Journal of Adolescence 34 (2011) 95–103
Contents lists available at ScienceDirect
Journal of Adolescence journal homepage: www.elsevier.com/locate/jado
Factors that influence the accuracy of adolescent proxy reporting of parental characteristics: A research note Heather Ridolfo a, *, Aaron Maitland b a b
Department of Sociology, University of Maryland, 2112 Art-Sociology Bldg., College Park, MD 20742, United States Joint Program in Survey Methodology, University of Maryland, 1218 Le Frak Hall, College Park, MD 20742, United States
a b s t r a c t Keywords: Adolescents Proxy reporting Socioeconomic status
Socioeconomic status (SES) is considered to be an important marker of physical, mental, and social outcomes. However, methods used to gather socioeconomic information vary widely in terms of both the indicators of SES used (e.g., education, occupation, income, etc.) and data collection strategies. In particular, studies of adolescents often rely on children’s proxy reports of measures of parents’ SES. There is disagreement in the literature regarding children’s ability to reliably serve as proxy respondents for their parents. Using data from the National Longitudinal Study of Adolescent Health, a nationally representative sample of adolescents and their parents, we assess the accuracy of children’s proxy reports of mothers’ education and receipt of public assistance. Results show that accuracy of children’s proxy reports varies by children’s age, gender, and race; however, social class largely accounts for race differences found. Ó 2010 The Association for Professionals in Services for Adolescents. Published by Elsevier Ltd. All rights reserved.
Introduction Examining the socioeconomic status (SES) of families is a common practice in studies of adolescents, as it can be very influential for a number of factors such as physical and mental health, as well as social and economic outcomes. Researchers can use a variety of indicators to measure SES. They may also use a variety of methods to collect this information (e.g., proxy versus self-reports). In particular, many studies of adolescents rely on children’s proxy reports of parents’ status characteristics (Ensminger et al., 2000), which may have implications for the accuracy of these measures. Proxy reporting is a common method used in survey data collection to maximize response rates, and reduce costs and time involved in data collection (Blair, Menon, & Bickart, 1991; Moore, 1988; Tourangeau, Rips, & Rasinski, 2000). However, numerous studies have found proxy reports to be less accurate than self-reports (Blair et al., 1991; Looker, 1989; Moore, 1988). The use of children as proxy respondents for parents may be especially problematic as the cognitive skills necessary to comprehend and respond to survey questions continue to develop throughout adolescence (de Leeuw, Borges, & Smits, 2004; Scott, 1997). Additionally, children are less likely than adults to have access to information regarding adult statuses (Rosenberg & Pearlin,1978). Therefore, adolescents (particularly younger adolescents) may have difficulty answering questions designed to capture their parents’ SES. Previous research has also found the accuracy of children’s proxy reports to vary depending on social characteristics of the child (i.e., race and gender); however, findings have been inconclusive (Looker, 1989). In this paper we address some of the inconclusive findings regarding the accuracy of children’s proxy reporting for their parents. Using a nationally representative sample of adolescents, we examine the accuracy of adolescents’ reports of two
* Corresponding author. Tel.: þ1 301 405 6417. E-mail address:
[email protected] (H. Ridolfo). 0140-1971/$ – see front matter Ó 2010 The Association for Professionals in Services for Adolescents. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.adolescence.2010.01.008
96
H. Ridolfo, A. Maitland / Journal of Adolescence 34 (2011) 95–103
indicators of their mothers’ SES (education and receipt of public assistance), and whether the accuracy of adolescents’ reports varies by adolescents’ age, race, and gender. We also add to this literature by exploring other factors that may influence the accuracy of adolescents’ proxy reports. Review of the literature Children as proxy respondents Proxy reporting in survey data collection is a popular method for reducing costs, time necessary for data collection, and the burden on interviewers. However, proxy reports are often found to be less accurate than self-reports (Blair et al., 1991; Looker, 1989; Moore, 1988). Differences found in proxy and self-reports may be due to differences in the cognitive processes proxy and self-respondents engage in when answering survey questions (Blair et al., 1991; Groves, 1989; Lee, Mathiowetz, & Tourangeau, 2004). For example, proxy respondents may be less likely than self respondents to engage in extensive cognitive processing when constructing answers to survey questions (Lee et al., 2004) and often have to rely on the use of partial information when answering survey questions about others (Tourangeau et al., 2000). The use of children as proxy respondents for adults may be especially problematic. Children’s cognitive processes continue to develop throughout adolescence (Piaget, 1929). Because memory and processing speed are still developing in children younger than age 16, it may be inappropriate to ask them questions designed for adults. It is generally agreed upon in the literature that children as young as seven can participate in surveys (de Leeuw et al., 2004; Scott, 1997); however, it is not until children reach the age of 16 that their cognitive functioning is considered to be fully developed and they are viewed as capable of using adult questionnaires (de Leeuw et al., 2004; Scott, 1997). In fact, studies that examined the accuracy of proxy reporting by age (whether measured by age in years or school grade) found older adolescents provided more accurate reports of indicators of parents’ SES than did younger adolescents (Ensminger et al., 2000; Kerckhoff, Mason, & Poss, 1973; Mare & Mason, 1980; Mason, Hauser, Kerckhoff, Poss, & Manton, 1976), with the exception of one study that found no age differences (Lien, Friestad, & Klepp, 2001). Studies have also found proxy reports to be more accurate when the information being solicited was relevant to the proxy respondent themselves (Blair et al., 1991). One problem with asking children to report on indicators of parents’ SES is that young children and adolescents may not have a clear understanding of adult statuses (Rosenberg & Pearlin, 1978). In comparison to adults, children are more likely to spend the majority of their time in homogeneous environments, where indicators of