<<

What Money Doesn’t Buy 479

What Money Doesn’t Buy What Money Doesn’t Buy: Class Resources and Children’s Participation in Organized Extracurricular Activities

Elliot B. Weininger, SUNY College at Brockport Annette Lareau, University of Pennsylvania Dalton Conley, New York University

ecent research suggests that participation in organized extracurricular activities by children and adolescents can have educational and occupational payoffs. This Rresearch also establishes that participation is strongly associated with social class. However, debate has ensued—primarily among qualitative researchers—over whether the association between class and activities stems exclusively from inequali- ties in objective resources and constraints or whether differing cultural orientations have a role. We address this debate using a nationally representative sample of chil- dren’s time diaries, merged with extensive information on their , to model par- ticipation in, and expenditures on, organized activities. While we cannot directly observe cultural orientations, we account for a substantially wider array of resources and constraints than previous studies. We find that, above and beyond these factors, maternal education has a consistently large effect on the outcomes we study. We ­discuss the plausibility of a cultural interpretation of this result, as well as alternative interpretations.

Introduction In the forty years since the publication of Inequality (Jencks 1972), the question of how social position influences children’s life experiences has occupied a great deal of sociologists’ attention. While scholars have focused for decades on issues such as inequality in family resources and school characteristics, in recent years interest has increased in the ways children spend their out-of-school time, and in particular, their participation in organized extracurricular activities. For example, some researchers have argued that organized activity participation

The authors gratefully acknowledge the generous support of the Spencer Foundation as well as the University of Pennsylvania. They appreciate the assistance of Melissa Velez. The authors are grateful to the anonymous reviewers and the editor for their helpful comments. Thanks to Josh Klugman, Maia Cucchiara, and Judith Levine for comments on earlier versions of this paper. All errors, however, are the sole responsibility of the authors.

© The Author 2015. Published by Oxford University Press on behalf of the Social Forces 94(2) 479–503, December 2015 University of North Carolina at Chapel Hill. All rights reserved. For permissions, doi: 10.1093/sf/sov071 please e-mail: [email protected]. Advance Access publication on 28 May 2015 480 Social Forces 94(2)

serves as an important selection factor in the recruitment of students to presti- gious colleges and universities (Kaufman and Gabler 2004; Stevens 2007) and in the hiring of candidates in certain elite professions (Rivera 2011). Additionally, numerous researchers have argued that participation in organized activities pro- vides children and adolescents with diffuse but important cognitive and non- cognitive skills that subsequently yield both educational and occupational payoffs (Bodovski and Farkas 2008; Covay and Carbonaro 2010; Lareau 2011). Others have asserted that participation positively affects psychosocial development (Linver,­ Roth, and Brooks-Gunn 2009). Not only is there evidence suggesting that extracurricular activities may be important in producing class outcomes, participation in such activity has been shown to be stratified by SES itself. Numerous studies of nationally representa- tive data have reported that participation rates rise sharply with their families’ social class or SES (Dumais 2006; Lareau and Weininger 2008; Covay and ­Carbonaro 2010). Meanwhile, qualitative scholars have debated the underlying mechanisms that produce this class gradient in extracurricular activities. On one side, some suggest that, among , the relation between social class and particular childrearing practices—including the propensity to enroll children in organized activities—is mediated by distinctive cultural repertoires (Lareau 2011). Others, however, challenge this assertion (Chin and Phillips 2004; Bennett,­ Lutz, and Jayaram 2012). They argue that family differences in material resources and objective circumstances produce these distinctive childrearing patterns. In this view, cultural orientations are of relatively little significance, and equalization of resources and circumstances would be sufficient to substantially diminish or eliminate differences in participation. Seen in this light, this debate belongs to the long-standing sociological inquiry into the mechanisms through which inequality is sustained in daily life, and to the more specific question of whether and how culture may a constitutive role in social stratification (Mayer 1998; see also Lamont, Small, and Harding 2010). In this paper, we seek to contribute to this debate over the factors structuring activity participation by analyzing a nationally representative data set containing detailed child-level time diaries and family-level questionnaire information. In addition to examining various measures of participation, we also address the previously unstudied issue of expenditures on organized activities. Taking up arguments made in the ethnographic literature, we model the effects of a much wider array of factors than previous quantitative work in the field, including numerous aspects of parents’ work and labor-market situations. Furthermore, we assess an argument stemming from this literature that asserts that “leveling” ­institutions—specifically, schools—equalize participation. To preview our results, we do not find support for the contention that class differences in activity participation or expenditures are solely the result of varia- tion in parents’ resources and the constraints they face. Instead, our analyses show that maternal education—which we view as a stand-in for cultural ­orientation—has large effects on both expenditures on and participation in orga- nized activities, even in the presence of extensive controls for resources and con- straints; moreover, the effects on participation are consistent across institutional settings. Of course, just as material circumstances could be proxies for cultural What Money Doesn't Buy 481 frames, maternal education could itself be a stand-in for some other unobserved factor like innate skill; however, if this were the case, our central point would still hold: the observed association is not reducible to material situations. In the con- clusion, we consider the mechanisms that might drive these effects. As we imply above, varying cultural dispositions are a plausible candidate, though not the only one.

Class, Constraints, and Culture A significant body of research has developed in recent years regarding the conse- quences of children’s participation in extracurricular activities. For example, researchers have suggested that—above and beyond academic considerations— undertaking particular organized activities directly influences one’s chances of admission to a prestigious college or university (Kaufman and Gabler 2004; see also Dumais 2009). Another recent study asserts that intensive participation in various activities constitutes an important informal hiring criterion for entry- level positions in elite law firms, investment banks, and management consulting firms (Rivera 2011). In addition to acting as a screening mechanism, activities participation may also provide children with skills and dispositions that will serve as a source of advantages later in life. For example, Lareau asserts that activities such as membership in a choir or on a sports team, music lessons, and scouts instill in young children “the ability to perform in public, in front of adults, including strangers” and to “work smoothly with acquaintances”; consequently, many organized activities reproduce, “in their organizational style … key aspects of the workplace” (2011, 61–63). For this reason, Lareau describes organized activities as a kind of “pre-employment training” (2011, 62), and views them as a source of “cultural capital” (see also Adler and Adler 1994). Other researchers have suggested that activity participation is predictive of educational outcomes such as standardized test performance and grades (Bodovski and Farkas 2008; Cheadle 2008). To be sure, absent randomized controlled interventions or natu- rally occurring experiments, we cannot know for sure whether some unobserved factor—like “grit”—is driving both activity participation and positive school ­outcomes. Nonetheless, the findings concerning the consequences of activity participation are suggestive enough that vigorous debates have arisen concerning the socio- logical antecedents of children’s participation. On one side, Lareau (2002, 2011) has argued that the relation between social structure and various childrearing behaviors (including enrollment of children in extracurricular activities) is medi- ated by a set of class-specific cultural orientations that confer distinct meanings on these behaviors. The relevant orientations correspond to families’ social class locations, but are not directly deducible from the constitutive characteristics of these ­locations—that is, from the attributes of their jobs or variation in their incomes. What Lareau terms “”—typical of middle-class parents in her data but rare among working-class or poor ones—views extracur- ricular activities as a means of cultivating children’s potential talents and skills. Working-class and poor parents, according to Lareau, tend to view childrearing through a distinct cultural lens that she terms the “accomplishment of natural 482 Social Forces 94(2)

