as they aged.

Variation across regions shows relatively little in the way of differences over the risk factors. Grade retention and low family income are somewhat more likely in the South than other regions, and the probability of speaking English less than "very well" is somewhat higher for children in the West, but there are not large systematic differences across regions.

Variation by area type, however, is very clear. For nearly every indicator of risk, children living in central cities have much higher levels of the risk factors than do children in either the central city balance ("suburbs") or in nonmetropolitan areas. This consistently high occurrence of risk factors for children in central cities may mean that their lives are far more volatile than their peers in other areas, and begins to point to the logical notion that many of these factors are not independent, but correlated with one another.

The table of individual risk factors help us to see that there are many different ways in which any child can be classified as "at risk", and choosing even one of these factors can result in demonstrating that many children have a potential problem. The consideration of all seven factors, however, allows us to answer two fundamentally important questions:

1. How many children experience at least ONE at-risk condition?
2. How many children experience MULTIPLE at-risk conditions?

Table 2 provides detailed information on the proportion of children who experience any, as well as multiple, at-risk conditions. The table shows these measures for the personal and familial factors alone and combined. Regarding the first question raised, while about 18% of all children have at least one personal risk factor, about 36% - twice as many - have at least one familial factor. When considering both types of factors 46% of all children - over 24 million, or close to one-half of the child population - report having at least one of these seven at-risk factors in their life.

Across the various subpopulations, there are some strong differences in the levels of risk children are exposed to. The most sizable of these differences would appear to be across race-ethnicity groupings. While only 35% of White children report at least one risk factor, 45% of Asian and Pacific Islander children, 62% of Hispanic children, and 72% of Black children, report at least one of the seven risk factors.

The other dimension which exhibits some strong variation in level is that of geographic area, where, following on the patterns observed in the analysis of individual risks, a far higher proportion of children living in central cities experience at least one risk factor (59%) than do those living in city-balance (39%) or nonmetropolitan areas (43%). This analysis will not attempt a decomposition of risk factors across the various subpopulation groups discussed, but, given that central city populations tend to be more highly comprised of minorities, it is quite possible that these higher risk levels in central cities simply reflect differential race-ethnicity composition, rather than some factor endemic to central city locales themselves. Nevertheless, the larger levels of children with at least some at-risk factors in these areas also likely means that these areas, their schools and other social service organizations, are being more heavily 'taxed' to accommodate the possible additional needs of these children.

We now turn to the second question asked earlier - that of multiple risks. While a large proportion of children do experience at least one of the seven at-risk factors, a sizable minority also experiences multiple factors. About 18% of all children are reported to experience more than one risk factor. As with the analysis of any single factor, examining the patterns across subgroups indicates that the greatest point of variation among children is with respect to the race-ethnicity groupings. Just 11% of White children were exposed to more than one risk factor. However, 18% of Asian and Pacific Islander children, 27% or Hispanic children, and 34% of Black children have more than one risk factor in their life. Examination of the regional data once again shows that over a quarter (27%) of the children living in central city areas have multiple risk factors, while children living in city-balance and nonmetro areas had substantially lower levels of multiple at-risk children.

One other demographic factor of note merits mention. In terms of both any and multiple factors of risk, there is very little difference in the levels across the three age groups. The fact that very young children are about as likely to have at least one risk factor as children nearly 10 years older on average (44% compared to 47%) is somewhat troubling on its own. However, these young children are also just as likely as the older group to have multiple risk factors (17% compared to 18%). While a single risk factor might be the result of a temporary circumstance, or addressed with a single remediation, children with multiple risk factors are likely to require more attention and need for intervention in order to address the variety of problems they may be encountering. The fact that the level of multiple risk exposure is just as great for very young as for older children, represents a possibly serious issue for child well-being.

While it is clear that many children experience more than one risk factor, it should also be apparent that this is because many risk factors are correlated with one another, at least to some degree. Table 3 looks at this phenomenon in a specific way, examining the number of total risk factors the individual has by the specific factors of risk.