stratification, such as education and income, have little salience (Rosenberg & Pearlin, 1978). Indeed, children’s ability to accurately report on indicators of parents’ SES has been found to vary by the status being reported. For example, in several studies children’s proxy reports were more accurate when reporting on parents’ occupation than parents’ education (Ensminger et al., 2000; Lien et al., 2001; Looker, 1989). Researchers believe this to be attributed to the higher salience of parents’ occupation to young children in comparison to other statuses such as education, which may not be salient until later years (Lien et al., 2001). As children age, however, they begin to move into more heterogeneous environments. By the time children reach high school they are more likely to be interacting with individuals of different social statuses on a regular basis, and as a result may become more aware of how their family’s SES diverges from others (Rosenberg & Pearlin, 1978), making these factors more salient. Additionally, as children age, parents may be more likely to engage in discussions with them about matters such as education and the family finances, therefore increasing children’s knowledge of parents’ education and income. Concern over children’s ability to answer questions regarding adult statuses, such as parents’ income, occupation, and education, has led some researchers to argue for the use of alternative measures of SES. For example, Wardle, Robb, and Johnson (2002) contend that children are better able to answer questions regarding material indicators of SES (e.g., ownership of home, car, computer, etc.), than questions regarding parents’ education and occupation. One study that examined agreement between parents’ and children’s reports of material indicators found fair to substantial agreement among the different alternative measures (Andersen et al., 2008). Another study, which explored the accuracy of children’s proxy reports using other alternative measures of SES, parents’ use of public assistance (i.e., welfare, food stamps, free/ reduced cost lunches), also found fair to substantial agreement between the child and the parent depending upon the specific measure examined and the age of the child (Ensminger et al., 2000). Gender and race differences – real or artificial findings? Some research also suggests that the accuracy of children’s proxy reports for their parents vary by the gender of the child; however, these findings are inconclusive. For example, in a study of sons’ and daughters’ retrospective reports of their fathers’ and mothers’ education and occupation, no gender differences were found for reports of fathers’ characteristics, however, sons provided less accurate reports of their mothers’ characteristics than did daughters (Corcoran, 1980). In another study, daughters were more accurate than sons when reporting for both mothers and fathers (West, Sweeting, & Speed, 2001), while other studies have found no gender differences in the accuracy of children’s proxy reports (Ensminger et al., 2000; Lien et al., 2001). While studies of gender differences in the accuracy of children’s proxy reports of parental characteristics have produced inconclusive findings, numerous studies have found race differences in the accuracy of children’s proxy reports. In two studies
H. Ridolfo, A. Maitland / Journal of Adolescence 34 (2011) 95–103
97
higher non-response rates were found among black adolescents in comparison to whites when reporting on indicators of parents’ SES (Kerckhoff et al., 1973; St. John, 1970); however, these differences were generally small and occurred within younger age groups. In other studies higher agreement was found among white adolescents and their parents in comparison to black adolescents and their parents (Cohen & Orum, 1972; Kerckhoff et al., 1973; Niemi, 1974). Additionally, in a review of the literature, Looker (1989) found white adolescents were more accurate when reporting fathers’ education than they are when reporting mothers’ education; however, the reverse was found to be true for blacks. While age differences in the accuracy of children’s proxy reports seems plausible given our knowledge of cognitive and social development, findings of gender and race differences in children’s proxy reporting are more puzzling. What other factors might explain these differences? Although no research to date has examined why gender differences are found in the accuracy of children’s proxy reporting, researchers speculate these differences may emerge because of differences in the amount of verbal contact between sons and daughters and their parents, and the differential effects mothers’ (and fathers) role modeling behavior may have on sons and daughters (Corcoran, 1980). In fact, the interpersonal relationship between the proxy respondent and the person for whom they are reporting is known to affect the accuracy of proxy reporting among adults (Groves, 1989). For example, studies have found that proxy respondents’ joint participation in activities and discussion of relevant issues with the person for whom they are reporting increases the accuracy of the proxy reports (Lien et al., 2001; Menon, Bickart, Sudman, & Blair, 1995). In terms of race differences in the accuracy of children’s proxy reports, other factors such as family structure and characteristics of the parent/family (i.e., social class and immigration status) may explain some of the differences found in previous studies. Children who grow up in single-parent households may have difficulty providing accurate proxy reports for parents who do not live with them (Looker, 1989). Because of the high proportion of both single female-headed and single male-headed households among blacks (Teachman, Tedrow, & Crowder, 2000), Hispanics, and American Indians (Schwede, 2007) in comparison to whites and Asians, children in these racial groups may be less accurate than their white and Asian peers when reporting on indicators of SES for their nonresident parents. Additionally, in a study of single and marriedmothers’ time use, Kendig and Bianchi (2008) found single mothers (and presumably single fathers) spend less time with their children than do married-mothers (and fathers); however, these differences diminished when SES is taken into account. The ability of children to provide accurate proxy reports for their parents may also be influenced by characteristics of the parents themselves. For example, in one study on children’s proxy reporting for their parents, non-response rates and misreporting was higher among children from lower SES groups in comparison to children from high SES groups (Wardle et al., 2002). The disproportionate rates of whites and Asians in higher SES groups compared to