growth.” The fundamental assumption is that children must be loved, cared for, and protected—but­ that otherwise, they will grow and thrive spontaneously. Thus, children are granted control over their “free” time, which is typically spent in informal play and socializing. This argument has been criticized. Also drawing on qualitative data, Chin and Phillips (2004) argue that variations in material resources and objective circum- stances are more important in explaining the association between class and activ- ity participation than differences in “ philosophies or values.”1 In particular, Chin and Phillips note that organized extracurricular activities often require significant financial resources to cover direct and indirect costs, thereby advantaging children from affluent families. Participation also demands significant availability of parental time and scheduling flexibility. They write that “even among parents who work the same number of hours per week, those who have professional jobs that allow them to set their own schedules and those who work nonoverlapping shifts are better able than are others to provide their children with varied experiences” (Chin and Phillips 2004, 187; see also Lareau and Weininger 2008). Thus, Chin and Phillips suggest that variations in economic resources, time, and work flexibility will account for much or all of the association between social class and activity participation. Bennett, Lutz, and Jayaram (2012) have developed a similar argument. Using a set of in-depth interviews, they contend that financial constraints, in the first instance, and the near-absence of neighborhood-based opportunities for partici- pation, in the second, hamper the desires of working-class parents to enroll their children in extracurricular activities. In making this argument, they assert the importance of various “objective” constraints by documenting the relative lack of class differences in activity participation that they observe in institutional set- tings that seek to programmatically ameliorate the impact of such constraints—in particular, schools, but also churches. Thus, they report that the working-class and middle-class adolescents in their data exhibit remarkably similar in-school activity profiles, but starkly different out-of-school ones. Researchers critical of Lareau’s argument therefore assert a general continuity in parents’ cultural orientations across social classes, with differences in resources and constraints creating sharp divergences in their ability to realize the goals that stem from these orientations, absent any “leveling” effect from equalizing institutions such as schools. Indeed, they claim that the impact of resources and constraints is particularly powerful, since working-class parents have additional motives for enrolling their children in organized activities—namely, protection. Thus, Bennett, Lutz, and Jayaram report that “working-class parents [but not middle-class par- ents] cited keeping children safe and away from trouble as an important reason for their children’s participation in structured activities” (2012, 142). In sum, there is an important debate in the literature regarding the role of ­cultural and economic factors in shaping children’s childrearing experiences. Of course, childrearing behaviors may stem from multiple influences, both cul- tural and material (e.g., Lareau 2011, 248ff). The debate, therefore, should be construed as one of emphasis. Within these parameters, however, it is an important one. We seek to advance this debate by carrying out a quantitative analysis of a nationally representative data set. This marks a departure from much of the What Money Doesn't Buy 483

­literature, which has relied on small-N, non-random samples. To be sure, the type of cultural dispositions invoked by Lareau to explain class differences in activity participation are not directly “observable” in secondary data. However, the data set we use—the Supplement (CDS) to the Panel Study of Income Dynamics (PSID)—does enable us to model the various constraints identi- fied in the qualitative literature, and to examine the contribution that each makes to the stratification of participation, while simultaneously asking what additional factors may play a role. Thus, we are able to undertake a thorough examination of the hypothesis that the class gap in organized activity involvement can largely be accounted for by economic factors and “objective” ­constraints. We also seek to extend the quantitative literature on activity participation (Dumais 2006; Cheadle 2008; Covay and Carbonaro 2010), where controls for “objective” constraints (beyond income) tend to be sparse. In contrast, our anal- yses account for a wide range of factors—pertaining to things such as parental work situations, labor-market experiences, and neighborhood conditions—that could plausibly be hypothesized to affect children’s participation in extracurricu- lar activities. Furthermore, in contrast to much of the quantitative literature— which measures organized activity participation through simple yes/no survey questions posed to children or one of their parents—we make use of a data set that contains time diaries for a representative sample of American children. As Hofferth (1999) has convincingly demonstrated, time diaries may be less affected by social desirability bias than survey questions when it comes to measuring behaviors that carry a strong normative inflection. In what follows, we use child-level data to investigate four outcomes. First, we examine a measure of expenditures on children’s organized extracurricular activ- ities, in order to determine to what degree factors other than financial resources come into play. More specifically, we assess the role that work and labor-market conditions, neighborhood characteristics, and other aspects of families’ circum- stances may play once income and wealth have been controlled. We know of no previous research that has addressed this question. Second, we analyze a measure of participation in organized activities derived from a set of children’s time diaries to directly identify the factors associated with differences in participation. Third, in order to assess the impact of equalizing institutions, we analyze a revised ver- sion of the time diary measure that reflects only organized activities that take place at the school. The arguments of Bennett, Lutz, and Jayaram (2012) imply that class differences in participation should be substantially attenuated in this setting, since the effects of “objective” constraints are largely neutralized by the institution. We also attempt to address the significance of equalizing institutions by examining a time diary measure that captures only nonreligious activities that take place on the weekend—that is, a set of activities that are unaffected by these institutions. Here, the arguments of Bennett, Lutz, and Jayaram would lead us to anticipate that the impact of financial resources will be exacerbated. Of course, in any of this analysis, objective measures of, say, family income may themselves be acting as proxies for cultural orientations; thus, we consider our approach of looking for residual class differences after accounting for the effects of material circumstances to be a relatively conservative approach to estimating the impact of parental cultural frames. 484 Social Forces 94(2)

Data, Measures, and Methods The PSID is a nationally representative panel of US households begun in 1968; since 1997, interviews have been carried out biennially. The sample is designed to have an evolving structure: as children in PSID households eventually break away from the families of origin and form new households, these too become part of the sample. The PSID is widely recognized for its scrupulous measures of house- hold finances: the standard interview records information on multiple sources of possible income (salary and wages, business profits, investment income and prof- its, bequests, alimony, pensions, government programs, etc.), as well as extensive data on family wealth. The PSID also contains a number of measures pertaining to labor-market attachment and work conditions that are directly relevant to debates over organized activity participation. In 1997, a complementary data-collection effort, the CDS, was initiated, focus- ing on a subset of PSID families that had children under 13 years old. A follow-up wave, which we draw on, collected data during the 2002–2003 school year (when the focal children were between the ages of 5 and 18). These data include inter- views with each child’s primary caregiver, as well as a set of time diaries for each child gathered on one randomly selected weekday and one randomly selected weekend day. The time diaries provide a detailed list of each child’s activities over the course of two 24-hour periods, recording how long each activity lasted, who was present with the focal child, and where each took place. Each family “contributed” up to two children to the sample. Our analyses restrict the sample to children who were at least 6 years of age and less than 15 years of age at the time of the 2002–2003 CDS data collection since these are approximately the ages covered by the various ethnographic studies. We also restrict the sample to black and white children. This decision is motivated by the fact that, because of its sampling strategy, the PSID has lacked members of groups whose demographic pres- ence is largely or mainly due to post-1968 immigration; while a “new immigrant” arm was added in 1997, numbers remain relatively low for these groups.2 Our sample is further limited to children living either in two- families or in families headed by a single mother, as other family types are too rare to yield reliable param- eter estimates. For the same reason, we restrict our analyses to children attending either a public or a private school. Finally, because compliance on the part of the research subjects was slightly lower for the time diaries than the questionnaires, we make use of two differently sized subsamples (N = 1,207 and N = 1,370). Our analyses employ the child-level weights provided by PSID staff.