The upper-left three cells of the table show the relationship between each of the three 'personal' at-risk factors and total at risk factors. As can be seen, persons who experience a disability or are retained are much more likely to experience multiple risk factors (26% and 28% respectively) than are persons who speak English less than "very well". Similarly, looking at the four right-most cells of the 'familial' risk factor row, it is clear that children en families with low income or neither parent working have very high rates of multiple risk (83% and 81%, respectively). Of course, these two factors themselves are likely to be highly correlated, thus creating at least some of this amplified affect.

When one examines the third panel of the table, the full range of both sets of factors become clear. Over 80% of children in low-income families or where no parent works also have at least one other risk factor, and 56% of the children in low-income families have three or more risk factors. As is often the case, these data speak strongly to the lasting negative impact of low income on children in ways that may not be quickly apparent as problems of family finances.

Interestingly, the single risk factor that is associated with the least additional risk is that of living in a household where both parents are not present. While 46% of the children who did not live with both parents had more than one risk factor, this is the lowest level associated with any of the seven factors examined.

This research has looked at a series of possible factors that could be used to define 'at-risk' conditions for school-age children in the United States. As has been demonstrated, using even a small set of seven indicators implies that a large proportion of children have at least one risk factor to deal with, and that many have multiple risk factors in their life.

But what does it mean to be "at risk"?

Concepts such as 'risk' and 'well-being' have, as part of their definition, an element of subjective evaluation that need to be kept cognizant in the minds of all who use them. What is perceived as 'risk' by one group or school of thought may be seen as 'opportunity' or 'challenge' by others. We have not begun to reach the point where we can conclusively say that a given factor does or does not constitute true risk to an individual's well-being or future. Virtually no factors assure, with certainty, that a future outcome is inevitable.

This proscription, however, will not keep researchers and policy analysts from continuing to try to determine how the population and its subgroups align along the various dimensions that attempt to define these concepts. Part of the research agenda for this field of study must focus on monitoring a wide variety of measures that tap into the multitude of dimensions of social, economic, psychic and whatever other dimensions of well-being we ultimately agree can and should be measured. Consistent measures, routinely applied, and integrated into larger research activities, will ultimately prove the value of each of these individual measures.

As this research proceeds, systematic patterns and findings will emerge. Ultimately, some factors will attain primacy, and others will fall to the wayside. Until then, the fact that we cannot say with certainty what risk is should not keep us from the certainty of trying to measure it.

Annie E. Casey Foundation. (2000). Kids Count Data Book 2000. Baltimore, MD.

Annie E. Casey Foundation. (1998). Kids Count Data Book 1998. Baltimore, MD

Barnett, Katherine P., Harvey F. Clarizio, and Karen A. Payette. (1996). Grade retention among students with learning disabilities. Psychology in the Schools, 33(4), 285-293.

Bryson, Ken. (1997). America's Children at Risk. Census Brief CENBR/97-2. U.S. Census Bureau. Washington D.C.

Clark, Reginald M. (1990). Why disadvantaged students succeed. Public Welfare, Spring, 17-23.

Kominski, Robert. (1989). "How Good is 'How Well'? An Examination of the Census English-Speaking Ability Question." Presented at the Annual Meeting of the American Statistical Association, Social Statistics Section. Washington, D.C.

McMillen, Marilyn, Phillip Kaufman, and Steve Klein. (1997). Dropout Rates in the United States: 1995, NCES 97-473. Washington, DC: U.S. Department of Education, National Center for Education Statistics.

Natriello, Gary, Edward L. McDill, and Aaron M. Pallas. (1990). Schooling Disadvantaged Children: Racing Against Catastrophe. New York: Teachers College Press.

Roderick, Melissa. (1995). Grade retention and school dropout: policy debate and research questions. Phi Delta Kappa Center for Evaluation, Development, and Research: Research Bulletin, 15.

Roderick, Melissa. (1994). Grade retention and school dropout: investigating the association. American Educational Research Journal, 31(4), 729-759.

1. In this analysis, four race-ethnicity groups are used. 'Whites' are defined as all White non-Hispanics, while persons of Hispanic origin are not taken from the counts of Black or Asian and Pacific Islanders. In the CPS, the vast majority (>95%) of all persons who are Hispanic ethnicity report their race as 'White'.