blacks, Hispanics, American Indians, and other races may account for the race differences found in the accuracy of children’s proxy reports (Bauman & Graf, 2003; Bishaw & Semega, 2008). Finally, with respect to Hispanic and Asian parents who might have immigrated to the United States, whether or not the parent was foreign born may play an important role. Foreign born parents may be more likely to complete education in a foreign school system in which the level of schooling does not directly coincide with the traditional levels of education in the United States, therefore making it difficult for their children to translate their parents’ education level to the US school system. Summary and research questions In sum, older children are better able to serve as proxy respondents for their parents than are younger children, due to advancements in cognitive and social development in later years of childhood. Additionally, some evidence suggests the accuracy of children’s proxy reporting can be improved through the use of alternative measures of SES, or measures that are more salient for the child. However, others studies have also found the accuracy of children’s proxy reports to vary by the gender and race of the child – findings that are less easily explained. Using a nationally representative sample we address two research questions: Does the accuracy of children’s proxy reporting of indicators of their mothers’ SES vary by the children’s age, gender, and race? If so, can we explain the gender and race differences found? Few studies of children’s proxy reports have examined reporting of mothers’ SES; however, as more women have entered the workforce and are the single heads of households, it is important to examine children’s reporting of their SES. We examine two indicators of mothers’ SES – a more traditional measure: mothers’ education, and an alternative measure: mothers’ receipt of public assistance. We hypothesize that gender differences in children’s proxy reports can be accounted for by examining the interpersonal relationship between children and their mothers. Children who have close relationships with their mothers should be better able to serve as proxy respondents for their mothers. Furthermore, the higher the mothers’ academic aspirations for their children, the more the mothers may discuss education with their children, and thus should lead to more accurate proxy reports of mothers’ education. Additionally, we hypothesize that race differences in children’s proxy reports can be explained by examining differences in family structure, social class and mothers’ nativity among the different groups. Data & methods The data used in this study are from Wave 1 of the National Longitudinal Study of Adolescent Health (Add Health). Add Health is a longitudinal study of health-related behaviors of adolescents and their outcomes in young adulthood. Data were
98
H. Ridolfo, A. Maitland / Journal of Adolescence 34 (2011) 95–103
first collected in 1994 using an in-school questionnaire, administered to a nationally representative sample of students in grades seven through 12. The adolescents were then surveyed a second time during Wave 1, using an in-home questionnaire approximately one year later. Parents of the adolescents were surveyed once, during the in-home portion of Wave 1. Mothers were the preferred respondents for the parent questionnaire. The original Wave 1 sample consisted of 20,745 adolescent respondents. During Wave 1 data collection, some respondents were selected outside of the sampling frame in order to gain a large enough sample to conduct genetic analyses. Because these respondents were selected outside of the sampling frame, sample weights could not be constructed for these individuals (Chantala & Tabor, 1999). Respondents who were missing weights were removed from our analytic sample, leaving a sample size of 18,924 adolescent respondents. The sample was then limited to adolescents who were living with their mothers and whose mothers also participated in the survey (n ¼ 11,988). Next, we removed respondents who were 19 years of age or older (n ¼ 11,808). Finally, respondents with incomplete or miscoded data were dropped from the analyses, leaving a sample of 11,333. Measures Dependent variables Two dependent variables were created based on concurrence between the child’s and mother’s reports of mother’s characteristics (see Appendix for information on measures). Concurrence between children’s and mothers’ reports of mothers’ education and concurrence between children’s and mothers’ reports of mothers’ receipt of public assistance were created based on whether or not the mothers’ and children’s reports of the characteristics concurred or not (1 ¼ concurrence, 0 ¼ divergence).1 Children who responded ‘‘don’t know’’ or refused to answer these questions were placed in the divergence group. Although it would be preferable to have an ‘‘external criterion’’ with which to compare both mother and child responses (Blair et al., 1991; Sudman & Bradburn, 1974), Add Health did not provide this type of measure. Therefore, we rely on mothers’ responses to be the ‘‘true’’ answer. Independent variables Age was measured using the child’s grade in school. Because age and grade in school (grades 7–12) were found to be highly correlated (r ¼ .927, p < .001), we chose to use grade in school over age in years because grade in school should be theoretically related to children’s thinking about parents’ education level. As children reach the end of high school they must begin to think about what they will do upon completion, and in doing so students may compare themselves and their life chances to that of their parents. Race of the child was measured using six dichotomous variables: whites, blacks, Hispanics, Asians, American Indians, and other/multi-racial. Gender of the child was a dichotomous variable (0 ¼ male, 1 ¼ female). Closeness to mother was measured using a scale of the strength and quality of the mother–child relationship. Responses to six questions regarding the child’s perception of their relationship with their mother were recoded so that higher scores indicated a closer relationship and then summed to create an index of closeness (alpha ¼ .849). Mothers’ academic aspiration for their children is a measure of the children’s perception of how disappointed their mothers would be if they did not graduate from high school. Family structure is a dichotomous measure of the children’s family structure, indicating whether they live in a two parent household or in a single-parent household. Adolescents were asked to report on all members of their household (0 ¼ single-parent household, 1 ¼ living with mother and father or mother and mother’s husband/partner). Family income was measured