Outcome Measures We define extracurricular organized activities to be those that are voluntary and adult led, and which have a fixed schedule or require some kind of formal enroll- ment. The main types of organized activities are sports (both team and individ- ual), religious activities (services, youth groups, religious education, and choirs), membership in community groups (mainly Boy Scouts and Girl Scouts), and cul- tural education (music and art lessons). Schools also provide some extracurricu- lar activities (in addition to sports) in the form of “clubs” and the like. What Money Doesn't Buy 485

The highly detailed codes applied to the CDS time diaries enable us to identify the various types of organized activities and count the amount of time devoted to them.3 We are also able to identify time devoted to traveling to and from these activities, and we include this in our measure as well. The rationale of keeping travel time in the measures is that such time is part of the temporal investments people have to make and offers a more complete account of the impact of activi- ties on family life. That said, exclusion of travel time does not substantively alter the results. Thus, we are able to construct a measure of the overall time children devote to organized activities. In order to apply an intuitive unit of measure to this variable, we have followed Hofferth and Sandberg’s (2001) procedure for computing a weekly estimate: the total amount of time from the weekday diary is multiplied by five, the total amount of time from the weekend diary is multi- plied by two, and the results are summed. Our variables suffer from measurement error, since many organized activities do not take place every day of the week. We address this by including controls for the weekday and the weekend day on which diary collection took place for each child in our models. These controls will account for the uneven distribution of activities across days of the week.4 In addition to our measures of overall participation, we also analyze a mea- sure of participation in school-based activities. Our intention is to see if we can adduce evidence for the proposition, articulated by Bennett, Lutz, and Jayaram (2012), that in institutional settings which seek to systematically ameliorate the impact of family resources, the relation between social class and activity par- ticipation will be substantially attenuated. As noted, the diaries include a set of codes that capture time devoted to “Before/After School Clubs” (e.g., band, debate team). However, the data contain very few instances of these activities; moreover, it is impossible for us to distinguish between school-sponsored sports and other organized sports. We therefore attempt to pick up school-sponsored activities by using the location codes included in the time diaries. Thus, our measure counts all organized activities identified as having taken place “at school.” This measure is most likely an overestimate, since some activities may utilize school facilities (e.g., a football field) without being sponsored by the school. Nevertheless, it is the best approximation available with the data. For reasons described below, this variable is not amenable to analysis when expressed in terms of a weekly metric, and must instead be dichotomized. How- ever, we retain the weekly estimate for descriptive purposes due to its intuitive nature. We also attempt to examine the role of equalizing institutions through a com- plementary strategy. Specifically, we analyze nonreligious organized activities that take place on the weekend. The underlying premise is that focusing solely on these activities should minimize the impact of the equalizing intuitions (i.e., school and church) that are central to Bennett, Lutz, and Jayaram (2012). Conse- quently, their argument would lead us to expect that, for these activities, the rela- tion to material resources and, perhaps, to objective constraints should be especially strong. Here again, our measure is not amenable to analysis when expressed in a weekly metric, and we must instead dichotomize. The weekly ver- sion is again retained for descriptive purposes. 486 Social Forces 94(2)

Finally, we use information provided by each child’s primary caregiver (almost always the mother) on expenditures on extracurricular activities. Respondents were asked whether their children participated in various types of extracurricular activities; in the event of an affirmative answer, they were then asked, “during the last 12 months, how much money did it cost for [the child] to be involved in” each type of activity. The types of activities covered include “athletic or sports teams”; “lessons, such as music, dance, or drama”; and “groups or programs in the community” (with explanatory follow-up indicating that this “includes scouts, service, or hobby clubs”).5 These questions cover expenditures in the major types of extracurricular activities; indeed, the activities that are not ­covered—religious ones and non-sports school-sponsored ones—are precisely those that are expected to incur few or no costs on the parts of participants, due to the effect of equalizing institutions.

Predictors We contribute to debates over the factors structuring children’s participation in extracurricular activities by including a comprehensive set of predictors in our models meant to account more fully than previous analyses for the potential role of multiple types of “objective” constraints. Material resources are captured by two family-level variables: income and wealth. We measure “permanent” income by computing a five-year average (1998–2002) of annual income from all sources, with values adjusted by the Consumer Price Index for All Urban Consumers (CPI-U). Because only biennial measurements are available, we capture “permanent” wealth by computing an average of CPI-U adjusted net wealth in 1999, 2001, and 2003. The annual wealth measures we use include housing equity; exploratory analyses indicated that these correlated more strongly with the outcome variables than wealth mea- sures that exclude equity (Conley 1999, 2001). On the basis of pilot results and prior studies, we use natural logarithms in our analyses, computed with non- positive values of wealth set to 1. Of course, wealth may be negatively influenced by children’s extracurricular activity enrollment to the extent that fees for such activities drain family resources. Meanwhile, income (and, by extension, wealth) may be influenced by children’s activities to the extent that enrollment in such activities affects parental labor supply—that is, making it possible for parents to work more. What we estimate, then, is a reduced-form relationship that captures income/wealth effects on children’s activities, children’s activities effects on income/wealth, and the joint determination of these measures by unmeasured variables such as cultural orien- tation. That said, we do attempt to factor out some of these pathways by control- ling directly for labor-force status. As noted earlier, Chin and Phillips (2004) emphasize the importance of paren- tal work flexibility in facilitating children’s participation in organized activities. Therefore, we include measures of the typical number of hours each parent worked at his or her main job in 2002. On the basis of exploratory analyses, we code this categorically, with dummy variables for the upper two terciles (the low- est tercile serves as the reference category).6 We also include a continuous variable­ What Money Doesn't Buy 487 for each parent indicating how much overtime, in hours, he or she worked in 2002—potentially a source of significant time constraint. These measures are again log transformed, with .10 added to values in order to remove 0s. Following Chin and Phillips (2004), we further control for membership in two occupational groups: “professionals, technical, and kindred workers” (per the 1970 census occupation codes) and the self-employed (per responses to a PSID survey ques- tion concerning the main job). Both are measured as of 2001. We include these measures for each parent. Members of the former group, we suspect, may have greater flexibility than other workers, whereas members of the latter group may have less flexibility. In keeping with Lareau and Weininger (2008), we also include dummy variables for union membership on the part of each parent. Finally, we control for labor-market status with a single dummy variable indicating that the mother was unemployed and looking for work at some point in 2002; unfortu- nately, we have too few cases of paternal unemployment to develop a reliable parameter estimate. Educational attainment is measured with dummy variables indicating high school completion, some college or an associate’s degree, and bachelor’s degree or higher; high school dropout serves as the reference category. Following Lareau and Weininger’s (2008) demonstration that paternal education does not predict participation in organized activities, we have included measures only for the mother’s educational attainment. Chin and Phillips (2004) describe parental edu- cation as a key “resource” enabling children’s participation, although they do not specify exactly how it is assumed to do so.7 However, the key assumption in our research strategy is that this is not, in fact, the case. Parental education does not get “spent down” like income, wealth or, in fact, time. Indeed, maternal education is our key proxy for maternal cultural orientation. Decades of research have documented that formal schooling is a critical socializing agent and conveys cul- tural repertoires (e.g., Dreeben 1968; Gracey 1975). Of course, the educational system acts as a sorting mechanism along cultural lines as well—particularly higher education, given its noncompulsory nature. Thus, it seems to us that if these variables exhibit significant predictive power net of material resources in a well-specified model, it will be important to consider the question of the associ- ated mechanisms. We return to this issue in the discussion section. We also control for characteristics of the community in which each family lives. First, much of the ethnographic literature suggests that living in a neighborhood perceived as dangerous can be a motivation for enrolling children in organized activities (Lareau 2011; Bennett, Lutz, and Jayaram 2012). In order to account for this, we use three questions from the survey. Respondents were asked, “How would you rate your neighborhood as a place to raise children?” The ordinal choice set is coded into three dummy variables, one capturing those who selected “Very Good,” another those who selected “Good,” and the third those who selected either “Fair” or “Poor”; the reference category is made up of those who chose “Excellent.” A second question asks, “How safe is it to walk around alone in your neighborhood after dark?” We code responses into two dummy variables—­ one capturing those who chose “Fairly Safe,” and the other those who chose “Somewhat Dangerous” or “Extremely Dangerous”; the reference category includes those who selected “Completely Safe.” A third question provides a rough 488 Social Forces 94(2)