using mothers’ self reported family income. Imputation provided values for the missing cases for the income variable using the mean income for each level of mother’s education. The top five percent of reported family income for the sample was recoded to the 95th percentile. Mothers’ education was measured using mothers’ reported highest level of schooling achieved. Mothers’ nativity is a dichotomous measure indicating if the mother was born in the U.S. Control variables We also controlled for several factors that may influence the accuracy of adolescents’ proxy reports. First, cognitive ability of the adolescent was measured using the Add Health Picture Vocabulary Test, which is an abridged version of the Peabody Picture Vocabulary Test (See Cheng & Udry, 2003; Halpern, Joyner, Udry, & Suchindran, 2001). Social desirability is a measure used to capture adolescents’ who try to present themselves in a favorable light (Regnerus & Uecker, 2007). It is important to control for social desirability when asking about potentially sensitive topics such as education and use of public assistance as respondents may want to present themselves (and their parents) favorably (Blair et al., 1991; Crowne & Marlowe, 1960). In addition, research has found blacks and Hispanics to score higher on social desirability scales than do whites (Dudley, McFarland, Goodman, Hunt, & Sydell, 2005). Parental presence during the interview was controlled as the presence of parents during the adolescent in-home interview may lead to bias in the data, since parents can easily correct adolescents’ responses. Interviewers noted whether or not a mother and/or a father were present during the interview.
1
Additional information on constructed measures can be found in the Appendix.
H. Ridolfo, A. Maitland / Journal of Adolescence 34 (2011) 95–103
99
Table 1 Distribution of children’s reports of mothers’ education given mothers’ reported education (percentages) n ¼ 11,333. Mothers’ reports Children’s reports
Less than HS
High school
Some college
College
Professional
Refused
Less than HS High school Some college College Professional DK/refused
75.5 12.3 1.6 0.5 0.1 10.0
6.5 80.0 6.9 2.4 0.3 3.8
2.0 24.2 54.5 15.0 1.5 2.9
0.6 2.9 7.3 75.4 11.8 1.9
0.4 0.9 1.6 35.9 60.4 0.8
0.0 0.0 0.0 0.0 0.0 0.0
Analysis plan Crosstab analyses were used to examine concurrence between children’s and mothers’ reports of these indicators. Kappa statistics were calculated to determine the likelihood that concurrence between children and their mothers occurred by chance alone. The kappa statistic ranges from .00 (complete disagreement) to 1.00 (perfect agreement) (Cohen, 1960; Landis & Koch, 1977). In addition, logistic regressions were used to examine the relationship between children’s demographic characteristics (age, gender, and race), the closeness of the mother–child relationship, mothers’ aspirations for their children,2 social class, family structure, mothers’ nativity and control variables on the accuracy of children’s proxy reports of mothers’ education and receipt of public assistance. Coefficients in the regression analysis were weighted for differential probabilities of selection and poststratification. Results Proxy reports of mothers’ education When we compared the average years of education held by mothers in this sample as reported by children (M ¼ 2.7) and mothers (M ¼ 2.7) there appears to be no difference between the two reports. Additionally, variation in the distribution of children’s and mothers’ reports of mothers’ education was minimal, with the exception of when mothers had completed some college. Thirty percent of mothers reported having some college education, whereas 20 percent of children reported their mothers had completed some college. We explored the distribution of children’s and mothers’ reports of mothers’ education further using crosstab analyses. Children’s reports of mothers’ education were found to be consistent with mothers’ report of their own education only about 69 percent of the time (kappa ¼ .598). Table 1 shows the distribution of children’s reports of mothers’ education given mothers’ report of their education. Concurrence between children and mothers appears to be higher when mothers have a high school degree (80 percent concurrence) or a college degree (75 percent concurrence) and concurrence appears to be lower when mothers have only some college experience (55 percent concurrence) and when mothers held a professional degree (60 percent concurrence). In addition, we examined how concurrence between children’s and mothers’ reports of mothers’ education varied by children’s age, gender, and race (see Table 2). Consistent with previous research, concurrence between children’s proxy reports and their mothers’ reports generally increased with age (each additional year in school). Concurrence was higher among daughters. Daughters’ reports of their mothers’ education were concurrent with their mothers 71 percent of the time, whereas sons’ reports of mothers’ education were concurrent with mothers’ reports 67 percent of the time. Concurrence was found to be greatest among whites (72 percent concurrence), followed by Hispanics (68 percent concurrence). For all other races, concurrence occurred 66 percent of the time or less. We were able to test these findings further with a regression analysis. This analysis also allowed us to explore why these gender and race differences in proxy reports of mothers’ education might be occurring. Results from the logistic regression analyses are presented in Table 3. The results from the first model were consistent with the crosstab analyses – for each additional year in school, children’s reports were 1.056 times more likely to concur with their mothers’ report of their education (p < .01). Girls’ reports were 1.238 times more likely to concur with their mothers’ reports than were boys’ reports (p < .01). When reporting on mothers’ education, concurrence was significantly higher among whites than blacks (OR ¼ .788, p < .01), Asians (OR ¼ .547, p < .01), and other/multi-racial respondents (OR ¼ .710, p < .05). No differences were found between whites and Hispanics or whites and American Indians. The second model introduced two indicators of interpersonal factors into the analyses. One indicator – children’s closeness to their mother – was found to be significantly associated with concurrence between children’s and mothers’ reports of mothers’ education (OR ¼ 1.018, p < .05). Mothers’ academic aspirations for their children were not significantly associated with concurrence between reports. The introduction of interpersonal factors into the model had little effect on the relationship between age, gender, and race of the children and concurrence between reports.