indication of neighborhood social capital by asking how likely it is that a neighbor­ would “do something if someone was trying to sell drugs to your children in plain sight.” We capture responses with a single dummy variable: those who indicated “Extremely Likely” form the reference category, with all others coded 1. Our models also account for the type of community each family lives in. Each PSID record contains a score on the Beale-Ross Urbanicity code. This is a county- level measure. We use a simplified version in which large metropolitan areas serve as the reference category. Separate indicators distinguish “fringe” (i.e., suburban) areas surrounding large metropolitan areas, mid-sized metropolitan areas, and all other types of communities (mostly towns and rural areas). Community type may be hypothesized to affect organized activity participation in at least two ways: through variation in the supply of opportunities and in the cost of opportunities. The data do not permit us to distinguish between these possibilities. Additionally, there are numerous other factors that may directly impact par- ticipation in organized activities. In particular, family composition is of obvious importance. We therefore include a dummy variable that identifies female-headed households and another that identifies families with a stepparent. We also include a continuous measure of the number of siblings under 18 the focal child has in the household. Beyond this, transportation may place constraints on children’s likelihood of participation. We therefore include dummy variables identifying families that own one vehicle and families that own two or more vehicles; no vehicle serves as the reference category. Finally, we also control for a number of factors that may confound associa- tions between the variables of interest. These include demographic characteris- tics, such as child’s age, mother’s age (which is highly correlated with father’s age), child’s gender (1 = male), and race (1 = black). We also include a simplified ordinal measure of the child’s overall health, as rated by the primary caregiver, with excellent as the reference category, and good and less than good indicated by dummy variables. We control for homeownership with a dummy variable coded as 1 if the family owned its own home (although note that any effects of this variable will be due to something other than equity value); another dummy variable captures the relatively small number of families that neither own nor rent their home (and who presumably live “doubled up” with others). School type is captured with a dummy variable indicating private school attendance.8 Addi- tionally, we include three dummies measuring region of the country: south, north- central, and west; the reference category is northeast. Finally, because most organized activities are not distributed randomly across days of the week, we employ four dummy variables to control for the day on which the weekday diary was collected (Monday is the reference category) and one dummy to control for the weekend day (Saturday is the reference category). These variables are used with the time diary outcomes, but not the expenditures measure.

Estimation Strategy All of the outcome variables we analyze are censored—that is, a substantial pro- portion of cases are massed on the limit value of 0. For expenditures on organized activities and total time devoted to organized activities, this can be addressed by What Money Doesn't Buy 489 using tobit regression to estimate model parameters (Breen 1996; Long 1997). The tobit estimator assumes a latent (i.e., unobserved) outcome variable, desig- nated y*, which is free to vary from negative to positive infinity. When this latent variable has a value at or above the censoring threshold, it is equal to the observed variable (designated y), but for cases in which it is below the censoring threshold, the observed variable is assumed to have taken the limit value (i.e., 0). A tobit coefficient represents the estimated linear effect of a unit change in the regressor on the latent variable (y*), holding covariates constant. However, the effect of a change in this regressor on the observed variable (y) is not linear. Therefore, in addition to the regression coefficients, we present the unconditional marginal effects of the regressors, evaluated at the means of all the predictors ([∂∂Ey|/xx]), for each tobit model. The marginal effects may be thought of as regression coef- ficients that have been weighted by the probability of being above the censoring threshold, given the values of the covariates. The other two outcomes we analyze—time devoted to organized activities at school and weekend time devoted to nonreligious organized activities—occur less frequently in the data, and as a result, the censoring is quite severe (the proportion of cases with any time is 10.3 percent for the former variable and 12.8 percent­ for the latter). In this situation, estimates from tobit equations would be dominated by the effect of the regressors on the probability of exceeding the censoring thresh- old (i.e., pr[y > 0]). In order to simplify, we therefore use the dichotomized versions of these variables—with cases for which any time in the relevant activities is observed coded as 1 and cases in which none is observed coded as 0—and estimate probit equations. Unfortunately, this makes the interpretation of the model param- eters slightly less intuitive. Coefficients from the equation analyzing activity par- ticipation at the school imply changes in the probability of participating on a randomly chosen weekday. Similarly, coefficients from the equation analyzing participation in nonreligious activities on the weekend imply changes in the prob- ability of participating on a randomly chosen weekend day. We report the esti- mated coefficients for these models, along with the marginal effects, again evaluated at the means of all the covariates ([∂=pr yx1| ])/∂x . Additionally, because many of the children in our data were sampled from the same families, we esti- mate robust standard errors with a clustering adjustment. We are aware that results may be sensitive to modeling specifications, so we have performed extensive sensitivity analyses to test the robustness of our find- ings. These include running models with multivariate imputation of missing val- ues; using OLS estimation in place of tobit/probit; and performing extreme bounds analyses (in which each regressor is excluded in turn to record the sensi- tivity of other parameters to the inclusion or exclusion of particular variables from the models). All of these show that the results described below are highly robust (at least within the PSID-CDS sample). The estimates from these alterna- tive approaches can be found in the online version of the paper.

Results Table 1 presents means and standard deviations for the variables used in the analysis. Because compliance was better with the questionnaires than with the 490 Social Forces 94(2)

Table 1. Means and Standard Deviations

Questionnaire Time diary sample sample Mean SD Mean SD Weekly time in OA – – 240.84 367.74 Weekly time in OA at school – – 50.69 173.70 Weekend time in nonrelig. OA – – 35.76 127.83 Annual expenditures on OA 305.57 637.14 – – Income and wealth Income (natural log) 10.71 0.84 10.73 0.83 Wealth (natural log) 9.32 3.89 9.38 3.85 Work and labor-market conditions Maternal work hours: Second tercile 0.47 0.50 0.47 0.50 Maternal work hours: Third tercile 0.20 0.40 0.19 0.40 Paternal work hours: Second tercile 0.29 0.45 0.35 0.48 Paternal work hours: Third tercile 0.33 0.47 0.28 0.45 Maternal union membership 0.09 0.29 0.09 0.29 Paternal union membership 0.10 0.30 0.11 0.31 Mother in professional occupation 0.17 0.38 0.18 0.38 Father in professional occupation 0.11 0.32 0.12 0.32 Mother self-employed 0.07 0.26 0.07 0.26 Father self-employed 0.08 0.27 0.07 0.26 Mother’s total overtime (natural log) –3.15 3.32 –3.17 3.28 Father’s total overtime (natural log) –3.30 3.26 –3.23 3.33 Mother unemployed in previous year 0.11 0.32 0.11 0.31 Maternal education High school degree 0.36 0.48 0.36 0.48 Some college/associate’s degree 0.31 0.46 0.31 0.46 Bachelor’s degree or higher 0.20 0.40 0.21 0.41 Community characteristics Neigh. very good 0.32 0.47 0.32 0.47 Neigh. good 0.21 0.41 0.21 0.40 Neigh. fair or poor 0.14 0.35 0.14 0.34 Neigh. fairly safe 0.55 0.50 0.55 0.50 Neigh. somewhat or extremely dangerous 0.15 0.36 0.15 0.35 Social capital (neighbor might not intervene) 0.36 0.48 0.36 0.48 Fringe county of large metro area 0.16 0.37 0.16 0.37 Mid-sized metro county 0.27 0.44 0.27 0.44 Other county 0.31 0.46 0.30 0.46 (Continued) What Money Doesn't Buy 491

Table 1. continued

Questionnaire Time diary sample sample Mean SD Mean SD Controls Black 0.44 0.50 0.43 0.50 Child’s age 10.47 2.57 10.44 2.54 Male 0.51 0.50 0.51 0.50 Number of siblings 1.37 0.95 1.36 0.96 Child’s health good 0.30 0.46 0.30 0.46 Child’s health less than good 0.14 0.35 0.14 0.35 Child attends private school 0.08 0.27 0.08 0.27 Single mother 0.34 0.47 0.32 0.47 Stepparent 0.08 0.26 0.07 0.26 Mother’s age 35.64 6.41 35.61 6.49 Own home 0.64 0.48 0.66 0.48 Other living arrangement 0.04 0.19 0.04 0.20 Own 1 vehicle 0.30 0.46 0.29 0.45 Own 2+ vehicles 0.59 0.49 0.60 0.49 North-central 0.26 0.44 0.24 0.43 South 0.45 0.50 0.45 0.50 West 0.15 0.36 0.17 0.37 Sunday diary collection – – 0.48 0.50 Tuesday diary collection – – 0.22 0.41 Wednesday diary collection – – 0.19 0.39 Thursday diary collection – – 0.20 0.40 Friday diary collection – – 0.20 0.40 N (children) 1,370 1,207 N (families) 1,030 916 time diaries, we have slightly different samples for the various outcome measures; hence, we provide two sets of descriptive statistics. However, there is little sub- stantive difference between them. As can be seen in column 1 of table 1, families spend an average of $306 (in 2002 dollars) per child per year. As column 2 indicates, our data imply that children between the ages of 6 and 14 devote an average of roughly four hours per week to organized activities during the school year. Just over a fifth of this occurs at a school, and about a half hour is devoted to nonreligious activities that take place on the weekend. The first two columns in table 2 address the previously unstudied question of expenditures on organized activities, measured in dollars per year. We would expect 492 Social Forces 94(2)

Table 2. Tobit Regressions of Annual Expenditures on Organized Activities and Total Weekly Time in Organized Activitiesa

OA expenditures Weekly time in OA B dy/dx B dy/dx Income and wealth Income (natural log) 252.14*** 151.66*** 6.81 3.49 Wealth (natural log) 2.77 1.67 19.79* 10.16* Work and labor-market conditions Maternal work hours: Second –29.23 –17.55 47.91 24.69 tercile Maternal work hours: Third –175.70* –100.87* –180.56* –85.43** tercile Paternal work hours: Second –154.98 –90.89 –22.67 –11.59 tercile Paternal work hours: Third –132.07 –78.75 3.77 1.93 tercile Maternal union membership –81.95 –47.97 56.38 29.79 Paternal union membership 230.07+ 148.34+ 157.65* 87.43* Mother in professional 20.89 12.63 –5.93 –3.04 occupation Father in professional 10.62 6.41 58.50 30.78 occupation Mother self-employed 331.52* 221.23* 118.20 64.70 Father self-employed –119.88 –69.28+ –171.89* –79.90* Mother’s total overtime –14.10 –8.48 21.32* 10.94* (natural log) Father’s total overtime –8.89 –5.35 –8.98 –4.61 (natural log) Mother unemployed in 318.51 212.25 193.24+ 110.33 previous year Maternal education High school degree 132.65 81.26 129.28 68.42 Some college/associate’s degree 317.70* 199.64* 244.98* 133.26+ Bachelor’s degree or higher 480.41** 314.43** 347.93* 197.22* Community characteristics Neigh. very good 17.82 10.74 –104.85 –52.55 Neigh. good –212.65** –120.38** –0.81 –0.41 Neigh. fair or poor –32.89 –19.56 –211.05+ –95.75+ Neigh. fairly safe –89.87 –54.28 55.81 28.52 Neigh. somewhat or 14.66 8.86 10.96 5.66 extremely dangerous (Continued) What Money Doesn't Buy 493

Table 2. continued

OA expenditures Weekly time in OA B dy/dx B dy/dx Social capital (neighbor might –167.36* –98.41** 57.11 29.66 not intervene) Fringe county of large metro 138.09 86.21 15.26 7.89 area Mid-sized metro county –225.53* –128.89** 12.82 6.61 Other county –199.02* –117.83* 158.63* 83.24* Controls Black –325.79*** –178.68*** –148.76 –71.44+ Child’s age 29.89* 17.98* 22.43* 11.51* Male –203.32*** –122.58*** –119.08** –61.28** Number of siblings –44.02 –26.48 86.84** 44.57** Child’s health good 14.39 8.68 –2.98 –1.53 Child’s health less than good –298.86** –162.03** –131.71 –62.82 Child attends private school 90.22 55.90 242.93** 140.87** Single mother –136.90 –79.85 –140.70 –68.38 Stepparent –35.47 –21.06 –516.52*** –186.86*** Mother’s age 1.71 1.03 0.85 0.44 Own home 217.25* 124.95** –75.24 –39.58 Other living arrangement 70.40 43.53 –9.13 –4.66 Own 1 vehicle 138.79 86.04 –215.89 –102.01 Own 2+ vehicles –30.96 –18.73 –164.61 –88.58 North-central –72.14 –42.75 51.48 26.90 South 67.35 40.88 206.95* 111.20* West 31.81 19.28 39.93 20.81 Sunday diary collection – – 214.91*** 110.33*** Tuesday diary collection – – 35.56 18.50 Wednesday diary collection – – 59.90 31.53 Thursday diary collection – – 42.21 22.01 Friday diary collection – – –162.74* –77.76* Constant –2884.87*** – –734.85 Log sigma 6.64*** 6.34*** Left-censored cases 517 664 N 1,370 1,207 ***p < .001 **p < .01 *p < .05 +p < .10 aT-tests based on robust standard errors adjusted for the clustering of children in families. Available from the authors on request. 494 Social Forces 94(2)

to find, obviously, that expenditures are highly sensitive to families’ material resources. But it may also be the case that labor-market and work conditions are consequential, since in some situations, constraints on participation will also lead to reductions in expenditures. Furthermore, this model is notable because expendi- tures are presumably unaffected by any equalizing institutions; consequently, one way to gauge the effect of these institutions will be to compare the patterns evident in this equation with those for the various measures of participation. The results demonstrate that material resources do indeed have a direct impact on these expenditures, with income exhibiting a large, positive effect. We find significant effects for some of the work and labor-market variables. Maternal work hours, for example, are negatively related to expenditures, suggesting that these may act as constraints on participation (and therefore, expenditures). How- ever, there is no indication that a professional occupation affects expenditures, and contrary to expectations, maternal self-employment has a statistically sig- nificant, positive effect. Among the community variables, we find that those living in mid-sized metropolitan areas and small towns and rural areas spend signifi- cantly less than their counterparts in large metropolitan areas. Also contrary to expectations, those living in neighborhoods with low levels of social capital spend substantially less than those who live in neighborhoods with high levels. Perhaps the most dramatic effect, beyond that associated with income, is the one for maternal education. Indeed, the marginals indicate a gap of $314 between the highest and lowest levels of attainment. In light of the fact that the overall mean for this variable is $306, this is a substantial effect. This result raises interesting questions concerning the underlying mechanisms, given that high-quality measures of material resources, work and labor-market conditions, demographic characteristics, and social context are included in the equation. Many of the control variables are also important predictors of expenditures. Among demographic characteristics, race, age, and gender are all significant. The families of children in less than excellent health spend less than others, while the families of those who attend private schools spend more than those of children who attend public schools. For reasons that are not entirely clear, homeowner- ship exhibits a large, positive effect. The second two columns of table 2 present results for our model overall of participation in organized activities. This model uses time diary data, with the outcome measured in minutes per week. The results indicate that “objective” fac- tors undoubtedly influence children’s participation. The tobit coefficient for the wealth variable has a significant direct effect on participation. Also consistent with arguments about the importance of constraints, we see a significant effect for the maternal work hours variable, with children whose mothers are in the highest tercile exhibiting decreased participation relative to those whose mothers are in the lowest. Furthermore, self-employment—at least on the part of the father—has a substantial negative effect on participation; indeed, the marginal effect (evaluated at the means of the covariates) of –80 minutes is quite substan- tial, given that the overall sample mean is approximately four hours per week of participation. Contrary to expectations, however, employment in a professional occupation is again non-significant once other variables are controlled. More- over, maternal overtime appears to have a positive effect. What Money Doesn't Buy 495