2
This variable was only examined in the analyses which looked at accuracy of proxy reporting on mothers’ education.
100
H. Ridolfo, A. Maitland / Journal of Adolescence 34 (2011) 95–103
Table 2 Concurrence of children’s and mothers’ report of mothers’ education and receipt of public assistance by children’s grade in school, gender, and race (percentages). Concurrence on education
Concurrence on public assistance
Grade in school 7 8 9 10 11 12
66 69 68 70 71 70
92 93 94 94 95 97
Gender Girls Boys
71 67
94 94
Race Whites Blacks Hispanics Asians American Indians Other/multi-racial
72 65 68 64 60 66
96 91 92 95 87 92
The third model introduced family structure, social class, and mothers’ nativity into the analysis. These variables were hypothesized to account for some of the race differences found in the initial model. While no effects of family structure or mothers’ nativity were found, social class was significantly related to concurrence of reports. The higher the mothers’ education the less likely children’s and mothers’ reports concurred (OR ¼ .713, p < .001), but the higher the family income the more likely children’s and mothers’ reports concurred (OR ¼ 1.003, p < .05). The introduction of these variables did account for some of the race differences found, with the differences between whites and blacks and whites and Asians remaining significant, although these differences decreased substantially. In fact, the upper limits of the confidence intervals on the odds ratios for all races in the model are now nearly one. This indicates an extremely small remaining race effect. The age and gender differences persisted. Proxy reports of mothers’ receipt of public assistance Children’s and mothers’ reports of mothers’ receipt of public assistance concurred substantially more than reports of mothers’ education (kappa ¼ .709). Children’s reports of mothers’ receipt of public assistance were found to be consistent with mothers’ reports about 94 percent of the time. When the mother does not receive public assistance children’s reports are consistent with their mothers about 97 percent of the time; however, when mothers do receive public assistance reports are consistent about 78 percent of the time.
Table 3 Results of logistic regression predicting the log odds of children providing reports of mothers’ education that are concurrent with their mothers’ own reports (n ¼ 11,333). Model 1
Model 2
Model 3
OR
C.I.
OR
C.I.
OR
C.I.
Age Female Black Hispanic Asian American Indian Other/multi-racial Closeness to mother Mothers’ aspirations for their children Family structure Mothers’ education Family income Mothers’ nativity
1.056** 1.238** 0.788** 0.842 0.547** 1.000 0.710*
(1.018–1.095) (1.093–1.402) (0.671–0.925) (0.704–1.009) (0.383–0.782) (0.461–2.178) (0.548–0.920)
1.061** 1.253*** 0.786** 0.846 0.548** 1.005 0.714* 1.017* 1.044
(1.023–1.102) (1.107–1.419) (0.670–0.924) (0.707–1.012) (0.383–0.785) (0.464–2.177) (0.549–0.928) (1.002–1.032) (0.980–1.112)
1.060** 1.271*** 0.840* 0.785* 0.654* 0.990 0.713* 1.019* 1.072*
(1.021–1.100) (1.119–1.443) (0.713–0.989) (0.652–0.946) (0.431–0.992) (0.467–2.097) (0.538–0.943) (1.003–1.034) (1.007–1.142)
1.102 0.713*** 1.003* 1.148
(0.960–1.264) (0.670–0.758) (1.000–1.006) (0.909–1.450)
Wald statistic
9.11 df ¼ (10,119)
8.25 df ¼ (12,117)
Note: Cognitive ability, social desirability and parental presence during the interview are controlled for in each model. *p < .05; **p < .01; ***p < .001.
12.51 df ¼ (16,113)
H. Ridolfo, A. Maitland / Journal of Adolescence 34 (2011) 95–103
101
Crosstabulations were used to examine the relationship between children’s age, gender, and race and concurrence with mothers’ reports of receipt of public assistance (Table 2). Concurrence among reports increased with each additional year of schooling, with children at the 12th grade level being most accurate. No differences were found between sons and daughters, and differences between racial groups were fairly small. Results from the regression analysis are presented in Table 4. As shown in Model 1, for each additional year in school, children’s reports were 1.179 times more likely to concur with their mothers’ reports of receipt of public assistance (p < .001). Gender was not found to be significantly related to the likelihood of concurrence between children and their mothers when reporting mothers’ receipt of public assistance. Black (OR ¼ .509, p < .001), Hispanic (OR ¼ .587, p < .01), American Indian (OR ¼ .331, p < .01), and other/multi-racial (OR ¼ .434, p < .001) children’s reports were significantly less likely than white children to provide reports that concurred with their mothers’ reports. The second model introduced one indicator of the interpersonal relationship between children and their mothers – the closeness of the mother–child relationship – and this was not found to be significantly associated with concurrence between reports. The third model introduced family structure and social class into the analysis. While no effect of family structure was found, social class was significantly related to concurrence between reports. The higher the family income the more likely children’s and mothers’ reports concurred (OR ¼ 1.032, p < .001). The addition of these variables eliminated the race differences found between whites and all other race groups, with the exception of other/multi-racial children (OR ¼ .486, p < .01). However, age differences persisted (OR ¼ 1.153, p < .001). Discussion The findings from this research indicate that adolescents are more accurate when reporting on their mothers’ receipt of public assistance than when reporting on their mothers’ education, and that some groups of children are more capable of providing accurate proxy reports for their mothers than are others. This study expands on the current literature regarding children’s proxy reporting in a number of ways. First, using a nationally representative sample we were able to address inconsistencies found in previous studies of proxy reporting that relied on smaller non-representative samples. The large sample allowed us to examine differences in proxy reports across multiple races, and demonstrated that differences in accuracy of reports are not limited to differences between whites and blacks only. Furthermore, we add to this literature by examining factors that contribute to gender and race differences found in the accuracy of children’s proxy reports. As in previous research we found gender and race differences in adolescents’ proxy reports of parental characteristics; however, our analysis examined why these differences are found. In terms of gender differences, previous research speculates that adolescents may be better able to report for their same-sex parent because of differences in verbal contact between sons and daughters and their parents, as well as differences in mothers’ and fathers’ role modeling behavior for their sons and daughters (Corcoran, 1980). In the current study gender differences were only found in children’s proxy reports of mothers’ education. While the strength of the interpersonal relationship between children and their mothers was significantly related to the accuracy of children’s proxy reports of mothers’ education, it did not matter for reports of mothers’ receipt of public assistance. Furthermore, it did not explain gender differences found in proxy reports of mothers’ education. Researchers have speculated that race differences may exist due to differences in family structure and social class (Looker, 1989; Wardle et al., 2002). While family structure was found to have no effect on the accuracy of children’s proxy reports, social class almost entirely explained the race differences found for both reports of mothers’ education and mothers’ receipt of public assistance. This study is limited in that it does not account for diversity within gender and race groups that may affect these groups’ ability to serve as reliable proxy respondents for their parents. Future research should consider how the accuracy of children’s Table 4 Results of logistic regression predicting the log odds of children providing reports of mothers’ receipt of public assistance that are concurrent with their mothers’ own reports (n ¼ 11,333). Model 1
Model 2
Model 3
OR
C.I.
OR
C.I.
OR
C.I.
Age Female Black Hispanic Asian American Indian Other/multi-racial Closeness to mother Family structure Mother’s education Family income
1.179*** 1.225 0.509*** 0.587** 0.603 0.331** 0.434***
(1.088–1.278) (1.000–1.499) (0.356–0.729) (0.411–0.837) (0.272–1.134) (0.168–0.653) (0.279–0.675)
1.184*** 1.234* 0.508*** 0.588** 0.604 0.330** 0.434*** 1.012
(1.093–1.282) (1.004–1.517) (0.355–0.726) (0.412–0.838) (0.273–1.340) (0.167–0.651) (0.279–0.676) (0.983–1.042)
1.153*** 1.226 0.752 0.813 0.574 0.581 0.486** 1.001 1.253 1.111 1.032***
(1.069–1.244) (0.998–1.507) (0.523–1.081) (0.562–1.178) (0.241–1.366) (0.298–1.132) (0.307–0.768) (0.973–1.030) (0.947–1.658) (0.984–1.255) (1.023–1.042)
Wald statistic
16.09 df ¼ (10, 119)
14.78 df ¼ (11, 118)
Note: Cognitive ability, social desirability and parental presence during the interview are controlled for in each model. *p < .05; **p < .01; ***p < .001.