Also noteworthy in this equation is the effect of maternal education. The mar- ginal effects (again, computed at the means of the covariates) increase by more than an hour for each degree category, and are particularly large for the two categories of college attenders, at 133 minutes and 197 minutes per week. Again, we consider the meaning of this effect in the discussion section. Some community characteristics have at least borderline significant effects. Those who ascribe the lowest quality rating to their neighborhood exhibit reduced participation. Those who live in small towns and rural areas exhibit modestly elevated participation, despite their lower expenditures—a pair of effects that may be suggestive of pricing differences. Among the control variables, we see that—consistent with their families’ elevated expenditures—students who attend private school devote substantially more time to organized activities than those who attend public schools. Additionally, the effect of living with a steppar- ent is notably large, highly significant, and negatively signed. The child’s age and gender remain significant, and the number of siblings is positively related to par- ticipation. Children living in the south exhibit higher participation than those living in the northeast. The first two columns of table 3 present the results of the probit regression of organized activities that take place at school. (In order to economize on space, coefficients for the control variables are not included in the table). The key ques- tion is whether we observe a substantial reduction of the effects of material resources and other sources of objective constraints when we restrict attention to activities situated within an equalizing institution. Contrary to expectations, however, there is a statistically significant, positive effect for the wealth variable here. Moreover, just as in the previous equation, we observe a negative effect at the highest level of maternal work hours. Additionally, we again observe substan- tial increases in the probability of participating as maternal education increases, with children whose mothers have a bachelor’s degree or higher 18.7 percent more likely to participate in a school-based organized activity on a given week- day than those whose mothers did not complete high school (with covariates held at their means). Thus, these results suggest that, in large measure, the same factors that are associated with participation in general also correlate with participation in the school setting. To be sure, there are notable differences between this model and the previous one. In particular, paternal self-employment is not significant in the equation for school participation, implying that provision of activities by schools may offset at least certain scheduling constraints stemming from work. Moreover, the mea- sure of paternal work hours becomes significant here, but with a positive sign: children whose fathers are in either of the higher terciles on work hours exhibit a substantially increased probability in the likelihood of school participation. While this might plausibly be construed as a sort of childcare arrangement, we find it too ambiguous to interpret, especially in light of the contrary (albeit smaller) effect associated with maternal hours. In the second two columns of table 3, we attempt to approach the same ques- tion from the opposite angle—that is, we restrict our analysis to nonreligious activities that take place on weekends, and which can therefore be assumed to be largely exempted from the impact of equalizing institutions. Thus, we might 496 Social Forces 94(2)

Table 3. Probit Regressions of Organized Activities at School and Weekend (Nonreligious) Organized Activitiesa,b

OA at school Weekend (nonrelig.) OA B dy/dx B dy/dx Income and wealth Income (natural log) –0.174 –0.024 0.106 0.019 Wealth (natural log) 0.064* 0.009* 0.053* 0.010* Work and labor-market conditions Maternal work hours: Second tercile –0.002 0.000 –0.112 –0.020 Maternal work hours: Third tercile –0.390+ –0.044* –0.469* –0.070** Paternal work hours: Second tercile 0.921*** 0.152** 0.087 0.016 Paternal work hours: Third tercile 0.738** 0.118** –0.018 –0.003 Maternal union membership 0.064 0.009 0.088 0.016 Paternal union membership 0.247 0.039 –0.101 –0.017 Mother in professional occupation –0.255 –0.031 –0.160 –0.027 Father in professional occupation 0.262 0.040 0.057 0.011 Mother self-employed 0.252 0.040 –0.176 –0.029 Father self-employed –0.088 –0.011 –0.532* –0.072** Mother’s total overtime (natural log) 0.049* 0.007* 0.007 0.001 Father’s total overtime (natural log) –0.032 –0.004 0.010 0.002 Mother unemployed in previous year 0.348 0.059 0.156 0.030 Maternal education High school degree 0.001 0.000 0.292 0.056 Some college/associate’s degree 0.466 0.072 0.407+ 0.080 Bachelor’s degree or higher 0.992** 0.187* 0.973** 0.226** Community characteristics Neigh. very good 0.355* 0.053* 0.019 0.003 Neigh. good 0.131 0.019 0.160 0.031 Neigh. fair or poor 0.020 0.003 –0.266 –0.042 Neigh. fairly safe 0.229 0.031 0.058 0.010 Neigh. somewhat or extremely –0.120 –0.015 0.520+ 0.119 dangerous Social capital (neighbor might not –0.222 –0.029 –0.012 –0.002 intervene) Fringe county of large metro area –0.471* –0.051** –0.196 –0.032 Mid-sized metro county –0.096 –0.013 0.072 0.013 Other county 0.053 0.007 0.044 0.008 Constant –1.871 –3.773* N 1207 1207 ***p < .001 **p < .01 *p < .05 +p < .10 aT-tests based on robust standard errors adjusted for the clustering of children in families. Available from the authors on request. bBoth models include all controls from the Expenditures equation in table 2. The School model also includes dummies for the weekday diary-collection day, while the Weekend model also includes a dummy for Sunday diary collection. What Money Doesn't Buy 497 anticipate that material resources would have an especially powerful effect in this domain, while work conditions would have minimal effects (since most people do not work on weekends). However, the results are remarkably similar to those from the previous two models. Familial wealth is once again significant and pos- itively signed (with an effect roughly comparable to the one it had in the school equation). Children whose mothers are in the highest tercile of the work hours variable continue to exhibit a lower likelihood of participation than those whose mothers are in the lowest tercile. And a large education effect is again evident, with the children of college-educated mothers (i.e., bachelor’s or higher) being 22.6 percent more likely to participate on a given weekend day than children whose mothers did not complete high school. To be sure, the importance of pater- nal work conditions shifts in the weekend equation, with work hours no longer significant, but self-employment again exhibits a large, negative effect, presum- ably because the self-employed are especially likely to work on weekends. The results for this equation may be affected by substitution effects—that is, because a large proportion of religious activities take place on weekends, some families may cut back on “secular” activities (the only ones counted in our out- come measure) in order to facilitate religious participation; as a result, the param- eter estimates we report could be influenced by uncontrolled differences in religiosity across families. However, our model does include a control for Sunday diary collection—the weekend day on which the vast majority of religious activi- ties take place.9 Furthermore, inclusion of an ordinal questionnaire item measur- ing the importance of religion to the primary caregiver has little effect on the other parameters in the model (results not shown).

Discussion Children’s participation in extracurricular organized activities has been increas- ingly recognized as a phenomenon that holds potentially significant implications for their future socioeconomic trajectories. Indeed, a substantial literature argues that participation in organized activities can provide children with subtle micro- interactional skills (or “capital”) that may have tangible payoffs in educational and occupational settings. Moreover, recent research suggests that participation in various organized activities is associated with both the college matching pro- cess and recruitment to certain types of elite occupations. In light of these find- ings, the class gradient in children’s participation becomes an important issue. Why is it that middle-class children are so much more likely to participate than their working-class and poor counterparts? On the one hand, Lareau (2011) has attributed substantial weight to cultural factors, arguing that class-specific cul- tural orientations toward childrearing lead parents to differentially structure their children’s out-of-school time, so that, on average, middle-class children exhibit substantially higher participation than their working-class or poor peers. On the other hand, Chin and Phillips (2004) and Bennett, Lutz, and Jayaram (2012) assert the existence of cross-class cultural continuity; observed variations in participation, they argue, derive largely from differences in the material resources parents can muster and the objective constraints that impede their chil- dren’s participation. 498 Social Forces 94(2)