21.43 df ¼ (14, 115)
102
H. Ridolfo, A. Maitland / Journal of Adolescence 34 (2011) 95–103
proxy reports is related to diverse characteristics within gender and racial groups. Also, because of a lack of an external criterion with which to compare mothers’ reports, this study makes the assumption that mothers’ reports of their education and receipt of public assistance are accurate. Convergence of responses alone does not necessarily mean that either mother or child is reporting accurately, as both the mother and child may have modified their response in order to present themselves in a more favorable light (Blair et al., 1991; Looker, 1989). Without objective measures of mothers’ education and receipt of public assistance, we are unable to know whether or not mothers’ or children’s responses are accurate. Additionally, more research is needed to identify whether the use of children’s proxy reports leads to any measurable bias in multivariate models (e.g., do children’s reports of their mothers’ education correlate with health measures differently than mothers’ own reports would?). Acquiring accurate proxy reports is in many cases important to attaining quality data, as well as being able to make sound claims when using measures of SES as predictor variables. This study demonstrated that the accuracy of children’s proxy reports of parental characteristics vary depending on the measures. Furthermore, age, gender, and race differences in the ability of children to provide accurate proxy reports are found; however, race differences can be eliminated or substantially decreased by controlling for social class in the analysis. Based on these findings, we would urge caution when using children’s proxy reports of indicators of parents’ SES. Researchers should consider controlling for the factors we’ve identified in this study that are significantly related to accuracy of proxy reports. While researchers may have to rely on children’s proxy reporting when data from the parent is missing, knowledge gained from this study and previous research should guide researchers when deciding if proxy reports can serve as valid substitutes for self-reports. Acknowledgements This research uses data from Add Health, a program project designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris, and funded by a grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, with cooperative funding from 23 other federal agencies and foundations. Special acknowledgment is due Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Persons interested in obtaining data files from Add Health should contact Add Health, Carolina Population Center, 123 W. Franklin Street, Chapel Hill, NC 27516-2524 (
[email protected]). No direct support was received from grant P01-HD31921 for this analysis. We would like to thank Brian Ward, Melissa Milkie, Stanley Presser, Steven Martin and Megan Henly for their help on this paper. Appendix. Constructed measures.
Label
Measures
Concurrence on mothers’ education
Mother: How far did you go in school? (1 ¼ less than a high school education 2 ¼ high school Child: How far in school did she go? (1 ¼ less than a high school education 2 ¼ high school graduate/GED or vocational school graduate, 3 ¼ vocational school beyond high school/some college, 4 ¼ college degree, 5 ¼ professional degree, 6 ¼ Don’t know/refused) (1 ¼ concurrence, 0 ¼ divergence)
Concurrence on mothers’ receipt of public assistance
Mother: Are you receiving public assistance, such as welfare? (0 ¼ no, 1 ¼ yes) Child: Does she receive public assistance, such as welfare? (0 ¼ no, 1 ¼ yes, 8 ¼ DK) (1 ¼ concurrence, 0 ¼ divergence)
Closeness to mother
How close do you feel to your mother? How much do you think she cares about you? (1 ¼ not at all, 5 ¼ very much) Most of the time, your mother is warm and loving toward you. When you do something wrong that is important, your mother talks about it with you and helps you understand why it is wrong. You are satisfied with the way your mother and you communicate with each other. Overall, you are satisfied with your relationship with your mother. (1 ¼ strongly agree, 5 ¼ strongly disagree
Family income
About how much total income, before taxes did your family receive in 1994? Include your own income, the income of everyone else in your household, and income from welfare benefits, dividends, and all other sources. $0–$100,000
Mothers’ education
Mother: How far did you go in school? (1 ¼ less than a high school education 2 ¼ high school graduate/GED or vocational school graduate, 3 ¼ vocational school beyond high school/some college, 4 ¼ college degree, 5 ¼ professional degree)
Social desirability
You never argue with anyone (1 ¼ strongly disagree, 5 ¼ strongly agree) You never get sad You never criticize other people For each item a dichotomous variable was created (1 ¼ strongly agree, 0 ¼ all else) We then summed these variables (0 ¼ low social desirability, 3 ¼ high social desirability)
H. Ridolfo, A. Maitland / Journal of Adolescence 34 (2011) 95–103
103