To date, most of the researchers involved in this debate have drawn on ethno- graphic data. We have instead deployed a relatively large, nationally representa- tive data set. Our analytic strategy has been to pursue an implication of the “materialist” arguments: as we read them, these studies imply that if material resources and objective constraints can be accounted for—either directly, or by restricting the empirical focus to institutional settings in which their effects are systematically ameliorated—then class differences in participation should be sharply attenuated or disappear altogether. We have also attempted to contribute to the emerging quantitative literature on organized activity participation, in which many of the factors that the ethnographic literature identifies as potentially important are not incorporated. Taken as a whole, our results clearly indicate that material resources and objec- tive constraints matter, at least in certain respects. This is most evident with the wealth and income variables, which consistently exhibit direct effects on our vari- ous measures of participation and on our measure of expenditures. Organized extracurricular activities usually cost money, and put simply, greater familial wealth or income facilitates both participation in and expenditures on these activ- ities. The importance of constraints is clear, as well, in the consistent impact that maternal work hours have on participation in organized activities: in keeping with the findings of Lareau and Weininger (2008), our results indicate that, ceteris pari- bus, children whose mothers spend the most time in employment exhibit reduced participation. Occupational effects are also apparent, although we do not observe the positive effect for professionals hypothesized by Chin and Phillips (2004); instead, we find evidence that children whose fathers are self-employed have lower participation than those whose fathers work in other types of jobs, net of other factors. Contextual effects are not consistently apparent, although in some models there are indications that living in a neighborhood perceived as being of low qual- ity or having low social capital leads to lower participation and lower expendi- tures, possibly due to a reduction in the opportunities they provide. Although we can affirm the importance of resources and constraints, two issues remain outstanding. First, we are unable to adduce any substantial evi- dence that equalizing institutions ameliorate social class differences in participa- tion. Second, the large net effects associated with maternal education in all of our models require interpretation. Our data indicate that relatively little organized activity participation takes place in the school setting, at least for children between the ages of 6 and 14. Moreover, the parameters from our models of in-school participation and week- end (nonreligious) participation, as presented in table 3, were remarkably similar. Together, these results would seem to suggest that, net of other factors, schools do not fulfill the dramatic equalizing function ascribed to them by Bennett, Lutz, and Jayaram (2012). This is consistent with McNeal’s analysis of the predictors of high school sophomores’ participation in school-sponsored extracurricular activ- ities, which reported that “higher SES students are significantly more likely to participate in each category of extracurricular activity [studied] except vocational activities” (McNeal 1998, 188). The other major finding to emerge from our analyses is the consistent effects of maternal education. Maternal education was consistently associated with large What Money Doesn't Buy 499 increases in participation or expenditures, raising the question of what the under- lying mechanism is. Why is it that children of highly educated mothers are so much more likely to participate in organized activities than children whose moth- ers have less education? Why is it that, at equivalent levels of wealth and income, families in which the mother is highly educated devote so much more money to children’s activities? To be sure, considerations of the mechanisms that bring about these effects are inevitably speculative. That said, we see various possibili- ties. First, we would argue that a cultural reading is one viable interpretation. This would entail the conjecture that institutions of higher education tend to select for and/or instill dispositions of the type described by Lareau (2011) as “concerted cultivation”—that is, a generalized assumption that children’s nascent skills and talents will be realized only if assiduously cultivated by adults. One appealing feature of this interpretation is that it implies the kind of consistency across contexts reported in table 3. Moreover, we note that other researchers (e.g., Cheadle 2009), when confronted with large education effects, have not hesitated to invoke the possibility of a cultural mechanism. Nevertheless, we also recognize that other interpretations are plausible. For example, maternal education may be correlated with uncontrolled variation in the characteristics of families’ personal networks. If these characteristics lead families to behave differently due to processes of mutual influence with regard to activity participation, then our education coefficients would be inflated. However, while we acknowledge that this type of confounding may exist, we find it implau- sible that it could account for the large educational effects we document. And, of course, personal network formation itself does not occur independently of culture (Lizardo 2006). That is, any network effect may, in fact, be a mediating factor rather than a competing one. Another plausible interpretation can be found in the work of Ramey and Ramey (2010), who observe that the number of seats in elite colleges and univer- sities has remained largely unchanged over the past two decades, while succeed- ing cohorts of high school graduates have increased in size and the wage premium has risen, leading to intensified competition for admission. In this context, the middle-class affinity for organized activities is part of a larger strategy of in children’s development that seeks to better their eventual college placement. By assumption, the “productivity” of this investment is steeply tied to the parents’ own educational attainment, thus suggesting that this factor—rather than income or wealth—is the key to explaining middle-class children’s immer- sion in extracurriculars. While this model undoubtedly has prima facie plausibility, we would note that it remains underdeveloped in important respects. In particular, as Rivera’s (2011) findings imply, the (occupational, and by extension educational) “payoff” to extracurriculars may vary across particular activities. Thus, at the micro level, the story that Ramey and Ramey (2010) tell would lead us to anticipate that when it comes to selecting activities, greater parental educational attainment would be associated with greater subjective concern about the expected educa- tional and/or occupational “profits” of different activities. However, in our ­reading, the ­qualitative literature to date is at best mixed on whether this type of 500 Social Forces 94(2)

consideration carries a great deal of subjective significance ex ante with highly educated parents.10 Finally, it could simply be the case that maternal education picks up unmea- sured differences in cognitive and non-cognitive skills, which are, in turn, passed on to offspring genetically (Conley et al. 2015). In the parental genera- tion, such skills may be proxied by educational attainment, and in the filial generation, they could be proxied by children’s activities. Indeed, a growing literature suggests that children are not passive recipients of parental socializa- tion but instead influence parents’ outlooks and behaviors in domains as dispa- rate as politics (Washington 2008; Oswald and Powdthavee 2010; Conley and Rauscher 2013) and parent-child activities (Cardona and Diewald 2014). It may be the case that high-energy, high-achieving children evince greater invest- ments from their parents in the form of extracurricular activities. In turn, it is possible that such children are disproportionately related to highly educated mothers. Yet, if shared genetic disposition was the only—or even main—mech- anism accounting for the observed relationship, we would also see it for father’s education, since genetic transmission would take place equally along the pater- nal line. However, the existing literature (Lareau and Weininger 2008) suggests that paternal education is substantially less predictive of children’s participa- tion than maternal education. Quantitative analysis of secondary data is, regrettably, unlikely to enable us to fully adjudicate among these possibilities (as well as any others). For this reason, further ethnographic research into the causes and consequences of the stratifica- tion of children’s participation in organized activities is imperative. Only by bringing to bear all the tools in the social scientific arsenal will we be able to disentangle the roles that social structure and culture play in setting children on their life trajectories.

Notes 1. The authors involved in these disputes appear to understand “culture” somewhat dif- ferently. However, space does not permit us to examine these differences here. 2. Note that many of the qualitative studies that motivate the current analysis focus on black-white comparisons, as well. 3. There are a few activities in the time diaries that are ambiguous with regard to their status as “organized”—for example, horseback riding may take place in either a for- mal setting or an informal one. We have excluded these ambiguous codes from our measure of time in organized activities, with one exception—playing a musical instru- ment, which, we assume, is almost always carried out in preparation for a lesson among children 6 to 14 years old. 4. We also ran models with controls for the season or month in which diary collection took place. However, these were not significant. 5. The reference to “teams” in the questions on sport unfortunately implies that partici- pation in “individualized” sports (such as karate) may not have been picked up. While it is impossible to know how severe any underreporting due to this may be, we are inclined to think that it is not too serious, as it is clear from responses to a follow-up question that some parents whose children participated in non-team sports used the What Money Doesn't Buy 501

question on “lessons” to register this participation (and therefore, presumably, to also provide information on the associated expenditures, ascribing them to “lessons”). 6. For families headed by a single mother, the father’s work hours are coded as 0, and the family is therefore included in the lowest tercile; a separate variable serves to distinguish these families from those in which the father is present but does not work or has low work hours. 7. Chin and Phillips (2004) refer in places to differences in parents’ “knowledge” of opportunities for activity participation and about “their children’s capabilities.” However, it is unclear whether (and if so, how) these forms of knowledge are assumed to be connected to their education. Further, such knowledge can also be considered part of the cultural frame that we seek to measure. 8. School type may be endogenous with respect to material resources and community type. We have therefore estimated nested versions of each of our models, with the second equation adding school type. However, the results of these analyses suggest that the relevant indirect effects are quite small. Therefore, in order to economize on space, we present only the full models, treating school type as a control variable. 9. The sample contains only five children whose primary caregiver self-identifies as ­Jewish. 10. In response to a query from a reviewer, we investigated whether the patterns we iden- tified in our analyses of participation varied when we looked at particular activities— that is, cultural lessons, sports, and so forth. However, the data are simply not ideal for answering this question. The number of children in the data set who participated in particular types of activity is relatively low. As a result, in most models the coeffi- cients of the key variables (i.e., income, wealth, maternal education, etc.) fail to attain statistical significance. (The exception is the equation for sports: here, we observe moderate or large effects for maternal education and numerous variables relating to the mother’s work situation). However, there is nothing in these results that is incon- sistent with our analyses of measures that aggregate across activity types.

About the Authors Elliot B. Weininger is Associate Professor of at SUNY College at Brockport. He has published on the theoretical foundations of the concept of social class, as well as cultural and social capital. More recent work has addressed the ways that parents select schools for their children in districts with school choice programs and the role of schooling considerations in families’ residential mobility. Annette Lareau is the Stanley I. Sheerr Professor in the Department of Sociol- ogy at the University of Pennsylvania. She is the author of Unequal Childhoods: Class, Race, and Family Life, as well as Home Advantage. She is also the co-editor (with Dalton Conley) of Social Class: How Does It Work? and (with Kimberly Goyette) of Choosing Homes, Choosing Schools. Lareau is a Past President of the American Sociological Association. Dalton Conley is University Professor at New York University. He holds appointments in NYU’s Sociology Department, School of Medicine, and the Wag- ner School of Public Service. He is also a Research Associate at the National Bureau of Economic Research (NBER). In a pro bono capacity, he is Dean of Arts and Sciences for the University of the People—a tuition-free institution commit- ted to expanding access to higher education. 502 Social Forces 94(2)

References Adler, Patricia A., and Peter Adler. 1994. “Social Reproduction and the Corporate Other: The Institutionaliza- tion of Afterschool Activities.” Sociological Quarterly 35(2):309–28. Bennett, Pamela R., Amy C. Lutz, and Lakshmi Jayaram. 2012. “Beyond the Schoolyard: The Role of Parent- ing Logics, Financial Resources, and Social Institutions in the Social Class Gap in Structured Activity Participation.” Sociology of Education 85(2):131–57. Bodovski, Katerina, and George Farkas. 2008. “‘Concerted Cultivation’ and Unequal Achievement in Elemen- tary School.” Social Science Research 37(3):903–19. Breen, Richard. 1996. Regression Models: Censored, Sample Selected, or Truncated Data. Thousand Oaks, CA: Sage Publications. Cardona, Andrés, and Martin Diewald. 2014. “Opening the Black Box of Primary Effects: Relative Risk Aver- sion and Maternal Time Investments in Preschool Children.” SFB 882 Working Paper Series 36. DFG Research Center (SFB) 882, From Heterogeneities to Inequalities, Bielefeld. Cheadle, Jacob E. 2008. “Educational Investment, Family Context, and Children’s Math and Reading Growth from Kindergarten through the Third Grade.” Sociology of Education 81(1):1–31. ______. 2009. “Parent Educational Investment and Children’s General Knowledge Development.” Social Science Research 38(2):477–91. Chin, Tiffani, and Meredith Phillips. 2004. “Social Reproduction and Child-Rearing Practices: Social Class, Children’s Agency, and the Summer Activity Gap.” Sociology of Education 77(3):185–210. Conley, Dalton. 1999. Being Black, Living in the Red: Race, Wealth, and Social Policy in America. Berkeley and Los Angeles: University of California Press. ______. 2001. “Decomposing the Black-White Wealth Gap: The Role of Parental Resources, Inheritance, and Investment Dynamics.” Sociological Inquiry 71:39–66. Conley, Dalton, and Emily Rauscher. 2013. “The Effects of Daughters on Partisanship and Social Attitudes toward Women.” Sociological Forum 28:700–18. Conley, Dalton, Benjamin W. Domingue, David Cesarini, Christopher Dawes, Cornelius A. Rietveld, and Jason D. Boardman. 2015. “Is the Effect of Parental Education on Offspring Biased or Moderated by Genotype?” Sociological Science 2:82–105. Covay, Elizabeth, and William Carbonaro. 2010. “After the Bell: Participation in Extracurricular Activities, Classroom Behavior, and Academic Achievement.” Sociology of Education 83(1):20–45. Dreeben, Robert. 1968. On What Is Learned in School. Reading, MA: Addison-Wesley. Dumais, Susan A. 2006. “Elementary School Students’ Extracurricular Activities: The Effects of Participation on Achievement and Teacher’s Evaluations.” Sociological Spectrum 26(2):117–47. ______. 2009. “Cohort and Gender Differences in Extracurricular Participation: The Relationship between Activities, Math Achievement, and College Expectations.” Sociological Spectrum 29(1):72–110. Gracey, H. L. 1975. “Learning the Student Role: Kindergarten as Academic Boot Camp.” In The Sociology of Education: A Source Book, edited by H. R. Stub, 82–95. Homewood, IL: Dorsey Press. Hofferth, Sandra L. 1999. “Family Reading to Young Children: Social Desirability and Cultural Biases in Reporting.” Paper presented at the Workshop on Time-Use Measurement and Research, Washington, DC, USA. National Research Council: Committee on National Statistics. Hofferth, Sandra L., and John F. Sandberg. 2001. “How American Children Spend Their Time.” Journal of and Family 63(2):295–308. Jencks, Christopher. 1972. Inequality: A Reassessment of the Effect of Family and Schooling. New York: Basic Books. Kaufman, Jason, and Jay Gabler. 2004. “Cultural Capital and the Extracurricular Activities of Girls and Boys in the College Attainment Process.” Poetics 32(2):145–68. Lamont, Michèle, Mario Luis Small, and David J. Harding. 2010. “Introduction: Reconsidering Culture and Poverty.” Annals of the American Academy of Political and Social Science 629:6–27. What Money Doesn't Buy 503

Lareau, Annette. 2002. “Invisible Inequality: Social Class and Childrearing in Black Families and White Families.” American Sociological Review 67(5):747–76. ______. 2011. Unequal Childhoods: Class, Race, and Family Life, Second Edition with an Update a Decade Later. 2nd ed. Oakland: University of California Press. Lareau, Annette, and Elliot B. Weininger. 2008. “Time, Work, and Family Life: Reconceptualizing Gendered Time Patterns through the Case of Children’s Organized Activities.” Sociological Forum 23(3):419–54. Linver, Miriam R., Jodie L. Roth, and Jeanne Brooks-Gunn. 2009. “Patterns of Adolescents’ Participation in Organized Activities: Are Sports Best When Combined with Other Activities?” 45(2):354–67. Lizardo, Omar. 2006. “How Cultural Tastes Shape Personal Networks.” American Sociological Review 71(5):778–807. Long, J. Scott. 1997. Regression Models for Categorical and Limited Dependent Variables. Thousand Oaks, CA: Sage Publications. Mayer, Susan E. 1998. What Money Can’t Buy: Family Income and Children’s Life Chances. Cambridge, MA: Harvard University Press. McNeal, Ralph B Jr. 1998. “High School Extracurricular Activities: Closed Structures and Stratifying Patterns­ of Participation.” Journal of Educational Research 91(3):183–91. Oswald, Andrew, and Nattavudh Powdthavee. 2010. “Daughters and Left-Wing Voting.” Review of Econom- ics and Statistics 92(2):213–27. Ramey, Garey, and Valerie A. Ramey. 2010. “The Rug Rat Race.” Brookings Papers on Economic Activity (Spring):129–76. Rivera, Lauren. 2011. “Ivies, Extracurriculars, and Exclusion: Elite Employers’ Use of Educational Credentials.” Research in Social Stratification and Mobility 29:71–90. Stevens, Mitchell L. 2007. Creating a Class: College Admissions and the Education of Elites. Cambridge, MA: Harvard University Press. Washington, Ebonya L. 2008. “Female Socialization: How Daughters Affect Their Legislator Fathers.” American Economic Review 98(1):311–32.