References Andersen, A., Krolner, R., Currie, C., Dallago, L., Due, P., Richter, M., et al. (2008). High agreement on family affluence between children’s and parent’s reports: international study of 11-year-old children. Journal of Epidemiology and Community Health, 62, 1092–1094. Bauman, K. J., & Graf, N. L. (2003). Educational attainment: 2000. Census 2000 Brief. Washington, DC: U.S.: Census Bureau. Bishaw, A., & Semega, J. (2008). Income, earnings, and poverty data from the 2007 American community survey. American Community Survey Reports. Washington, DC: U.S.: Census Bureau. Blair, J., Menon, G., & Bickart, B. (1991). Measurement effects in self vs. proxy responses to survey questions: an information-processing perspective. In P. P. Biemer, R. M. Groves, L. E. Lyberg, N. A. Mathiowetz, & S. Sudman (Eds.), Measurement errors in surveys (pp. 145–166). Hoboken, New Jersey: John Wiley & Sons, Inc. Chantala, K., & Tabor, J. (1999). Strategies to perform a design-based analysis using the add health data. Chapel Hill: Carolina Population Center, University of North Carolina. Cheng, M. M., & Udry, J. R. (2003). How much do mentally disabled adolescents know about sex and birth control? Adolescent and Family Health, 3, 28–38. Cohen, J. (1960). Coefficient of agreement for nominal scales. Educational and Psychological Measurement, 20, 37–46. Cohen, R. S., & Orum, A. M. (1972). Parent-child consensus on socioeconomic data obtained from sample surveys. Public Opinion Quarterly, 36, 95–98. Corcoran, M. (1980). Sex differences in measurement error in status attainment models. Sociological Methods & Research, 9, 199–217. Crowne, D. P., & Marlowe, D. (1960). A new scale of social desirability independent of psychopathology. Journal of Consulting Psychology, 24, 349–354. Dudley, N. M., McFarland, L. A., Goodman, S. A., Hunt, S. T., & Sydell, E. J. (2005). Racial differences in socially desirable responding in selection contexts: magnitude and consequences. Journal of Personality Assessment, 85, 50–64. Ensminger, M. E., Forrest, C. B., Riley, A. W., Kang, M., Green, B. F., Starfield, B., et al. (2000). The validity of measures of socioeconomic status of adolescents. Journal of Adolescent Research, 15, 392–419. Groves, R. M. (1989). Survey errors and survey costs. New York: John Wiley and Sons. Halpern, C. T., Joyner, K., Udry, J. R., & Suchindran, C. (2001). Smart teens don’t have sex (or kiss much either). Journal of Adolescent Health, 26, 213–225. Kendig, S. M., & Bianchi, S. M. (2008). Single, cohabiting, and married mothers’ time with children. Journal of Marriage and Family, 70, 1228–1240. Kerckhoff, A. C., Mason, W. M., & Sandomirsky Poss, S. (1973). On the accuracy of children’s reports of family social status. Sociology of Education, 46, 219–247. Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical data. Biometrics, 33, 159–174. Lee, S., Mathiowetz, N. A., & Tourangeau, R. (2004). Perceptions of disability: the effect of self- and proxy response. Journal of Official Statistics, 20, 671–686. de Leeuw, E., Borgers, N., & Smits, A. (2004). Pre-testing questionnaires for children and adolescents. In S. Presser, J. M. Rothgeb, M. P. Couper, J. T. Lessler, E. Martin, J. Martin, & E. Singer (Eds.), Methods for testing and evaluating survey questionnaires (pp. 409–429). Hoboken, New Jersey: John Wiley & Sons, Inc. Lien, N. C., Friestad, C., & Klepp, K.-L. (2001). Adolescents’ proxy reports of parents’ socioeconomic status: how valid are they? Journal of Epidemiological Community Health, 55, 731–737. Looker, D. E. (1989). Accuracy of proxy reports of parental status characteristics. Sociology of Education, 62, 257–276. Mare, R. D., & Mason, W. M. (1980). Children’s reports of parental socioeconomic status: a multiple group measurement model. Sociological Methods & Research, 9, 178–198. Mason, W. M., Hauser, R. M., Kerckhoff, A. C., Poss, S. S., & Manton, K. (1976). Models of response error in student reports of parental socioeconomic characteristics. In W. H. Sewell, R. M. Hauser, & D. L. Featherman (Eds.), Schooling and achievement in American society (pp. 443–494). New York: Academic Press. Menon, G., Bickart, B., Sudman, S., & Blair, J. (1995). How well do you know your partner? Strategies for formulating proxy-reports and their effects on convergence to self-reports. Journal of Marketing Research, 32, 75–84. Moore, J. (1988). Self/proxy response status and survey response quality. A review of the literature. Journal of Official Statistics, 4, 155–172. Niemi, R. (1974). How family members perceive each other: Political and social attitudes in two generations. New Haven, CT: Yale University Press. Piaget, J. (1929). The child’s conception of the world. London: Routledge & Kegan Paul. Regnerus, M. D., & Uecker, J. E. (2007). Religious influences on sensitive self-reported behaviors: the product of social desirability, deceit, or embarrassment? Sociology of Religion, 68, 145–163. Rosenberg, M., & Pearlin, L. I. (1978). Social class and self-esteem among children and adults. The American Journal of Sociology, 84, 53–77. Schwede, L. (2007). A new focus: Studying linkages among household structure, race ethnicity, and geographical levels with implications for census coverage. Statistical Research Division Research Report Series (Survey Methodology #2007-38). Washington, DC: U.S. Census Bureau. Scott, J. (1997). Children as respondents: methods for improving data quality. In I. E. Lyberg, P. Biemer, M. Collins, E. De Leeuw, C. Dippo, N. Schwarz, & D. Treman (Eds.), Survey measurement and process quality. New York: John Wiley & Sons. St. John, N. (1970). The validity of children’s reports of their parents’ education level: a methodological note. Sociology of Education, 43, 255–269. Sudman, S., & Bradburn, N. M. (1974). Response effects in surveys: A review and synthesis. Chicago: Aldine. Teachman, J. D., Tedrow, L. M., & Crowder, K. D. (2000). The changing demography of America’s families. Journal of Marriage and the Family, 62, 1234–1246. Tourangeau, R., Rips, L. J., & Rasinski, K. (2000). The psychology of survey response. New York: Cambridge University Press. Wardle, J., Robb, K., & Johnson, F. (2002). Assessing socioeconomic status in adolescents: the validity of a home affluence scale. Journal of Epidemiology and Community Health, 56, 595–599. West, P., Sweeting, H., & Speed, E. (2001). We really do know what you do: a comparison of reports from 11 year olds and their parents in respect of parental economic activity and occupation. Sociology, 35, 539–559.